Essays on Macroeconomic Dynamics, 
Credit Intermediation and Financial 
Stability 
 
Umang Rawat 
 
 
Christ’s College, University of Cambridge  
Faculty of Economics 
July 2017 
 
 
This dissertation is submitted for the degree of 
Doctor of Philosophy 
 
 
 
Contents
Preface 1
Summary 2
Introduction 4
1 Household and Firm Leverage in Business Cycle Model 15
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2 Literature: Theory and Evidence . . . . . . . . . . . . . 19
1.3 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3.1 Patient Households (P) . . . . . . . . . . . . . . . . . . . 23
1.3.2 Impatient Households (I) . . . . . . . . . . . . . . . . . . 26
1.3.3 Entrepreneurs . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.3.4 Firms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.3.5 Capital Goods Producer . . . . . . . . . . . . . . . . . . . 33
1.3.6 Market Clearing and Competitive Equilibrium . . . . . 34
1.4 Understanding the effects of financing constraints
in the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.4.1 Impatient Households . . . . . . . . . . . . . . . . . . . . . 35
1.4.2 Entrepreneurs . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.4.3 Interaction of households and firm credit constraints . 37
1.5 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.5.1 Methodology and Data . . . . . . . . . . . . . . . . . . . . 38
1.5.2 Calibrated Parameters and Prior Distribution . . . . . 39
1.5.3 Posterior Distributions . . . . . . . . . . . . . . . . . . . . 42
1.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.6.1 Model Comparison . . . . . . . . . . . . . . . . . . . . . . 42
1.6.2 Posterior Impulse Response Analysis . . . . . . . . . . . 44
1.6.3 Comparison of results from first and second order ap-
proximation . . . . . . . . . . . . . . . . . . . . . . . . . . 49
i
1.6.4 Robustness: Prior sensitivity . . . . . . . . . . . . . . . . 50
1.6.5 Variance Decomposition . . . . . . . . . . . . . . . . . . . 50
1.7 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . 52
Appendix 1.1: Data and Sources 64
Appendix 1.2: Model Equations 65
Appendix 1.3: Steady State 71
Appendix 1.4: Prior and Posterior 78
Appendix 1.5: Model Variables and Parameters 80
2 Securitization, Shadow Banking and Regulation 84
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.2 Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . 88
2.3 Institution, Entities and Instruments . . . . . . . . . . 93
2.4 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
2.4.1 Households . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.4.2 Entrepreneurs . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2.4.3 Commercial Banks . . . . . . . . . . . . . . . . . . . . . . 102
2.4.4 Shadow Banks . . . . . . . . . . . . . . . . . . . . . . . . . 108
2.4.5 Consumption good producer . . . . . . . . . . . . . . . . 111
2.4.6 Capital good producer . . . . . . . . . . . . . . . . . . . . 112
2.4.7 Deposit insurance agency . . . . . . . . . . . . . . . . . . 112
2.4.8 Competitive Equilibrium . . . . . . . . . . . . . . . . . . 113
2.5 Baseline Parameterization . . . . . . . . . . . . . . . . . . . 114
2.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
2.6.1 Total factor productivity shock: shock to the aggre-
gate productivity . . . . . . . . . . . . . . . . . . . . . . . 117
2.6.2 Bank risk shock: shock to the volatility of banks’ id-
iosyncratic risk . . . . . . . . . . . . . . . . . . . . . . . . . 117
ii
2.6.3 Robustness Checks . . . . . . . . . . . . . . . . . . . . . . 120
2.6.4 Government credit policy . . . . . . . . . . . . . . . . . . 121
2.6.5 Capital Regulation . . . . . . . . . . . . . . . . . . . . . . 124
2.7 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . 126
Appendix 2.1: Model Equations and Competitive Equilibrium 137
Appendix 2.2: Model Variables and Parameters 143
3 Can Low interest rates be harmful: An assessment of bank risk
taking channel in Asia-Pacific 147
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
3.2 Literature: Theory and Evidence . . . . . . . . . . . . . 150
3.3 Model Specification . . . . . . . . . . . . . . . . . . . . . . . . 157
3.4 Data Sample and Estimation Issues . . . . . . . . . . . 161
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
3.5.1 Annual Model . . . . . . . . . . . . . . . . . . . . . . . . . 168
3.5.2 Quarterly Model . . . . . . . . . . . . . . . . . . . . . . . . 173
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Appendix 3.1: Data, Sources and Transformation 187
List of Tables
1 Calibrated Parameters . . . . . . . . . . . . . . . . . . . . . . . . 39
2 Key Steady State Results . . . . . . . . . . . . . . . . . . . . . . 41
3 Prior and Posterior Distribution of the Structural Parameters 42
4 Prior and Posterior Distribution of the Shock Parameters . . 43
5 Model Comparison: Results from estimation . . . . . . . . . . 43
6 Posterior median of baseline model: Prior sensitivity . . . . . 51
7 Variance Decomposition (in percent) . . . . . . . . . . . . . . . 63
10 Calibrated Parameters . . . . . . . . . . . . . . . . . . . . . . . . 128
iii
13 Summary Statistics of the Variables used in Annual Regres-
sion, 2000-2011 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
14 Descriptive Statistics by Country (mean values), 2000 - 2011 177
15 Correlation Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
16 Results: Annual Model with Taylor Interest Rate Gap (ig1)
and Different Bank Risk Measures . . . . . . . . . . . . . . . . . 179
17 Results: Annual Model with Natural Interest Rate Gap (ig2)
and Different Bank Risk Measures . . . . . . . . . . . . . . . . . 180
18 Results: Annual Model with Taylor Interest Rate Gap (ig1)
- without ir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
19 Results: Annual Model with Natural Interest Rate Gap (ig2)
- without ir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
20 Results: Annual Model with Taylor Interest Rate Gap (ig1)
- restricted sample . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
21 Results: Annual Model with Natural Interest Rate Gap (ig2)
- restricted sample . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
22 Results: Quarterly Model with Idiosyncratic Bank Risk (br1) 185
List of Figures
1 Impulse responses to the estimated non-housing technology
shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2 Impulse responses to the estimated housing preference shock 55
3 Impulse responses to the estimated housing technology shock 56
4 Impulse responses to the estimated investment shock . . . . . 57
5 Impulse responses to the estimated risk shock . . . . . . . . . 58
6 Impulse responses to the estimated preference shock . . . . . 59
7 Impulse responses to the estimated labour supply shock . . . 60
8 Impulse response functions under first order v/s second order
approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
iv
9 Impulse response functions under first order v/s second order
approximation (cont.) . . . . . . . . . . . . . . . . . . . . . . . . . 62
10 TFP shock: One percent drop in TFP . . . . . . . . . . . . . . 129
11 Bank risk shock: 0.1 percentage point increase in the volatil-
ity of bank’s idiosyncratic risk . . . . . . . . . . . . . . . . . . . 130
12 Bank risk shock: Comparison with different size and persis-
tence of shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
13 First v/s second order approximation . . . . . . . . . . . . . . . 132
14 TFP shock: Baseline and Robustness scenario . . . . . . . . . 133
15 Risk shock: Baseline and Robustness scenario . . . . . . . . . 134
16 Risk shock: Response under baseline and active credit policy 135
17 Procyclical capital requirement: Response to a negative pro-
ductivity shock under Basel I and Basel II . . . . . . . . . . . . 136
v
Acknowledgments
It would not have been possible to write this doctoral thesis without the help and
support of the kind people around me, to only some of whom it is possible to give
particular mention here.
I would like to thank the Gates Trust for the financial support, particularly in
the award of the Gates Cambridge scholarship, that made this research possible.
I would like to express my gratitude to my supervisors Prof. Hashem Pesaran
and Dr. Pontus Rendahl for the useful comments, remarks and engagement through
the learning process of this doctoral thesis. I would also like to thank Dr. Hsiao
Chink Tang and Dr. Arief Ramayandi for their useful comments and guidance in
writing chapter three of this thesis.
Finally, I would like to thank my loved ones, my parents, wife and friends, who
have supported me throughout the process, both by keeping me harmonious and
helping me putting pieces together. I will be grateful forever for your love.
For any errors or inadequacies that may remain in this work, of course, the
responsibility is entirely my own.
vi
Preface
I hereby declare that this dissertation is the result of my own work and includes
nothing that is an outcome of work done in collaboration except where specifically
indicated in the text. It is not substantially the same as any other work that I have
submitted or will be submitting for a degree or diploma or other qualification at the
University of Cambridge or any other University or similar institution. This thesis
does not exceed the prescribed word limit of 60,000 words.
Umang Rawat
July 2017
1
Summary
This dissertation consists of three chapters. In the first chapter, we study the role
of financial frictions on the demand side of the economy. In particular, we study the
interaction between firm and household credit constraints over the business cycle.
We construct a real business cycle model with explicit modeling of price and quan-
tity side of housing. This allows us to include both firm and household financing
frictions. The model is estimated for the U.S economy using quarterly data on key
macroeconomic variables over the period 1970 - 2006. Household and firm financial
accelerators operate primarily through movement in house and capital prices respec-
tively. We find clear evidence of the operation of a financial accelerator mechanism,
whereby shocks to the economy are amplified most in the presence of both types
of frictions, as opposed to just firm or household frictions. Over the business cy-
cle, total factor productivity shocks in the non-housing sector explain about half of
the volatility of GDP and consumption. However, cyclical variations in housing in-
vestment and housing prices are predominantly explained by housing preference and
housing technology shocks. Finally, spillovers from household financing frictions are
mostly concentrated in consumption. However, they also affect business investment
via its impact on the demand for capital and consequently its price.
The second chapter focuses on financial frictions on the supply side. We study the
role of bank capital in the transmission of shocks to the economy. Given the evolu-
tionary change in the financial services industry and the growth of shadow banking in
the decades prior to the global recession, we characterize credit intermediation with
a heterogeneous banking sector comprised of traditional retail and shadow bank-
ing. We approach the shadow banking system from a regulation perspective wherein
commercial banks have incentives to transfer loans from on- to off-balance sheet to
gain regulatory relief. Since bank capital is costly, banks cover part of their funding
needs by loan sale in the secondary market. Furthermore, these transferred loans
are bundled together and converted into liquid asset backed securities. Commercial
banks’ effective return is subject to their monitoring effort, which is unobservable
and hence introduces a moral hazard problem in loan sale. This limits the amount of
2
loan sold in the secondary market. We find that loan sale and securitization enhances
credit intermediation in normal times and improves the resilience of the system to
productivity shocks. However, it also exposes the economy to shocks emerging in
the financial system. In response to financial market shocks, the government via its
backstop program, can ameliorate its impact on the economy. Finally, we compare
the model economy with Basel I and Basel II capital requirement and find that busi-
ness cycle fluctuations are amplified under Basel II regime. Furthermore, in response
to a negative productivity shock there is a transfer of loans from on to off balance
sheet under Basel II rules with procyclical capital constraints. This points towards
a need for countercyclical capital requirement as being implemented under Basel III
accord.
In the third chapter, we focus on the question of trade off between price and
financial stability goals for the conduct of monetary policy. The recent crisis has
generated renewed interest in Hayekian theory and Minsky’s instability hypothesis,
which claims that accommodative monetary policy can be harmful for an economy by
promoting excessive risk taking – the so called risk taking channel of monetary policy
transmission. Risk Taking Channel has been documented for the U.S and Euro area
and we investigate the presence of this in Asia. Using annual and quarterly data on
publicly listed banks in Asia, we find that when interest rates are too low - lower
than a benchmark - bank risk increases. Furthermore, there is also a case for greater
supervision and capital stringency to alleviate risk taking.
3
Introduction
This dissertation is on understanding macroeconomic propagation in the presence
of financial frictions. Following the global financial crisis of 2007, there has been a
renewed interest in understanding the demand and more specifically the supply side
of credit intermediation. Moreover, the macro-financial linkages and interconnect-
edness of macroeconomic and financial stability have come to the forefront. This
dissertation, through three main chapters, provides an insight into macroeconomic
dynamics in the presence of financial frictions on demand and supply side of credit
intermediation. It also explores the linkages between the real and the financial econ-
omy and the dual goals of macroeconomic and financial stability.
The recent financial turmoil has called for a need to understand the role of finan-
cial frictions in propagating shocks and to develop approaches to embed these fric-
tions in the macro models of central banks. In chapter 1, we study the macroeconomic
dynamics resulting from financial frictions on the demand side, more specifically on
households and firms. Existing literature looks at financial frictions on household and
firm side in isolation, thereby missing any interaction between household and firm
leverage in propagation of shocks. We construct a model with heterogeneous pro-
duction of housing and non-housing that allows for studying the dynamics between
residential and non-residential investment. This also allows us to analyze the nature
of interaction between firm and household financing constraints over the business
cycle.
In chapter 2, we focus on the supply side dynamics of credit intermediation.
Following the global recession of 2007, many papers have attempted to understand
the role played by the financial system in credit generation and growth. We add to
the existing literature by incorporating a heterogeneous banking sector, comprising of
traditional commercial banking and shadow banking. Shadow banking emerged as a
product of growth in innovation, financial technology and regulation in the decades
prior to the crisis. Our model provides a more realistic setup to understand the
macro-financial linkages between the financial sector and real economic activity. It
also allows us to analyze the link between macroeconomic stability, financial stability
4
and regulation.
Finally in chapter 3, we delve deeper into the question of tradeoff between macroe-
conomic and financial stability. In particular, we look at the link between monetary
policy and financial stability. Policymakers and researchers have questioned the rea-
sons behind the crisis, trying to provide some explanations on the forces behind the
fragility of the global financial system. Among the many potential causes of the cri-
sis, a contentious one amongst economists is the contribution of the accommodative
monetary policy to the build up of the crisis. Some economists argue that a pro-
longed period of extremely low interest rates and lax liquidity conditions encouraged
financial institutions to take on more risk i.e. the so-called ‘risk taking’ channel of
monetary policy transmission. In chapter 3, we contribute to the empirical literature
on the risk taking channel by analyzing the risk taking behavior of banks in Asian
economies. This also fills a notable gap in the existing literature that is mostly
confined to the developed western economies, particularly of the U.S and Europe.
Chapter 1 focuses on the role of demand side frictions in propagation of shocks and
macroeconomic volatility. Financial frictions on the firm side have been widely docu-
mented starting from Bernanke and Gertler (1989), Bernanke, Gertler and Gilchrist
(henceforth BGG) (1999), Kiyotaki and Moore (1997). More recently papers includ-
ing Iacoviello (2005), Iacoviello and Neri (2010) and Justiniano et. al. (2013) have
explored housing market dynamics and their spillovers to the macroeconomy. The
composition of households’ balance sheets and assets such as housing, play a signif-
icant role in shaping the transmission mechanism. In this chapter, we add to the
existing literature, by jointly studying the interaction between household and firm
debt dynamics in a business cycle model. Studying household and firm financial fric-
tions in isolation assumes that they are independent of each other. However, recent
literature suggests that household balance sheet shocks have a substantial effect on
aggregate economic activity and employment. For example, Mian and Sufi (2009,
2011) and Haltenhof (2014) document the role of household financing constraints on
aggregate economic activity and employment.
We begin with a multi-sector production economy with housing and non-housing
5
goods. On the supply side of the model, firms engage in joint production of housing
and non-housing goods. The non-housing sector produces consumption goods using
capital and labour and the housing sector produces new homes using capital, labour,
intermediate business input and land. Firms hire labour and purchase capital using
their net worth and by borrowing from the financial intermediary as in BGG (1999).
The demand side of the model embodies financial frictions for both household and
firms. Financial frictions for households appear in the form of borrowing based on
collateralisable assets (housing). For firms, our model generates endogenous external
finance premium as a function of the firms’ leverage.
We use quarterly data for the U.S on six key macroeconomic variables (consump-
tion, non-residential investment, residential investment, house prices, hours worked
(consumption sector) and hours worked (housing sector)) for the period 1970 – 2006
to estimate the model using Bayesian techniques.
We find merit in including both household and firm financial frictions in a general
equilibrium model. As compared to a model with only firm or household financial
frictions, joint analysis of both leads to a greater amplification of shocks to GDP,
consumption and investment. For example, the response of output and consumption
to a non-housing technology shock is more than twice as large in a model with both
kind of frictions as compared to a model with only one of the two frictions. Firms’
financial accelerator operates via the procyclical impact of capital prices on its net
worth. As the net worth increases with rising capital prices, the external finance
premium facing firms falls. Household financial accelerator on the other hand oper-
ates primarily through the movement in house prices, which affects their borrowing
capacity. Our model finds two unique ways in which firms’ and households’ financial
constraints interact with each other. First, is via the loan market equilibrium condi-
tion: as the financial constraint on households tighten, their demand for loans fall,
causing a downward pressure on interest rate. This reduces the cost of borrowing
external funds for the firms. This would suggest a negative relationship between firm
and household credit frictions as seen in Solomon (2010). The second effect operates
through the impact of house prices on housing demand and investment. As house
prices decrease (due to a negative housing demand shock for instance), the demand
6
for housing and housing investment decreases. This further reduces the demand for
housing capital and capital price. A decline in capital price worsens the net worth
position of firms and therefore their terms of credit. Thus, a fall in households’
borrowing capacity in this case also worsens firms credit condition. Overall in our
model, the second effect dominates, providing evidence to the claim that there is a
positive interaction between firm and household credit market constraints.
We also find very strong household wealth effects in our model, wherein shocks
to housing wealth (house prices) have a substantial effect on household non-housing
consumption. This is in line with the results of Mian and Sufi (2011) who find
that money extracted from increased home equity is not used to purchase new real
estate or pay down high credit card balance, suggesting that borrowed funds maybe
used for real outlays. Finally, we also look at the factors affecting cyclical variation
in aggregate variables at business cycle frequencies. We find that a productivity
shock in the non-housing sector accounts for about half of the variation in GDP
and consumption and about 15 percent of the variation in investment. Variations in
housing investment and prices are however primarily explained by housing technology
(supply) and housing preference (demand) shocks. Finally, financial shocks in the
form of investment and entrepreneur risk shock are prime candidate for explaining
variation in business investment.
The financial accelerator and the collateral constraint frameworks discussed in
chapter 1 assume that borrowers could obtain funds directly from lenders without
much of a role for the financial intermediaries. Introducing a banking sector into
DSGE models provides an additional avenue for incorporating financial frictions
specifically linked to the costs of intermediation, balance sheets of financial inter-
mediaries, and their risk management behavior. Furthermore, credit channels and
financial intermediaries embedded in macroeconomic models help explain dynamics
of the business cycle capturing the inherent procyclicality of the financial system.
The banking sector has been ignored in most DSGE models developed in the lit-
erature, except for some recent papers. In these, the analysis is mostly focused on
commercial/retail banking without any emphasis on non bank financial intermedia-
tion, comprising of the so-called shadow banking, which grew in prominence prior to
7
the great recession. The principal motivation for chapter 2 is therefore, to explicitly
model the behavior of financial intermediaries, whereby we allow for a heterogeneous
setup with both commercial and shadow banking.
We begin our analysis by understanding the reasons for the growth of shadow
banking prior to the financial crisis. The growth in innovation, financial technology,
regulation and competition in the decades prior to the crisis, led to an evolutionary
change in the whole financial services industry. In particular, there was a marked
shift from the usual originate-to-hold model of commercial banking to originate-
to-distribute model of financial intermediation. While in the former model credit,
maturity and liquidity transformation occurred on the balance sheets of a bank, in
the latter, instead of holding loans on the balance sheets, banks could easily transfer
them off the balance sheets. This led to the emergence of a new financial entity -
Special Purpose Vehicles (SPV) or Off Balance Sheet Entities (OBSE) where these
loans are transferred. Even though shadow banking comprises of a myriad of financial
structures and institutions, we focus on OBSE as the sole entity of shadow banking
in our model. This choice makes the model tractable and also allows us to focus
on the existence of shadow banking from a regulatory perspective, which forms the
central theme of this chapter.
We motivate the existence of shadow banking from both demand and supply side.
On the supply side of the system, commercial banks could not compete with finance
companies and broker-dealers that are not subject to as tight capital constraints
as the former is. Thus, the presence of regulatory arbitrage marked a shift from
traditional loan issuance and funding (originate-to-hold) to an originate-to-distribute
model whereby loans are shifted from bank balance sheet to shadow banks. These
loans were then pooled, underwritten and sold as rated asset backed securities linked
to the loan portfolio. The originate-to-distribute model helped traditional banks
to manage risks, provided regulatory benefits and also freed up capital, which was
used to issue further loans to the private sector. On the demand side, we look at the
market for the products of shadow banking i.e. the asset-backed securities. Managed
funds and other institutional investors are important risk takers through investment
in securities and other market debt instruments. Greenwood, Hanson, and Stein
8
(2012) and Pozsar (2011) argue that the total balances of the institutional cash
investors significantly exceed the (inelastic) supply of short-term government debt
and insured deposits. Therefore, institutional investors, such as managed funds who
do not have any access to safe, short term and interest earning investment, became
major players as shadow bank depositors.
The banking system in our model consists of retail and shadow banking. Retail
banks collect deposits, raise bank capital and make loans (a part of which is sold in
the secondary market). The effective return on bank portfolio depends on the moni-
toring effort exerted by banks. We further assume that the level of bank monitoring
services is unobservable so that the bank and loan buyers cannot write contracts that
are contingent on the level of the effort. Therefore, loan sales involve a moral hazard
problem, namely, that the bank may not exert the monitoring effort at the most
efficient level. This limits the amount of loan sold in the secondary market. Banks
are also subject to idiosyncratic risk similar to entrepreneurs in the BGG framework.
This introduces aggregate default in the banking sector as banks whose idiosyncratic
risk falls below a certain threshold cannot service their deposits and hence declare
default. The returns from defaulting banks are expropriated by the deposit insur-
ance agency, who service the deposits from failed banks by levying lump sum taxes
on households. The banks are therefore subject to a minimum capital regulatory
constraint to limit the losses incurred due to failing banks. An important motiva-
tion for incorporating bank balance sheets with capital into the model framework is
that it facilitates analysis of financial stability issues and assessing the implications
of prudential regulatory measures affecting bank capital. The shadow bank on the
other hand is not subject to regulatory constraint. The shadow bank undertakes the
securitization process whereby the transferred loans are converted into marketable
liquid asset-backed securities. To avoid adding too many agents in our model, we
designate households as the final investors in these securities.
Our results indicate that the inclusion of shadow banking enhances credit inter-
mediation and resilience of the financial system to productivity shocks. However, it
increases the vulnerability of the economy to shocks emerging in the financial system.
The recent financial crisis brought to light the perverse effect of financial shocks on
9
the supply side of credit intermediation. In particular, an increase in the riskiness
of commercial banks, leads to a shift of larger proportion of loans to the secondary
market thereby increasing the moral hazard associated with bank monitoring and
lowering the effective return on bank portfolio.
We further analyze the efficacy of government backstop programs in containing
the crisis. In this regard, we evaluate the impact of direct loan purchase by the
government. Active government credit policy is effective in ameliorating the adverse
impact of risk shock on the economy.
Finally, we draw our focus to the macro-financial linkages of prudential regula-
tion. In analyzing the impact of capital regulation on macroeconomic dynamics, we
compare the Basel I and Basel II framework. We find that the impact of a neg-
ative technology shock on output, investment and capital price is amplified under
procyclical Basel II regime. Furthermore, an increase in capital requirement under
Basel II leads to a higher securitization and shift of loans from on to off balance
sheets. This leads to a greater tendency to move to the unregulated sector in re-
sponse to a negative shock under Basel II regime. Furthermore, it also reduces the
overall return on bank portfolio by increasing the moral hazard problem associated
with bank monitoring.
To reiterate, chapter 1 focuses on the financial accelerator or broad credit channel
of monetary policy transmission wherein financial friction on the demand side am-
plifies shocks affecting the real economy. Chapter 2 reviews the influence of capital
regulation on bank-behavior. A key issue here is the degree to which capital regula-
tion both reflects and influences the measurement, management and pricing of risk.
It summarizes what is known about the “bank capital” channel, i.e. the impact on
the transmission mechanism that operates through the threat of breaching minimum
capital requirements. It further analyzes the evolution of shadow banking from reg-
ulation perspective and provides a theoretical model to understand the interaction
between traditional and shadow banking as well as the link between macroeconomic
and financial stability. In chapter 3, we further this discussion on the link or trade-
offs between macroeconomic and financial stability. Here, we study a more recently
10
understood channel of monetary transmission – the ‘bank risk-taking channel’.
During the pre-crisis period, central banks mostly disregarded the financial sta-
bility aspect, since the conventional wisdom suggested that the design of monetary
policy was to maintain price stability. Developments in the credit transfer techniques
that came with financial innovation were often regarded as contributing to financial
stability (Duffie (2008), Altunbas et. al. (2014)). However, following the global
financial crisis, the view that monetary policy actions may have consequences on
financial stability has gained popularity. Further, questions regarding the role of fi-
nancial stability considerations in monetary policy decisions and the ways to modify
the existing monetary policy frameworks taking into account macro imbalances, have
gathered strong prominence.
In light of these developments, the debate over the relationship between monetary
policy and financial stability has intensified. The question of how monetary policy
affects banks’ risk-taking incentives is key to the aforementioned policy debate. This
led to the emergence of the theory of risk-taking channel of monetary policy trans-
mission (Borio and Zhu (2012), Rajan (2006)). In short, risk-taking channel posits
that an expansionary monetary policy for a prolonged period of time can have an
impact on risk perceptions or attitudes of banks. It relates to how changes in in-
terest rates affect either risk perceptions or risk-tolerance of financial intermediaries.
In this case, the result is not only an increase in lending in line with the traditional
transmission mechanisms, but the risk-taking channel also implies an increase in
riskiness of lending, i.e. a deterioration in the quality of portfolios. Thus the risk
taking channel posits that monetary policy actions could contribute to the buildup
of financial imbalances via its impact on risk attitudes, which could eventually result
in a financial crisis.
The risk-taking channel of monetary policy transmission, even though fairly new,
has garnered substantial attention, as manifested in recent papers – both theoretical
and empirical. Empirical literature on the risk-taking channel has thus far focused
mostly on the US and the euro area. Chapter 3 therefore fills an important gap in this
literature by examining the impact of low interest rates in 10 Asian economies (the
11
People’s Republic of China (PRC); Hong Kong, China; India; Indonesia; the Republic
of Korea; Malaysia; the Philippines; Singapore; Taipei,China; and Thailand) from
2000 to 2011. Emerging Asia is an interesting case because during the global financial
crisis a persistent interest rate differential remained between the relatively higher
interest rate in the emerging economies and that of the advanced economies. The
search-for-yield in this case led to a surge in capital flows into the emerging economies
buoying domestic economic and financial conditions. The appropriate stabilization
strategy would be to tighten policy (let interest rates rise), yet this was seldom done
in order to discourage further capital inflows and hold the exchange rate relatively
steady. This has led to lower than appropriate interest rates in these economies.
We empirically test for the existence of the risk-taking channel by analyzing the
panel of banks in Asia using both annual and quarterly data. For annual data, we
employ four different risk indicators incorporating both market based and accounting
based measures of risk.
1. Idiosyncratic bank risk from an estimated CAPM model using daily indi-
vidual bank return index and daily stock market return index for each country
2. The Z-score, which measures bank soundness in terms of insolvency risk or
distance to default is calculated with data obtained from Bankscope. In essence, the
z-score measures the lower bound in which expected returns need to drop in order
to exhaust the bank’s equity.
3. The ratio of non-performing loans to total loans.
4. Volatility of return on bank assets, measured by standard deviation of
return on bank assets.
Furthermore we use two interest rate gaps for monetary policy stance as per Al-
tunbas et al. (2014) in calculating the deviation of policy rate from some benchmark
level;
1. Taylor gap, computed as the difference between the actual nominal interest
rates and that generated from a Taylor-type equation.
2. Natural gap, which is the difference between the real short-term interest
rate and the natural rate (or more accurately the trend rate).
If the bank risk-taking channel is operational, we expect the coefficient on interest
12
rate gap to be negative, meaning low interest rates (lower than some benchmark)
increase bank risk-taking. We also include other variables to account and control for
other determinants of bank risk taking. These include macroeconomic controls (out-
put growth, yield curve, exchange rate volatility, stock market index and debt/GDP
ratio), bank specific controls (size, profitability, liquidity and capitalization) and reg-
ulatory controls (capital stringency, supervisory power and market discipline). Given
the dynamic nature of bank risk we utilize a dynamic panel data model, which is
estimated using a System GMM approach along the lines of Arellano- Bover (1995)/
Blundell-Bond (1998).
We find evidence of an operational bank risk-taking channel. In particular, we
find that when interest rates are “too low”—lower than a benchmark—bank risk in-
creases. This result is robust to other factors, which might have influenced banks’
risk-taking behavior, namely, general economic conditions, bank specific character-
istics, country specific market risk, and the regulatory environment. There is also
some evidence suggesting the positive role of regulatory environment in ameliorating
bank risk taking. We find the increase in bank risk in response to lax monetary
policy to be lower in the case of Asia as compared to advanced economies like the
US and euro area (for example Altunbas et. al (2014)), however the estimates are
closer to findings in some other emerging market economies (Ozsuca and Akbostancı
(2016)). The results are robust to various specifications, for example considering a
subsample by excluding China, Hong Kong and Singapore from the sample where
interest rates are not the main policy instrument.
The effects of short-term interest rate on banks’ risk tolerance can be very brisk
therefore we add to our analysis of an annual data model by including a specification
for quarterly data. Due to lack of accounting data at quarterly frequency, we have
to rely on a market-based indicator of bank risk (idiosyncratic bank risk from an
estimated CAPM model) computed from quarterly data on individual banks from
Bloomberg. Quarterly model offers more nuanced results wherein we can distinguish
between the two opposing ways in which interest rates can affect bank risk taking
as suggested by Jimenez et. al. (2014). In the short term (contemporaneous effect),
low interest rates, by reducing the probability of default on existing loans, lead to
13
a fall in the risk of banks’ portfolio. Over time, however, banks start to undertake
more risky operations and lending to projects with a higher probability of default.
Thus, the results from this chapter suggest that the conduct of monetary policy
is not independent from the goal of achieving financial stability and therefore it
is important to explicitly model financial stability considerations in macro models.
Furthermore, for emerging market economies in particular, which are not insulated
from low interest rates in advanced economies, it might be instructive to let the
exchange rate move more flexibly. If not, they might find themselves in a position
where the low interest rates in advanced economies can lead to low interest rates
in these economies too. This would not only encourage bank risk taking but also
interfere with the domestic stabilization goals.
14
1 Household and Firm Leverage in Business Cycle
Model
This chapter studies the role of credit and demand side frictions in business cycle fluc-
tuations. In particular, we model the interaction between firm and household credit
constraints over the business cycle. Using data on the U.S to estimate the model with
Bayesian techniques, we find a clear evidence of an operating financial accelerator
mechanism, whereby shocks to the economy are amplified by the presence of credit
frictions. Inclusion of both firm and household financing constraints, as opposed to
just one of them, amplifies the impact of exogenous shocks on GDP, investment and
consumption. Regarding the importance of various shocks in explaining macroeco-
nomic fluctuations, we find that total factor productivity shock in non-housing sector
explains more than 40 percent of the fluctuations in GDP and consumption, however
housing demand and supply shocks explain most of the cyclical variation in housing
investment and housing prices. Investment shocks are an important driver in explain-
ing variation of non-housing and housing investment. Finally, we find substantial
spillovers from housing market to the real economy, predominantly to non-housing
consumption, due to the wealth effect of house prices.
1.1 Introduction
The view that financial structure and the performance of credit markets may be
important to understand the macroeconomy dates back to Gurley and Shaw (1955).
Thereafter it has been reconsidered in the last 50 years in a growing number of stud-
ies. The common feature of these contributions is the idea that the way in which
agents finance their activities, have access to financial markets and choose contrac-
tual arrangements is mostly relevant to understanding the business cycle. The recent
financial turmoil of 2007 has further brought to light the importance of financial prop-
agation in macroeconomic volatility. In particular, the housing market has come to
the spotlight. Financial propagation, often called the financial accelerator mecha-
15
nism, implies that the presence of credit market frictions can amplify the impact of
exogenous shocks on business cycles by affecting the financial health of borrowers
(firms and households).
Traditionally, most papers have analyzed the importance of credit constraints on
the firm or the entrepreneur side. Bernanke, Gertler and Gilchrist, henceforth BGG
(1999) and Carlstrom and Fuerst (1997) study frameworks in which credit market im-
perfections arise because there is asymmetric information between entrepreneurs and
banks. This asymmetry translates into external borrowing being more expensive than
internal financing. The costly state verification approach proposed by these studies
is used extensively in the literature thereafter (Christiano et. al. (2014); Christensen
and Dib (2008) among others). However, more recently there has been an emerging
interest in modeling relation between household debt and the macroeconomy, partic-
ularly in the U.S (Iacoviello and Neri (2010), Justiniano, Primiceri and Tambalotti
(2013)). Since household debt in the U.S. is held primarily in the form of mortgages,
these models impose a collateral constraint that limits debt to a fraction of home
values. As a consequence, house prices play a crucial role in the dynamics of debt, a
connection that is evident in the data. Studying firm and household debt in isolation
however assumes that they are independent of each other. If both household and
firm credit constraints independently generate their own financial accelerator, there
is a possibility of a positive interaction between the two. This paper by integrating
both household and firm debt in a dynamic stochastic general equilibrium (DSGE)
model aims at answering few related questions - what is the interplay between credit
market frictions faced by firms and households in generating business cycle fluctua-
tions? what is the role played by productivity and financial shocks over the business
cycle?
Poor household financial conditions can affect firms’ activity via various channels.
First, through the impact of a negative shock to household balance sheet on aggre-
gate economic activity and employment. Mian and Sufi (2012) examine the role of
the aggregate demand channel in the decline in U.S employment from 2007 to 2009.
A drop in aggregate demand driven by shocks to household balance sheet caused a
substantial decline in the U.S employment over this period. They further found that
16
employment losses in response to balance sheet shocks are higher in the U.S counties
with high household leverage. Therefore, worse household financial condition by af-
fecting the aggregate economic activity, can accentuate the impact of firm financial
constraints through the usual financial accelerator channel. Another channel relates
to the impact of households’ terms of borrowing on the durable goods industry. The
decade prior to the great recession witnessed an increased availability of consumer
loans collateralised with residential properties. Mian and Sufi (2011) estimate that
the average U.S homeowner extracted 25 to 30 cents for every dollar increase in home
equity from 2002 to 20061. Changes to consumers’ access to credit affects employ-
ment, particularly in industries producing goods that are purchased with consumer
loans. Haltenhof et al. (2014) find that between 2007 and 2010 the decline in home
equity extraction explains about 10 percent of the decline in employment in the
durable goods industries. Their results also suggest that changes in lending stan-
dards on commercial and industrial loans and the availability of home equity lines
of credit affect notably the manufacturing employment and net firm dynamics. The
evidence though does not necessarily represent a causal link from worse household fi-
nancial frictions to tightening of firm financial constraints; it points towards the need
to better understand the interaction and dynamics of firm and household leverage
and their impact on propagation mechanisms.
We construct our model on the lines of Iacoviello and Neri (2010) with hetero-
geneous production sector and introduce an active role for evolution of firms’ net
worth and their financial leverage. On the supply side of the model, firms engage
in joint production of housing and non-housing goods. The non-housing sector pro-
duces consumption good using capital and labour and the housing sector produces
new homes using capital, labour, intermediate business input and land. Firms hire
labour and purchase capital using their net worth and borrowing from the financial
intermediary as in BGG (1999). Firms are ex-post heterogeneous and differ in their
realization of the idiosyncratic shock. This results in an external finance premium
1Mian and Sufi (2011) find that money extracted from increased home equity is not used to
purchase new real estate or pay down high credit card balance, suggesting that borrowed funds
maybe used for real outlays (i.e. consumption or home improvement).
17
on their borrowing as a positive function of their leverage i.e. higher firm leverage
leads to a higher premium on external funds. The demand side comprises of hetero-
geneous households following Iacoviello (2005) differing in their degree of impatience.
Consumption and housing enter households’ utility and housing can also be used as
collateral for loans. In equilibrium, patient households are lenders providing funds
to impatient households and firms. Impatient households borrow a proportion of
their expected value of housing stock. The joint production of non-housing and
housing sector produces interesting dynamics vis-a-vis residential and nonresidential
investment. Furthermore, changes in house prices significantly affect the borrowing
capacity of impatient households affecting the dynamics of both firm and household
behavior.
We use quarterly data for the U.S for the period 1970 - 2006 on six key macroe-
conomic variables (consumption, non-residential investment, residential investment,
house prices, hours worked in the consumption sector and hours worked in the hous-
ing sector to estimate the model using Bayesian techniques. We find that although
the non-housing technology shock is the most important driving force for variation
in GDP and consumption, housing preference and housing technology shock together
account for more than 50 percent of the variation in housing investment and housing
prices. Investment and risk shocks are an important determinant of the cyclical vari-
ation in macroeconomic aggregates, particularly business investment. Further, we
find that housing prices and housing investment are strongly procyclical as observed
in the data.
We document a clear evidence of an existing financial accelerator at both firm
and household level. As compared to financial accelerator at only firm or household
level, inclusion of both types of financial frictions further amplifies the response of
GDP, consumption and investment (both residential and non-residential) to exoge-
nous shocks. This implies that a model containing both firm and household frictions
incorporates additional channels in which demand side frictions impact business cycle
dynamics and macroeconomic volatility. Further in our model, firm and household
financing constraints interact in two important ways. First, higher financing frictions
for households reduce their demand for loans. This, through the loan market equi-
18
librium condition decreases the interest rate and causes a downward pressure on the
external financing costs for firms. Second, a fall in house prices and housing demand
(that is amplified in the model with household collateral constraints), reduces the
demand for housing capital, which further deteriorates capital price and therefore
the net worth of firms. This increases the external finance premium facing firms.
The two effects work in opposite direction. Overall the second effect dominates and
worse household financing conditions tend to tighten firm financing constraints.
The rest of the paper is structured at follows. We first give an overview of
the existing literature in the following section. Section 1.3 discusses the model with
household and firm leverage following the existing work by BGG (1999) and Iacoviello
(2010). Section 1.4 describes the data and estimation technique. In section 1.5, we
present the results from bayesian estimation and finally conclude in section 1.6.
1.2 Literature: Theory and Evidence
Early line of literature relating to financial market frictions, attempts at deriving
credit constraints from microeconomic considerations in a partial equilibrium frame-
work. Some of these contributions toward a theory of credit constraints based on
first principles are by Jaffee and Russell (1976) and Stiglitz and Weiss (1981). These
contributions are based on the impossibility to enforce debt contracts (i.e. the ex-
istence of incentives to default) and adverse selection or moral hazard phenomena,
respectively.
A more recent and growing body of literature in this area has investigated the
role of financial market imperfections using models where agents are allowed to select
the best contractual arrangements. These papers rely on either costly state verifica-
tion/agency costs or limited commitment/collateral obligations to introduce finan-
cial frictions in a fully-fledged general equilibrium model. In Bernanke and Gertler
(1989), entrepreneurs undertake a costly screening activity to assess the character-
istic of a project. Since these characteristics are privately observed, lenders face an
adverse selection problem in financing entrepreneurs’ project. Following this paper,
various authors have employed costly state verification in dynamic models in which
19
financial frictions on firm side may amplify output fluctuations in response to exoge-
nous shocks. Examples include Carlstorm and Fruest (1997) and BGG (1999). BGG
(1999) present a model of financial accelerator in a quantitative general equilibrium
framework with price stickiness and investment lags. They allow for heterogeneity
among firms to capture the fact that borrowers have differential access to capital
markets. Firms are ex-ante similar but differ in the ex-post realization of the id-
iosyncratic shock. The financial contract between an intermediary and a risk-neutral
entrepreneur in this framework implicitly results in a risk premium, which is an in-
creasing function of the leverage of the firm. Under reasonable parametrisations of
the model, they find that the financial accelerator works to amplify and propagate
shocks to the macroeconomy and has a significant influence on business cycle dynam-
ics. Following this, there have been attempts to estimate the BGG model for various
economies, most notably the U.S and Euro area. Christiano, Motto and Rostagno
(2003, 2008) estimate a DSGE model with financial accelerator by calibrating pa-
rameters related to financial frictions and get a significant impact of credit market on
the economy. Christensen and Dib (2008) estimate the BGG model for the U.S using
maximum likelihood and find evidence in favor of the financial accelerator model.
Gilchrist, Ortiz, and Zakrajsek (2009) incorporate a high information-content credit
spread constructed directly from the secondary-market prices of outstanding corpo-
rate bonds into the Bayesian ML estimation. This credit spread serves as a proxy for
the unobservable external finance premium. Using the U.S data for 1973-2008, they
find that increases in the external finance premium causes significant and protracted
declines in investment and output. Virginia Queijo von Heideken (2009) modified the
DSGE financial accelerator model developed by BGG (1999) by adding frictions such
as price indexation to past inflation, sticky wages, consumption habits and variable
capital utilization. Using Bayesian methods to estimate the model for the US and
euro area, he finds that financial frictions are relevant in both cases.
An alternative way to explain credit constraints and financial propagators is via
limited commitment and unenforceability of contracts. When studying optimal (indi-
vidually rational) contracts with limited commitment, the existence of collateralisable
assets may play a very important role. This is because a high value of collateralisable
20
asset increases borrower’s net worth, which plays an important role in enhancing his
debt capacity and financial propagation. The latter may be especially strong since
the value of a collateral is typically low in situations of financial distress. Kiyotaki
and Moore (1997) include collateralisable asset in the form of land in their model.
The role of land is twofold, not only is it a factor of production, but it also serves
as collateral for loans. They show that small, temporary shocks to technology or
income distribution can generate large, persistent fluctuations in output and asset
prices.
Traditionally the literature on credit markets has focused on financial frictions
at the firm level. There have been some empirical studies documenting the impor-
tance of financial constraints at the household level, for e.g.. Zeldes (1989), Jappelli
and Pagano (1989), Campbell and Mankiw (1989), and Carroll and Dunn (1997).
However, the focus has hardly been to study the business cycle dynamics of these
frictions. More recently, Iacoviello and Neri (2010), Mian and Sufi (2009, 2011, 2012),
Favara and Imbs (2015) and Justiniano, Primiceri and Tambalotti (2013) have stud-
ied the evolution of household debt, housing market spillovers and the credit boom
and bust of 2000s. These papers however focus entirely on the housing leveraging
and deleveraging ignoring frictions at the firm level. We add to the existing liter-
ature by providing a joint analysis of financing constraints on both household and
entrepreneur side. Earlier work with financial frictions at both household and firm
level include Iacoviello (2005) and Gerali et. al. (2010). They rely on quantity
borrowing constraints tied to the value of collateral for both household and firms,
however they do not examine the dynamics of modeling both types of financial fric-
tions as opposed to just one. Solomon (2010) is the only other study that provides a
model of interaction of consumer and household debt over the business cycle. How-
ever, he does not explicitly model the quantity and price side of housing sector. We
construct a model with heterogeneous production sector comprising separate housing
and non-housing production such that both housing quantity and prices are endoge-
nous. Unlike in Solomon (2010) we are able to capture two salient features of housing
and firm financial friction as documented in empirical literature. First, we find that
inclusion of both household and firm financial constraints amplifies the response of
21
GDP, consumption and investment as opposed to having only firm or household fric-
tions. In contrast, Solomon (2010) find that adding household financial frictions to
an environment with firm financing frictions dampens the overall effect of shocks on
aggregate variables, which is at odds with the data and empirical evidence. Secondly,
in our model residential and non-residential investment generally tends to co-move
in the same direction however Solomon (2010) finds a strong negative counterfactual
comovement between residential and nonresidential investment.
1.3 The Model
The Economy consists of two types of households and of entrepreneurs. Households
consume, work and accumulate housing. The two households differ in the degree of
their impatience i.e they discount future utility at differing rates. We call the ones
with a higher discount factor as the patient household and the others as impatient
(following Iacoviello (2005)). Patient and impatient households are both available in
unit mass each. In equilibrium patient households are lenders and impatient ones are
borrowers who borrow against the value of housing stock as collateral. On the supply
side, firms operate the housing and non-housing sector. The non-housing sector uses
labour supplied from households and capital to produce wholesale consumption and
intermediate business goods. The housing sector produces new homes using capital,
labour, intermediate business good and land. Entrepreneurs provide physical capital
to firms that is used in the production of consumption good and housing.
The output produced by firms is used up in consumption, investment (housing
and non-housing goods), goods used up in capital utilization and in bank moni-
toring. Capital producers combine investment goods with used capital purchased
from entrepreneurs to produce new capital. This new capital is then purchased by
entrepreneurs. Entrepreneurs make these purchases using their own resources —
net worth, or equity, which they accrue by compounding the net proceeds of their
activity from one time to the next — as well as bank loans. Heterogeneity among
entrepreneurs exists in the form of a different idiosyncratic shock faced by each one of
them (thus ex-ante each entrepreneur is same). This arrangement is similar to that
22
in BGG (1999). There exists a financial intermediary, which channels the savings
of the patient households to impatient household and entrepreneurs. Two types of
one-period financial instruments are provided by the bank - deposits and loans. The
patient households purchase a positive amount of deposits and do not borrow while
the impatient households borrow a positive amount of loan (tied to housing collat-
eral). For firms, following BGG (1999), we have a financial contract, which minimizes
the expected agency costs of the borrowers arising from idiosyncratic shocks.
1.3.1 Patient Households (P)
There is a continuum of measure 1 of agents in each of the two groups (patient and
impatient). A representative patient household maximizes the expected utility:
max
cp,t,hp,t,dp,t,kb,t,pla,t,lpc,t,lph,t
E0
1X
t=0
tpzt

log(cp,t) + Jh,t log hp,t  &t
(1 + p)
(l1+⌘ppc,t + l
1+⌘p
ph,t )
(1+p)
(1+⌘p)

(1.1)
The expected utility depends on the current individual consumption cp,t , housing
services hp,t and hours worked in the non-housing, lpc,t and the housing sector, lph,t2.
Traditionally, productivity shocks are hailed as the prominent driver of fluctuations
in real business cycle models. We add additional shocks to the model to see their
relative contribution in driving business cycle fluctuations. Three such shocks are
listed below. First, movements in zt captures shock to inter-temporal preferences
that governs how households weigh current utility relative to future utility. Several
recent papers have used time preference shocks to explain business cycle dynamics
(Hall (2013), Christiano, Eichenbaum and Rebelo (2011), Iacoviello and Neri (2010),
Christensen and Dib (2008), Smets and Wouters (2003)) as well as asset pricing
puzzle (Albuquerque, Eichenbaum and Rebelo (2012)). Second, movements in &t
captures shock to intratemporal preference - it affects how agents value utility from
consumption or housing relative to utility from leisure. Many recent quantitative
2See appendix 1.5 for a complete list of model variables and parameters.
23
macroeconomic models allow for such intratemporal preference shocks. An unob-
served demand shock plays an important role in explaining aggregate fluctuations in
Rotemberg and Woodford (1997). They interpret the shock as a combination of a
labour preference shock and a shock to government spending. Erceg et al. (2000)
and Smets and Wouters (2003) also have a quantitatively important labour pref-
erence shock in their models of monetary policy. More recently, Galí and Rabanal
(2004) find that a labour preference shock explains fifty-seven percent of the variance
of output and seventy percent of the variance of hours in their estimated dynamic
stochastic general equilibrium model. We assume that both zt and &t follow mean
zero AR(1)s in the log.
logzt = ⇢zlogzt1 + "z,t;
log&t = ⇢& log&t1 + "&,t,
where "x,t is i.i.d N(0, 2x); for x 2 (z, &).
Few things to point out: First, zt does not show up in the labour supply equilib-
rium condition; higher zt increases the marginal utility of both consumption and the
marginal disutility of labour, but these cancel out. Second, since zt is mean revert-
ing, when zt is high we will have a relatively low value of Etzt+1. This means that
an increase in zt is isomorphic to a temporary reduction in p, i.e. you are relatively
more impatient. Third, &t shows up in the labour supply equilibrium condition in a
way analogous to a distortionary tax rate on labour income.
Finally, movements in Jh,t can be viewed as a housing preference shock. This
shock is included to study the business cycle dynamics in response to shifts in hous-
ing demand. Iacoviello and Neri (2010) provide two possible interpretation of this
shock. “One interpretation is that the shock captures, in a reduced form way, cycli-
cal variations in the availability of resources needed to purchase housing relative to
other goods or other social and institutional changes that shift preferences toward
housing. Another interpretation is that fluctuations in Jh,t could proxy for random
24
changes in the factor mix required to produce home services from a given housing
stock.” The U.S housing market, and particularly the housing finance market, has
undergone several institutional changes in the past decades. These factors, including
the rise in mortgage finance and home ownership rates3 has affected households’ de-
mand for housing and are possible sources of household demand shocks. In addition,
the preference shock could also capture shifters of housing demand not explicitly
included in our stylized model, for instance the changing demographic profile of the
U.S population. Jh,t denotes the weight attached to housing in the utility function,
which also has an AR(1) representation with i.i.d normal innovations.
logJh,t = ⇢hlogJh,t1 + (1 ⇢h)logJh + "h,t,
where "h,t is i.i.d N(0, 2h). Jh > 0 denotes the weight attached to housing in the
steady state.
The log-log specification of preferences for consumption and housing is borrowed
from the literature (Davis and Heathcote (2005), Fisher (2007) and Iacoviello and
Neri (2010)). The specification of the disutility of labour (p, ⌘p  0) follows Hor-
vath (2000) and Iacoviello and Neri (2010). This allows for less than perfect labour
mobility across sectors, wherein ⌘p = 0 would imply that hours in the two sectors
are perfect substitutes. p is the inverse of Frisch wage elasticity of labour supply.
Patient households maximize their utility subject to the budget constraint (in
real terms):
cp,t+ qh,t(hp,t (1 h)hp,t1) + dp,t+ pla,tlat+ kb,t  wpc,tlpc,t+wph,tlph,t+Rt1dp,t1
+pb,tkb,t + (pla,t +Rla,t)lat1 + lumpt (1.2)
3Home ownership rates in the United States were constant around 64/65 percent between 1970
and 1995, and rose at a constant pace up to 69 percent at the peak of the housing boom in 2005.
Rising home ownership can by ’keeping up with the joneses’ hypothesis affect consumers demand
for housing.
25
Thus, patient household chooses consumption cp,t, housing stock hp,t priced at qh,t,
deposits dp,t, intermediate business input kb,t priced at pb,t, land lat priced at pla,t and
hours worked lpc,t and lph,t in the non-housing and housing sector respectively. h is
the housing depreciation rate, Rt1 is the gross real return on the deposits between
period t 1 and t, wpc,t and wph,t the real wages earned in non-housing and housing
sector respectively. lumpt are transfers from entrepreneurs (discussed later).
1.3.2 Impatient Households (I)
Impatient households neither hold deposits nor own retail firms. They have the
same intra-period preferences over housing, consumption and leisure as the patient
households and are subject to the same preference shocks, but they have a lower
discount factor than the lenders (patient households) i.e p > i. A representative
impatient household maximizes the expected utility:
max
ci,t,hi,t,lic,t,lih,t,bi,t
E0
1X
t=0
tizt

log(ci,t) + Jh,t log hi,t  &t
(1 + i)
(l1+⌘iic,t + l
1+⌘i
ih,t )
(1+i)
(1+⌘i)

(1.3)
The expected utility again depends on consumption, housing and hours worked.
The budget constraint now is:
ci,t + qh,t(hi,t  (1 h)hi,t1) +Rt1bi,t1  wic,tlic,t + wih,tlih,t + bi,t (1.4)
Therefore, for impatient households, resources spent on consumption, accumula-
tion of housing stock and reimbursement of past borrowing have to be financed with
wage income and new borrowing (bi,t). They further face a borrowing constraint4:
4The assumption p > i implies that the borrowing constraint faced by impatient borrowers is
binding for small shocks near the steady state.
26
bi,t  mEt qh,t+1hi,t
Rt
(1.5)
This means that the borrowers cannot borrow more than the net present expected
value of the collateralisable housing stock. Thus, m reflects the amount of loans
given by the intermediary against a given amount of housing stock. A high value of
m means loose credit constraints for impatient households. At this stage we do not
include defaults explicitly in the case of household loans. However, implicitly if we
assume that the price (or value) of houses does not fall below the amount borrowed
i.e mEt
qh,t+1hI,t
Rt+1
, lenders can simply usurp the collateral and are effectively protected
from default.
1.3.3 Entrepreneurs
Entrepreneurs are assumed to be risk-neutral and have finite horizons. Specifically,
we assume that each entrepreneur has a constant probability  of surviving to the
next period (implying an expected lifetime of 1/(1  )). We assume the birth rate
of entrepreneurs to be such that the number of entrepreneurs is constant. In each
period t, entrepreneurs acquire physical capital, which is used in the production of
consumption goods and housing. Entrepreneurs that die in period t are not allowed
to purchase capital. A fraction ⇥ of their total net worth is consumed upon exit,
and the remaining fraction of their net worth is transferred as a lump-sum payment
to patient households. New entrepreneurs entering in period t+1 receive a ‘start-up’
transfer of net worth, W e.
Each entrepreneur is indexed by j. Each period, entrepreneurs purchase physical
capital, Kt,j, at the real price, Qk,t, which is financed by entrepreneurial wealth, or
net worth, Nt,j and borrowing, Bt,j = Qk,tKt,j  Nt,j. At this point, the state of
an entrepreneur is summarized by its net worth, Nt,j. The purchased capital, Kt,j
is transformed into !Kt,j, where ! is a log-normally distributed random variable
across all entrepreneurs with a cumulative distribution function denoted by Ft(!).
The assumption about ! implies that entrepreneurial investments in capital are risky.
27
The mean and standard deviation of log! are µ$ and t respectively. The random
variable, !, is observed by the entrepreneur, but can only be observed by the bank if
it pays a monitoring cost. The standard deviation, t, is the realization of a stochastic
process, which we refer to below as the risk shock. This shock captures the notion
that the riskiness of entrepreneurs varies over time and follows a stationary AR(1)
process.
logt = ⇢logt1 + (1 ⇢)log + ",t,
where  is the steady state standard deviation and ",t is i.i.d N(0, 2r).
The total payoff in period t+1 to an entrepreneur with idiosyncratic productivity
shock, !, can be written as Rk,t+1Qk,t!t+1Kt,j where,
Rk,t+1 =
rk,t+1 + (1 k)Qk,t+1
Qk,t
,
where rk,t+1 is marginal productivity of capital at t + 1 and the expression also
incorporates capital gain on holding a unit of capital from period t to t+1 where k
is capital depreciation rate.
As in BGG, entrepreneurs can self finance only a fraction of the capital stock.
They need external finance to complement their net worth as a source of funding.
Entrepreneurs obtain external finance from the bank in the form of bank loans.
The standard debt contract that they enter foresees that entrepreneurs above an
endogenously determined cut off value, !¯t+1 pay gross interest, on their bank loan.
!¯t+1Rk,t+1Qk,tKt,j = RF,tBt,j (1.6)
Where RF,t is contractual gross interest rate on the loan taken from the bank in
period t. Therefore, we can write the threshold value of idiosyncratic shock, !¯t+1,
above which the entrepreneur can repay its loan as:
!¯t+1 =
RF,tBt,j
Qk,tKt,j
.
1
Rk,t+1
28
We define xt =
Qk,tKt,j
Nt,j
as the entrepreneurial leverage, defined as the ratio of
entrepreneurs’ purchase to net worth at t. We have left off the indexing j on xt, !¯t+1,
and RF,t. This is to minimize notation and is a reflection of the fact (see below)
that the equilibrium value of these objects is independent of the state variable (Nt,j).
When entrepreneurs defaults on their loan, following the CSV approach, we assume
that the lender must pay a cost if s/he wishes to observe the borrower’s realized
return on capital, which it then seizes. This auditing cost is interpretable as the cost
of bankruptcy (including for example auditing, accounting, and legal costs, as well as
losses associated with asset liquidation and interruption of business). The monitoring
cost is assumed to equal a proportion µe of the realized gross payoff to the firm’s
capital, i.e., the monitoring cost equals µeRk,t+1Qk,t!t+1Kt,j. Thus, under certain
realizations of the idiosyncratic shocks, entrepreneurs will go bankrupt, especially
when their ex-ante leverage xt is high.
We follow BGG notation for the division of return from entrepreneurial invest-
ment between entrepreneurs and banks. The share of gross return (gross of verifica-
tion costs) accruing to the bank is:
G(!¯t+1) =
ˆ !¯t+1
0
!t+1f(!t+1)d!t+1 + !¯t+1
ˆ 1
!¯t+1
f(!t+1)d!t+1 (1.7)
The part of these returns that comes from defaulted loans is:
G(!¯t+1) =
ˆ !¯t+1
0
!t+1f(!t+1)d!t+1 (1.8)
Therefore, the share of returns accruing to entrepreneurs can be written as 1 
G(!¯t+1). The share (net of verification costs) that the bank appropriates can be
written as G(!¯t+1) µeG(!¯t+1).
Entrepreneurs maximize their expected return subject to a participation con-
straint for the bank. In particular, they choose the leverage, xt, and default threshold
!¯t+1 to maximize:
max
xt,!¯t+1
Et [(1 G(!¯t+1))Rk,t+1xtNt,j]
29
Subject to banks’ participation constraint:
Et[(G(!¯t+1) µeG(!¯t+1))]Rk,t+1Qk,tKt,j = Rt(Qk,tKt,j Nt,j),
which states that the return to banks should cover their cost of raising funds. The
object on the left is the return received by banks from lending to entrepreneurs. Using
the definition of leverage, xt =
Qk,tKt,j
Nt,j
, we can rewrite the participation constraint
as:
Et[(G(!¯t+1) µeG(!¯t+1))]Rk,t+1xt = Rt(xt  1) (1.9)
The (xt, !¯t+1) combinations which satisfy (1.9) define a menu of state (t + 1)-
contingent standard debt contracts offered to entrepreneurs. Entrepreneurs select the
contract that maximizes their objective. Since Nt,j does not appear in the constraint
and appears only as a constant of proportionality in the objective, it follows that all
entrepreneurs select the same (xt, !¯t+1) regardless of their net worth. As mentioned
earlier, a fraction  of the entrepreneurs die at the end of each period and are
replaced by the same amount of new entrepreneurs. New entrepreneurs entering in
period t + 1 receive a ‘start-up’ transfer of net worth, W e. Entrepreneurs who die
consume a constant fraction, ⇥, of their net worth and transfer the rest to patient
households.
Aggregation - The quantity of raw capital purchased by entrepreneurs in period
t is given by summing the capital across all entrepreneurs. Market clearing in the
capital services require that supply of capital, Kt must equal firms’ demand for
capital (discussed later).
Kt =
P
Kt,j
The aggregate entrepreneurial net worth at the end of period t can be written
as (1  G(!¯t+1))Rk,t+1Qk,tKt. Once a fraction (1  ) of the entrepreneurs exit the
scene and the newly entering entrepreneur receive lump sum payments W e, the law
of motion of entrepreneurial net worth, Nt+1 can be written as:
30
Nt+1 = (1 G(!¯t+1)Rk,t+1Qk,tKt +W e,
where W e is the ‘start-up’ transfer of net worth received by new entering en-
trepreneurs in period t+1. The aggregate net worth can be further written as:
Nt+1 = {Rk,t+1Qk,tKt [Rt+µe
´ !¯t+1
0 !t+1f(!t+1)d!t+1Qk,tKt
Qk,tKt Nt ](Qk,tKtNt)}+W
e
(1.10)
The object in braces in (1.10) represents receipts by entrepreneurs active in period
t+1 minus their payments to banks. The object in square brackets is the gross rate
of return paid by entrepreneurs to banks. The external finance premium on loans
between period t and t+1 can therefore be defined as:
efpt+1 = µe
´ !¯t+1
0 !t+1f(!t+1)d!t+1Qk,tKt
Qk,tKt Nt (1.11)
The remaining two financial variables to determine are the aggregate quantity
of debt extended to entrepreneurs in period t, Bt, and their state-contingent inter-
est rate, RF,t. The aggregate borrowing is given by the sum of borrowing of all
entrepreneurs.
Bt =
P
Bt,j = Qk,tKt Nt
Finally, the interest rate takes the form (aggregating equation 1.6 and solving):
RF,t = !¯t+1Rk,t+1(
xt
xt  1)
The entrepreneurs that exit the economy in t+1 consume a fraction ⇥ of their
net worth. Therefore, their aggregate consumption is given by:
ce,t+1 = ⇥
(1 )

(Nt+1 W e) (1.12)
31
Where (Nt+1W
e)
 is the average net worth held by an entrepreneur who exits
the economy at the end of the current period and (1  ) is the share of exiting
entrepreneurs. Finally, the lump sum transfers, lumpt, made to patient households
is (1⇥) (1) (Nt+1 W e).
1.3.4 Firms
Firms produce wholesale consumption goods and housing using two CRS production
technologies. They hire labour from households, rent capital from entrepreneurs,
purchase intermediate business goods, kb,t, at price, pb,t and use capital for the pro-
duction of wholesale goods Yt and new housing IHt. They undertake the following
static optimization:
max
Kc,t1,Kh,t1,lic,t,lpc,t,lih,t,lph,t,kb,t,lat1
Yt+qh,tIHt{
X
s=c,h
wps,tlps,t+
X
s=c,h
wis,tlis,t+rkc,tKc,t1+
(1.13)
rkh,tKh,t1 +Rla,tlat1 + pb,tkb,t}
Where Kt = Kc,t + Kh,t; Kc,t is the business capital and Kh,t is the housing
capital and rkc,t and rkh,t are marginal productivity of capital in the two sectors.5
The production technology for non-housing and housing are respectively:
Yt = Ac,tK
↵
c,t1(l
µ
ic,tl
1µ
pc,t )
1↵ (1.14)
IHt = Ah,t(l
µ
ih,tl
1µ
ph,t )
1✓⌧Kh,t1k
✓
b,tla
⌧
t1 (1.15)
Where Ac,t and Ah,t is non-housing and housing sector specific aggregate technol-
ogy that follows a stationary AR(1) process of the form:
5Since capital is homogeneous (there is no difference between non-housing and housing capital),
in equilibrium the return on capital is equalized rkc,t = rkh,t = rk,t.
32
lnAc,t = ⇢ac lnAc,t1 + "ac,t;
and
lnAh,t = ⇢ah lnAh,t1 + "ah,t
"ac,t and "ah,t are normally distributed with zero mean and standard deviation
ac and ah respectively.
1.3.5 Capital Goods Producer
At the beginning of each period, capital good producer buys final goods and the
stock of undepreciated capital from the firms. We introduce convex investment costs
whereby It units of output is converted into ⌘k,t(1  S(Gt))It units of new capital
(sold at real price Qk,t).
Where ⌘k,t is the investment shock, Gt = It/It1 and the function S(Gt) =
'x(Gt  1)2 satisfies S 0, S 00  0. Also in steady state S(.) = S 0(.) = 0.
The capital goods producer maximize expected discounted profits:
max
It
Et
1X
k=0
Dt,t+k[Qk,t+k.⌘k,t(1 S( It+k
It+k1
))It+k  It+k],
where Dt,t+k = p
MUc,t+k
MUc,t
is the real stochastic discount factor, MUc,t is the
marginal utility of consumption. Capital evolves according to the following relation:
Kt+1 = (1 k)Kt + ⌘k,t(1 S(Gt))It (1.16)
The investment shock is a source of exogenous variation in the efficiency with
which the final good can be transformed into physical capital, and thus into tomor-
row’s capital input. Justiniano, Primiceri, and Tambalotti (2011) show that this vari-
ation might stem from technological factors specific to the production of investment
goods, as in Greenwood, Hercowitz, and Krusell (1997), but also from disturbances
33
to the process by which these investment goods are turned into productive capital.
Here, as in Justiniano, Primiceri and Tambaloti (2010), we ignore that distinction
and maintain an agnostic stance on the ultimate source of these disturbances. The
investment shock follows a stochastic process:
log ⌘k,t = ⇢⌘log ⌘k,t1 + "⌘,t,
where "⌘,t is i.i.d N(0, 2⌘).
1.3.6 Market Clearing and Competitive Equilibrium
Patient households and impatient households are each present in unit mass. There-
fore, their aggregate consumption is the sum of consumption of representative patient
and impatient household. The aggregation for entrepreneurs, including their aggre-
gate consumption, is discussed in 1.3.3. Combining the budget constraints of the
model’s agents gives us the resource constraint of the economy.
Yt = Ct + It + kb,t + µeG(!¯t)Rk,tQk,t1Kt1 (1.17)
Where Ct = cp,t+ ci,t+ ce,t. The goods market equilibrium governs that the non-
housing output is either consumed, invested, used as a factor in housing production
or lost in monitoring services. Where, µG(!¯t)Rk,tQk,t1Kt1 is the aggregate output
used up in monitoring costs. Investment is already net of adjustment costs described
in the earlier section.
Housing production is equal to the change in housing stock net of depreciation.
IHt = Ht  (1 h)Ht1 (1.18)
Where Ht = hp,t+hi,t is the aggregate housing stock. Equilibrium in the housing
market requires that the supply of housing is equal to the demand from patient and
impatient households. Furthermore, loan market equilibrium implies:
34
dp,t = bi,t +Bt (1.19)
Equilibrium in capital market requires that the supply from entrepreneurs is equal
to the capital demanded by firms. The labour market is in equilibrium when firms’
demand for labour equals supply at the competitive wage level. Finally, GDP (net
of monitoring costs) is defined as the value added of the two sectors:
GDPt = Yt  kb,t + qIHt (1.20)
A competitive equilibrium consists of a set of prices and interest rates {pb,t, pla,t, qh,t,
qk,t, Rla,t, Rt, rk,t, wic,t, wih,t, wpc,t, wph,t} for all possible states and for all t  0 such
that all markets clear when all agents i.e. patient households, impatient households,
entrepreneurs, firms and capital goods producer solve their maximization problems
while taking these prices and interest rates as given.
1.4 Understanding the effects of financing constraints
in the model
Before going into estimation of the model, in this section we discuss some theoret-
ical insights on how financial frictions affect dynamics of the model. We first look
at households and entrepreneur in partial equilibrium and then discuss how their
financing constraints interact with each other.
1.4.1 Impatient Households
Impatient households first order conditions for the choice of consumption and housing
is given by:
zt
ci,t
= Et(
zt+1iRt
ci,t+1
) + tRt
35
zt
ci,t
 Et(tm¯ qh,t+1qh,t ) = Et(
izt+1qh,t+1(1 h)
qh,tci,t+1
) +
ztJh,t
qh,thi,t
Finally, the borrowing constraint is given by:
Rtbi,t  mEtqh,t+1hi,t
The marginal utility from consumption is affected by an additional term, tRt,
where t is the lagrange multiplier on the borrowing constraint and equals the in-
crease in lifetime utility that would stem from borrowing Rt today, consuming the
proceeds, and reducing consumption by the appropriate amount next period. Since
agents are constrained from borrowing more, but not from saving more, t enters the
equation with a positive sign. Therefore, at the constrained optimum, the marginal
utility from consuming one unit today is always greater than or equal to the marginal
utility from waiting until tomorrow to consume the extra amount.
Next, we look at how the borrowing constraint affects the first order condition for
choice of housing by impatient household. If the borrowing constraint never binds, i.e.
t = 0, then the purchasing cost of housing - the marginal cost of current resources or
the marginal utility from consumption- is equated to the marginal utility of housing
and its discounted resale value (as is the case for patient households, see appendix
1.2). However, when the borrowing constraint is binding, the purchasing cost of
housing is lower because it can be used as collateral. Thus, borrowing constraint
affects the inter-temporal allocation of resources as well as the within period one,
since housing is both a good and collateral.
Any shock that increases house price increases the borrowing capacity of impa-
tient households on impact. The effect of the shock in future periods depends on
how borrowers adjust loan demand and housing in response to the shock. The im-
patient household behaves like the consumption smoothing patient household with
a bias towards debt financed consumption instead of saving. For a given amount of
desired housing, a consumption smoothing borrower reacts to an increase in wealth
by increasing savings. This increase in saving can be accomplished through a combi-
36
nation of reduced borrowing and increased accumulation of collateral in the form of
housing. At the same time, an increase in the value of collateral encourages higher
borrowing. Impatient household increase their demand for housing as the purchasing
cost of housing is lower with higher collateral values. Furthermore, their consump-
tion of non durable good also increases since the marginal utility from consuming
one extra unit today is greater than the utility from waiting until tomorrow when
the borrowing constraint is binding.
1.4.2 Entrepreneurs
The external finance premium faced by entrepreneurs is given by equation 1.11 in
section 1.3.3.
efpt+1 = µe
´ !¯t+1
0 !t+1f(!t+1)d!t+1Qk,tKt
Qk,tKt Nt
Therefore @efpt+1@Qk,t < 0 i.e. the external finance premium is decreasing in the price
of capital in a neighbourhood of the steady state. Any shock that increases the
price of capital will tend to reduce the external finance premium thereby increasing
investment.
1.4.3 Interaction of households and firm credit constraints
Here, we highlight two channels by which firm and household financial frictions
interact with each other. First is through the loan market equilibrium condition.
dp,t(Rt, ⇤) = bi,t(Rt, qh,t+1, ⇤) + Bt(Rt, Qk,t, ⇤)
A rise in house price increases households demand for loans for any given risk
free interest rate. This puts an upward pressure on the risk-free interest rate. From
the bank loan participation constraint above, this increases financing cost for en-
trepreneurs. Therefore, loosening of households borrowing constraint in this case
37
would lead to higher financing cost for entrepreneurs i.e. a negative interaction
between household and firm financial frictions.
The second channel operates through impact of house prices on the demand for
capital. An increase in house price raises the marginal product of capital, qh,tIHtKh,t1 ,
in housing production. This increases the demand for capital in housing production
and consequently housing investment. There is a shift of resources from production
of consumption good to housing i.e. the use of capital in consumption good falls
while more capital is diverted to production of housing good. The impact on total
capital depends on which effect dominates. If total capital increases, the price of
capital and therefore entrepreneurs net worth rises as well. This leads to a decline
in entrepreneurs financing spread i.e. a positive interaction between household and
firm financial frictions.
1.5 Estimation
1.5.1 Methodology and Data
A second order approximation of the model is used, which is computed using Dyane
version 4.3.3. The solution takes form of a state-space model that is used to com-
pute the likelihood function. We use Bayesian approach, whereby we specify the
prior distribution for parameters and their posterior distribution is computed by
Metropolis-Hastings algorithm. We use six observables: real consumption, real busi-
ness investment, real housing investment, real house prices, hours worked in the
consumption sector and hours worked in the housing sector. The sample period runs
from 1970:1 to 2006:4. To make the data in accordance with the variables in our
model, we remove the trend from consumption, business non-residential investment,
residential investment and house prices using the HP filter. Also, we demean the
data series for hours worked in the consumption and housing sector6.
6Details on data series transformation along with their sources are presented in appendix 1.1.
38
Table 1 – Calibrated Parameters
Parameter Description Value
p Patient household discount factor 0.99
i Impatient household discount factor 0.97
k Capital depreciation rate 0.025
h Housing depreciation rate 0.015
↵ Share of capital in non-housing production 0.33
 Share of capital in housing production 0.10
⌧ Share of land in housing production 0.10
✓ Share of intermediate goods in housing production 0.10
Jh Steady state share of housing in the utility function 0.15
m LTV ratio 0.80
µe Fraction of realized profits lost in bankruptcy 0.30
V ar(log!) Variance of log of idiosyncratic productivity 0.25
⇥ Fraction of net worth consumed by entrepreneurs upon exit 0.10
W e Transfer from households to entering entrepreneurs 0.009
 Survival rate of entrepreneurs 0.97
1.5.2 Calibrated Parameters and Prior Distribution
We calibrate some of the model’s parameters to get reasonable values for key steady-
state values and ratios, while the rest of the parameters are estimated. Table 1
reports the calibrated values. The patient households discount factor is set to 0.99
to get the steady state real interest rate close to 4 percent. For impatient households,
we set the discount factor to be slightly lower at 0.97 to ensure that the borrowing
constraints bind for them. Jh, the weight attached to housing in the utility function
is set to have a value of 0.15. This together with the depreciation rate for housing,
h, pins down the ratio of residential investment to total output at about 6 percent
as observed in the data.
The capital share of production in the non-housing sector, ↵, and the capital
depreciation rate, k, is set to get reasonable value of the steady state I/GDP ratio.
In the housing sector, we choose a capital share of  = 0.10 and a land share
of ⌧ = 0.10 following Davis and Heathcote (2005) and Iacoviello and Neri (2010).
39
Finally the share of intermediate business goods, ✓, is set at 0.10 following Iacoviello
and Neri (2010).
Next we turn to parameters relating to financial frictions. The first parameter
is the loan to value ratio m. This parameter reflects the typical LTV for credit
constrained households who would borrow maximum against their housing stock.
Following Iacoviello and Neri (2010), we set this value at 80%. The values of the
parameters that control firm financial frictions (e.g.. , µe, F (!¯) and V ar(log(!))
are primarily determined to match, RF R, leverage ratio and the rate of return on
capital. Following popular literature (BGG (1999) and Christensen and Dib (2008)),
we use a value of 0.97 for the survival rate of entrepreneurs, , implying an expected
working life for entrepreneurs of 36 years. Bankruptcy cost parameter, µe, is set at
0.3 that is within the range of 0.20 - 0.36 that Carlstrom and Freust (1997) defend as
empirically relevant. F (!¯) is set at 0.75 quarterly percent value implying an annual
bankruptcy rate of 3 percent (as in BGG). There is considerable uncertainty about
the spread between the cost of external finance, RF and the return on households
deposits, R. BGG (1999) measure the external finance premium as approximately the
historical average spread between the prime lending rate and the six- month Treasury
bill rate, which amounts to 200 basis points. Levin, Natalucci and Zakrajsek (2004)
report a spread of 227 basis points for the median firm included in their sample. De
Fiore and Uhlig (2005) report that the spread between the prime rate on bank loans
to business and the commercial paper is 298 basis points over the period 1997-2003.
Carlstrom and Fuerst (1997) report a somewhat lower spread of 187 basis points.
Our calibrations produce a spread of 210 basis points. Finally, the calibrated values
of fraction of net worth consumed by entrepreneurs upon exit, ⇥, and the transfer
from households to entering entrepreneurs,W e, are borrowed from Christiano, Motto
and Rostagno (CMR, 2014).
Table 2 reports the key steady state steady state ratios. Consumption share of
GDP is 75% and the business investment share is 18.5%. Housing investment is close
to 6% percent of GDP. Furthermore, the chosen calibrated values imply a ratio of
business capital to GDP of about 1.51 and housing wealth to GDP of about 1.1 and
the ratio of business loans to GDP of 1.4. The rate of return on capital at 11.5
40
Table 2 – Key Steady State Results
Annual real interest rate 4⇥R 1 4.1%
Consumption share C/GDP 75%
Business investment share I/GDP 18.5%
Housing investment share qhIH/GDP 6.5%
Housing wealth qhH/GDP 4.34
Business capital in non-housing sector kc/GDP 6.05
Business capital in housing sector kh/GDP 0.11
leverage ratio K/N 1.29
Rate of return of Capital Rk 11.5%
Cost of external finance RF 6.2%
Firm borrowing/GDP B/GDP 1.40
percent is slightly higher than in the data (10.32 percent) as documented in Verona
et. al. (2014). Finally, the leverage ratio at 1.29 is very close to the value in Verona
et. al (2014) who document a range of 1.21 - 1.77 based on US data over the sample
1998:Q4 to 2003:Q4.
The priors used in the study are presented in Tables 3 and 4. The choice of priors
for the estimated parameters is usually determined by the theoretical implications
of the model and evidence from previous studies. For the standard deviation of
the shocks we use an inverse gamma distribution. For the parameters reflecting
persistence of the shocks, we use a beta distribution with a prior mean of 0.8 and a
standard error of 0.1. The inverse of labour supply elasticity for patient and impatient
households,p and i have a gamma prior with 0.5 mean and 0.1 standard deviation.
This implies an elasticity of hours aggregator of 2. For parameters reflecting the
inverse elasticity of substitution across hours in the two sectors, ⌘p and ⌘i, we set a
normal prior with a mean of 1, as estimated by Horvath (2000) and Iacoviello and
Neri (2010). Finally, the prior mean of the labour income share of the collateral
constrained households, µ, is set to 0.35, with a standard error of 0.05. This is close
to Iacoviello (2005) estimate of the wage share of collateral constrained agents of
around 36%, using a limited information approach.
41
Table 3 –Prior and Posterior Distribution of the Structural Parameters
Prior distribution Posterior distribution
Parameter Distribution Mean SD Mean 2.5% Median 97.5%
p Gamma 0.5 0.1 1.0961 0.7467 1.0890 1.4250
i Gamma 0.5 0.1 0.3953 1.2585 0.3885 0.5252
⌘p Normal 1 0.1 1.0947 0.9701 1.0935 1.2177
⌘i Normal 1 0.1 1.0251 0.8634 1.0256 1.1841
µ Beta 0.35 0.05 0.1873 0.1486 0.1866 0.2259
'x Gamma 5 2.5 16.88 14.12 16.82 19.56
1.5.3 Posterior Distributions
Table 3 and 4 gives the posterior values for the parameters. The posterior estimates
of labour supply elasticity, p and i are 1.09 and 0.39. The estimated mean of ⌘p
and ⌘i showing substitution between hours in the two sectors is close to 1 implying
relatively low preference for mobility across sectors. The value of µ, the labour
income share of credit constrained households in non-housing sector is around 19
percent, which is lower than the prior mean however close to the value estimated by
Iacoviello and Neri (2010). The estimated adjustment cost parameter is about 17,
which is in the range found in other estimated models with financial frictions - from
around 6 in De Graeve (2008) to around 30 in CMR (2014). Finally with regards to
the shocks, all shocks are fairly persistent.
1.6 Results
1.6.1 Model Comparison
We estimate several alternative versions of the model to see the impact of household
and firm financial frictions and the interaction between them. The baseline model
above is compared with (i) Only housing frictions - a model without firm financial
frictions i.e. restricting the value to V ar(log!) to zero. This ensures that firms’
probability of default is zero and hence lending to entrepreneurs is risk-less. The
spread between the prime lending rate and risk-less rate in this case is fixed at
42
Table 4 – Prior and Posterior Distribution of the Shock Parameters
Prior distribution Posterior distribution
Parameter Distribution Mean SD Mean 2.5% Median 97.5%
⇢ac Inv. gamma 0.8 0.1 0.7967 0.7500 0.7975 0.8437
⇢ah Inv. gamma 0.8 0.1 0.8641 0.8216 0.8652 0.9091
⇢h Inv. gamma 0.8 0.1 0.9125 0.8596 0.9168 0.9712
⇢z Inv. gamma 0.8 0.1 0.8182 0.6714 0.8290 0.9657
⇢& Inv. gamma 0.8 0.1 0.9951 0.9910 0.9956 0.9992
⇢⌘ Inv. gamma 0.8 0.1 0.8281 0.8002 0.8365 0.8563
⇢ Inv. gamma 0.8 0.1 0.6054 0.5490 0.6066 0.6446
ac Inv. gamma 0.01 0.05 0.0449 0.0404 0.0448 0.0493
ah Inv. gamma 0.01 0.05 0.0403 0.0364 0.0402 0.0443
h Inv. gamma 0.01 0.05 0.0784 0.0385 0.08 0.1327
z Inv. gamma 0.01 0.05 0.0085 0.0021 0.0062 0.0170
& Inv. gamma 0.01 0.05 0.0442 0.0394 0.0441 0.0490
r Inv. gamma 0.01 0.05 0.9927 0.8794 0.9889 1.0809
⌘ Inv. gamma 0.01 0.05 0.5044 0.4407 0.5026 0.5696
Table 5 – Model Comparison: Results from estimation
Marginal density/Model Baseline Only house Only firm No credit
frictions frictions frictions
Modified Harmonic Mean 1655.03 1360.34 1544.11 1385.23
Laplace Approximation 1655.38 1360.65 1545.09 1385.95
the steady state obtained from the baseline model. Therefore, the usual net worth
dynamics and financial accelerator effects in response to shocks is missing from this
model; (ii) Only firm frictions - a model without household leverage constraint i.e.
restricting the share of impatient households in the economy (µ = 0); and (iii) No
credit frictions - in this model both firm and households are not credit constrained
(V ar(log!) = 0, µ = 0). The four models are estimated using data from 1970Q1 to
2006Q4 and the marginal likelihood (see Table 5) suggests that the data favors our
baseline model with both household and firm financial frictions.
43
1.6.2 Posterior Impulse Response Analysis
The impulse response functions are based on averaging 1000 simulated series for
each variable. Dynare generates IRFs using the approximate policy functions as
follows: (i) It draws a series of all the exogenous shocks for a number of 100 + T
periods, where T is the number of periods shown in the IRF graphs, here T = 40,
(ii) It performs a simulation H1 of all the variables of interest using this realization
of the shocks, (iii) It changes the sequence of exogenous shocks, by adding a one-
standard deviation of the shock, to the realization of period 101, (iv) It performs a
new simulation H2 of all variables based on the new sequence of shocks, and (v) It
calculates the IRF from H2-H1. These steps are then repeated a number of times
(we choose 1000 replications) and the produced IRFs are averages of the series from
these 1000 experiments.
Non housing technology shock The response of aggregate variables to technol-
ogy shock is plotted in figure 1. A positive technology shock leads to an increase in
output and its components – consumption and investment, both residential and non-
residential. An estimated one standard deviation (0.04 percent) increase in aggregate
productivity increases GDP and consumption by 0.04 and 0.05 percent respectively
in the baseline model. Business investment increases by 0.02 percent while housing
investment increases by 0.002 percent. The firm financial accelerator works in its
well understood way. A positive technology shock leads to an increase in the price of
capital and hence firms’ net worth. This improves firms’ cost of funding, in partic-
ular, driving down the external finance premium leading to an increase in business
investment and output. We notice that the amplification of output, consumption
and business investment is maximum in the presence of both types of frictions. The
response of GDP and consumption in the presence of both household and firm fric-
tion is more than twice as compared to the frictionless economy. The amplification in
business investment is even more pronounced with firm financing frictions accounting
for most of it.
We now turn to households. A positive technology shock drives up both resi-
44
dential investment and house prices. The interaction between firm and household
frictions operates in different way for different variables. The effect is in the same di-
rection (more amplification) for output, consumption and business investment. This
is driven by the increase in the price of capital and housing, which is maximum in
the model with both frictions. The price of capital increases by 0.06 percent in the
baseline model while the increase is only about 0.02 percent in a frictionless setup.
Similarly, housing price increases by over 0.04 percent in baseline model and only
0.02 percent in frictionless economy. However, household financial frictions dampen
the response of housing investment in the initial periods after a shock. This result
stems from the consumption and labour supply decisions of the patient and impa-
tient households. A positive technology shock increases the wealth (by an increase in
house prices) and income (an increase in wages). Both patient and impatient house-
holds increase their consumption of production good in response to rising house
prices thereby displaying the importance of wealth effects. Patient household re-
spond to higher wealth and income by consuming more consumption good, selling
housing stock and increasing their supply of labour to both housing and non-housing
production. Impatient household increase their demand for housing in response to
loosening borrowing constraint. They further also reduce their labour supply (to
both consumption and housing production) leading to an overall reduction of labour
in the housing sector. In contrast, in models without impatient households labour
supply to housing production increases, translating into a higher housing investment.
The increase in house prices and housing investment leads to a higher demand of
housing capital, thus amplifying the demand for total capital boosting capital price
and non-residential investment. The peak response of total capital at 0.008 percent in
the baseline model is more than three times than in frictionless economy. The increase
in capital price and firm’s net worth is most pronounced in the model with both firm
and household financial accelerator. Therefore, while allowing for household financial
frictions leads to lower residential investment, it amplifies the response of business
investment to technology shocks. We see a similar amplification at play for output
and consumption, suggesting a positive interaction between financial accelerators
driven by firm and household financing constraints.
45
Housing demand shock (housing preference shock) We analyze the impact
of an estimated one standard deviation (0.1 percent) housing preference shock in
figure 2. In response to the shock, house prices rise by 0.02 percent in the baseline
model thereby leading to an increase in housing investment (also by 0.02 percent).
As the collateral value increases with increasing house prices, borrowing capacity
of impatient households increase. This raises both the borrowing and consumption
of impatient households. On the other side of the ledger, lenders need to mobilize
the extra resources consumed by the borrowers. They do so in response to a higher
interest rate, which induces them to consume less, sell part of their housing stock, and
work harder. The effect of lenders dominate and overall consumption falls however
the fall is lower in the model with household collateral constraints reflecting the
importance of wealth effects due to higher house prices.
Overall total housing demand is higher in the model with collateral effects, driven
primarily by the demand from impatient households. Their demand for housing in-
crease by over 0.05 percent in the baseline model in contrast to 0.04 percent in a
model with only housing frictions. Higher demand translates into a higher increase
in housing investment in the baseline (0.02 percent) as compared to the frictionless
model (0.015 percent), reflecting the role played by collateral effects in the amplifica-
tion process. The elasticity of residential investment to house prices in this model is
less than that found by Iacoviello and Neri (2010). They note that, the combination
of flexible housing prices and sticky wages in construction is required to have a high
sensitivity of residential investment to changes in demand conditions.
We next turn to the impact on firms. Given the loan market equilibrium (dp,t =
bi,t +Bt), an increase in impatient household borrowing puts an upward pressure on
interest rate. This therefore raises the cost of funding for entrepreneurs. The overall
impact on real business investment however is a cumulative effect of two forces. On
the one hand, capital in the housing sector Kh increases, with the increase being
amplified under collateral constraints. Housing capital increases by 0.04 percent in
the baseline model in contrast to 0.03 in frictionless economy. At the same time,
business capital declines with the fall being more under the presence of housing
financing shocks (as interest rates rise more). As mentioned earlier, the overall
46
impact depends on which effect dominates and whether the total capital stock rises
or falls. The highest increase in business investment is in the model with only
collateral constraints with a shift of resources (capital) from non-housing to housing
sector. The model with both frictions generates slightly positive business investment
after the first few quarters. While housing constraints play a major role in transfer
of resources from consumption to housing sector, the overall impact of this shock on
business investment is only marginal in the baseline model.
Housing technology shock A positive housing technology shock (plotted in fig-
ure 3) leads to a rise in housing investment and a fall in house prices (driven by a fall
in construction costs). A one standard deviation (0.05 percent) increase in housing
technology raises housing investment by 0.1 percent and reduces housing price by
0.01 percent in the baseline model. Once again the responses are amplified under the
presence of housing credit frictions. For instance, the increase in housing investment
in the baseline model is more than twice the impact in frictionless setting. With
regards to housing v/s non-housing capital, there is a shift of resources from non
housing to housing, which is further amplified in the presence of household financial
frictions. Therefore, there is an increase in housing capital Kh, which increases by 0.1
percent in baseline model, nearly twice the size of increase in frictionless case. How-
ever, non-housing capital Kc falls. Overall, these opposing forces nearly cancel each
other with the effect on business investment being small. While the change in total
capital is negligible in the case of no credit frictions, it moves in opposite directions
in the model with only firm and household friction respectively. In a model with
both firm and household frictions, the increase in housing capital is higher than the
fall in non-housing capital, leading to positive effect on net worth and capital price.
This acts as an amplification mechanism and increases business investment. Thus,
while looking at housing frictions in isolation would result in a negative comovement
between business and housing investment, treating them together in a model leads
to an increase in both in response to a positive housing technology shock.
A fall in house prices reduces the borrowing capacity and therefore the demand
for housing and consumption by impatient households. Patient households buy more
47
houses to maintain equilibrium in the housing market and reduce their consumption
of nondurable goods. Overall, there is a fall in consumption, with the fall mostly
driven by the presence of household financial frictions. Furthermore, housing technol-
ogy shocks also creates a tradeoff between wholesale production in non-housing and
housing sector. With a fall in consumption and business investment (in some cases),
there is a fall in economic activity in non-housing production. However, overall GDP
increases driven by an increase in housing production.
Investment shock A positive shock to marginal productivity of investment (Fig-
ure 4) increases real interest rate, giving households an incentive to save more and
postpone consumption. With lower consumption, the marginal utility of income in-
creases, shifting labour supply to the right - an inter-temporal substitution effect.
Along an unchanged labour demand schedule, this supply shift raises hours and out-
put. The investment shock leads to an increase in capital and a fall in capital price
depressing net worth. This causes investment to gradually fall back to its steady
state. The increase in capital and therefore business investment in higher in the
model with firm financing frictions. A one standard deviation (0.5 percent) positive
investment shock increases business investment by 0.05 percent in the baseline model
(more than twice the increase (0.02 percent) in frictionless setup). As the capital and
housing price decrease, impatient households hit their borrowing constraint - forcing
them to reduce consumption of both final consumption good and housing. The con-
sumption therefore falls more in the model with housing constraints. Overall, firm
and household credit constraints do not seem to have any important interaction in
response to an investment shock.
Other shocks The other three shocks included in the model are - risk shock,
preference shock and labour supply shock. A one standard deviation (0.99 percentage
point) risk shock decreases output and consumption by close to 0.3% (figure 5).
Business investment falls by 0.08% while housing investment falls by over 0.15%
on impact. An increase in entrepreneur’s idiosyncratic risk leads to higher default.
48
This causes a decline in their net worth and a fall in capital price and investment.
While risk shocks have quantitatively large effects, housing collateral constraints
do not play an important role in driving the response to this shock. As for the
responses of aggregate variables to preference and labour supply shock, our findings
resemble those of estimated DSGE models that do not include a housing sector
(Smets andWouters (2003), Christensen and Dib (2008)). Figure 6 shows the impulse
responses to a one standard deviation (0.009%) positive preference shock, which
raises the marginal utility of consumption and therefore the opportunity cost of
holding deposits (savings). As households divert deposits towards consumption, the
return on deposits (the risk-free real interest rate) rises. The responses of output and
consumption are almost identical in the models with and without financial frictions.
Finally, figure 7 shows the effects of a negative labour supply shock. The qualitative
effects of this supply shock on GDP, consumption and investment are very similar
to those of a negative productivity shock. The main qualitative differences are that
employment also falls in line with output. Again the results are not affected by
adding housing constraints to the model.
1.6.3 Comparison of results from first and second order ap-
proximation
In this section, we compare impulse responses generated from first order v/s second
order approximation of the model around the steady state. Figure 8 and 9 show
responses of key variables to the seven shocks described above. In figure 8, responses
to housing preference, housing technology and nonhousing technology shocks are un-
changed under first v/s second order approximation of the model. A similar trend
is seen in figure 9 where responses to preference shock, labour supply shock and
investment shock are mostly invariant to the order of approximation. Therefore, the
response of key macroeconomic variables to these shocks does not exhibit significant
non-linearities. The only exception is risk shock (figure 9) where non-linearities play
a big role - GDP and consumption under second order approximation fall about six
times more than under first order approximation. A one standard deviation (0.99
49
percentage point) increase in the risk reduces GDP by 0.05% and 0.3% under first
and second order approximation, respectively. It is important to note that second
order approximation may still not capture the non-linearities well as it is approxi-
mation within the neighbourhood of the deterministic steady state. To be able to
characterize the non-linearities in the model, we would require a global solution to
the model. Although this would certainly be a fruitful avenue for future research, it
is beyond the scope of this paper.
1.6.4 Robustness: Prior sensitivity
To check whether the results are robust to prior selection, we estimate the baseline
model again with same prior means but double the standard deviation. The results
are presented in table 6. The posterior median is largely robust to increasing standard
deviation of the prior distribution.
1.6.5 Variance Decomposition
The results from variance decomposition are presented in table 7. Non-housing tech-
nology shock explains nearly half of the variation in GDP, about 40 percent variation
in consumption and 18 percent variation in nonresidential investment. However, its
contribution towards explaining variation in other aggregate variables, particularly
relevant to housing market is low suggesting that other shocks are important in our
model too. Housing demand (housing preference) and supply (housing technology)
shocks together account for more than 70 percent of the variation in housing invest-
ment and about 50 percent of the variation in house prices. Investment shock also
plays an important role, explaining about 50 percent of the variance of non-housing
investment and 10 percent of the variance of housing investment. It also explains
close to 20 percent of the variation in consumption and house prices and 15 percent
of the variation in output. Finally, risk shock explains more than 20 percent of the
variation in output, consumption and non-housing investment. Thus, total factor
productivity shock together with investment and risk shocks explain majority of the
variation in GDP, consumption and nonresidential investment at business cycle fre-
50
Table 6 – Posterior median of baseline model: Prior sensitivity
Baseline Sensitivity w.r.t prior std. dev
Structural Parameters Structural Parameters
Parameter Mean Std. dev Median Mean Std. dev Median
p 0.5 0.1 1.0890 0.5 0.2 1.6235
i 0.5 0.1 0.3885 0.5 0.2 0.3236
⌘p 1 0.1 1.0935 1 0.2 1.1756
⌘i 1 0.1 1.0256 1 0.2 1.0247
µ 0.35 0.05 0.1866 0.35 0.1 0.1506
'x 5 2.5 16.82 5 5 16.05
Shock Parameters Shock Parameters
⇢ac 0.8 0.1 0.7975 0.8 0.2 0.7894
⇢ah 0.8 0.1 0.8652 0.8 0.2 0.8435
⇢h 0.8 0.1 0.9168 0.8 0.2 0.9533
⇢z 0.8 0.1 0.8290 0.8 0.2 0.8728
⇢& 0.8 0.1 0.9956 0.8 0.2 0.9976
⇢⌘ 0.8 0.1 0.8365 0.8 0.2 0.8299
⇢ 0.8 0.1 0.6066 0.8 0.2 0.6108
ac 0.01 0.05 0.0448 0.01 0.1 0.0450
ah 0.01 0.05 0.0402 0.01 0.1 0.0435
h 0.01 0.05 0.08 0.01 0.1 0.0614
z 0.01 0.05 0.0062 0.01 0.1 0.0062
& 0.01 0.05 0.0441 0.01 0.1 0.0410
r 0.01 0.05 0.9889 0.01 0.1 0.9723
⌘ 0.01 0.05 0.5026 0.01 0.1 0.5031
51
quency. However, dynamics of the housing market are predominantly explained by
housing demand and supply shocks.
1.7 Concluding Remarks
In this paper, we develop and estimate a dynamic stochastic general equilibrium
model of firm and household leverage. We introduce a multi sector production econ-
omy with housing and non-housing goods. The demand side of the model embodies
financial frictions for both household and firms. Financial frictions for households
appear in the form of borrowing based on collateralisable asset (housing). For firms,
our model generates an endogenous external finance premium as a function of firm’s
leverage. The fluctuations of the asset prices (housing and capital) affect the value
of the collateral and hence, the borrowing capacities of these agents. We see that
the presence of these financial frictions amplifies the impact of shocks that originate
elsewhere in the economy. We observe a strong financial accelerator mechanism in
operation, wherein inclusion of both housing and firm financial frictions amplify the
shocks hitting the economy as opposed to just one. There are two important ways
in which firm and household financial frictions interact with each other. First is
via the loan market equilibrium condition, a tighter credit constraint on households
encourage them to reduce their demand for loans and increase their labour supply.
This reduces the financing costs faced by firms. Secondly, an increase in housing
demand and investment (that is amplified with collateral constraints) increases the
demand for housing capital thereby boosting capital price. An increase in capital
price improves the net worth position of firms and therefore reduces the external
finance premium they face.
We also notice interesting dynamics with regards to resources being devoted to
housing v/s non-housing production. Shocks related to housing market that increase
the demand and price of housing cause capital to be diverted from non-housing to
housing production. There is an increase in housing capitalKh and a fall in capital in
the consumption sector, Kc, with the overall effect on business investment depending
on which effect dominates. We further find the effect of housing market frictions to
52
be predominantly concentrated on consumption and housing investment. There is
a substantial spillover effect of house prices on non-housing consumption pointing
towards the importance of wealth effects on household consumption.
At cyclical frequencies, we observe that both housing prices and housing in-
vestment are strongly procyclical as in the literature (Iacoviello and Neri (2010)).
Further, business and housing investment tend to co-move. Finally, we find that
non-housing technology shocks are most important in explaining variations in GDP,
consumption and business investment. However housing demand and supply shocks
account for more than 50% of the cyclical volatility of housing investment and hous-
ing prices.
53
Figure 1 – Impulse responses to the estimated non-housing technology
shock
0 10 20 30 40
0
0.01
0.02
0.03
0.04
0.05
GDP
0 10 20 30 40
0
0.01
0.02
0.03
0.04
0.05
Consumption
0 10 20 30 40
−5
0
5
10
15
20
x 10−3 Business Investment
0 10 20 30 40
0
2
4
6
x 10−3 Housing Investment
0 10 20 30 40
−2
0
2
4
6
8
x 10−3 Business Capital
0 10 20 30 40
−2
0
2
4
6
8
x 10−3 Housing Capital
0 10 20 30 40
0
0.02
0.04
0.06
0.08
Net Worth
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
0.08
Capital Price
0 10 20 30 40
0
0.01
0.02
0.03
0.04
0.05
Housing Price
0 10 20 30 40
0
2
4
6
8
x 10−3 Total Capital
0 10 20 30 40
−0.03
−0.02
−0.01
0
0.01
Housing Patient
0 10 20 30 40
0
0.05
0.1
0.15
0.2
Housing Impatient
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−3 Real Interest Rates
0 10 20 30 40
0
0.05
0.1
0.15
0.2
Borrowing Impatient
0 10 20 30 40
−6
−4
−2
0
2
x 10−3Labour − Consumption
0 10 20 30 40
−6
−4
−2
0
2
4
x 10−3 Labour − Housing
 
 
Baseline Only Housing Frictions Only Firm Frictions No Credit Frictions
Note: IRFs to the estimated one standard deviation (0.04%) non-housing technology
shock. Based on averaging one thousand simulation of a second-order approximation
of the model around the steady state. Y-axis denotes percentage deviation from the
steady state i.e. a 0.04% increase in non-housing technology increases GDP by 0.04%
in the baseline model. X-axis is time in quarters.
54
Figure 2 – Impulse responses to the estimated housing preference shock
0 10 20 30 40
−2
0
2
4
6
8
x 10−4 GDP
0 10 20 30 40
−6
−4
−2
0
2
x 10−4 Consumption
0 10 20 30 40
−4
−2
0
2
4
x 10−4 Business Investment
0 10 20 30 40
−0.01
0
0.01
0.02
0.03
Housing Investment
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−4 Business Capital
0 10 20 30 40
−0.01
0
0.01
0.02
0.03
0.04
Housing Capital
0 10 20 30 40
−6
−4
−2
0
2
x 10−3 Net Worth
0 10 20 30 40
−4
−3
−2
−1
0
1
x 10−3 Capital Price
0 10 20 30 40
−5
0
5
10
15
20
x 10−3 Housing Price
0 10 20 30 40
−5
0
5
10
15
x 10−5 Total Capital
0 10 20 30 40
−15
−10
−5
0
5
x 10−3 Housing Patient
0 10 20 30 40
0
0.02
0.04
0.06
Housing Impatient
0 10 20 30 40
−5
0
5
10
15
x 10−4 Real Interest Rate
0 10 20 30 40
0
0.02
0.04
0.06
0.08
Borrowing Impatient
0 10 20 30 40
−6
−4
−2
0
2
4
x 10−4 Labour Consumption
0 10 20 30 40
−5
0
5
10
15
20
x 10−3 Labour Housing
 
 
Baseline Only Housing Frictions Only Firm Frictions No Credit Frictions
Note: IRFs to the estimated one standard deviation (0.1%) housing preference shock.
Based on averaging one thousand simulation of a second-order approximation of the
model around the steady state. Y-axis denotes percentage deviation from the steady
state i.e. a 0.1% increase in demand for housing increases GDP by 0.0007% in the
baseline model. X-axis is time in quarters.
55
Figure 3 – Impulse responses to the estimated housing technology
shock
0 10 20 30 40
−2
0
2
4
6
x 10−3 GDP
0 10 20 30 40
−15
−10
−5
0
5
x 10−4 Consumption
0 10 20 30 40
−1
−0.5
0
0.5
1
x 10−3 Business Investment
0 10 20 30 40
−0.05
0
0.05
0.1
0.15
Housing Investment
0 10 20 30 40
−20
−15
−10
−5
0
5
x 10−4 Business Capital
0 10 20 30 40
−0.05
0
0.05
0.1
0.15
Housing Capital
0 10 20 30 40
−1
0
1
2
3
4
x 10−3 Net Worth
0 10 20 30 40
−1
0
1
2
3
x 10−3 Capital Price
0 10 20 30 40
−0.015
−0.01
−0.005
0
Housing Price
0 10 20 30 40
−4
−2
0
2
4
x 10−4 Total Capital
0 10 20 30 40
0
0.005
0.01
0.015
Housing Patient
0 10 20 30 40
−0.03
−0.02
−0.01
0
0.01
0.02
Housing Impatient
0 10 20 30 40
−10
−5
0
5
x 10−4 Real Interest Rates
0 10 20 30 40
−0.04
−0.03
−0.02
−0.01
0
0.01
Borrowing Impatient
0 10 20 30 40
−5
0
5
10
15
x 10−4 Labour Consumption
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
Labour Housing
 
 
Baseline Only Housing Frictions Only Firm Frictions No Credit Frictions
Note: IRFs to the estimated one standard deviation (0.05%) housing technology
shock. Based on averaging one thousand simulation of a second-order approximation
of the model around the steady state. Y-axis denotes percentage deviation from the
steady state i.e. a 0.05% increase in housing technology increases GDP by 0.006%
in the baseline model. X-axis is time in quarters.
56
Figure 4 – Impulse responses to the estimated investment shock
0 10 20 30 40
0
0.01
0.02
0.03
0.04
GDP
0 10 20 30 40
−0.1
−0.05
0
0.05
0.1
Consumption
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
Business Investment
0 10 20 30 40
−0.04
−0.02
0
0.02
Housing Investment
0 10 20 30 40
0
0.02
0.04
0.06
0.08
0.1
Business Capital
0 10 20 30 40
−0.1
−0.05
0
0.05
0.1
Housing Capital
0 10 20 30 40
−0.4
−0.3
−0.2
−0.1
0
0.1
Net Worth
0 10 20 30 40
−0.3
−0.2
−0.1
0
0.1
Capital Price
0 10 20 30 40
−0.1
−0.05
0
0.05
0.1
Housing Price
0 10 20 30 40
0
0.02
0.04
0.06
0.08
0.1
Total Capital
0 10 20 30 40
−0.05
0
0.05
0.1
0.15
Housing Patient
0 10 20 30 40
−0.15
−0.1
−0.05
0
0.05
Housing Impatient
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
Real Interest Rates
0 10 20 30 40
−0.2
−0.15
−0.1
−0.05
0
0.05
Borrowing Impatient
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
Labour Consumption
0 10 20 30 40
−0.04
−0.03
−0.02
−0.01
0
0.01
Labour Housing
 
 
Baseline Only Housing Frictions Only Firm Frictions No Credit Frictions
Note: IRFs to the estimated one standard deviation (0.5%) investment shock. Based
on averaging one thousand simulation of a second-order approximation of the model
around the steady state. Y-axis denotes percentage deviation from the steady state
i.e. a 0.5% increase in the efficiency with which final good is converted into capital,
increases GDP by 0.03% in the baseline model. X-axis is time in quarters.
57
Figure 5 – Impulse responses to the estimated risk shock
0 10 20 30 40
−0.4
−0.3
−0.2
−0.1
0
0.1
GDP
0 10 20 30 40
−0.3
−0.2
−0.1
0
0.1
0.2
Consumption
0 10 20 30 40
−0.1
−0.05
0
0.05
0.1
Business Investment
0 10 20 30 40
−0.2
−0.15
−0.1
−0.05
0
0.05
Housing Investment
0 10 20 30 40
−0.01
−0.005
0
0.005
0.01
Business Capital
0 10 20 30 40
−0.6
−0.4
−0.2
0
0.2
Housing Capital
0 10 20 30 40
−2
−1.5
−1
−0.5
0
0.5
Net Worth
0 10 20 30 40
−1.5
−1
−0.5
0
0.5
Capital Price
0 10 20 30 40
−0.3
−0.2
−0.1
0
0.1
0.2
Housing Price
0 10 20 30 40
−0.01
−0.005
0
0.005
0.01
Total Capital
0 10 20 30 40
−0.08
−0.06
−0.04
−0.02
0
0.02
Housing Patient
0 10 20 30 40
−0.6
−0.4
−0.2
0
0.2
Real Interest Rate
0 10 20 30 40
−0.2
−0.1
0
0.1
0.2
0.3
Labour Consumption
0 10 20 30 40
−0.1
−0.05
0
0.05
0.1
Labour Housing
 
 
Baseline Only Firm Frictions
Note: IRFs to the estimated one standard deviation (0.99 percentage point) risk
shock. Based on averaging one thousand simulation of a second-order approximation
of the model around the steady state. Y-axis denotes percentage deviation from the
steady state i.e. a 0.99 percentage point increase in the standard deviation of firms’
idiosyncratic shock reduces GDP by 0.3% in the baseline model. X-axis is time in
quarters.
58
Figure 6 – Impulse responses to the estimated preference shock
0 10 20 30 40
−1
−0.8
−0.6
−0.4
−0.2
0
x 10−3 GDP
0 10 20 30 40
−5
0
5
10
x 10−4 Consumption
0 10 20 30 40
−20
−15
−10
−5
0
5
x 10−4 Business Investment
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−3 Housing Investment
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−4 Business Capital
0 10 20 30 40
−0.015
−0.01
−0.005
0
Housing Capital
0 10 20 30 40
−10
−5
0
5
x 10−3 Net Worth
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−3 Capital Price
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−3 Housing Price
0 10 20 30 40
−8
−6
−4
−2
0
x 10−4 Total Capital
0 10 20 30 40
−1
0
1
2
3
x 10−3 Housing Patient
0 10 20 30 40
−0.02
−0.015
−0.01
−0.005
0
Housing Impatient
0 10 20 30 40
0
0.5
1
1.5
x 10−3 Real Interest Rate
0 10 20 30 40
−0.025
−0.02
−0.015
−0.01
−0.005
0
Borrowing Impatient
0 10 20 30 40
−8
−6
−4
−2
0
2
x 10−4 Labour Consumption
0 10 20 30 40
−6
−4
−2
0
2
x 10−3 Labour Housing
 
 
Baseline Only Housing Frictions Only Firm Frictions No Credit Frictions
Note: IRFs to the estimated one standard deviation (0.009%) preference shock.
Based on averaging one thousand simulation of a second-order approximation of
the model around the steady state. Y-axis denotes percentage deviation from the
steady state i.e. a positive 0.009% preference shock reduces GDP by 0.0006%. X-axis
is time in quarters.
59
Figure 7 – Impulse responses to the estimated labour supply shock
0 10 20 30 40
−0.02
−0.015
−0.01
−0.005
0
GDP
0 10 20 30 40
−0.02
−0.015
−0.01
−0.005
0
Consumption
0 10 20 30 40
−0.015
−0.01
−0.005
0
Business Investment
0 10 20 30 40
−0.04
−0.03
−0.02
−0.01
0
0.01
Housing Investment
0 10 20 30 40
−6
−4
−2
0
2
x 10−3 Business Capital
0 10 20 30 40
−0.03
−0.02
−0.01
0
Housing Capital
0 10 20 30 40
−0.03
−0.02
−0.01
0
Net Worth
0 10 20 30 40
−0.03
−0.02
−0.01
0
0.01
Capital Price
0 10 20 30 40
−15
−10
−5
0
5
x 10−3 Housing Price
0 10 20 30 40
−5
−4
−3
−2
−1
0
x 10−3 Total Capital
0 10 20 30 40
−5
0
5
10
x 10−3 Housing Patient
0 10 20 30 40
−0.04
−0.03
−0.02
−0.01
0
Housing Impatient
0 10 20 30 40
0
0.5
1
1.5
2
x 10−3 Real Interest Rate
0 10 20 30 40
−0.05
−0.04
−0.03
−0.02
−0.01
0
Borrowing Impatient
0 10 20 30 40
−0.025
−0.02
−0.015
−0.01
−0.005
0
Labour Consumption
0 10 20 30 40
−0.04
−0.03
−0.02
−0.01
0
0.01
Labour Housing
 
 
Baseline Only Housing Frictions Only Firm Frictions No Credit Frictions
Note: IRFs to the estimated one standard deviation (0.04%) negative labour supply
shock. Based on averaging one thousand simulation of a second-order approximation
of the model around the steady state. Y-axis denotes percentage deviation from the
steady state i.e. a negative 0.04% labour supply shock reduces GDP by 0.015%.
X-axis is time in quarters.
60
Figure 8 – Impulse response functions under first order v/s second
order approximation
Impulse response to the estimated housing preference shock
0 10 20 30 40
0
1
2
3
4
5
6
7
8
x 10−4 GDP
0 10 20 30 40
−5
−4
−3
−2
−1
0
1
2
x 10−4 Consumption
0 10 20 30 40
−3
−2
−1
0
1
2
3
4
x 10−4 Business Investment
0 10 20 30 40
−0.005
0
0.005
0.01
0.015
0.02
0.025
Housing Investment
Impulse response to the estimated housing technology shock
0 10 20 30 40
0
1
2
3
4
5
6
x 10−3 GDP
0 10 20 30 40
−12
−10
−8
−6
−4
−2
0
2
x 10−4 Consumption
0 10 20 30 40
−2
−1
0
1
2
3
4
5
6
x 10−4 Business Investment
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
Housing Investment
Impulse response to the estimated nonhousing technology shock
0 10 20 30 40
0
0.01
0.02
0.03
0.04
0.05
GDP
0 10 20 30 40
0
0.01
0.02
0.03
0.04
0.05
Consumption
0 10 20 30 40
−5
0
5
10
15
20
x 10−3 Business Investment
0 10 20 30 40
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
x 10−3 Housing Investment
 
 
Second Order First Order
Note: IRFs to the estimated one standard deviation shock based on first and second
order approximation (averaging one thousand simulation) of the model around the
steady state. X-axis is time in quarters.
61
Figure 9 – Impulse response functions under first order v/s second
order approximation (cont.)
Impulse response to the estimated investment shock
0 10 20 30 40
0.01
0.015
0.02
0.025
0.03
0.035
GDP
0 10 20 30 40
−0.06
−0.04
−0.02
0
0.02
0.04
Consumption
0 10 20 30 40
−0.02
0
0.02
0.04
0.06
0.08
Business Investment
0 10 20 30 40
−0.02
−0.015
−0.01
−0.005
0
0.005
0.01
0.015
0.02
Housing Investment
Impulse response to the estimated risk shock
0 10 20 30 40
−0.35
−0.3
−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
GDP
0 10 20 30 40
−0.3
−0.2
−0.1
0
0.1
0.2
Consumption
0 10 20 30 40
−0.1
−0.08
−0.06
−0.04
−0.02
0
0.02
0.04
Business Investment
0 10 20 30 40
−0.2
−0.15
−0.1
−0.05
0
0.05
Housing Investment
Impulse response to the estimated preference shock
0 10 20 30 40
−6
−5
−4
−3
−2
−1
0
x 10−4 GDP
0 10 20 30 40
−2
0
2
4
6
8
10
x 10−4 Consumption
0 10 20 30 40
−14
−12
−10
−8
−6
−4
−2
0
2
x 10−4 Business Investment
0 10 20 30 40
−6
−5
−4
−3
−2
−1
0
1
x 10−3 Housing Investment
Impulse response to the estimated labour supply shock
0 10 20 30 40
−0.016
−0.014
−0.012
−0.01
−0.008
−0.006
−0.004
−0.002
0
GDP
0 10 20 30 40
−0.02
−0.015
−0.01
−0.005
0
Consumption
0 10 20 30 40
−0.012
−0.01
−0.008
−0.006
−0.004
−0.002
0
Business Investment
0 10 20 30 40
−0.035
−0.03
−0.025
−0.02
−0.015
−0.01
−0.005
0
0.005
Housing Investment
 
 
Second Order First Order
Note: IRFs to the estimated one standard deviation shock based on first and second
order approximation (averaging one thousand simulation) of the model around the
steady state. X-axis is time in quarters.
62
Table 7 – Variance Decomposition (in percent)
Variable/Shock "ac "ah "h "z "& "⌘ "
GDP 44.50 3.29 4.28 1.96 10.51 15.26 20.2
C 42.24 1.94 0.85 1.40 6.73 21.14 25.7
I 18.37 0.01 0.26 4.11 7.42 47.35 22.48
IH 0.78 48.79 28.73 3.56 2.02 9.88 6.24
qh 14.55 30.32 21.99 3.16 2.19 19.28 8.51
63
Appendix 1.1: Data and Sources
Aggregate Consumption: Real personal consumption expenditure (seasonally ad-
justed, billions of chained 2009 dollars), divided by the civilian non-institutional
population (CNP16OV). Source: Bureau of Economic Analysis (BEA).
Business Investment : Real private fixed investment - Nonresidential (seasonally
adjusted, billions of chained 2009 dollars), divided by the civilian non-institutional
population (CNP16OV). Source: Bureau of Economic Analysis (BEA).
Residential Investment : Real private fixed investment - Residential (seasonally
adjusted, billions of chained 2009 dollars), divided by the civilian non-institutional
population (CNP16OV). Source: Bureau of Economic Analysis (BEA).
Real House Prices: Census Bureau house price index deflated with the implicit
price deflator for the non-farm business sector.
Source: Census Bureau, http://www.census.gov/const/price_sold_cust.xls.
Hours worked in Consumption-good Sector : Total Non-farm Payrolls (Series ID:
PAYEMS in Saint Louis Fed - FRED) less all employees in the construction sector
(Series ID: USCONS), times average weekly hours of production workers (Series ID:
CES0500000005), divided by CNP16OV. Demeaned. Source: BLS.
Hours worked in Housing Sector : All employees in the construction sector (Series
ID: USCONS in Saint Louis Fed - FRED), times average weekly hours of construction
workers (Series ID: CES2000000005), divided by CNP16OV. Demeaned. Source:
BLS.
64
Appendix 1.2: Model Equations
Here, we summarize the equations describing the equilibrium of the model. The
competitive equilibrium of our model is a set of endogenous variables Bt, cp,t, ci,t, ce,t,
dp,t, efpt, GDPt, hi,t, hp,t, Ht, It, IHt, Kt, kb,t, kc,t, kh,t, t, lat, xt, lic,t, lih,t, lpc,t, lph,t,
Nt, !¯t, pb,t, pla,t, qh,t, qk,t, rk,t, rkc,t, rkh,t, Rla,t, Rt, RF,t, Rk,t, wic,t, wih,t,wpc,t, wph,t, ⇣e,t,
Yt satisfying the relations (1.21) to (1.62), given the exogenous stochastic processes
Ac,t, Ah,t, zt, &t, Jh,t, ⌘k,t, t and the initial conditions bi.t1, dp,t1, hi,t1, hp,t1,
Ht1, Kt1, kc,t1, kh,t1, lat1, Nt1, rt1, It1, qk,t1. The endogenous variables
jointly satisfy the following equations that include first-order conditions for patient
households, impatient households, entrepreneurs, firms, capital goods producers, and
the market clearing conditions.
A competitive equilibrium consists of a set of prices and interest rates {pb,t, pla,t, qh,t,
qk,t, Rla,t, Rt, rk,t, wic,t, wih,t, wpc,t, wph,t} for all possible states and for all t  0 such
that all markets clear when all agents i.e. patient households, impatient households,
entrepreneurs, firms and capital goods producer solve their maximization problems
while taking these prices and interest rates as given.
Patient Households
The budget constraint of patient households is:
cp,t+ qh,t(hp,t (1 h)hp,t1) + dp,t+ pla,tlat+ kb,t  wpc,tlpc,t+wph,tlph,t+Rt1dp,t1
+pb,tkb,t + (pla,t +Rla,t)lat1 (1.21)
The first order conditions faced by the patient household can be summarized by:
zt
cp,t
= Et(
pzt+1Rt
cp,t+1
) (1.22)
65
ztqh,t
cp,t
=
ztJh,t
hp,t
+ pEt(
zt+1qh,t+1(1 h)
cp,t+1
) (1.23)
pb,t  1
cp,t
= 0 (1.24)
ztpla,t
cp,t
= Et(
pzt+1
cp,t+1
(pla,t+1 +Rla,t+1)) (1.25)
wpc,t
cp,t
= &tl
⌘p
pc,t(l
1+⌘p
pc,t + l
1+⌘p
ph,t )
(p⌘p)
(1+⌘p) (1.26)
wph,t
cp,t
= &tl
⌘p
ph,t(l
1+⌘p
pc,t + l
1+⌘p
ph,t )
(p⌘p)
(1+⌘p) (1.27)
Impatient Households
The budget constraint of the impatient households is:
ci,t + qh,t(hi,t  (1 h)hi,t1) +Rt1bi,t1  wic,tlic,t + wih,tlih,t + bi,t (1.28)
The first order conditions for impatient households are as follows7:
zt
ci,t
= Et(
zt+1iRt
ci,t+1
) + tRt (1.29)
ztqh,t
ci,t
= Et(
izt+1qh,t+1(1 h)
ci,t+1
+ tm¯qh,t+1) +
ztJh,t
hi,t
(1.30)
wic,t
ci,t
= &tl
⌘i
ic,t(l
1+⌘i
ic,t + l
1+⌘i
ih,t )
(i⌘i)
(1+⌘i) (1.31)
7t is the Lagrange multiplier on the borrowing constraint faced by impatient households.
66
wih,t
ci,t
= &tl
⌘i
ih,t(l
1+⌘i
ic,t + l
1+⌘i
ih,t )
(i⌘i)
(1+⌘i) (1.32)
bi,t =
m¯qh,t+1hi,t
Rt
(1.33)
Entrepreneurs
The law of motion of aggregate entrepreneurial net worth8 is:
Nt+1 = {Rk,t+1Qk,tKt [Rt+µe
´ !¯t+1
0 !t+1f(!t+1)d!t+1Qk,tKt
Qk,tKt Nt ](Qk,tKtNt)}+W
e
(1.34)
Their aggregate borrowing is:
Bt = Qk,tKt Nt (1.35)
The first order conditions for entrepreneur yields:
G0(!¯t+1)) + ⇣e,t(G0(!¯t+1) µeG0(!¯t+1)) = 0 (1.36)
(1 G(!¯t+1))Rk,t+1 + ⇣e,t[(G(!¯t+1) µeG(!¯t+1))Rk,t+1 Rt] = 0 (1.37)
Where ⇣e,t is the Lagrange multiplier associated with the participation constraint:
Et[(G(!¯t+1) µeG(!¯t+1))]Rk,t+1xt = Rt(xt  1) (1.38)
The aggregate consumption of entrepreneurs that exit the economy in t+1 is
given by:
8The aggregation of net worth, capital and borrowing is discussed in section 1.3.3.
67
ce,t+1 = ⇥
(1 )

(Nt+1 W e) (1.39)
The external finance premium for loans between period t and t+1 is:
efpt+1 = µe
´ !¯t+1
0 !t+1f(!t+1)d!t+1Qk,tKt
Qk,tKt Nt (1.40)
The state contingent interest rate, RF,t is given by:
RF,t = !¯t+1Rk,t+1.(
xt
xt  1) (1.41)
and
Rk,t+1 =
rk,t+1 + (1 k)Qk,t+1
Qk,t
(1.42)
Firms
The production technology for non-housing and housing are respectively:
Yt = Ac,tK
↵
c,t1(l
µ
ic,tl
1µ
pc,t )
1↵ (1.43)
IHt = Ah,t(l
µ
ih,tl
1µ
ph,t )
1✓⌧Kh,t1k
✓
b,tla
⌧
t1 (1.44)
The first order conditions are:
wic,t =
µ(1 ↵)Yt
lic,t
(1.45)
wpc,t =
(1 µ)(1 ↵)Yt
lpc,t
(1.46)
rkc,t =
↵Yt
Kc,t1
(1.47)
68
wih,t =
µ(1   ✓  ⌧)qh,tIHt
lih,t
(1.48)
wph,t =
(1 µ)(1   ✓  ⌧)qh,tIHt
lph,t
(1.49)
rkh,t =
qh,tIHt
Kh,t1
(1.50)
Rla,t =
⌧qh,tIHt
lat1
(1.51)
pb,t =
✓qh,tIHt
kb,t
(1.52)
Capital goods producer
The evolution of capital stock is given by:
Kt+1 = (1 k)Kt + ⌘k,t(1 S(Gt))It (1.53)
and the inter-temporal profit maximization with respect to It yields:
Qk,t(1 S(Gt)XtS 0(Gt)) + Et[Dt,t+1Qk,tS 0(Gt+1)I
2
t+1
I2t
] =
1
⌘k,t
(1.54)
Market clearing
Patient and impatient households are available in unit mass each. Housing in-
vestment in the economy is given by:
IHt = Ht  (1 h)Ht1 (1.55)
Ht = hp,t + hi,t (1.56)
69
Equilibrium in the housing market implies that the housing stock (undepreciated
old stock and the new housing investment) is equal to the sum of demand for housing
from patient and impatient household.
Resource constraint (goods market equilibrium) is obtained by adding budget
constraints for all agents.
Yt = Ct + It + kb,t + µeG(!¯t)Rk,tQk,t1Kt1 (1.57)
Loan market equilibrium is:
dp,t = bi,t +Bt (1.58)
Total land is normalized to unity.
lat = 1 (1.59)
Total capital is the sum of capital in non-housing and housing sector whereby
equilibrium in capital services market requires that the supply from entrepreneurs is
equal to the capital demanded by firms.
Kt = Kc,t +Kh,t (1.60)
GDP (net of monitoring costs) is defined as the value added of the two sectors.
GDPt = Yt  kb,t + qIHt (1.61)
Equilibrium in the labor market requires that labor demanded by firms is equal
to the labor supplied by households (both patient and impatient) at their respective
competitive real wage. Finally, equilibrium in the market for capital requires that
the marginal product is same across both the sectors, i.e.
rk,t = rkc,t = rkh,t (1.62)
70
Appendix 1.3: Steady State
Entrepreneurs and Net Worth
In equilibrium, optimization by capital goods producers implies capital price of
one.
qk = 1
Following popular studies BGG(1999), Christensen and Dib (2008), we pin down
the steady state capital to net-worthK/N ratio. This together with the entrepreneurs
loan contract gives the values of !¯ and RK,t. Therefore:
rk = Rk  (1 k)
Furthermore, in equilibrium marginal return of capital in the consumption and
housing good production is equalized.
rkc = rkh = rk
The Euler equations for the demand of capital from non-housing and housing
sector can then be used to derive the following ratios:
S0 =
Kc
Y
=
↵
(Rk  (1 k))
S1 =
Kh
qhIH
=

Rk  (1 k)
Finally, setting land la = 1, we have
Rla = ⌧qhIH
Households
71
Marginal utility of consumption and housing are equal to 1/c and Jh/h respec-
tively. This gives the steady state level of real interest rate as
R =
1
p
The Euler equation of hp yields
S2 =
qhhp
cp
=
Jh
(1 p(1 h))
Euler equation for hi and bi
S3 =
qhhi
ci
=
Jh
1 i(1 h)m(p  i)
 =
(p  i)
ci
From the above ratios, we have:
Kc = S0Y
Kh = S1qhIH
qhhp = S2cp
qhhi = S3ci
From housing market clearing condition, we get:
h(qhhp + qhhi) = qhIH
The loan market equilibrium yields:
dp = pmqhhi + (K N)
72
Using K = Kc +Kh, we get
S4 =
K
Y
= S0 + S1h(S2
cp
Y
+ S3
ci
Y
)
The budget constraints of the two types of households solve:
cp + hqhhp =
X
wl + ⌧qhIH + dp(R 1)
ci + hqhhi =
X
wl +mqhhi(p  1)
Eliminating hp and hi, we get:
h(S2cp + S3ci) = qhIH
cp + hS2cp =
X
wl + ⌧qhIH + (pmS3ci + S4Y (1 nk))(R 1)
ci + hS3ci =
X
wl +mS3ci(p  1)
The Euler equations for labour demand provide following steady state relations:
wic =
µ(1 ↵)Y
lic
wpc =
(1 µ)(1 ↵)Y
lpc
wih =
µ(1   ✓  ⌧)qhIH
lih
wph =
(1 µ)(1   ✓  ⌧)qhIH
lph
73
Further, we saw above that we need the total wage bill earned by each group to
compute the steady state:
wpclpc + wphlph = (1 µ)((1 ↵)Y + (1   ✓  ⌧)qhIH)
wiclic + wihlih = µ((1 ↵)Y + (1   ✓  ⌧)qhIH)
h(S2cp + S3ci) = qhIH
cp+hS2cp = (1µ)((1↵)Y+(1✓⌧)qhIH)+⌧qhIH+(pmS3ci+S4Y (1nk))(R1)
ci + hS3ci = µ((1 ↵)Y + (1   ✓  ⌧)qhIH) +mS3ci(p  1)
Using the formula for qhIH, we have:
cp+ hS2cp = (1 µ)((1↵)Y + (1  ✓ ⌧)h(S2cp+S3ci)) + ⌧h(S2cp+S3ci)+
(pmS3ci + (S0Y + S1h(S2cp + S3ci))(1 nk))(R 1)
ci + hS3ci = µ((1 ↵)Y + (1   ✓  ⌧)h(S2cp + S3ci)) +mS3ci(p  1)
These two equations can be used to derive consumption/output rations:
cp(1+hS2(1µ)(1✓⌧)hS2⌧hS2S1hS2(1nk)(R1))ci((1µ)(1✓⌧)hS3+⌧hS3
+(pmS3 + S1hS3(1 nk))(R 1)) = (1 µ)((1 ↵)Y + S0Y (1 nk)(R 1)
74
ci(1+hS3µ(1✓⌧)hS3mS3(p1))cp(µ(1✓⌧)hS2) = µ(1↵)Y
Solving for cp/Y, ci/Y and qhIH/Y , we get:
cp
Y
=
⌥2⌥6 +⌥3⌥4
⌥1⌥4 ⌥2⌥5
ci
Y
=
⌥1⌥6 +⌥3⌥5
⌥1⌥4 ⌥2⌥5
qhIH
Y
= h(S2
cp
Y
+ S3
ci
Y
)
Where
⌥1 = 1 + hS2  (1 µ)(1   ✓  ⌧)hS2  ⌧hS2  S1hS2(1 nk)(R 1)
⌥2 = (1 µ)(1   ✓  ⌧)hS3 + ⌧hS3 + (pmS3 + S1hS3(1 nk))(R 1)
⌥3 = (1 µ)((1 ↵) + S0(1 nk)(R 1)
⌥4 = 1 + hS3  µ(1   ✓  ⌧)hS3 mS3(p  1)
⌥5 = µ(1   ✓  ⌧)hS2
⌥6 = µ(1 ↵)Y
Next we derive the levels of the variables starting with the hours worked. We use
the equilibrium in the labour market equating labour demanded to labour supply:
(1 µ)(1 ↵)Y = cpl1+⌘ppc (l1+⌘ppc + l1+⌘pph )
(p⌘p)
(1+⌘p)
(1 µ)(1   ✓  ⌧)qhIH = cpl1+⌘pph (l1+⌘ppc + l1+⌘pph )
(p⌘p)
(1+⌘p)
75
This gives:
lph
lpc
=
⇢
(1   ✓  ⌧)qhIH
(1 ↵)Y
 1
1+⌘p
Substituting this in the first equation yields:
(1 µ)(1 ↵)Y
cp
=
⇢
1 +
(1   ✓  ⌧)qhIH
(1 ↵)Y
 (p⌘p)
(1+⌘p)
l1+pph
We can now substitute for cpY and
qhIH
Y to get the level of lph.
lph =
8<:
(1µ)(1↵)Y
cp
[1 + (1✓⌧)qhIH(1↵)Y ]
(p⌘p)
(1+⌘p)
9=;
1
1+p
We can now use this to get lpc.
lpc =
lphn
(1✓⌧)qhIH
(1↵)Y
o 1
1+⌘p
lih and lic can be similarly derived. Once we have the levels of hours worked we
can find non-housing output as follows:
Y =
✓
Kc
Y
◆ ↵
1↵
lµicl
1µ
pc = (S0)
↵
1↵ lµicl
1µ
pc
We get the value of qhIH using:
qhIH =
qhIH
Y
.Y
Subsequently we find Kh and kb using:
Kh =
qhIH
Rk  (1 k)
76
kb = ✓qhIH
We can now compute housing production:
IH = (lµihl
1µ
ph )
1✓⌧Kh,k
✓
b
Also, the price of housing can be computed:
qh =
qhIH
IH
Finally, the levels of consumption of the two types of households:
cp = {⌥2⌥6 +⌥3⌥4
⌥1⌥4 ⌥2⌥5}.Y ;
ci = {⌥1⌥6 +⌥3⌥5
⌥1⌥4 ⌥2⌥5}.Y,
and their housing stock:
hp = S2
cp
qh
hi = S3
ci
qh
77
Appendix 1.4: Prior and Posterior
0 0.5 1 1.5 2
0
2
4
φ1       
0 0.2 0.4 0.6 0.8 1
0
2
4
φ2       
0.6 0.8 1 1.2 1.4 1.6
0
2
4
η1       
0.4 0.6 0.8 1 1.2 1.4 1.6
0
2
4
η2       
0.1 0.2 0.3 0.4 0.5
0
5
10
15
µ            
5 10 15 20 25
0
0.1
0.2
φ
x
         
0.5 0.6 0.7 0.8 0.9
0
5
10
ρ
ac
      
0.5 0.6 0.7 0.8 0.9 1
0
5
10
ρ
ah      
0.5 0.6 0.7 0.8 0.9 1 1.1
0
5
10
ρh       
0.5 0.6 0.7 0.8 0.9 1
0
50
100
150
ρζ   
0.5 0.6 0.7 0.8 0.9
0
10
20
ρη
k
0.2 0.4 0.6 0.8 1 1.2
0
2
4
ρ
z
       
0.4 0.5 0.6 0.7 0.8 0.9
0
5
10
ρ
r
       
 
 
Posterior Prior
78
0.01 0.02 0.03 0.04 0.05 0.06
0
50
100
ε
ae
      
0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055
0
50
100
150
ε
ah      
−0.05 0 0.05 0.1 0.15 0.2 0.25
0
50
100
εh       
0.01 0.02 0.03 0.04 0.05 0.06
0
50
100
εζ   
0.1 0.2 0.3 0.4 0.5 0.6 0.7
0
50
100
εη
k
−0.01 0 0.01 0.02 0.03 0.04 0.05
0
50
100
150
ε
z
       
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
0
50
100
ε
σ
  
 
 
Posterior Prior
79
Appendix 1.5: Model Variables and Parameters
Variable Notation
Patient household consumption cp
Impatient household consumption ci
Patient household housing hp
Impatient household housing hi
Weight of housing in the utility function Jh
Patient household labour; consumption good lpc
Patient household labour; housing lph
Impatient household labour; consumption good lic
Impatient household labour; housing lih
Deposits; patient households dp
Intermediate business input kb
Price of intermediate input pb
Land la
Price of land pla
Return on land Rla
Patient household real wage rate; consumption good sector wpc
Patient household real wage rate; housing wph
Impatient household real wage rate; consumption good sector wic
Impatient household real wage rate; housing wih
Impatient household borrowing bi
Deposit return R
House prices qh
Entrepreneurs’ profit diverted to households Lump
Entrepreneur j’s net worth Nj
80
Entrepreneur j’s purchase of capital Kj
Entrepreneur j’s bank loan Bj
Aggregate Net worth N
Aggregate capital K
Aggregate loans B
Capital price Qk
Return on capital Rk
Marginal product of capital rk
Entrepreneurs’ idiosyncratic shock !e
Threshold value of idiosyncratic shock (for entrepreneurs survival) !¯e
Contractual gross return on bank loan RF
Entrepreneurial leverage x
Bank’s share of gross return on entrepreneurial investment e(.)
Share of return from defaulted loans Ge(.)
Startup transfer of net worth to new entrepreneurs W e
External finance premium efp
Entrepreneurs’ consumption ce
Consumption good Y
Non housing productivity Ac
Housing productivity Ah
Housing investment IH
Investment I
Business capital Kc
Housing capital Kh
Total housing stock H
Total consumption C
81
Convex investment costs S(.)
Parameter Notation
Patient household discount factor p
Impatient household discount factor i
Degree of labour substitution between sectors; patient households ⌘p
Degree of labour substitution between sectors; impatient households ⌘i
Inverse labour supply elasticity, patient household p
Inverse labour supply elasticity, impatient household i
Survival probability of firms 
Entrepreneur bankruptcy cost µe
Proportion of net worth consumed by exiting entrepreneurs ⇥
Investment adjustment cost parameter 'x
Capital depreciation rate k
Housing depreciation rate h
Share of capital in non-housing production ↵
Share of capital in housing production 
Share of land in housing production ⌧
Share of intermediate goods in housing production ✓
LTV ratio m¯
labour income share of impatient households µ
Non housing technology shock; persistence ⇢ac
Housing technology shock; persistence ⇢ah
Housing preference shock; persistence ⇢h
Time preference shock; persistence ⇢z
labour supply shock; persistence ⇢&
82
Investment shock; persistence ⇢⌘
Risk shock, persistence ⇢
Non housing technology shock; standard deviation ac
Housing technology shock; standard deviation ah
Housing preference shock; standard deviation h
Time preference shock; standard deviation z
labour supply shock; standard deviation &
Investment shock; standard deviation ⌘
Risk shock, standard deviation r
83
2 Securitization, Shadow Banking and Regulation
We construct a dynamic stochastic general equilibrium model with an active shadow
banking system to study the interaction between traditional and shadow banking. Se-
curitization is motivated by incentives for commercial banks to shift loans from on
to off balance sheet entities for regulatory arbitrage. Bank equity is costly, therefore
commercial banks cover part of their funding needs by selling loans in the secondary
market. These loans are then bundled together to provide marketable liquid deposit
like securities. Commercial banks’ return is subject to their monitoring effort. Their
incentives to monitor reduces as a larger share of loans is sold in the secondary mar-
ket. Thus, bank monitoring is subject to a moral hazard that limits loan sale in the
secondary market. We find that while securitization enhances the resilience to pro-
ductivity shocks, it exposes the economy to additional vulnerabilities emanating from
financial shocks. Government via its credit intermediation policy can play a signif-
icant role in containing the impact of financial shocks. Finally, we also look at the
macroeconomic and financial impact of capital regulation. In particular, we compare
fixed (Basel I) with procyclical (Basel II) capital requirement. In response to a neg-
ative technology shock, commercial banks transfer a higher proportion of loans from
on to off balance sheet i.e. from regulated commercial to unregulated shadow bank-
ing sector under Basel II. Furthermore, bank monitoring falls leading to a decline in
overall return on bank portfolio.
2.1 Introduction
Decades prior to the global financial crisis were marked by a changing financial land-
scape with a shift from traditional commercial banking towards securitized banking.
Commercial banking facilitated credit, maturity and liquidity transformation by issu-
ing short-term deposits to provide long-term loans. However, growth in innovation,
financial technology, regulation and competition led to an evolutionary change in
the whole financial services industry (see e.g. Pozsar (2008), Rosen (2009) and Blair
(2010)). A series of regulatory changes and innovations eroded competitive advan-
84
tage of banks and led to the growth of shadow banking9. Shadow banking system
undertook securitized banking by issuance of tradable securities against collateral of
an underlying pool of assets, including mortgages and business loans. As shadow
banking and securitization grew in prominence, it exposed the economy to financial
vulnerabilities leading up to systemic instability culminating in credit crunch and
global financial crisis. This paper provides a theoretical framework for understand-
ing the role and operations of shadow banking, associated risks, and key regulatory
and broader policy implications.
Securitization is one of the most important financial innovations that occurred
in the last part of the century. Several reasons have been put forward by academic
studies for the growth of securitization. For example, a liquidity hypothesis suggests
that securitization allows banks to transform illiquid financial claims into tradable
ones via a pooling and pass-through mechanism (Greenbaum (1986), DeMarzo and
Duffie (1999), DeMarzo (2003) and Martin-Oliver et. al. (2007)). With the new
funds raised, they can increase lending. In this setup, soon after the loan is granted,
it is packaged into a bundle of other mortgages, given a risk assessment by a rating
agency and sold out through asset backed securities (ABS). According to a hypothe-
sis of regulatory arbitrage, securitization allows banks to shift loans off balance sheet
and thus reduces the costly bank capital required as per minimum regulatory capital
ratios (Greenbaum and Thankor (1987), Pennacchi (1998)). Calomiris and Mason
(2004) and Ambrose, Lacour-Little and Sanders (2005) show that securitization of
credit cards and mortgages, respectively, respond to regulatory capital arbitrage.
Similarly, Thomas (2001) finds that securitization alleviates the regulatory burden.
A specialization hypothesis argues that securitization facilitates separation of loan
9The term shadow banking was coined by McCulley (2007) and the Financial Stability Board
(FSB, 2011) defines shadow banking broadly as “credit intermediation involving entities and ac-
tivities outside the regular banking system”. Other authors give complementary definitions that
emphasize different aspects of shadow banking for example Adrian and Ashcraft (2012) say it is
“a web of specialized financial institutions that channel funding from savers to investors through
a range of securitization and secured funding techniques”. Pozsar, Adrian, Ashcraft and Boesky
(2013) emphasize that “shadow banks [unlike] traditional banks lack access to public sources of liq-
uidity such as the Federal Reserve’s discount window, or public sources of insurance such as Federal
Deposit Insurance”.
85
origination, servicing, and bond/ABS administration and thus capitalizes on spe-
cialization in each step of intermediation (Greenbaum (1986) and Hess and Smith
(1988)). Finally, as security issuers may possess private asymmetric information re-
garding asset returns, they can use pooling and tranching to diversify risk (DeMarzo
2005).
The main purpose of this paper is to provide a model with securitization that
deepens our understanding of the interaction between traditional banking and shadow
banking. In our model, the first three hypotheses listed above are main drivers of
securitization. Commercial banks cannot compete with shadow banks that are not
subject to tight capital constraints as the former. Thus, the presence of regulatory ar-
bitrage leads to a shift from traditional loan issuance and funding (originate-to-hold)
to a originate-to-distribute model. Instead of holding loans on the balance sheet, the
originator could easily sell and transfer them off balance sheet. This creates space for
a new financial entity - shadow banks - where these loans are transferred.10 As per
the specialization hypothesis, these shadow banks act as the workhouse of securiti-
zation where issued loans are underwritten and sold as rated asset backed securities
(ABS) linked to the loan portfolio. Unlike the illiquid loans, which cannot be resold
(other than to the shadow banks), the ABS underwritten by the shadow banks are
viewed as liquid securities as per the liquidity hypothesis.
Thus, the model by adding shadow banks and ABS issuance to a canonical busi-
ness cycle model, allows for a non-trivial role for credit intermediation in the economy.
Traditional banks finance their lending to entrepreneurs by issuing deposits, equity
and via loan sales. Since these banks are subject to regulatory capital requirement,
loan sales allow them to finance loans less expensively than costly equity issuance.
However, the extent of banks’ loan sale is limited by a moral-hazard problem, as
shown in Pennacchi (1998) that arises from the diminished incentive by banks to
efficiently monitor and service loans after they have been sold. This defines the size
of shadow banking system in the economy.
The second main contribution of the paper is to encompass the most relevant
10To provide regulatory relief, an explicit sale of the loans is required from the commercial bank
balance sheet to shadow banks’ balance sheet.
86
aspects of the macroprudential concern in a single model. Regulation is at the
heart of our model setup, which provides incentives for loan sale in the secondary
market and growth of shadow banking. The model allows for a study of the most
relevant interlinkages between the real and the financial sectors and thereby facilitates
the interaction between financial and macroeconomic stability. The construction
of this DSGE model is very distant from the typical models in use prior to 2007,
which were seen as ignoring the relevance of financial intermediation and financial
stability. The model therefore provides analytical tools to properly consider the
impact of prudential policies on macroeconomic performance and financial stability.
Furthermore, the model provides for an interaction between asset quality and leverage
where the moral hazard problem faced by the banks in monitoring links off balance
sheet finance and the quality of bank loan portfolio.
We find that securitization dampens the response of the economy to negative
total factor productivity shocks by improving the resilience of the financial system
however shocks emerging within the financial system expose the economy to sustained
vulnerabilities. An increase in bank risk leads to transfer of higher proportion of loans
to the secondary market. This increases the moral hazard problem facing commercial
banks who respond by reducing the monitoring effort thereby lowering the effective
return on bank portfolio. We find that government via its credit intermediation
policy, in form of direct loan purchase, can play a positive role in containing the
impact of such shocks in the economy. Such policies are discussed elsewhere in the
literature including Gertler and Karadi (2011) and Meeks et. al. (2014) who find
similar positive effects. Finally, we also look at the macroeconomic and financial
impact of capital regulation. We find that there is a transfer of loans from on to off
balance sheet i.e. from regulated commercial to unregulated shadow banking sector
under Basel II rules with procyclical capital requirement.
The notion of shadow banking encompasses a wide universe of financial activities
and entities, not all of which are captured in this paper. In particular, we take a
stylized approach wherein shadow banks are the sole workhouse of securitization.
The simplification aligns well with our aim of looking at the securitization dynamics
arising from regulatory arbitrage. Furthermore, we do not model any type of risk
87
transfer motive behind shadow banking, or the sponsorship relationship between
traditional banks and shadow banks11. The modeling of these phenomena is beyond
the scope of this paper. Instead, we focus on shadow banking emerging from reg-
ulatory arbitrage and the interaction between securitization, shadow banking and
macroeconomic instability.
The remainder of the paper is structured as follows. We begin with a brief
overview of the literature in next section. In section three, we will analyze the
development of specific institutions of the shadow banking system. Furthermore,
we will examine instruments and transaction used by shadow banking institutions,
e.g. ABS issuance. Thereafter, section four describes the general equilibrium model
including households, entrepreneurs, firms and the financial system. Section five
present model calibration and section six explains results from simulations and policy
experiments. Finally, we conclude in section seven.
2.2 Related Literature
This paper draws from different strands of literature related to supply side dynamics
of credit growth, bank capital channel and the recently growing work on shadow
banking.
The borrower or demand side mechanisms relating to credit growth and business
cycle have been extensively studied starting from the financial accelerator mecha-
nism of Bernanke, Gertler, and Gilchrist (1999). The recent financial crisis however
brought to light the important role played by the financial system in credit gener-
ation and growth. Following that, there has been a surge in papers attempting to
model the supply side dynamics of credit growth. Some notable examples include
Gerali et. al. (2010), Dib (2010), Meh and Moran (2010), Gertler and karadi (2011),
and Gertler et. al. (2012). Gerali et. al. model an imperfectly competitive bank-
ing sector subject to exogenous bank capital requirement for the euro area and find
that the shocks to the banking sector explain the major part of fall in output in
11For risk transfer see, for example, Hobijn and Ravenna (2010). For sponsorship relationship
and specific models on special purpose vehicles see, for example, Parlatore (2016).
88
2008. Dib (2010) embed an active banking sector with interbank market and bank
capital into a standard financial accelerator model. Meh and Moran (2010) analyze
the bank capital channel of transmission by introducing a moral hazard problem
between banks and investors which can be mitigated by banks investing their own
capital in entrepreneurial projects. In a similar vein, Gertler and Karadi (2011) intro-
duce an agency problem between banks and depositors that imposes an endogenous
constraint on intermediary’s leverage ratio and thereby linking overall credit flows to
bank equity capital. Gertler et. al. (2012) go a step further and allow banks to raise
outside equity as an alternative to short term debt (deposits) to provide loans. One
common feature of all these studies is the inclusion of bank capital either in the form
of exogenous capital requirements or endogenous bank leverage ratios. This paper is
similar to the extent that we also look at the supply side dynamics of financial inter-
mediation and bank capital requirements. However, all these papers take banks to
represent the entire financial system. This paper allows for heterogeneity, namely in
the form of a stylized shadow banking sector that focuses on securitization, thereby
generating an additional source of dynamics.
The financial stability issues around shadow banking, and securitization in par-
ticular, have by now been widely discussed (Adrian and Shin, 2008). Rosen (2009)
sketches the evolution of the U.S. financial system from traditional banking to shadow
banking. Furthermore, the author focuses on the role of shadow banking and de-
scribes the crisis as a logical outcome of increasing interconnectedness and leverage
of financial intermediaries, overconfident investors and rapid growth in structured fi-
nance. Pozsar et al. (2013) and Pozsar (2008) provide a detail catalogue of the types
of shadow banks. They map and describe shadow banking as a daisy chain of finan-
cial intermediaries that conduct credit intermediation. Stein (2010) provides a short
overview of the securitization process, how it developed in the event of the crisis and
how it is conducted within the shadow banking system. Furthermore, the article also
explains the fragility of the securitization market during the crisis and yields some
regulatory approaches of securitization and suggests implications for future regula-
tion. Poschmann (2012) explores the nexus between traditional and shadow banking
and provides a stylized descriptive framework of a banking sector with NBFI (Non
89
Bank Financial Institutions) in a macroeconomic context. Claessens et. al. (2012)
focus on securitization and collateral intermediation function of the shadow banking
system in a descriptive set up. In particular, they posit that the securitization func-
tion of shadow banking provides the wholesale/institutional investors with deposit
like safe investments12. At the same time it also gives banks an access to highly
liquid (high rated) securitized debt, which can be used for repo funding and to in-
crease leverage. The securitization process also provides banks with an opportunity
to bypass regulatory capital constraints by shifting loans to off balance sheet entities
(shadow banks); (Poschmann (2012) and Stein (2010)).
The existing literature concentrates on descriptive analysis: how the system de-
veloped before and within the crisis and what are the important aspects that led
to the growth of certain parts of the financial system. Until now, few papers have
attempted to model shadow banking in a macroeconomic context. We aim to fill this
gap in the literature by embedding a traditional and shadow banking sector in an
otherwise standard business cycle model.
Verona, Martins, and Drumond (2014) introduce shadow banks as a distinct class
of financial intermediaries into a DSGE model. In their model, shadow banking dif-
ferentiates bond financing from bank finance with the former done by shadow banks
and latter by commercial banks. The model however, does not feature securitization,
and shadow banks have no direct interaction with the commercial banking system.
Goodhart, Kashyap, Tsomocos, and Vardoulakis (2012) study a two period general
equilibrium model with heterogeneous households, banks and shadow banks. The
authors generate a role for shadow banking by assuming lower risk aversion amongst
non-banks than amongst banks. They find that when borrowers default the shadow
banks pass losses back to the rest of the financial system and kick off a fire sale
dynamic that further tightens the financial constraints. Hobijn and Ravenna (2010),
introduce an adverse selection problem into a New Keynesian monetary policy model.
12Global corporate short-term savings grew from less than $50 billion in 1990 to more than $750
billion in 2007 and over $1.2 trillion by the end of 2010 (Pozsar et. al. (2013)). Together, corporate
and asset management (“institutional”) cash pools have grown over the past decade by a factor of
30 to equal almost half of traditional deposits (Claessens et. al. (2012)).
90
The asymmetric information held by borrowers leads to an endogenous sorting of
loans into those directly held by originators, and those sold into securitization pools
of differing qualities13. Meeks et al. (2014) introduce in the framework of Gertler and
Karadi (2011) a shadow banking sector that funds itself from traditional banks, and
assume that traditional banks have a weaker friction when investing in shadow bank
liabilities. They show that endogenous constraints on shadow bank’s net worth can
amplify the shocks due to limited ability of commercial banks to securitize. Ferrante
(2015) consider a model with traditional and shadow banking intermediation where
banks can finance risky projects and affect their quality by exerting costly screen-
ing effort. They find that while shadow banking enhances credit intermediation, it
also makes the economy more fragile. None of these studies approach the growth of
shadow banking from regulation perspective, a gap that we fill in this paper.
This paper builds on the work of Christiano et. al. (2014). They extend
the framework of BGG (1999) with costly state verification in the market for en-
trepreneurial loans. The returns from entrepreneurial investment are subject to an
idiosyncratic variable whose cross-sectional standard deviation is regarded as risk.
They model this risk to be time varying. They find that when risk is high, the credit
spread is high, and credit extended to entrepreneurs is low. Furthermore, risk shocks
are found as the most important driver of business cycles. While the focus of their
paper is to study the importance of entrepreneurial risk shocks in driving business
cycle fluctuations, we augment their framework with a heterogeneous banking sector
to study the propagation of financial shocks, in particular, risk shock in the banking
system. We build on their model in following important ways 1.) Since the focus of
this paper is to study the regulatory arbitrage motive behind shadow banking, we
develop a banking sector that is subject to a regulatory capital constraint. Capital
regulation and macroprudential policies have an economically important role as they
arise to tackle two key distortions. First distortion stems from banks’ limited liabil-
13We refrain from any endogenous sorting of loans due to adverse selection however that would
further exacerbate the situation. In our model, commercial bank transfer a representative pool of
loans into off balance sheet rather than a sorting based on quality. Therefore, securitization in this
model is motivated by a desire to receive regulatory relief rather than to transfer risky loans off
balance sheet.
91
ity and the existence of deposit insurance. Deposit insurance provides incentives for
banks to take on risk at the expense of the deposit insurance agency. The second
distortion arises from the assumption that depositors suffer some transaction costs
in the event of a bank failure despite the presence of deposit insurance. Clerc et al.
(2015) study macroprudential policies in a macroeconomic model with default in a
similar setting. 2.) We introduce shadow banks in the model to facilitate loan sale
in the secondary market, which are then securitized. Traditional banks are subject
to capital requirements and therefore find it profitable to shift some of the loans
to shadow banks, which are not regulated. The size of these loan sales is based on
a moral-hazard problem similar to Pennacchi (1988). Traditional banks undertake
costly monitoring that affects the return on its investment. This loan monitoring is
unobservable to loan buyers in secondary market, which creates diminished incentive
for banks to efficiently monitor and service loans after they have been sold. This
limits the amount of loans sold in the secondary market and therefore the size of
the shadow banking sector. 3.) Finally, shadow banks finance their operations by
underwriting asset backed securities. The underwriting services are usually provided
by investment banks, much as a corporation would use an investment bank to under-
write the sale of bonds or stocks14. The market for loan purchase and underwriting
securities by shadow banks is assumed to be monopolistically competitive. This is
consistent with empirical evidence for the U.S for bond underwriting. Fang (2005)
shows that the largest five investment banks underwrite more than 60 percent of all
deals, and the largest fifteen banks account for roughly 95 percent of all deals. We
expect a similar pattern for ABS underwriting as well.
Our model belongs to the class of DSGE models and we solve it with standard
methods, i.e. perturbation methods. This is different from a relatively recent body
of work, which uses a continuous time methodology to solve full dynamics of models
with financial frictions. For example, Brunnermeier and Sannikov (2012) build on
14Typically, a shadow bank would hire an investment bank to sell its securities. The investment
bank (the underwriter) acts as an intermediary between the issuing entity and the ultimate investors.
See Ellis, Michaely, and O’Hara (2000) for an in-depth analysis of the underwriting process in bond
finance.
92
the BGG model in a continuous time frame. The model has two types of agents
- productive experts and less productive households. Because of financial frictions,
the wealth of experts is important for their ability to buy physical capital and use
it productively. They find that the systems reaction to shocks is highly nonlinear.
While the system is resilient to most shocks near the steady state, unusually large
shocks are strongly amplified. Furthermore. most of the effects are asymmetric and
only arise in a downturn.
2.3 Institution, Entities and Instruments
This section provides a brief overview of the institutions and financial instruments
that constitute the shadow banking system with an emphasis on mapping the fi-
nancial structures to the stylized set up of the model. Following Pozsar (2008) and
Poschmann (2012) we classify entities and instruments as those involved in the is-
suance of loans (risk originators), creation of securities (loan warehousing and ABS
issuance) and institutions that invest in these instruments (risk bearers).
Risk originators comprise of depository institutions like the commercial banks,
investment banks and broker-dealers. In this paper, we only focus on depository
institutions, in particular, on the commercial banks which accounts for 80 to 90% of
the total assets of depository institutions in the United States. Deposits dominate
the liability side of depository institutions, however they are also required to hold a
minimum regulatory capital in accordance with the Basel rules.
As regards to the shadow banks, we take a narrow view and concentrate on the
securitization process involving ABS issuance. Securitization denotes a financing
process where illiquid assets from the risk originators are pooled and transformed
into liquid financial instruments. The securitization process is conducted by shadow
banks. They are thinly capitalized and loan sale involves a true sale of assets from risk
originators to shadow banks. This transaction allows the originator to free capital
and therefore, reduce capital requirements. Shadow banks then issue ABS to fund
purchase of loans in the secondary market. ABS is a collective term and encompasses
93
traditional ABS, mortgage backed securities (MBS), Collateralized debt obligations
(CDO) and asset backed commercial papers (ABCP) depending on the underlying
pool of assets. Traditional ABS mostly consists of trade and credit card receivables,
consumer credits and lease contracts. MBS, a large part of ABS, are ordinarily
based on a pool of residential and commercial mortgages. CDOs are securitized
loan and bond portfolios, including CLO - collateralized loan obligations and CBO -
collateralized bond obligations.
Finally, we have the risk bearers who invest in the securitized products and serve
as wholesale or shadow bank depositors. Traditionally, these are the institutional
investors comprising mostly of mutual funds and money market funds. These insti-
tutional investors raise money from individuals, households and businesses, through
the sale of fund shares, which is then invested in mainly short term high credit quality
and liquid debt instruments like ABS. To simplify the analysis and avoid introduc-
ing too many new agents in the model, we treat household as final buyers of these
securities15.
2.4 The Model
The Economy consists of seven types of agents: households, entrepreneurs, commer-
cial banks, shadow banks, consumption good producer, capital good producer and
deposit insurance agency. The economic modeling of households, entrepreneurs, con-
sumption good producer and capital goods producer parallels that in Christiano et.
al (2014) wherein inspite of some differences in modeling assumptions, the dynamics
are similar16.
15Alternatively, the model can also be closed by adding a new class of agent i.e. the wholesale
depositors. They would be the depositors in shadow banking system by purchasing ABS securities.
Wholesale depositors on the other hand, would issue mutual fund shares to households to raise
money.
16In particular, there are following differences in modeling assumption for these agents. In Chris-
tiano et. al. (2014), households are assumed to build raw capital (subject to convex adjustment
costs) in the economy instead of competitive capital good producer. The is done to limit the number
of agents in the economy and does not alter the dynamics. Entrepreneurs have the same dynamics
as in Christiano et. al (2014). However, since we work with a real business cycle model, nominal
94
Households. Households are risk averse and infinitely lived and derive utility from
a storable consumption good. They provide labour to firms and funding in the form
of deposits, bank equity and securities, to the financial sector. The consumption
good acts as a numeraire in our economy and its price is normalized to one.
Entrepreneurs. Entrepreneurs are risk neutral agents specialized in owning and
maintaining the stock of physical capital, which they rent in each period to the
firms involved in the production of the consumption good. At the beginning of each
period, entrepreneurs buy capital using their net worth and borrowed loans from the
commercial banks as in BGG. These loans are subject to limited liability and default
risk, and recovering residual returns from bankrupt entrepreneurs requires banks to
pay verification costs.
Consumption good producer. There is a perfectly competitive consumption good
producing sector which combines rented capital from entrepreneurs and labour from
households to produce consumption good using a standard cobb-douglas production
sector.
Capital goods producers. Capital goods producer are perfectly competitive and
owned by the households. These firms optimize intertemporally in response to
changes in the price of capital and are subject to investment adjustment costs in
conversion of consumption goods into capital. This sector is not directly affected by
financial frictions.
The rest of the agents in the economy relating to the financial system i.e. com-
mercial banks and shadow banks are specific to our model and are described below.
Commercial banks. Commercial banks are perfectly competitive and risk neutral
in the credit markets when providing loans to the firms. We differ from BGG and
Christiano et. al. by introducing macroprudential regulation on bank balance sheet.
Bank funding in our model comprise of debt in the form of deposits from households,
equity by issuing bank capital shares and loan sales. Banks are subject to regulatory
capital requirement that provides incentives for loan sales to minimize costs arising
from costly bank capital. However, the extent of banks’ loan sale is limited by a
frictions - both in production of final good and supply of labour - are not included.
95
moral-hazard problem, as shown in Pennacchi (1998) that arises from the diminished
incentive by banks to efficiently monitor and service loans after they have been sold.
In addition, banks are faced by an idiosyncratic risk, which in equilibrium implies that
some banks cannot service their deposit payments and thus are subject to defaults.
Shadow banks. Shadow banks are not subject to regulation, are thinly capitalized
and are buyers of loans in the secondary market. They are funded by underwriting
ABS against loan pools. Unlike commercial banks which specialize in credit inter-
mediation i.e. providing loans to final borrowers (firms), shadow banks specialize in
securitization and cannot directly provide loans to the final borrowers.
Deposit insurance agency. Deposit insurance agency recoups earnings from failed
banks (net of monitoring costs) and service household deposits. It is assumed to
ex-post balance its budget by imposing lump sum taxes on households.
2.4.1 Households
The households derive utility from consumption Ct, and leisure, 1  Lt, where Lt
is the hours worked17. A representative household maximizes the following utility
function.
max
Ct,Lt,Dt,ABSt,Zt
Et
1X
t=0
t(logCt +
⌘
(1 + ')
(1 Lt)(1+')) (2.1)
The parameter 0 <  < 1 is the discount factor, ⌘ > 0 is the weight attached
to leisure in the utility function and ' > 0 is the inverse of Frisch wage elasticity
of labour supply. Households can either save in the form of one-period deposits Dt
with retail bank that pays a non-contingent risk-less rate eRt, invest in bank shares
Zt, or buy asset backed securities, ABSt, from shadow banks with return rate Rm,t.
A representative household enters period t with Dt1 units of deposits, Zt1 units
of bank shares and ABSt1 units of shadow bank deposits, which pays a gross real
return of eRt, Rz,t, and Rm,t, respectively between periods t 1 and t.
17See appendix 2.2 for a complete list of model variables and parameters.
96
Every period t, households provide labour to firms in return for real wage income
wtLt;wt is the economy wide real wage. In addition, they also receive lump sum
transfers, Lumpe,t from exiting entrepreneurs, as well as profits from shadow banks
(⇧I,t). The overall household income is allocated to consumption, commercial and
shadow bank deposits.
The household budget constraint can be written as:
Ct+Dt+ABSt+Zt  wtLt+ eRtDt1+Rm,tABSt1+Rz,tZt1+⇧I,t+Lumpe,t⇥zZ2t
2
TDt
(2.2)
Where
eRt = Rt1(1 ⌦PDt) (2.3)
The term, ⇥z Z
2
2 , represents costs associated with equity transactions, with ⇥z >
0 denoting an adjustment cost parameter. Rt1 is the fixed (gross) interest rate
received at t on savings deposited at banks in t-1. The principal and interest of
bank deposits are fully guaranteed by a deposit insurance agency (DIA) that, for
simplicity, is assumed to ex post balance its budget by imposing a lump sum tax TDt
on households. However, in case a bank fails, households incur a linear transaction
cost ⌦ in recovery of the funds. PDt is the fraction of deposits in banks that fail
in period t. This transaction cost introduces a link between banks’ probability of
default and their funding cost via deposits. Following Clerc et. al. (2015), it is
modeled in a way that does not affect the incentives of individual banks to take
excessive risks at the expense of the DIA.
A representative household chooses consumption (Ct), labour (Lt), deposits (Dt),
bank share (Zt) and securities (ABSt) to maximize the lifetime utility function sub-
ject to individual period budget constraint.
97
2.4.2 Entrepreneurs
Entrepreneurs are assumed to be risk-neutral and have finite horizons: Specifically,
we assume that each entrepreneur has a constant probability  of surviving to the
next period (implying an expected lifetime of 1/(1  )). The assumption of finite
horizons for entrepreneurs is intended to capture the phenomenon of ongoing births
and deaths of firms, as well as to preclude the possibility that the entrepreneurial
sector will ultimately accumulate enough wealth to be fully self-financing. A fraction
J of the total net worth owned by those entrepreneurs who close business is consumed
upon exit, and the remaining fraction of their net worth is transferred as a lump-sum
payment to households. We assume the birth rate of entrepreneurs to be such that
the fraction of agents who are entrepreneurs is constant. New entrepreneurs entering
in period t+ 1 receive a ‘start-up’ transfer of net worth, W e18.
There is a large number of entrepreneurs, each indexed by j. An entrepreneur’s
state in period t is its level of net worth, Nt,j. In each period, an entrepreneur
combines its net worth, Nt,j, with a bank loan, Bt,j to purchase new, installed physical
capital, Kt,j from the capital producer (Bt,j + Nt,j = Qk,tKt,j), where Qk,t is the
price of capital. The entrepreneur then experiences an idiosyncratic shock, !e, to its
stock of capital. The purchased capital, Kt,j is transformed into !eKt,j where !e is
a log-normally distributed random variable which is independently and identically
distributed across entrepreneurs with an expected value of one, and density and
cumulative distribution function denoted by f(.) and F (.), respectively. The shock
therefore is a simple way to generate a non-trivial default rate on entrepreneurial
loans. The setup for entrepreneurs is similar to as studied in the previous chapter.
The gross return per unit of capital obtained in period t+1 out of capital owned
in period t is:
Rk,t+1 =
rk,t+1 + (1 )Qk,t+1
Qk,t
(2.4)
18In BGG (1999), newly entering entrepreneurs have positive net worth by supplying a unit of
labour inelastically to be used in the production of consumption good. We follow later papers, for
example Christiano, Motto and Rostagno (2014) who assume lump-sum transfer of start-up capital
for newly entering entrepreneurs.
98
Where  is the depreciation rate of capital and rk is the marginal return on
capital.
Let RF,t be the contractual gross interest rate on the loan taken from the bank
in period t. Then the total contracted repayments due to banks at period t+1 is
RF,tBt,j. The gross return that an entrepreneur obtains on the capital investment
undertaken in the previous period is Rk,t+1Qk,t!et+1Kt,j. We assume limited liability
for entrepreneurs whereby they cannot be held liable for any contracted repayments
above the gross return, in which case they default on their bank loans. In particular,
the entrepreneur will repay loan at t+ 1 when
Rk,t+1Qk,t!
e
t+1Kt,j  RF,tBt,j
Which provides the threshold value of idiosyncratic shock, !¯et+1, above which the
entrepreneur can repay its loan.
!¯et+1 =
RF,tBt,j
Qk,tKt,j
.
1
Rk,t+1
(2.5)
We define xt =
Qk,tKt,j
Nt,j
as the entrepreneurial leverage, defined as the ratio of
entrepreneurs’ purchase to net worth at t. We have left off the indexing j on xt, !¯t+1,
and RF,t. This is to minimize notation and is a reflection of the fact (see below)
that the equilibrium value of these objects is independent of the state variable (Nt,j).
When an entrepreneur defaults on its loan, following the Costly State Verification
(CSV) approach, we assume that the lender must pay a cost if he or she wishes to
observe the borrower’s realized return on capital, which it then seizes. This audit-
ing cost is interpretable as the cost of bankruptcy (including for example auditing,
accounting, and legal costs, as well as losses associated with asset liquidation and
interruption of business). The monitoring cost is assumed to equal a proportion
µe of the realized gross payoff to the firm’s capital, i.e., the monitoring cost equals
µeRk,t+1Qk,t!et+1Kt,j. Thus, under certain realizations of the idiosyncratic shocks,
entrepreneurs will go bankrupt, especially when their ex-ante leverage xt is high.
We follow BGG notation for the division of return from entrepreneurial invest-
99
ment between entrepreneurs and banks. The share of gross return (gross of verifica-
tion costs) accruing to the bank is:
Ge(!¯et+1) =
ˆ !¯et+1
0
!et+1f
e(!et+1)d!
e
t+1 + !¯
e
t+1
ˆ 1
!¯et+1
f e(!et+1)d!
e
t+1 (2.6)
The part of these returns that comes from defaulted loans is:
Ge(!¯et+1) =
ˆ !¯et+1
0
!et+1f
e(!et+1)d!
e
t+1 (2.7)
Therefore, the share of returns accruing to entrepreneurs can be written as 1 
Ge(!¯et+1). The share (net of verification costs) that the bank appropriates can be
written as Ge(!¯et+1) µeGe(!¯et+1).
Entrepreneurs maximize their expected return subject to participation constraint
for the bank. In particular, they choose the leverage, xt, and default threshold !¯et+1
to maximize:
max
xt,!¯et+1
Et [(1 Ge(!¯et+1))Rk,t+1xtNt,j]
Subject to banks’ participation constraint:
Et[(Ge(!¯et+1) µeGe(!¯et+1))]Rk,t+1xt = eRF,t+1(xt  1) (2.8)
Where eRF,t+1 is the return on a well diversified portfolio of bank loans and will
be discussed in detail in the next section. The main difference from chapter one is
that the cost of raising funds for commercial banks is no longer the risk free rate.
The participation constraint captures the fact that banks would provide loans to
entrepreneurs only if the return is sufficient to cover their cost of raising funding.
Aggregation - The quantity of raw capital purchased by entrepreneurs in period
t is given by summing the capital across all entrepreneurs. Market clearing in the
capital services require that supply of capital, Kt must equal firms’ demand for
capital (discussed later).
100
Kt =
P
Kt,j
The aggregate entrepreneurial net worth at the end of period t can be written
as (1  G(!¯t+1))Rk,t+1Qk,tKt. Once a fraction (1  ) of the entrepreneurs exit the
scene and the newly entering entrepreneur receive lump sum payments W e. The law
of motion of entrepreneurial net worth, Nt+1 can be written as:
Nt+1 = (1 G(!¯t+1)Rk,t+1Qk,tKt +W e,
where W e is the ‘start-up’ transfer of net worth received by new entering en-
trepreneurs in period t+1. The aggregate net worth can be further written as:
Nt+1 = {Rk,t+1Qk,tKt [Rt+µe
´ !¯t+1
0 !t+1f(!t+1)d!t+1Qk,tKt
Qk,tKt Nt ](Qk,tKtNt)}+W
e
(2.9)
The object in braces in (2.9) represents receipts by entrepreneurs active in period
t+1 minus their payments to banks. The object in square brackets is the gross rate
of return paid by entrepreneurs to banks. The external finance premium on loans
between period t and t+1 can therefore be defined as:
efpt+1 = µe
´ !¯et+1
0 !
e
t+1f
e(!et+1)d!
e
tQk,tKt
Qk,tKt Nt
The remaining two financial variables to determine are the aggregate quantity
of debt extended to entrepreneurs in period t, Bt, and their state-contingent inter-
est rate, RF,t. The aggregate borrowing is given by the sum of borrowing of all
entrepreneurs.
Bt =
P
Bt,j = Qk,tKt Nt
Finally, the interest rate takes the form (aggregating equation 2.5 and solving) :
101
RF,t = !¯
e
t+1Rk,t+1(
xt
xt  1)
The entrepreneurs that exit the economy in t+1 consume a fraction ⇥ of their
net worth. Therefore, their aggregate consumption is given by:
Ce,t+1 = ⇥
(1 )

(Nt+1 W e) (2.10)
Where (Nt+1W
e)
 is the average net worth held by an entrepreneur who exits
the economy at the end of the current period and (1  ) is the share of exiting
entrepreneurs. Finally, the lump sum transfers made to patient households is (1 
⇥) (1) (Nt+1 W e).
2.4.3 Commercial Banks
Commercial banks are risk neutral and perfectly competitive in the credit market.
They collect deposits, Dt, from households and also raise funds in the form of bank
capital, Zt. To provide loans Bt to the final borrowers (entrepreneurs), banks must
fulfill a certain regulatory capital requirement. This capital requirement is defined
as the ratio of bank capital to bank assets. However, in practice banks hold more
capital than the minimum requirement, which is motivated by the presence of gains
associated with bank capital buffer. To gain relief from the regulatory constraint,
banks can shift a part of the loan, Bs,t, away from their balance sheet to the shadow
banks. Therefore, commercial banks not only originate loans for the primary bor-
rowers, but also sell loans in the secondary market19. We assume that banks sell a
portion, bt, of the return on all loans, retaining the portion (1 bt). Thus, a repre-
sentative bundle of loan is sold in the secondary market. Bundling is important as it
helps banks to overcome an adverse selection problem when they come to sell loans.
If private information about individual loan cannot be credibly communicated to the
seller, no creditor in the secondary market would be willing to buy an individual
19For details on bank funding via loan sales see, for example, Pennacchi (1988), Berger and Udell
(1993) and Gorton and Pennacchi (1995).
102
loan under the fear of being sold the worst class of loans. Bundling of loans destroys
private information and thereby ensures that a fair mix of loans is transferred and
not just lemons20. Therefore, the shift of loans off balance sheet is motivated by
regulatory reasons rather than the desire to remove the worst loans from commercial
banks’ balance sheet.
Furthermore, banks’ effective return on its portfolio is subject to its monitoring.
One way of thinking of this is that bank managers can divert a part of return for
their own personal benefit. In such a case, monitoring of bank managers would be
beneficial in increasing the final return to the bank equity holders. Alternatively,
one could also think of a possibility where the manager of bank may shirk since the
final return is accrued to shareholders. In such a case, an inefficiency would arise
where a part of return would be lost. Banks may then undertake costly effort to
increase the final return. This could be in the form of effort (e) exerted in screening
or monitoring21. We adopt the standard assumption that the level of bank credit
services is unobservable so that the bank and loan buyers cannot write contracts
that are contingent on the level of these services. Therefore, loan sales involve a
moral hazard problem, namely, that the bank may not exert the monitoring effort
at the most efficient level. Importantly, monitoring is costly, since it entails a non-
pecuniary convex cost c(et) = ⌧12 (e
2
t + ⌧2et), where ⌧2 is allowed to be negative,
meaning that there could be some benefits from screening. However, we consider
calibrations where c0(et) > 0, meaning that it is costly for financial intermediaries to
20The idea that the purpose of bundling is to destroy private information is also found in DeMarzo
(2005). In general, private information may exist on either the side of the market. DeMarzo
considers the case of sellers who specialize in originating and marketing assets, but do not have a
comparative advantage in valuing or holding them. Pooling of these assets reduces the information
asymmetry as it prevents sophisticated buyers, such as specialist brokers, from cherry-picking assets.
21A monitoring role for intermediaries is modeled in Gorton and Haubrich (1990), and Gorton
and Kahn (1994). Further, Campbell and Kracaw (1980) and Boyd and Prescott (1986) explain
the existence of financial intermediaries as providing efficient credit evaluation services. We assume
that monitoring is required to ensure bank managers do not shirk/divert return for their own
gain/operate inefficiently. Alternatively, monitoring could be introduced to ensure that banks
perform credit evaluation on borrowers i.e. entrepreneurs to ensure they undertake projects with
high returns. This would however require that different projects are available to entrepreneurs and
without proper incentives, entrepreneurs may deliberately choose a project with a lower ex post
return (see Holmstrom and Tirole (1997), Chen (2001), Meh and Moran (2010)).
103
increase their screening effort.
Commercial bank’s balance sheet at any time t can be represented as below:
Balance Sheet: Commercial Bank
Assets Liabilities
Loans (Bt) Deposits (Dt)
Bank capital (equity) (Zt)
Loans (sold to shadow banks) (Bs,t)
The bank’s objective is to maximize the return on shareholders’ equity, which
is given by the positive part of the difference between returns from its on balance
sheet loans and the repayments due to its deposits and payment for monitoring. Let
e eRF,t+1 denote the realized return on a well diversified portfolio of loans. The return
is assumed to be a simple linear function of the monitoring effort (i.e. a fraction
(1 - e) is either diverted or lost in inefficiency costs). Similar to the entrepreneurs,
banks are also subject to an idiosyncratic risk. The idiosyncratic portfolio return
shock !b, which can be interpreted as a shock to each individual banks’ ability to
extract payoffs from its loans, is i.i.d across banks and is assumed to have log-normal
distribution with a mean of 1 and a distribution function F (!b). The mean and
standard deviation of log!b are µ$b and t respectively. The standard deviation,
t, is the realization of a stochastic process, which we refer to below as the risk
shock. This shock captures the notion that the riskiness of banks varies over time
and follows a stationary AR(1) process.
logt = ⇢logt1 + (1 ⇢)log + ",t,
where  is the steady state standard deviation and ",t is i.i.d N(0, 2r).
Thus, individual bank return !bt+1et eRF,t+1 (and therefore bank default) depends
on monitoring effort, aggregate loan return eRF,t+1 (itself driven by firms’ default
rates) and the bank-idiosyncratic shock !bt+1. The monitoring effort therefore plays
104
an important role as the effective return on bank loans (and therefore the cost of
bank funding in a competitive environment where banks are risk neutral) is directly
related to monitoring. For a given effective return, a fall in bank monitoring would
lead to an increase in eRF,t+1. This implies a higher cost of funding for the final
borrowers i.e. entrepreneurs since eRF,t+1 enters banks participation constraint in
equation (2.8) in entrepreneurs optimization problem.
When a bank fails, its equity is written down to zero and its deposits are taken
over by a deposit insurance fund that pays out all deposits in full. The presence
of deposit insurance ensures that the gross interest rate paid on the deposits, Rt is
uniform across all banks. The deposit insurance fund replenishes itself by taking
over failed banks’ loan portfolio minus resolution costs, which are assumed to be a
fraction, µb, of the total bank assets (net of screening costs).
Commercial bank is also subject to a regulatory constraint such that:
Zt = tBt(1 bt)  ¯Bt(1 bt) (2.11)
Where kt is the bank capital ratio and ¯ is the regulatory minimum capital re-
quirement. In reality, banks often keep higher capital as it reflects a healthy banking
sector and thus reduces the cost of issuing equity. It also captures the view that
maintaining a positive capital buffer generates some benefits—it represents a signal
that the bank’s financial position is strong and reduces the intensity of regulatory
scrutiny. This, in turn, reduces the pecuniary cost associated with the preparation
of data and documents required by the supervision authority22. The gain associated
with banks being more capitalized i.e (t > ) is modeled in a reduced form way as
⌧z(
t
 )
0.5, where ⌧z > 0. We assume the gain to be concave, which implies that the
benefits from capital buffers diminish over time.
We begin by analyzing the impact of fixed capital requirement, as in Basel I.
In section 2.6.5, implications of procyclical Basel II are also discussed. The capital
requirement here plays a role to counteract the effect of limited liability in the banking
22See for example, Ayuso et al. (2004), in which capital buffers reduce the probability of not
complying with capital requirements.
105
sector. As in Clerc et. al. (2015), the presence of deposit insurance provides an
implicit subsidy to the defaulting banks, which can lead to perverse incentives for
banks to take excessive risks. A higher capital requirement reduces leverage, forcing
banks to get funded with a larger share of equity, which is more costly. It also
reduces the probability of bank default and hence the subsidy implied by the deposit
insurance, leading to a higher lending rate and more restricted access to credit for
entrepreneurs.
Unlike in Clerc et. al., we also allow banks to fund themselves by selling loans in
the secondary market. This helps the bank to relax the capital requirement constraint
to an extent. This is reflected by the factor (1 bt) in equation (2.11). However, the
extent of loan sales is limited by the moral-hazard problem described earlier, since
loan monitoring is not observable by loan buyers. The gains from monitoring are
assumed to be linear however the cost of monitoring is convex implying a decreasing
marginal profit from monitoring.
Let !¯bt+1 denote the threshold realization of !bt+1 below which the bank fails
because the realized return on its loan portfolio is lower than the deposit repayment
obligation:
(1 bt)!¯bt+1et eRF,t+1Bt = RtDt (2.12)
The left hand side shows the return on on-balance sheet loans net of the screening
costs. Since a portion bt of the loan is sold to shadow banks in the secondary market
(i.e. Bs,t = btBt), only a fraction (1  bt) is available to service debt and equity
repayments. The balance sheet of the commercial bank reads as:
Bt +
⌧1
2
(e2t + ⌧2et)Bt = Dt + Zt + Ps,tbtBt (2.13)
Where the term ⌧12 (e
2
t + ⌧2et)Bt is the funds used in costly monitoring services.
Therefore, the resources required to fund and monitor the loan can come from either
deposits, equity finance or via sale of loans in the secondary market. Ps,t is the price
of loans in the secondary market. The balance sheet constraint can be simplified as
106
Dt = Bt(1 +
⌧1
2 (e
2
t + ⌧2et) t(1 bt) Ps,tbt). Therefore:
!¯bt+1 =
Rt
et eRF,t+1(1 bt)(1 + ⌧12 (e2t + ⌧2et) t(1 bt) Ps,tbt) (2.14)
Thus, the threshold below which the bank fails is decreasing in bank capital ratio,
t, and the average bank return (on a well diversified portfolio), eRF,t. A higher capital
ratio and higher average return therefore implies lower bank default. The threshold
is increasing in the fraction of bank loans diverted to secondary loan market, bt, the
return on bank deposits, Rt, and the screening cost per unit of loan.
The profit function of commercial banks can be written as:
⇡z,t+1 = (1bt)et eRF,t+1Bt ˆ 1
!¯bt+1
!bt+1f
b(!bt+1)d!
b
t+1
ˆ 1
!¯bt+1
f b(!bt+1)d!
b
t+1RtDtRz,t+1Zt+⌧z(
t  

)0.5
The first term is the return from banks’ on-balance sheet loans that do not default.
The second term is the payment for return on deposits. This is multiplied with the
probability of bank survival i.e.
´1
!¯bt+1
f b(!bt+1)d!
b
t+1. The third term is the required
return on bank capital and the fourth term denotes concave gains from keeping
capital buffers. Using equation (2.12), the profit function can be simplified to:
⇡z,t+1 = (1bt)et eRF,t+1Bt[ˆ 1
!¯bt+1
!bt+1f
b(!bt+1)d!
b
t+1!¯bt+1
ˆ 1
!¯bt+1
f b(!bt+1)d!
b
t+1]Rz,t+1Zt+⌧z(
t  

)0.5
Following BGG , we can again define:
Gb(!¯bt+1) =
ˆ !¯bt+1
0
!bt+1f
b(!bt+1)d!
b
t+1 + !¯
b
t+1
ˆ 1
!¯bt+1
f b(!bt+1)d!
b
t+1
Gb(!¯bt+1) =
ˆ !¯bt+1
0
!bt+1f
b(!bt+1)d!
b
t+1
107
Gb(!¯bt+1) denotes the share of total bank assets that belong to banks which end up
in default. Thus, µbGb(!¯bt+1) is the total cost of bank default expressed as a fraction
of total bank assets where µb is the fraction of assets lost in bankruptcy.
Banks choose total bank loans, Bt, the fraction of bank loans diverted to OBSE,
bt, efficient level of bank capital ratio, t, and the monitoring effort, et to maximize
the following profit function subject to equation (2.11) and (2.14):
max
Bt,bt,t,et
Et (1 bt)et eRF,t+1Bt(1 Gb(!¯bt+1))Rz,t+1Zt + ⌧z(t   )0.5 (2.15)
Finally, the probability of bank default, which also enters the households’ budget
constraint is given by:
PDt =
ˆ !¯bt+1
0
f b(!bt+1)d!
b
t+1 (2.16)
2.4.4 Shadow Banks
Shadow banks solely serve the purpose of securitization and to provide regulatory
relief to commercial banks. In reality shadow banks comprise of myriad range of in-
stitutions and instruments. However, here we characterize the system as special off
balance sheet entities that purchase a part of loan portfolio of commercial banks and
refinance these loans through underwriting of structured debt instruments (ABCP,
ABS etc.). We assume a continuum of shadow banks of measure one. Unlike com-
mercial banks, they are not subject to regulatory capital requirement. Their balance
sheet can be represented as:
Balance Sheet: Shadow Bank
Assets Liabilities
Bs,t ABSt
108
Where Bs,t is the amount of loan sold to the shadow bank. These loans are then
underwritten into marketable asset backed securities, ABSt. According to literature,
the competition in underwriting industry in the U.S can be described as monopolistic
competition (see for example, Fang (2005)).
Loan sale in secondary market There is a continuum of shadow banks, indexed
by z ✏ [0, 1], and they have some market power when conducting their services. A
commercial bank seeking to raise an amount of borrowing, Ps,tBSt, through loan sale
would allocate it among different shadow banks, Ps,t(z)BSt(z), so as to maximize
the total proceeds. In period t, commercial banks choose how much loan to sell to
shadow bank z, BSt(z) by solving the following problem:
maxBSt(z)
ˆ 1
0
Ps,t(z)BSt(z)dz,
subject to a Dixit-Stiglitz aggregator:
BSt = {
ˆ 1
0
(BSt(z))
"s1
"s dz} "s"s1 ,
where Ps,t(z) is the price paid by the zth shadow bank and "s > 1 is the elasticity
of the demand for funds. The first order condition yields the following commercial
bank’s demand for funds:
BSt(z) = (
Ps,t(z)
Ps,t
)
"s
BSt,
where Ps,t is the average sale price of loans in the secondary market, defined as:
Ps,t = {
ˆ 1
0
(Ps,t(z))
1"sdz} 11"s
Therefore, when the price paid by the z-th shadow bank falls relative to the
average price, the commercial bank wishes to sell less loans to that shadow bank.
109
Shadow bank profit maximization The loan sale in previous section is financed
by underwriting of asset backed securities. Households are the ultimate investors in
these securities23. To keep the analysis as simple as possible, we follow the recent
DSGE banking literature and assume perfect competition in the market for house-
holds’ funding (see e.g. Kobayashi, 2008). At the end of period t, the z-th shadow
bank thus chooses the price Ps,t(z) to solve the following profit maximization prob-
lem:
max
Ps,t(z)
et eRF,t+1BSt(z)Rm,t+1Ps,t(z)BSt(z),
subject to
BSt(z) = (
Ps,t(z)
Ps,t
)"s,BSt
Deriving the first order conditions and imposing symmetric equilibrium yields:
et eRF,t+1 = "s
"s  1Ps,tRm,t+1 (2.17)
As mentioned before, loans purchased by shadow banks are financed by under-
writing marketable liquid ABS24 wherein, ABSt = Ps,tBSt. The average return on
ABS underwritten by shadow banks, Rs,t+1 is therefore:
Rs,t+1 =
et eRF,t+1BSt
Ps,tBSt
,
23The process of securitization whets the appetite of institutional investors for liquid deposit like
safe investments. To avoid introducing too many agents in the model, we allow households to be
the ultimate investors in ABS. Alternatively, another agent institutional investors can be added to
the model who would mobilize resources from households in return of mutual fund shares. These
funds can then be used by institutional investors to buy ABS. This would however not alter the
dynamics of the model.
24The shadow bank (the underwriter) acts as an intermediary between the ultimate borrower and
the ultimate investors. In bond finance, the most common type of underwriting arrangement is the
firm commitment underwriting, according to which the underwriter buys the entire stock of bonds
from the firm and resells it to investors at a higher price (i.e., at a lower interest rate). This spread
represents the shadow bank’s profits. We use a similar setup for ABS underwriting. See Ellis et al.
(2000) for an in-depth analysis of the underwriting process.
110
where numerator is the return on loan purchases and denominator is the price
paid by shadow banks for these loans25. Therefore, equation (2.17) can be written
as:
Rs,t+1 =
"s
"s  1Rm,t+1
Households’ maximization problem suggests that the return on shadow bank de-
posits, Rm,t is equal to the return on deposits (i.e. Rm,t+1 = eRt+1). Thus, the cost
of issuing ABS is a markup, "s"s1 , over the deposit rate. The profits of the shadow
banks in period t+1 is given by:
⇧I,t+1 = (Rs,t+1  eRt+1)ABSt
2.4.5 Consumption good producer
Consumption good producer uses capital Kt rented from entrepreneurs and labour
Lt hired from households to produce output Yt with the following constant returns
to scale production function:
Yt = AtK
↵
t1L
1↵
t ; 0 < ↵ < 1
The aggregate technology parameter, At evolves as:
ln(At) = ⇢a ln(At1) + "a,t; ⇢a < 1,
"a,t is normally distributed with zero mean and standard deviation a.
25Since we assume that shadow banks receive a fraction of loan from all banks, they can per-
fectly diversify all the risk and are thus guaranteed a risk-less return in aggregate (still subject to
monitoring effort of the commercial bank). Therefore, the value added of shadow bank is to allow
regulatory arbitrage in the model, which has been hailed in the literature as one of the key factors
for the growth of securitization and shadow banking.
111
2.4.6 Capital good producer
At the beginning of each period, capital good producer buys final good and the
stock of undepreciated capital from the firms. We introduce convex investment costs,
therefore at time t, It units of output is converted into (1  S(Xt))It units of new
capital26. Xt = It/It1 and the function S(.) satisfies S 0, S 00  0. Also in steady
state S(.) = S 0(.) = 0.
S(Xt) = 'x(Xt  1)2
The capital goods producer maximizes expected discounted profits:
max
It
Et
P1
k=0Dt,t+k[Qt+k(1 S( It+kIt+k1 ))It+k  It+k],
where Dt,t+k is the real stochastic discount factor. Capital therefore evolves
according to the following relation:
Kt+1 = (1 )Kt + (1 S(Xt))It
2.4.7 Deposit insurance agency
Deposit insurance agency absorbs the costs of servicing the deposit payments of
banks that default. This is in turn covered by imposing lump-sum taxes, TDt, on
households where
TDt = [!¯
b
t  Gb(!¯bt ) + µbGb(!¯bt )]et1 eRF,t(Qk,t1Kt1 Nt1)(1 bt1)
26In the absence of investment adjustment costs, the capital price Qt is constant and equals 1.
Investment adjustment costs generate capital price variability, which contributes to the volatility
of entrepreneurs’ net worth.
112
2.4.8 Competitive Equilibrium
Households are present in unit mass therefore aggregate household consumption
is equal to consumption of a representative household. Aggregation across en-
trepreneurs’ (described in section 2.4.2) yields aggregate entrepreneurial consumption
(Ce,t). Further, aggregate resources lost in entrepreneurial bankruptcy are given by
µeGe(!¯et )Rk,tQk,t1Kt1.
Combining the budget constraints of the model’s agents gives us the resource
constraint of the economy:
Yt = Ct + It + Ce,t + µeG
e(!¯et )Rk,tQk,t1Kt1 + ⌦PDtRt1Dt1 (2.18)
+µbG
b(!¯bt )et1 eRF,t(Qk,t1Kt1 Nt1) + (1 et1) eRF,tBt1(1 Gb(!¯bt ))
Total output Yt, is equal to the aggregate consumption demand of households,
Ct, plus the resources absorbed in the production of new capital, It, plus aggregate
consumption of entrepreneurs who exit in period t, plus the resources lost in the
recovery by commercial banks of the proceeds associated with defaulted bank loans,
in transaction costs by depositors at failed banks and by the deposit insurance agency
in the recovery of assets from banks that default plus the loss in effective bank return
due to diversion/inefficiency by bank managers.
In addition labour market equilibrium requires that labour demand from firms is
equal to household’s labour supply at the competitive real wage rate. Loan market
equilibrium implies that amount borrowed by entrepreneurs is equal to loan supply
from banks that is financed by deposits, bank capital or sale of loan in the secondary
market. Market for capital services require that supply from entrepreneurs is equal to
the capital demanded by firms. Finally, equilibrium in the ABS market requires that
ABS issued by shadow banks is equal to the demand for securities by households.
113
A competitive equilibrium consists of a set of prices and interest rates {Qk,t, Rt, Rm,t,
Rk,t, wt} for all possible states and for all t  0 such that all markets clear when
the agents in the model i.e. households, entrepreneurs, commercial banks, shadow
banks, consumption good producer and capital goods producer solve their maximiza-
tion problems while taking these prices and interest rates as given.
2.5 Baseline Parameterization
The model is calibrated to replicate key features of the US economy and the param-
eter values are given in table 9. The parameters fall into two broad categories. The
first consists of the standard, more widely understood and estimated macroeconomic
parameters, for which values are borrowed from previous studies. The second are the
parameter values more specific to the financial system presented in this model and
deserve special mention. We consider two model, one is the case of securitization in
which financial sector of the model imitates features of the U.S. in the decade prior
to the financial crisis. The other case is that of no-securitization. In an absence of
any counterfactual, we draw out inferences from the features of the U.S. financial
system of the early 1990’s as in Meeks et. al. (when there was some securitization,
but lending was predominantly done by banks).
As regards the macroeconomic parameters, household’s discount factor, , is set
to 0.99 to match the historical average of risk free interest rate, Rt. Parameter ,
the inverse labour supply elasticity is set to 0.5, as is common in the literature.
The share of capital in production, ↵ and its depreciation rate,  is set at 0.33 and
0.025 respectively, which is standard in the literature. The calibrated values for
these parameters are same as those used in chapter 1. The probability of survival
of firms to next period, , is set at 0.97 (as in BGG(1999) and Christensen and Dib
(2008)), which implies an expected working life for entrepreneurs of 36 years. The
values of other parameters controlling firm financial frictions (e.g.. µe, F ($), and e)
are determined to match, RF  R, leverage ratio and the rate of return on capital.
Bankruptcy cost parameter, µe, is set at 0.3 that is within the range of 0.20 - 0.36
that Carlstrom and Freust (1997) defend as empirically relevant. F (!¯) is set at 0.75
114
quarterly percent value implying an annual bankruptcy rate of 3 percent (as in BGG).
The variance of the entrepreneurial risk shock, 2e , is set at 0.26 following Christiano,
Motto, Rostagno (2014). There is considerable uncertainty about the spread between
the cost of external finance, RF and the return on households deposits. BGG (1999)
measure the external finance premium as approximately the historical average spread
between the prime lending rate and the six- month Treasury bill rate, which amounts
to 200 basis points. Levin, Natalucci and Zakrajsek (2004) report a spread of 227
basis points for the median firm included in their sample. De Fiore and Uhlig (2011)
report that the spread between the prime rate on bank loans to businesses and the
commercial paper is 298 basis points over the period 1997-2003. Our calibrations
produce a spread of 300 basis points. Finally, the calibrated values of fraction of net
worth consumed by entrepreneurs upon exit, ⇥, and the transfer from households
to entering entrepreneurs, W e, is borrowed from Christiano, Motto and Rostagno
(2014).
Capital requirements are set at a benchmark level of 8 percent for corporate
loans (compatible with the full weight level of Basel I and the treatment of non-
rated corporate loans in Basel II and III). Parameter determining the probability of
default of banks is chosen so as to make their baseline steady-state values equivalent
to annual rates of 2 percent for banks. The bankruptcy cost parameters imply
losses of 10 percent of face value of deposits for depositors at failed banks and of
30 percent of asset value for creditors repossessing assets from defaulting borrowers
(Clerc et. al. (2015)). Parameters ⌧1, ⌧2 and ⌧z are set to hit the following targets
- a steady state bank capital ratio of 10 percent i.e. higher than the regulatory
constraint, share of bank loans sold and securitized, bt, at 0.3 following Meeks et. al.
(2014) who find that the share of securitized asset based on call report data on bank
asset sold and securitized at 30 percent and an annualized return on bank equity
of 15 percent, matching the evidence reported by Berger (2003). Adjustment cost
parameter associated with household purchase of bank equity, ⇥z, is calibrated at
0.36 to ensure that household demand for bank equity is in accordance with banks’
steady state capital ratio. Finally, the steady state value of the elasticity of funds in
the secondary loans market, "s is calibrated to ensure a risk spread of 50 basis point
115
(above the deposit return) in the ABS market, as in Meeks et. al. (2014)27.
2.6 Results
In this section, we compare two versions of our model economy - with and without
securitization. In particular, the no securitization economy is characterized by the
share of loans sold in the secondary market, bt = 0. Furthermore, since banks do not
sell loans in the secondary market they have a higher incentive to monitor resulting
in a higher steady state monitoring effort as compared to the baseline model. There
is a range of evidence from various markets that borrowing rates fell amongst assets
that could be sold and securitized. For example, in the market for corporate loans,
Nadauld and Weisbach (2012) cite a reduction in the yield of 15 basis points. The
steady state lending rate in the no-securitization model is therefore targeted at an
annualized 20 basis point higher than in the baseline model. The results presented
below are based on second order approximation of the model and are computed
using Dyane 4.3.3. The impulse response functions are based on averaging 1000
simulated series for each variable. Dynare generates IRFs using the approximate
policy functions as follows: (i) It draws a series of all the exogenous shocks for
a number of 100 + T periods, where T is the number of periods shown in the
IRF graphs, (ii) It performs a simulation H1 of all the variables of interest using
this realization of the shocks, (iii) It changes the sequence of exogenous shocks, by
adding a one-standard deviation of the shock, to the realization of period 101, (iv) It
performs a new simulation H2 of all variables based on the new sequence of shocks,
and (v) It calculates the IRF from H2-H1. These steps are then repeated a number of
times (we choose 1000 replications) and the produced IRFs are averages of the series
from these 1000 experiments. We begin our analysis by looking at the response of the
model economy to (i) a standard productivity shock; and (ii) shocks emerging in the
27The 2000–2007 average ABS spread over comparable swap rates for high-quality securitizations
varied from around 7 basis points for credit card and auto receivables, to 25 basis points for large
equipment and 70 basis points for non-conforming mortgages. Meeks et. al. (2014) adopt a rough
mid-point of 50 basis points for the steady state ABS spread in their model.
116
financial system - a risk shock, which increases the volatility of banks’ idiosyncratic
risk.
2.6.1 Total factor productivity shock: shock to the aggregate
productivity
Figure (10) reports the effect of a one percent decline in total factor productivity.
We compare the results from our baseline model to a model with no shadow bank-
ing. A decline in aggregate productivity results in a drop in capital return in both
models. This adversely affects capital price and net worth of the banks in both cases.
However, the impact is muted under the baseline model. While capital price and net
worth falls by about 2 percent in no shadow banking model, the fall is less than half
of that in a model with shadow banking. Similarly, fall in investment is lower in the
baseline model. The dynamics of the two models differ due to differing response of
monitoring effort and the ability to sell loans in the baseline model. In response to
a negative productivity shock, the return on deposits increase. This increases the
cost of funding for shadow banks thereby decreasing their overall issuance of ABS,
resulting in a lower proportion of loans sold in the secondary market. As a larger
share of commercial bank loans is now kept on balance sheet, they have greater in-
centives to monitor reflected by an increase in bank monitoring of about 0.5 percent.
Therefore, the initial increase of interest rate on bank loans is reversed more quickly
in the baseline model. The existence of shadow banking has a dampening effect on
the fall in investment and overall credit intermediation.
2.6.2 Bank risk shock: shock to the volatility of banks’ idiosyn-
cratic risk
Next, we look at the impact of a financial shock. Figure (11) shows the dynamic
effects of a 0.1 percentage point increase in the standard deviation of the idiosyncratic
shock to banks’ performance (with a high persistence of 0.9), which we interpret as
a shock to bank risk. The size of the shock is chosen such that it causes the loan
interest rate to increase by about 0.25 basis point (100 basis point increase of the
117
annualized interest rate). Later, we also check robustness with a lower persistence
and larger size of the shock. The effects of the shock are amplified in the economy
with shadow banking (baseline model). An increase in bank risk implies a higher
probability of bank default. In a model without shadow banking, banks resort to
higher monitoring to increase the effective return on their portfolio. This significantly
contains the impact of risk shock on loan interest rate and consequently investment
and loan supply. However, in a model with shadow banking, banks respond to a
greater risk by shifting loans to the secondary market as can be seen by an increase
of nearly 6 percent in the loan share sold. Lower proportion of on-balance sheet
loan leads to reduced monitoring (a fall in monitoring of over 1 percent) effectively
downgrading the quality of bank portfolio. A fall in monitoring increases interest
rate charged on loans (to ensure the effective return can cover the cost of funding
for banks) as described in section 2.4.3. Finally, this results in an amplified and
prolonged fall in output, investment, capital price and banks’ net worth vis-a-vis a
model with only commercial banks. Overall credit intermediation in the baseline
model falls about tenfold than in the no shadow banking economy. The peak decline
in loans in the baseline model is over 5 percent in contrast to about 0.5 percent
in a model without shadow banking. We see similar dynamics for capital, capital
price and investment. Capital in the baseline model falls by just under 2 percent,
while investment falls by about 5 percent. In contrast, the impact on these variables
in the economy without shadow banking is negligible in relative terms. Output in
no-shadow banking economy falls by just over 0.5 percent, primarily driven by the
fall in consumption. Furthermore, the impact is short-lived and output goes back to
steady state in about 10 quarters. In the baseline model, not only is the peak impact
magnified (output declines by 1 percent) but it’s also more persistent. Finally, as
the quality of loan portfolio falls, loan price in secondary market falls by nearly 1
percent. This also leads to a fall in the ABS return. Therefore, as commercial banks’
risk increases a larger portion of loans is sold into the secondary market increasing
the fragility of the system. It also significantly constrains the credit intermediation
process consequently causing an amplified decline in investment and output.
Next, we look at the impact of risk shock under different shock size and persis-
118
tence. The baseline scenario (shock of 0.1 percentage point increase, 0.9 persistence)
is compared with three additional scenarios in figure (12). Additional scenarios in-
clude (i) scenario 1: shock size of 0.2 percentage point, 0.9 persistence; (ii) scenario
2: shock size of 0.1 percentage point, 0.5 persistence; and (iii) scenario 3: shock
size of 0.2 percentage point, 0.5 persistence. Comparing baseline model and sce-
nario 1 shows that the impact of risk shock is highly non-linear. Doubling the size
of shock from 0.1 to 0.2 percentage point leads to larger than three-fold decrease
in key variables - output, investment, loans, capital, capital price and net worth -
as compared to the baseline scenario. The persistence of shock has an important
role to play as well. The impact of baseline model and scenario 3 are very similar.
Baseline has a smaller risk shock that scenario 3, however that is counterbalanced
by a lower persistence in scenario 3. Overall, the impact is amplified as the size or
the persistence of shock increases. A large enough risk shock augmented with high
persistence is therefore capable to create a crisis like scenario (scenario 2). In such
a scenario, credit intermediation is seriously hampered reflected in a decline of loans
by 20 percent. Capital price falls by about 8 percent and entrepreneurs net worth
falls by 10 percent. A large part of loan portfolio is sold in the secondary loan market
at low prices as the overall quality of loan portfolio falls with declining monitoring
effort. Finally, investment falls by over 15 percent and peak impact on output is a
decline of 4 percent.
Second v/s first order approximation The results presented above are based
on a second order approximation of the economy around the steady state. Under
first order approximation, unconditional expectations of the endogenous variables
are equal to their non- stochastic steady state values. Therefore, risk or uncertainty
shock - in essence shock to the variance of the shock do not affect the dynamics of
the model. Figure 12 shows that responses to risk shocks are highly non-linear.
To further assess the non-linearity of the system, figure 13 compares impulse
responses under first and second order approximation of the model around the de-
terministic steady state. In the top panel, responses to one percent technology shocks
are larger under second order approximation. The amplification is most pronounced
119
for monitoring effort and loan sale in secondary market, which consequently leads
to larger decline in capital and loans for second order approximation of the model.
In particular, peak decline in capital and investment under second order approxima-
tion is nearly twice of that seen under first order approximation. The lower panel,
which reports impulse responses to risk shocks, shows a similar pattern however the
non-linearities are even more pronounced here. The response under second order is
significantly larger, so much so that results under first order approximation appear
negligible in comparison. In general, the impact under second order approximation
is about tenfold of the responses under first order. For example, while for first order
approximation output falls by less than 0.1 percent in response to risk shock (of 0.1
percentage point), the decline is nearly 1 percent when second order approximation
of the model is used. Similarly, peak decline in investment is over 4 percent when
the model is solved using second order in comparison to about 0.4 percent when first
order approximation is used. Again the amplification, under second order approxi-
mation, is driven by significantly larger decline in bank monitoring and higher share
of bank loans sold in secondary market.
2.6.3 Robustness Checks
In section 2.6.2 (figure 11 and 12), we documented the response of key variables
to bank risk shock. Figure 12 shows that the magnitude of variation in economic
variables in response to the shock depends to a large extent on the size and persistence
of the shock. In this section, we assess the robustness of our results to the choice of
calibrated parameters. The baseline calibration is guided by (i) standard parameters
for which there is agreement in the literature; and (ii) parameters which are specific
to the model and are calibrated to match certain steady state values. The solution
under the baseline parameterization is stable and unique i.e. the Blanchard-Kahn
condition is satisfied (the number of eigenvalues larger than one in modulus is equal to
the number of forward looking variables). Further, in response to temporary shocks
the system converges back to the steady state. For robustness, we take the standard
parameters as given and increase the calibrated values of parameters specific to the
120
model by 10 percent. In particular, for this robustness exercise we increase by 10
percent the parameters related to the financial system (e, b, µe, µb,⌦, ⌧1, ⌧2, ⌧z, "s).
Figure 14 shows the response of model economy to aggregate productivity shock
under the baseline and robustness scenario. The results are largely robust to the
change in calibrated parameters. Further, the solution is again unique and stable
(Blanchard-Kahn conditions are satisfied). Similarly, impulse responses to risk shock
are also robust to the measured change in calibrated parameters (see figure 15).
2.6.4 Government credit policy
The problems experienced in credit and interbank markets at the onset of the sub-
prime crisis in August 2007 were swiftly followed by policy actions from the central
bank and government, including cuts in official interest rates and enhanced provision
of liquidity. Overtime as the crisis intensified, policy makers have taken recourse
to various unconventional monetary and fiscal policy tools. In the United States,
official backstops for the banking system have been a prominent component of the
policy response (Pozsar et al., 2010). These have included a policy of providing
long-term liquidity through the TALF28, and outright purchases of Agency MBS
and debt funded by central bank reserves. In this section, we describe the operation
and impact of government backstop programs.
Credit Policy We have previously characterized the credit intermediation process
via lending by both the traditional retail banks and the shadow banks. We now
suppose that the central bank is willing to facilitate lending. It can do this by lending
directly to firms i.e. loan purchase. Therefore, overall credit intermediated can be
written as the sum of private (Bp,t) and government (Bg,t) assisted intermediation.
28Under TALF, the Federal Reserve Bank of New York (NY Fed) lent up to $1 trillion (originally
planned to be $200 billion) on a non-recourse basis to holders of certain AAA-rated ABS backed
by newly and recently originated consumer and small business loans. In the United Kingdom, the
SLS (Special Liquidity Scheme) was also aimed at long-term liquidity provision. It allowed banks
to undertake swaps of securitized assets for Treasury bills, which could then be used as collateral
to obtain secured funding in wholesale markets.
121
Bt = Bp,t +Bg,t (2.19)
To conduct credit policy, the central bank issues government debt to households
that pays the risk-less rate Rt and then lends the funds to non-financial firms. The
return from these loans is assumed to be equal to the return on a well diversified
portfolio, eRF,t+1. Thus for households, government debt is a substitute for bank
deposits. We suppose that government intermediation involves efficiency costs: in
particular, the central bank credit involves an efficiency cost of ⌧g per unit supplied.
This deadweight loss could reflect the costs of raising funds via government debt.
It might also reflect costs to the central bank of identifying preferred private sector
investments. On the other hand, the government always honors its debt: thus, unlike
the case with private financial institutions there is no agency conflict that inhibits
the government from obtaining funds from households. Put differently, unlike private
financial intermediation, government intermediation is not balance sheet constrained.
The overall government intermediation costs therefore are:
Gt = ⌧g ⇤Bg,t (2.20)
If we define Tt as government transfers and assume that at every period the
government is going to fund its asset purchases through risk-free bonds then:
⌧g ⇤Bg,t = Tt + ( eRF,t+1 Rt)Bg,t
Government intermediation policy takes the form of buying a proportion 'i of
the steady state value of loans. In particular, we suppose that at the onset of a
crisis, which we define loosely to mean a period where credit spreads rise sharply,
the central bank injects credit in response to movements in credit spreads, according
to the following feedback rules:
'l,t = 'l + ⌫lEt[( eRF,t+1 Rt) (R˜F R)] (2.21)
122
Where 'l is the steady state fraction of publicly intermediated assets and R˜FR is
the steady state premia. In addition, the feedback parameter ⌫i is positive. According
to this rule, the central bank expands credit as the spread increases relative to its
steady state value.
Risk shock with credit policy We now look at the response to risk shock (as
in figure 11) with and without government credit policy. The results from this
exercise are presented in figure 16. As mentioned before, risk shock leads to a decline
in investment, output and capital price, which is amplified by the usual financial
accelerator channel. Total credit intermediation in the economy falls with loans
falling by about 6 percent in the baseline economy. An increase in banks’ probability
of default reduces bank monitoring effort leading to higher proportion of loans sold
and securitized. The effective return on the portfolio of banks falls further leading
to a reduction in credit intermediation and consequently investment and output. At
its peak investment falls by over 4 percent and output by 1 percent in the baseline
economy.
We evaluate the impact of central bank backstop programs in forms of direct loan
purchase. Loan purchases go a long way in ameliorating the impact of this shock
on the economy. We calibrate the steady state fraction of publicly intermediated
assets 'i at 2.5 percent and the feedback parameter is set to ensure that government
intermediation as a proportion of total capital in the economy increases to around
5 percent. Direct loan purchases stabilize the economy through various channel.
First, since the loan sold to government does not suffer from monitoring inefficiency,
overall effective return on loans does not fall as much as in the baseline economy.
This translates to a damped response of loan interest rate to the shock preventing
the interest rate spread from soaring. While in the baseline economy loan spread
increases by about 150 basis point (annualized), the impact is muted with credit
policy (less than 100 basis point increase in the risk spread). Second, government
intervention helps in stabilizing the capital price in the economy and thus limits
the amplification process. Capital price declines by 2 percent in baseline model and
only about 1 percent with an active credit policy. This significantly dampens the
123
response of firms’ net worth and consequently output and investment in the economy
with active credit policy. It also limits the maximum fall in total credit intermediated
from 6 percent to about 3 percent. The fall in output and investment is also reduced
by half when credit policy is in place. Furthermore, government through its credit
policy can also significantly shorten the duration of the impact of risk shocks on key
macro economic variables. While with no policy response the impact of risk shock
is long lived, active credit policy significantly lowers the number of periods (here
quarters) it takes for the economy to return to equilibirum.
2.6.5 Capital Regulation
The financial crisis has reinforced interest in macroprudential tools and policies that
might be applied by policymakers to reduce the risks of financial boom and bust
cycles and thereby lead to a more stable path of real economic growth. In addition
to the attention on the role of financial intermediaries brought forward by the fi-
nancial crisis, the introduction of more risk-sensitive capital requirements (i.e., the
Basel II capital adequacy framework) has reinforced the concerns that financial in-
termediation by itself might have substantial feedback effects on the real economy.
In this section, we aim to analyze the impact of capital regulation on macroeconomic
dynamics. In particular, we look at the procyclicality of the economy under Basel
II.
Procyclical Impact of Capital Regulation In this section, we compare the
impulse responses for 2 alternative scenario which we call Basel I and Basel II. Next
we describe the form of minimum capital constraints in these two alternative regimes.
The capital requirement used in the paper until now is in accordance with Basel I
accord. We assumed that minimum tier 1 regulatory capital was fixed at 8% of the
risk weighted assets and used that to compute the constraints on commercial bank
leverage. Basel II however calls for a procyclical capital regime. Therefore, we can
redefine the minimum capital requirement as:
124
Capt = C0A
c1
t (C0 > 0)
For Basel I c1 = 0, C0 = 0.08.
Let us first discuss the calibration of capital requirement ratio. Under the pro-
cyclical regulation regime, the capital requirement ratio increases (decreases) when
the economy is in a downturn (boom), implying that c1 < 0. We calibrate C0 and
c1 by using Basel II risk-weight formula29. Note that Basel II requires banks to cal-
culate the risk weight for each loan based on the formula. We follow Covas et. al.
(2010) and calibrate c1 = 8.30.
The response of the economy to one percent negative productivity shock under
the two alternative regulation regimes is shown in figure 17. In response to a negative
productivity shock, minimum capital requirement under Basel II increases i.e. the
capital buffer reduces. This leads to a higher increase in the lending rate under Basel
II regime. Overtime bank capital increases and lending rate falls. Bank capital ratio
and required return on bank equity are not significantly affected in the Basel I regime.
A higher increase in lending rate leads to a much higher fall in loans under Basel II
than in Basel I, which due to the financial accelerator mechanism leads to a higher
fall in capital price, net worth of entrepreneurs and investment. Overall loans fall
by 7 percent under basel II as opposed to about 1 percent under Basel I. Similarly,
capital price and net worth decline by over 2 percent under Basel II while the fall in
case of Basel I is under 1 percent. Investment decline of about 6 percent under Basel
II is sixfold of the decline under basel I. Finally, output reduction is also amplified
under Basel II regime. Following a fall in total loans both on and off balance sheet
29The internal ratings based (IRB) approach uses the probability of default (PD), the loss given
default (LGD), the exposure at default, and maturity for each exposure (M) to calculate the bank’s
capital requirement for each loan (see Basel Committee on Banking Supervision (2004)). The capital
requirement ratio is derived from the formula: Capitalt = LGD[N(N
1(PDt)+
p
Ct⇥N1(0.999)p
1Ct ) 
PDt]⇥ ( 1+(M2.5)⇥bt11.5⇥bt ),
where N(.) is the standard normal distribution function and C the correlation factor.
30Covas et. al. (2010) calculate the average risk weight (i.e., the average capital requirement
ratio) by using average values for PD, LGD and M. LGD is taken to be 40%. and they use the
Moody’s default rate series for PD.
125
loans decline in fixed capital requirement case. However in the basel II model, an
increase in capital requirement leads to a higher share of securitization and shift of
loans from on to off balance sheets as can be seen from the loan share sold in the
secondary market. Furthermore, there is also a reduction in bank monitoring which
has similar effects as described earlier. Therefore, there are two perverse effects in
response to a negative productivity shock under Basel II with procyclical capital
constraints. First, there is a greater tendency to move to the unregulated shadow
banking sector and second, banks face higher incentives to reduce monitoring effort
thereby lowering the effective return on their portfolio.
2.7 Concluding Remarks
The financial crisis that started in 2007 exposed a number of flaws in the financial
system. Many of these were associated with financial instruments that were issued
by the shadow banking system. The growth of the shadow banking system as an
alternative to traditional banking had been going on for a number of years, but ac-
celerated rapidly with the expansion of securitization in recent years. Even though
the crisis brought forward the ills of the shadow banking system, securitization is
still significant and important in the credit intermediation process. Keeping that
in mind, regulatory authorities are now seeking to reform the financial system and
shadow banking in particular. This paper provides guidance on the interaction be-
tween traditional and shadow banking in the economy. We analyze both the credit
intermediation process as well as the impact of current regulatory regimes.
In particular we build a Dynamic Stochastic General Equilibrium model with an
active role for bank capital and regulation. In addition we also embed features of the
shadow banking system. For ease of exposition, we only focus on off balance sheet
entities as the shadow banks, which are specialized financial institutions designed to
transfer a part of loans from banks balance sheet to offer regulatory capital relief.
Shadow banks bundle these loans to offer liquid securities.
Our analysis highlights that originate to distribute model makes the economy
vulnerable to financial market shocks. An increase in bank risk leads to a sale of
126
larger share of loans in the secondary market in the presence of shadow banks. This
generates perverse incentives for banks to lower their monitoring effort, reducing the
effective return on their portfolio. The overall impact is a substantial fall in credit
intermediation, investment and output in contrast to an economy without shadow
banking. Government intervention in the form of direct loan purchase proves to be
an effective measure in containing the impact of a negative shock emanating within
the financial system.
We also study the role of bank capital regulation. Higher capital provides gain
in the form of lower borrowing costs as it signals a healthy banking system. We
analyze the behavior of the economy under two alternative regulatory regimes, Basel
I with fixed capital requirement and Basel II with procyclical capital constraints. In
response to a negative productivity shock, we see a fall in economic activity, which
is amplified under Basel II. Furthermore, under Basel II a negative technology shock
leads to a shift in the economic activity from regulated traditional banking to the
unregulated shadow banking sector. Therefore, there is a need to investigate counter
cyclical capital requirement as being implemented under Basel III accord.
Finally, while this paper provides a model of shadow banking in the presence of
regulatory arbitrage it does not provide micro-foundations for the shadow banking
system - including pricing of risk in the shadow banking, and a possible run on the
shadow banks. It is difficult to pin down multiple and complex determinants of ABS
finance in a DSGE model – we therefore defer that task to more stylized models of
finance.
127
Table 10 – Calibrated Parameters
Parameter Description Value
Households
 Household discount factor 0.99
' Inverse labour supply elasticity 0.5
⇥z adjustment cost parameter, bank equity transactions 0.36
⌘ weight attached to leisure in utility function 9.06
Production
 Capital depreciation rate 0.025
↵ Share of capital in production 0.33
 Survival probability of firms 0.97
'x Investment adjustment cost parameter 12
Entrepreneurs
2e Variance of entrepreneurial idiosyncratic shock 0.255
µe Entrepreneur bankruptcy cost 0.3
⇥ Proportion of net worth consumed by exiting entrepreneurs 0.10
W e Transfer from households to entering entrepreneurs 0.009
Financial System
 Capital requirement 0.08
2b Variance of banks’ idiosyncratic shock 0.046
µb Banks’ bankruptcy cost 0.3
⌦ Depositor cost of bank default 0.1
⌧1 Monitoring cost parameter 1.41
⌧2 Monitoring cost parameter 0.997
⌧z Coefficient on bank capital buffer 0.0022
"s Elasticity of funds in ABS market 298.2
Shocks
⇢a, ⇢ Baseline persistence of technology and risk shock 0.9
128
Figure 10 – TFP shock: One percent drop in TFP
0 20 40 60
−1.5
−1
−0.5
0
Output
 
 
Baseline No Shadow Banking
0 20 40 60
−1.5
−1
−0.5
0
Consumption
0 20 40 60
−2
−1.5
−1
−0.5
0
0.5
Investment
0 20 40 60
−2
−1.5
−1
−0.5
0
0.5
Capital Price
0 20 40 60
−2.5
−2
−1.5
−1
−0.5
0
Net Worth
0 20 40 60
−1
−0.5
0
0.5
1
Bank Capital Ratio
0 20 40 60
−0.5
−0.4
−0.3
−0.2
−0.1
0
Capital
0 20 40 60
0
0.05
0.1
0.15
0.2
Deposit Rate
0 20 40 60
−0.05
0
0.05
0.1
0.15
Bank Equity Return
0 20 40 60
−2
−1.5
−1
−0.5
0
0.5
Capital Return
0 20 40 60
−0.05
0
0.05
0.1
ABS Return
0 20 40 60
0
0.05
0.1
0.15
0.2
0.25
Loan Interest Rate
0 20 40 60
−0.2
0
0.2
0.4
0.6
Bank Monitoring
0 20 40 60
−0.2
0
0.2
0.4
0.6
Loan Price
0 20 40 60
−4
−3
−2
−1
0
ABS
0 20 40 60
−4
−3
−2
−1
0
1
loan share sold
Note: IRFs to one percent decrease in total factor productivity (TFP). Based on
averaging one thousand simulation of a second-order approximation of the model
around the steady state. Y-axis denotes percentage deviation from the steady state
i.e. a one percent drop in TFP decreases output by one percent in the baseline model.
X-axis is time in quarters.
129
Figure 11 – Bank risk shock: 0.1 percentage point increase in the
volatility of bank’s idiosyncratic risk
0 50 100 150
−1
−0.5
0
0.5
Output
 
 
Baseline No Shadow Banking
0 50 100 150
−1
−0.5
0
0.5
Consumption
0 50 100 150
−6
−4
−2
0
2
Investment
0 50 100 150
−3
−2
−1
0
1
Capital Price
0 50 100 150
−3
−2
−1
0
1
Net Worth
0 50 100 150
−15
−10
−5
0
5
10
Bank Capital Ratio
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Capital
0 50 100 150
−8
−6
−4
−2
0
2
Loans
0 50 100 150
−1
−0.5
0
0.5
Bank Equity Return
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Capital Return
0 50 100 150
−1
−0.5
0
0.5
ABS Return
0 50 100 150
−0.6
−0.4
−0.2
0
0.2
0.4
Loan Interest Rate
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Bank Monitoring
0 50 100 150
−1.5
−1
−0.5
0
0.5
Loan Price
0 50 100 150
−5
0
5
10
ABS
0 50 100 150
−5
0
5
10
15
loan share sold
Note: IRFs to 0.1 percentage point increase in the standard deviation of banks’
idiosyncratic risk. Based on averaging one thousand simulation of a second-order
approximation of the model around the steady state. Y-axis denotes percentage
deviation from the steady state i.e. a 0.1 percentage point increase in bank risk
decreases output by one percent in the baseline model. X-axis is time in quarters.
130
Figure 12 – Bank risk shock: Comparison with different size and per-
sistence of shock
0 50 100 150
−4
−3
−2
−1
0
Output
 
 
Baseline Scenario 1 Scenario 2 Scenario 3
0 50 100 150
−4
−2
0
2
Consumption
0 50 100 150
−20
−15
−10
−5
0
Investment
0 50 100 150
−8
−6
−4
−2
0
2
Capital Price
0 50 100 150
−10
−5
0
5
Net Worth
0 50 100 150
−60
−40
−20
0
20
40
Bank Capital Ratio
0 50 100 150
−8
−6
−4
−2
0
Capital
0 50 100 150
−25
−20
−15
−10
−5
0
Loans
0 50 100 150
−4
−3
−2
−1
0
1
Bank Equity Return
0 50 100 150
−8
−6
−4
−2
0
2
Capital Return
0 50 100 150
−3
−2
−1
0
1
2
ABS Return
0 50 100 150
0
0.5
1
1.5
Loan Interest Rate
0 50 100 150
−10
−5
0
5
Bank Monitoring
0 50 100 150
−6
−4
−2
0
2
4
Loan Price
0 50 100 150
−20
0
20
40
ABS
0 50 100 150
−20
0
20
40
60
loan share sold
Note: IRFs to bank risk shock under different scenarios. Baseline: persistence 0.9,
shock size 0.1 percentage point; Scenario 1: persistence 0.9, shock size 0.2; Scenario
2: persistence 0.5, shock size 0.1; Scenario 3: persistence 0.5, shock size 0.2. Based
on averaging one thousand simulation of a second-order approximation of the model
around the steady state. Y-axis denotes percentage deviation from the steady state
i.e. a 0.1 percentage point increase in bank risk decreases output by one percent in
the baseline model. X-axis is time in quarters.
131
Figure 13 – First v/s second order approximation
Impulse response to one percent drop in aggregate productivity
0 20 40 60
−1.4
−1.2
−1
−0.8
−0.6
−0.4
−0.2
0
Output
 
 
Second Order First Order
0 20 40 60
−1.5
−1
−0.5
0
0.5
Investment
0 20 40 60
−0.8
−0.6
−0.4
−0.2
0
0.2
Capital Price
0 20 40 60
−1.4
−1.2
−1
−0.8
−0.6
−0.4
−0.2
0
Net Worth
0 20 40 60
−0.5
−0.4
−0.3
−0.2
−0.1
0
Capital
0 20 40 60
−0.8
−0.6
−0.4
−0.2
0
0.2
Loans
0 20 40 60
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Bank Monitoring
0 20 40 60
−4
−3
−2
−1
0
1
Loan Share Sold
Impulse response to 0.1 percentage point increase in volatility of bank’s idiosyncratic risk
0 50 100 150
−1
−0.8
−0.6
−0.4
−0.2
0
Output
 
 
Second Order First Order
0 50 100 150
−5
−4
−3
−2
−1
0
Investment
0 50 100 150
−2.5
−2
−1.5
−1
−0.5
0
0.5
Capital Price
0 50 100 150
−2.5
−2
−1.5
−1
−0.5
0
Net Worth
0 50 100 150
−2
−1.5
−1
−0.5
0
Capital
0 50 100 150
−7
−6
−5
−4
−3
−2
−1
0
Loans
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Bank Monitoring
0 50 100 150
−2
0
2
4
6
8
10
12
Loan Share Sold
Note: IRFs to one percent decrease in total factor productivity (TFP) and 0.1 per-
centage point increase in bank risk based on first and second order approximation
(averaging one thousand simulations) of the model around the steady state. X-axis
is time in quarters.
132
Figure 14 – TFP shock: Baseline and Robustness scenario
0 20 40 60
−1.5
−1
−0.5
0
Output
 
 
Baseline Robustness
0 20 40 60
−1.5
−1
−0.5
0
Consumption
0 20 40 60
−1.5
−1
−0.5
0
Investment
0 20 40 60
−1
−0.5
0
0.5
Capital Price
0 20 40 60
−1.5
−1
−0.5
0
Net Worth
0 20 40 60
−1.5
−1
−0.5
0
0.5
1
Bank Capital Ratio
0 20 40 60
−0.5
−0.4
−0.3
−0.2
−0.1
0
Capital
0 20 40 60
−0.8
−0.6
−0.4
−0.2
0
Loans
0 20 40 60
−0.05
0
0.05
0.1
0.15
Bank Equity Return
0 20 40 60
−1
−0.5
0
0.5
Capital Return
0 20 40 60
−0.05
0
0.05
0.1
ABS Return
0 20 40 60
0
0.01
0.02
0.03
Loan Interest Rate
0 20 40 60
−0.2
0
0.2
0.4
0.6
0.8
Bank Monitoring
0 20 40 60
−0.2
0
0.2
0.4
0.6
Loan Price
0 20 40 60
−5
−4
−3
−2
−1
0
ABS
0 20 40 60
−6
−4
−2
0
2
loan share sold
Note: IRFs to one percent decrease in total factor productivity (TFP) under baseline
and robustness scenario. Robustness scenario involves increasing financial sector
calibrated parameters (e, b, µe, µb,⌦, ⌧1, ⌧2, ⌧z, "s) by 10 percent each. Based on
averaging one thousand simulation of a second-order approximation of the model
around the steady state. Y-axis denotes percentage deviation from the steady state
i.e. a one percent drop in TFP decreases output by one percent in the baseline model.
X-axis is time in quarters.
133
Figure 15 – Risk shock: Baseline and Robustness scenario
0 50 100 150
−1
−0.8
−0.6
−0.4
−0.2
0
Output
 
 
Baseline Robustness
0 50 100 150
−1
−0.5
0
0.5
Consumption
0 50 100 150
−6
−4
−2
0
2
Investment
0 50 100 150
−3
−2
−1
0
1
Capital Price
0 50 100 150
−2.5
−2
−1.5
−1
−0.5
0
Net Worth
0 50 100 150
−15
−10
−5
0
5
10
Bank Capital Ratio
0 50 100 150
−2
−1.5
−1
−0.5
0
Capital
0 50 100 150
−8
−6
−4
−2
0
Loans
0 50 100 150
−1
−0.5
0
0.5
Bank Equity Return
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Capital Return
0 50 100 150
−1
−0.5
0
0.5
ABS Return
0 50 100 150
0
0.1
0.2
0.3
0.4
Loan Interest Rate
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Bank Monitoring
0 50 100 150
−1.5
−1
−0.5
0
0.5
Loan Price
0 50 100 150
−5
0
5
10
ABS
0 50 100 150
−5
0
5
10
15
loan share sold
Note: IRFs to 0.1 percentage point increase in the standard deviation of banks’ id-
iosyncratic risk under baseline and robustness scenario. Robustness scenario involves
increasing financial sector calibrated parameters (e, b, µe, µb,⌦, ⌧1, ⌧2, ⌧z, "s) by 10
percent each. Based on averaging one thousand simulation of a second-order approx-
imation of the model around the steady state. Y-axis denotes percentage deviation
from the steady state i.e. a 0.1 percentage point increase in bank risk decreases
output by one percent in the baseline model. X-axis is time in quarters.
134
Figure 16 – Risk shock: Response under baseline and active credit
policy
0 50 100 150
−1
−0.5
0
0.5
Output
 
 
Baseline Credit Policy
0 50 100 150
−1
−0.5
0
0.5
Consumption
0 50 100 150
−6
−4
−2
0
2
Investment
0 50 100 150
−3
−2
−1
0
1
Capital Price
0 50 100 150
−2.5
−2
−1.5
−1
−0.5
0
Net Worth
0 50 100 150
−20
−10
0
10
20
30
Bank Capital Ratio
0 50 100 150
−2
−1.5
−1
−0.5
0
Capital
0 50 100 150
−8
−6
−4
−2
0
Loans
0 50 100 150
−1
−0.5
0
0.5
ABS Return
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Capital Return
0 50 100 150
−0.2
0
0.2
0.4
0.6
Loan Spread (Rl − R)
0 50 100 150
0
0.1
0.2
0.3
0.4
Loan Interest Rate
0 50 100 150
−2
−1.5
−1
−0.5
0
0.5
Bank Monitoring
0 50 100 150
−1.5
−1
−0.5
0
0.5
Loan Price
0 50 100 150
−20
−10
0
10
ABS
0 50 100 150
−5
0
5
10
15
Loan Share Sold
Note: IRFs to 0.1 percentage point increase in the standard deviation of banks’
idiosyncratic risk under baseline and active credit policy. Based on averaging one
thousand simulation of a second-order approximation of the model around the steady
state. Y-axis denotes percentage deviation from the steady state i.e. a 0.1 percentage
point increase in bank risk decreases output by one percent in the baseline model.
X-axis is time in quarters.
135
Figure 17 – Procyclical capital requirement: Response to a negative
productivity shock under Basel I and Basel II
0 50 100 150
−1.5
−1
−0.5
0
Output
 
 
Basel I Basel II
0 50 100 150
−1.5
−1
−0.5
0
0.5
Consumption
0 50 100 150
−6
−4
−2
0
2
Investment
0 50 100 150
−3
−2
−1
0
1
Capital Price
0 50 100 150
−4
−3
−2
−1
0
Net Worth
0 50 100 150
−10
0
10
20
30
Bank Capital Ratio
0 50 100 150
−2.5
−2
−1.5
−1
−0.5
0
Capital
0 50 100 150
−8
−6
−4
−2
0
Loans
0 50 100 150
−0.5
0
0.5
1
Bank Equity Return
0 50 100 150
−3
−2
−1
0
1
Capital Return
0 50 100 150
−0.1
0
0.1
0.2
0.3
0.4
ABS Return
0 50 100 150
0
0.1
0.2
0.3
0.4
Loan Interest Rate
0 50 100 150
−2
−1
0
1
2
Bank Monitoring
0 50 100 150
−1
−0.5
0
0.5
1
1.5
Loan Price
0 50 100 150
−10
−5
0
5
ABS
0 50 100 150
−10
−5
0
5
10
Loan Share Sold
Note: IRFs to one percent decrease in total factor productivity (TFP) under fixed
and procyclical capital requirement. Based on averaging one thousand simulation of
a second-order approximation of the model around the steady state. Y-axis denotes
percentage deviation from the steady state i.e. a one percent drop in TFP decreases
output by one percent in the baseline model. X-axis is time in quarters.
136
Appendix 2.1: Model Equations and Competitive Equilib-
rium
Here, we summarize the equations describing the equilibrium of the model. The
competitive equilibrium of our model is a system of endogenous variables ABSt, Bt,
Ct, Ce,t, Dt, et, efpt, It, Kt, t, Lt, bt, xt, Nt, !¯et , !¯bt , PDt, Ps,t, Qk,t, Rt, RF,t, Rm,t,
Rk,t, Rs,t, eRF,t, eRt, Rz,t, TDt, Zt, wt, Yt, GDPt and ⇣e,t, satisfying the relations (2.22)
to (2.54), given exogenous stochastic processes At,bt and the initial conditions Bt1,
Dt1, et1, Ft1, It1, Kt1, t1, bt1, xt1, Nt1, PDt1, Ps,t1, Qk,t1, Rt1, Zt1.
The endogenous variables jointly satisfy the following equations that include first-
order conditions for households, entrepreneurs, commercial banks, shadow banks,
firms, capital goods producers, and the resource constraint.
The goods market equilibrium is given by the aggregate resource constraint. Loan
market equilibrium requires that the amount borrowed by entrepreneurs is equal to
loan supply from banks that is financed by deposits, bank capital or sale of loan
in the secondary market. Market for capital services require that the supply from
entrepreneurs is equal to the capital demanded by firms. Labour market is equili-
brated at the competitive wage when labour supplied by households is equal to firms’
labour demand. Finally, equilibrium in the ABS market requires that ABS issued by
shadow banks - which in turn depends on the loan sale in secondary market - is equal
to the demand for securities from households. A competitive equilibrium consists of
a set of prices and interest rates {Qk,t, Rt, Rm,t, Rk,t, wt} for all possible states and
for all t  0 such that all markets clear when the agents in the model i.e. households,
entrepreneurs, commercial banks, shadow banks, consumer goods producer and cap-
ital goods producer solve their maximization problems while taking these prices and
interest rates as given.
Households:
Households maximize their present value of utility by choosing Dt, ABSt , Zt and
Lt The first order conditions are:
137
Ct = Et(
Ct+1eRt+1 ) (2.22)
Ct = Et(
Ct+1
Rm,t+1
) (2.23)
wt = ⌘Ct(Lt)
' (2.24)
Ct = (1 +⇥zZt).Et(
Ct+1
Rz,t+1
) (2.25)
eRt = Rt1(1 ⌦PDt) (2.26)
Entrepreneurs:
The law of motion of aggregate entrepreneurial net worth is:
Nt+1 = {Rk,t+1Qk,tKt[ eRF,t+1+µe ´ !¯et+10 !et+1f e(!et+1)d!etQk,tKt
Qk,tKt Nt ](Qk,tKtNt)}+W
e
(2.27)
Their aggregate borrowing is:
Bt = Qk,tKt Nt (2.28)
The first order conditions for entrepreneur yields:
Ge0(!¯et+1)) + ⇣e,t(Ge
0
(!¯et+1) µeGe0(!¯et+1)) = 0 (2.29)
(1 Ge(!¯et+1))Rk,t+1 + ⇣e,t[(Ge(!¯et+1) µeGe(!¯et+1))Rk,t+1  eRF,t+1] = 0 (2.30)
138
Where ⇣e,t is the Lagrange multiplier associated with the participation constraint.
Et[(Ge(!¯et+1) µeGe(!¯et+1))]Rk,t+1xt = eRF,t+1(xt  1) (2.31)
The consumption of entrepreneurs that exit the economy is given by:
Ce,t = ⇥
(1 )

(Nt W e) (2.32)
The state contingent interest rate, RF,t is given by:
RF,t = !¯
e
t+1Rk,t+1.(
xt
xt  1) (2.33)
and
Rk,t+1 =
rk,t+1 + (1 )Qk,t+1
Qk,t
(2.34)
The external finance premium on loans between period t and t+1 therefore can
be defined as:
efpt+1 = µe
´ !¯et+1
0 !
e
t+1f
e(!et+1)d!
e
tQk,tKt
Qk,tKt Nt (2.35)
Commercial Banks:
Banks’ default threshold is given by:
!¯bt+1 =
RtDt
et eRF,t+1Bt(1 bt) (2.36)
The first order conditions of commercial banks’ maximization problem are:
Rz,t+1 =
et eRF,t+1(1 Gb(!¯bt+1))
t
(2.37)
139
et eRF,t+1(1 Gb(!¯bt+1)) Gb0(!¯bt+1).et eRF,t+1(!¯bt+1 + Rt(1 bt)(tPs,t) +Rz,t+1t = 0
(2.38)
Rz,t+1(1 bt)Bt  Gb0(!¯bt+1)(1 bt)RtBt =
⌧z
2
(
t  

)0.5.
1

(2.39)
[(1 Gb(!¯bt+1)) + Gb
0
(!¯bt+1).!¯
b
t+1] eRF,t+1 = Gb0(!¯bt+1)Rt[⌧12 (2et + ⌧2)(1 bt) ] (2.40)
The probability of bank default, which also enters the households’ budget con-
straint is given by:
PDt =
ˆ !¯bt+1
0
f b(!bt+1)d!
b
t+1 (2.41)
and the total bank equity is given by:
Zt = tBt(1 bt) (2.42)
Finally, the commercial banks’ balance sheet states:
Bt +
⌧1
2
(e2t + ⌧2et)Bt = Dt + Zt + Ps,tbtBt (2.43)
Shadow banks:
The average return on ABS sold by shadow banks is:
Rs,t+1 =
et eRF,t+1
Ps,t
(2.44)
The supply of ABS based on loan sale in secondary market is:
140
Ps,tBs,t = Ps,tbtBt = ABSt (2.45)
and their profit maximization condition yields:
Rs,t+1 =
"s
"s  1Rm,t+1 (2.46)
Consumption Good Producer:
The production function of consumption good producer is:
Yt = AtK
↵
t1L
1↵
t (2.47)
The firms maximize their profits subject to the production function which gives
the following first order conditions:
rk,t = ↵
Yt
Kt1
(2.48)
wt = (1 ↵)Yt
Lt
(2.49)
Capital Good Producer:
The evolution of capital stock is given by:
Kt+1 = (1 k)Kt + (1 S(Gt))It (2.50)
and the inter-temporal profit maximization with respect to It yields:
Qk,t(1 S(Gt)XtS 0(Gt)) + Et[Dt,t+1Qk,tS 0(Gt+1)I
2
t+1
I2t
] = 1 (2.51)
where Dt,t+1 is the real stochastic discount rate.
Deposit Insurance Agency:
141
The total costs to the deposit insurance agency are given by:
TDt = [!¯
b
t  Gb(!¯bt )) + µbGb(!¯bt )]et1 eRF,t(Qk,t1Kt1 Nt1)(1 bt1) (2.52)
Resource Constraint:
Yt = Ct + It + Ce,t + µeG
e(!¯et )Rk,tQk,t1Kt1 + ⌦PDtRt1Dt1 (2.53)
+µeG
b(!¯bt )et1 eRF,t(Qk,t1Kt1 Nt1) + (1 et1) eRF,tBt1(1 Gb(!¯bt ))
Finally, we define GDP as net of bankruptcy costs due to default.
GDPt = Ct + It + Ce,t (2.54)
142
Appendix 2.2: Model Variables and Parameters
Variable Notation
Household Consumption C
labour L
Deposits D
Asset backed securities/shadow bank deposits ABS
Bank shares Z
Real wage rate w
Riskless return R
Commercial bank deposit return eR
Shadow bank deposit return Rm
Shadoe banks’ return on asset backed securities Rs
Bank share return Rz
Shadow bank profit diverted to households ⇧I
Entrepreneurs’ profit diverted to households Lumpe
Lump sum tax on households TD
Fraction of deposits that default PD
Entrepreneur j’s net worth Nj
Entrepreneur j’s purchase of capital Kj
Entrepreneur j’s bank loan Bj
Aggregate Net worth N
Aggregate capital K
Aggregate loans B
Capital price Qk
Return on capital Rk
Marginal product of capital rk
143
Entrepreneurs’ idiosyncratic shock !e
Variance of entrepreneurial idiosyncratic shock 2e
Threshold value of idiosyncratic shock (for entrepreneurs survival) !¯e
Contractual gross return on bank loan RF
Entrepreneurial leverage x
Bank’s share of gross return on entrepreneurial investment e(.)
Share of return from defaulted loans Ge(.)
Startup transfer of net worth to new entrepreneurs W e
External finance premium efp
Entrepreneurs’ consumption Ce
Off balance sheet loan Bs
Share of loans sold in secondary market b
Bank manager’s monitoring effort e
Convex monitoring cost c(e)
Bank return on a well diversified portfolio fRF
Banks’ idiosyncratic shock !b
Variance of banks’ idiosyncratic shock 2b
Threshold value of idiosyncratic shock (for bank survival) !¯b
Bank capital ratio 
Loan price in secondary market Ps
Bank’s profit ⇡z
Depositors share of gross return on loans b(.)
Share of bank assets of defaulted banks Gb(.)
Loan price paid by zth shadow bank in secondary market Ps(z)
Loan sold to zth shadow bank in secondary market BS(z)
Return on household funding to shadow banks Rm
144
Output Y
Total factor productivity A
Investment I
Convex investment costs S(.)
Government lending; credit policy Bg
Private lending; credit policy Bp
Government intermediation costs; credit policy G
Government transfers; credit policy T
Minimum capital requirement; Basel II Cap
Parameter Notation
Household discount factor 
Weight attached to leisure in the utility function ⌘
Inverse labour supply elasticity '
Adjustment cost parameter, bank equity transactions ⇥z
Depositor cost of bank default ⌦
Survival probability of firms 
Entrepreneur bankruptcy cost µe
Proportion of net worth consumed by exiting entrepreneurs ⇥
Monitoring cost parameter; bank effort ⌧1
Share of capital in production ↵
Monitoring cost parameter; bank effort ⌧2
Banks’ bankruptcy cost µb
Capital requirement ¯
Coefficient on bank capital buffer ⌧z
Elasticity of funds in secondary loan sale market "s
145
Investment adjustment cost parameter 'x
Capital depreciation rate 
Technology shock; persistence ⇢a
Risk shock; persistence ⇢
Technology shock; std. deviation a
Risk shock; std. deviation r
146
3 Can Low interest rates be harmful: An assessment
of bank risk taking channel in Asia-Pacific
Events surrounding the global financial crisis have brought to light the potential
role of monetary policy in precipitating the crisis. Numerous studies on advanced
economies have documented a significant negative relationship between interest rates
and bank risk-taking. This paper also finds the presence of the risk-taking chan-
nel based on a panel of publicly listed bank data in Asia. Using both annual and
quarterly data, “too low” interest rates are found to lead to an increase in bank
risk-taking.31
3.1 Introduction
Of the many causes of the global financial crisis, one that strikes a particular discord
among central bankers is the notion that an accommodative monetary policy can be
harmful to the economy. For many, it is unsettling to realize that the role of monetary
policy, which has long been considered a stabilizer or stimulant of short-term business
cycles—the solution—could in fact be the problem. Still, the literature that impli-
cates the role of monetary policy is not completely new or unknown. Fisher (1933),
Hayek (1939), and Kindleberger (1978) are some studies that highlight easy monetary
conditions as classical ingredients in boom-bust business fluctuations. Hayek (1939)
of the Austrian business cycle theory fame espouses that excessive credit expansion
accommodated by lax monetary policy precedes and lays the groundwork for reces-
sions. Meanwhile, Minsky’s (1992) financial instability hypothesis posits that during
periods of high economic growth, the income-debt arrangements of economic units
shifts from hedge finance (completely hedged debt) toward speculative and Ponzi
finance (highly unstable leverage and higher risk). This tendency is heightened by
protracted low interest rates when a growing number of investors seek higher yields
31This chapter was started during my short stint at the Asian Development Bank (ADB) under
the guidance of Dr. Hsiao Chink Tang and Dr. Arief Ramayandi. However, all analytical work,
estimations and analysis of results was done by me. I continued to work on the chapter even after
I left ADB. Dr. Hsiao Chink Tang provided useful comments and suggestions.
147
despite increasing risk. Thus, loose monetary and credit policy play an important
role in the evolution from financial stability to crisis.
This paper focuses exclusively on the risk-taking behavior of financial interme-
diaries in response to loose monetary conditions. A prolonged period of relatively
low interest rates can induce financial imbalances by reducing risk aversion of banks
and other investors. In the context of the global financial crisis, it is said to have
contributed to the credit boom, asset price increases, a decline in risk spreads, and a
search-for-yield that eventually saw the bust in the US subprime and housing mar-
kets, the collapse of major financial institutions, and ultimately, the Great Recession
(Hume and Sentance 2009; Taylor 2009; Diamond and Rajan 2009). This, until re-
cently, missing link in the monetary policy transmission mechanism—known as the
risk-taking channel (Borio and Zhu 2012; Rajan 2006)—relates to how changes in in-
terest rates affect either risk perceptions or risk-tolerance of financial intermediaries.
As a relatively recent issue of monetary transmission mechanism, risk-taking
channel does not have a specific definition, but indeed, it is a common term used
for various mechanisms at work, which are all mutually inclusive. While this new
monetary policy channel has its gray areas at the time being, it deserves close explo-
ration for a fuller understanding the link between the monetary policy and financial
stability and to draw clearcut policy conclusions. The findings regarding the risk-
taking channel have potentially important implications for the conduct and design
of monetary policy, as a better understanding of risk taking channel may provide an
insight for monetary authorities to adjust their policies in order to mitigate the ad-
verse consequences of their polices on bank risk-taking and in turn, avoid the buildup
of risks in the financial system. If policymakers understand banks’ risk-taking in-
centives and focus on the potential impact of their polices on bank risk, they may
find answers to when and how to be more cautious and what factors they should
take into account in their policy design. Furthermore, understanding the risk-taking
channel would provide us comprehension regarding the macroeconomic implications
of bank supervision and regulation as well.
Empirical literature on the risk-taking channel has thus far focused mostly on the
148
US and the euro area. This study therefore fills an important void in examining the
impact of low interest rates in 10 Asian economies (the People’s Republic of China
(PRC); Hong Kong, China; India; Indonesia; the Republic of Korea; Malaysia; the
Philippines; Singapore; Taipei,China; and Thailand) from 2000 to 2011. Since a
meaningful study of bank risk taking would require sufficiently large and relatively
well developed banking sector, we restrict our focus to developed and emerging mar-
ket economies in Asia. The study also sheds light on the bank specific characteristics
which may have an impact on bank risk. Furthermore, in our computation of bank
risk instead of relying on one particular risk measure as done by most studies on the
risk-taking channel, we employ alternative risk measures to cover different risk-taking
behavior. In particular, both market based and accounting based indicators of bank-
risk are considered. We control for a number of factors that may have an impact on
banks’ risk such as macroeconomic performance, financial deepening, stock market
returns, regulatory environment as well as bank specific characteristics, namely size,
capitalization and profitability of banks. Based on both quarterly and annual data,
we find the presence of the risk-taking channel in Asia. In addition, we find more
nuanced results using the quarterly data. In particular, we find contemporaneous
impact of low interest rates reduces bank riskiness mostly likely in lowering the de-
fault probability of existing loans. Yet over time, banks become more aggressive and
tend to lend less to creditworthy borrowers. As a result, the overall risk level of
banks rises.
There are some important caveats in our study that should be be asserted before
going into details of the analysis. First, we do not make any inferences on the
optimality of risk choices of banks, as from a theoretical viewpoint, it may be optimal
for a bank to engage in riskier projects when interest rates are low and further,
it may also be the socially optimal outcome of monetary policy during recession
periods as well. Thus, the results from the study should be interpreted as changes
in the risk appetite/tolerance of banks in response to interest rates being lower than
some benchmark level. Second, and more important there is a part of literature
suggesting that risk-taking principally refers to new risk i.e. new loans. In effect,
the focus then is on the incentives of banks to undertake risky projects ex-ante.
149
It is therefore important to distinguish between realized and new risk to draw an
accurate inference regarding the relationships between monetary policy and bank
risk taking. Such a comprehensive analysis requires the use of detailed data on
individual bank loans from credit registers, which provides information on lending
standards, loan performance etc, which is missing for the case of Asia. Furthermore,
data on individual loan borrower characteristics is confidential in most cases and
available for very few countries that maintain a credit register. This is reflected in
the handful of studies that make use of such detailed data to study the relationship
between monetary policy and bank risk-taking (Jimenez et al. (2014); Ioannidou et
al. (2015) and Lopez et al. (2011)). Finally, the paper does not explicitly take into
account the issue of implicit government support. There has been significant debate
about the role of implicit guarantees in exacerbating moral hazard and undue risk-
taking. The effect of implicit government support would be more pervasive under
an extended period of lower-than-usual interest rates. Capturing this effect would
require detailed data to proxy the expectation of external government support, which
to our knowledge is not available for Asia.
The rest of this paper is organized as follows. Section two reviews the theoretical
and empirical literature on the bank risk-taking channel. Section three discusses the
model specification of the bank risk-taking channel with a focus to delineate it from
other channels of monetary policy transmission mechanism. Section four delves on
data and estimation issues. Results are presented in section five, while section six
concludes.
3.2 Literature: Theory and Evidence
The relatively new concept of the risk-taking channel of monetary policy transmission
mechanism is reflected in the scarcity of its theoretical contributions. Some elements
of the theory of risk-taking channel can be traced in the theoretical propositions of
some previous work such as Keeley (1990); Allen and Gale (2000); Dell’ Ariccia and
Marquez (2006); and Rajan (2006). Although some of the mechanisms have been
150
discussed previously, the term ’risk-taking channel’ of monetary policy was coined
by Borio and Zhu (2012) in which they point to the potential of low interest rates to
increase bank risk-taking.
More specifically, low interest rates can lead banks to take on more risk in sev-
eral ways. First, low interest rates boost prices and collateral values of assets on
banks’ balance sheets, which in turn modify banks’ estimates of probabilities of de-
fault, losses in the case of default, and overall volatility of bank returns32. Measured
volatility tends to decrease in rising markets33, which relaxes banks’ budgetary con-
straints and leads to even more risk-taking (Borio and Zhu 2012). According to
Adrian and Shin (2009), low short-term rates may lead to more risk-taking because
they improve banks’ profitability and relax their budgetary constraints.
Second, low interest rates may influence banks to search-for-yield (Rajan 2006).
In a low interest rate environment, the incentives of asset managers to engage in more
risky projects rise for a number of reasons. Primarily, this mechanism predominantly
works through the relationship between the low levels of short-term interest rates
and sticky target rate of returns. Financial institutions often enter into long-term
contracts committing them to produce high nominal rates of return. In a period
of low interest rates, these contractual rates may exceed the yields available on
safe assets. To earn higher returns, banks therefore will have to search for higher
yields from the more risky assets. Moreover, a similar mechanism could be in place
whenever managerial compensation is linked to absolute yields. In a low interest rate
environment, lower yields on safe assets imply a lower compensation for managers
that choose to invest in safe assets, giving managers higher incentives to invest in
more risky assets. In all cases, the effect of the channel becomes stronger as the
resulting gap between the market and target rates becomes larger.
32This is in close spirit to the familiar financial accelerator mechanism, which argues that increases
in collateral values reduce borrowing constraints (Bernanke et al. 1996). Adrian and Shin (2009)
claim that the risk-taking channel is distinct but complementary to the financial accelerator because
it focuses on amplification mechanisms due to financing frictions in the lending sector
33The literature suggests that short-term volatility is directional, that is, it tends to fall in rising
markets and rise in falling markets. For instance, for equity returns, see Schwert (1989) and, for
bonds, Borio and McCauley (1996).
151
Third way through which the risk taking channel operates is through the effect
of communication policies and the reaction function of the central bank (Borio and
Zhu (2012). In this process, capability of central banks for managing short term
inflation expectations and inflation rate is especially important. In this context,
transparency and predictability of monetary policies decrease both the uncertainty
related to the variation in inflation and short and long term interest rates and accord-
ingly in financial market prices. Fourth, banks’ perception that the central bank will
ease monetary policy in bad economic times can also lower the expectations of large
downside risks—the moral hazard problem. This implies that changes in interest
rates can have an asymmetric impact on risk-taking, that is, reductions encourage
risk-taking by more than the equivalent increases in discouraging risk-taking. This
expectation became more important in the recent period. When banks expose to a
shock threatened the stability of the financial system, and they expect from central
bank to lower the interest rate aggressively, they will incline to undertake higher risk.
Accordingly, essential reason of the formation of moral hazard is the commitment
stating interest rates will be implicitly kept low instead of having low interest rates
(De Nicolo et al, 2010). Fifth, Dell’Ariccia and Marquez (2006) point out that low
interest rates reduce adverse selection in credit markets, which decrease banks’ in-
centives for screening loan application. Finally, Akerlof and Shiller (2009), in turn,
suggest that, due to money illusion in periods of low interest rates, investors take
higher risks to increase returns.
All of the above mentioned mechanisms are the candidate driving forces behind
the risk-taking channel. Although being diverse, they may tend to work at the same
time as well. Furthermore, it should be noted that none of these proposed explana-
tions is more important than the other, as there is no conclusive evidence regarding
the relative importance of them. In part, this is due to the lack of theoretical models,
which reveal the details of either potential mechanism and allows the precise under-
standing of their characteristics. The risk-taking channel is a relatively recent area
of monetary economics; hence the theoretical literature is still being developed and
is rather limited for the time being. There are only a handful studies that present
formal models where several mechanisms of the risk-taking channel act together. We
152
summarize some of these theoretical studies here. However, since our main focus is
on the empirical analysis of risk-taking channel we list the empirical contributions
in greater detail. Dubecq et al. (2015) provide a model with risk-shifting where
the level of interest rates affects the risk perception of some investors and risk expo-
sure by others. They argue that situation of uncertainty with respect to regulatory
constraints may cause market participants to form wrong inferences on risks. Dell’
Ariccia et al. (2010) use a static model to assess the impact of prolonged easy mon-
etary policy on bank risk-taking. In their model, banks’ risk appetite increase in
prolonged periods of lax monetary conditions, however the net effect of monetary
policy depends on the balance of the interest rate pass-through, risk shifting and
capital structure. Giavazzi and Giovannini (2010) model the interactions between
monetary policy and liquidity transformation. They claim that optimal monetary
policy consists of a modified Taylor rule, which takes into account the possibility of
liquidity crises and thus adjusts to changes in risk-taking. Diamond and Rajan (2012)
present a model where the fiscal authority is able to affect real interest rates, arguing
that ex ante regulation may not be effective if the extent of ex ante promises are
hard to observe. They suggest raising real interest rates in normal times to prevent
banks from making excessive liquidity promises. Cao and Illing (2015) model how
financial intermediaries’ incentives for liquidity transformation are affected by the
central bank’s reaction to financial crisis. They demonstrate that interest rate pol-
icy as financial stabilizer is dynamically inconsistent, and imposing ex ante liquidity
requirements can increase efficiency.
Although the risk-taking channel of monetary transmission is still not well-understood,
an increasing number of empirical studies have been produced to analyze whether
there is a relationship between low interest rates and bank risk-taking and attempt
to clarify characteristics of the risk-taking channel. Most single or multiple-country
empirical studies have tended to find the presence of risk-taking channel. There are
two broad types of studies: those using macro data that try to capture the link
between monetary policy and risk behavior, and others using micro data to provide
micro-level panel evidence for the impact of interest rate changes on individual bank’
153
risk-taking behavior34. The list of papers that use micro data to study bank behav-
ior has been increasing rapidly in the recent past. Our work belongs to this line of
research and the remaining part of this section summarizes these studies. The first
empirical investigation on bank risk taking behavior in response to monetary policy
conditions belongs to Jimenez et al. (2014) who examined the case of Spanish banks
before the introduction of the euro. Using a variety of duration models and time to
default as a measure of risk, they find that low interest rates affect the risk of bank
loan portfolio in two opposing ways. In the short-term, low interest rates reduce the
probability of default of the outstanding loans; in the medium term, however, banks
act more aggressively, that is, they lend to borrowers with a worse credit history and
grant more loans with a higher probability of default. In the same vein, Gaggl and
Valderrama (2010) look at Austrian banks and find that the expected default rates
of banks’ business loans increased during the period of low refinancing rates from
2003 to 2005.
On the other hand, Ioannidou et al. (2015) take a slightly different approach by
examining the quantity and quality of loans, as well as the price of loans. Based
on data from 1999 to 2003, they find that when the federal funds rate was reduced,
Bolivian banks not only increased the number of risky loans, they also reduced the
rates they charged to risky borrowers relative to the rates for less risky ones35. Using
a similar methodology, in addition to the loan and bank specific characteristics as
per Ioannidou et al. (2015), Paligorova and Santos (2017) also control for borrower
specific characteristics and find over the period of 1990 to 2010, when the US federal
funds rates were low, banks demanded relatively lower spreads on their loans to
riskier borrowers than to safer ones.
Tabak et al. (2010) uses individual bank- level data for commercial banks oper-
ating in Brazil over the period from 2003 to 2009 in order to analyze the risk-taking
34Studies using macro data include, for instance, Angeloni et al. (2015); Eickmeier and Hoffman
(2013) and Bekaert et al. (2013). These papers employ vector autoregression (VAR) to provide
time series evidence on the risk taking channel. Another group of studies utilize both macro and
micro level data in their analysis, for example De Graeve et al. (2008); De Nicolo et al. (2010).
35Bolivia is a highly dollarized economy with the exchange rate pegged to the US dollar. Hence,
the US federal funds rate was used in their study.
154
channel of monetary policy transmission. Their results indicate that lower interest
rates lead to an increase in banks’ credit risk exposure, supporting the existence of
the risk-taking channel. Furthermore, liquidity and bank size are found to have a
positive relation with risk. Lopez et al. (2011) employs a dataset from the Credit
Register from Colombia over the period 2000-2008. This paper finds a statistically
significant link between interest rates and banks’ risk taking. Lower interest rates in-
crease the probability of default on new loans and reduce that on outstanding loans.
They find that small and highly leveraged banks are more willing to take risks. Using
53 quarterly bank level data for Turkey over the period 2002-2012, Özsuca and Ak-
bostancı (2016) find evidence that low levels of interest rates have a positive impact
on banks’ risk-taking behavior. Specifically, low short term interest rates reduce the
risk of outstanding loans; however short term interest rates below a benchmark level
increase risk-taking of banks. Delis et al. (2011) examine the impact of US monetary
policy on bank risk-taking by using two alternative micro datasets: quarterly bal-
ance sheet data from Call Reports and data on new loans from the syndicated loan
market. They present empirical evidence that low interest rates tend to decrease
loan portfolio risk of a bank in the short-run, but increase it in the long-run.
In terms of multi-country studies, Maddaloni and Peydro (2011) use credit stan-
dards surveys and find periods of low short-term rates are associated with softening of
lending standards by banks in the euro area and the US. Altunbas et al. (2014) also
find low interest rates over protracted periods lead to an increase in bank risk as cap-
tured by banks’ expected default frequencies. They use publicly available quarterly
information of 600 banks from 1999 to 2008, of which about two-thirds are US listed
entities, the rest, European. Using a database of listed banks from the European
Union and United States developed by Altunbaş et al. (2014), Gambacorta (2009)
states that when interest rates are low for an extended period banks’ EDFs tend to
increase. The results confirm the existence of a risk-taking channel. All other things
being equal, liquid and well capitalized banks are found to be less risky. This result
differs from Jimenez et al. (2014) and Ioannidou et al. (2015). Delis and Kouretas
(2011) use a similar panel generalized method of moment (GMM) methodology but
with annual euro bank data from 2001 to 2008 and find substantive evidence of bank
155
risk-taking.
In general, there is enough international evidence to suggest the presence of an
operational risk-taking channel. Notably, most of the existing empirical literature on
the risk-taking channel provide evidence for the U.S and Euro area, whereas very few
studies have investigated the role of monetary policy in emerging market economies.
Specifically, these include Ioannidou et al. (2015) for Bolivia, Tabak et al. (2010)
for Brazil, Lopez et al. (2011) for Colombia and Özsuca and Akbostancı (2016) for
Turkey and all of them present empirical evidence on the existence of such a channel.
So far there has not been any study examining the risk-taking channel in Asia, a gap
we intend to fill. Our paper is closest to Altunbas et al. (2014), but with the major
exception of being Asian focused. In addition to using quarterly data as per Altunbas
et al. (2014), which helps provide a more nuanced analysis of the dynamics of bank
risk-taking, we also use annual data as they present a longer time-series and larger
cross-sectional dimension. Unlike Altunbas et al. (2014), however, we explicitly
derived our bank risk dependent variable, whereas their main bank risk indicator is
obtained from a proprietary data provider.
For many emerging economies, it is instructive to note that low interest rates
abroad especially in advanced economies can also translate to lower than appropriate
interest rates in the domestic economy (Bank for International Settlements (BIS)
2012; Taylor 2012). The centrality of the US dollar in the world trading system
means that many emerging economies tend to peg their currencies to the US dollar
(McKinnon 2013; McKinnon and Schnabl 2004). With capital mobility, this implies
that their monetary policies will typically track the monetary policy of the US.
A particular problem arises when the emerging economies’ business cycles do not
synchronize with the US business cycle. BIS (2012) highlights a specific scenario
during the global financial crisis, when despite the decline in world interest rates,
persistent interest rate differentials remained between the relatively higher interest
rates in the emerging economies and that of the advanced economies. The search-for-
yield in this case led to a surge in capital flows into the emerging economies buoying
domestic economic and financial conditions. The appropriate stabilization strategy
would be to tighten policy (let interest rates rise), yet this was seldom done in order
156
to discourage further capital inflows and hold the exchange rate relatively steady.
3.3 Model Specification
This paper follows the modeling spirits of Altunbas et al. (2014) and Delis and
Kouretas (2011), which is based on the literature of the determinants of bank risk
and monetary policy transmission mechanism. In essence, besides the interest rate
variables of focus, the model also includes a set of bank specific controls, and a set of
structural, macro, and regulatory controls pertinent to a particular country. These
controls are aimed to better delineate the impact of bank risk-taking channel from
other monetary policy transmission channels.
The model with annual data is:
bri,j,t = ↵bri,j,t1+irj,t+igj,t+grj,t+✓ycj,t+⇢cgj,t+⌧erj,t+smj,t+1szi,j,t+
2cpi,j,t1 + 3lqi,j,t1 + 4pri,j,t1 + µ1spj,t + µ2mdj,t + µ3csj,t + ⇣cei,t + "i,t
On the other hand, the quarterly model is:
bri,j,t = ↵bri,j,t1+
P1
k=0 kirj,tk+
P1
k=0 kigj,tk+
P1
k=0 kgrj,tk+
P1
k=0 ✓kycj,tk+P1
k=0 ⇢kcgj,tk+
P1
k=0 ⌧kerj,tk+
P1
k=0 ksmj,tk+1szi,j,t+2cpi,j,t1+3lqi,j,t1+
µ1spj,t + µ2mdj,t + µ3csj,t + ⇣cei,t + "i,t
bri,j,t is a measure of bank risk of bank i in country j at time t. For the annual
model, we use four measures of bank risk: (1) the idiosyncratic risk of a simple
capital asset pricing model (CAPM), br1; (2) the ratio of non-performing loans to
total loans, br2; (3) the z-score, br3 ; and (4) the standard deviation of the return
on assets, br4. For the quarterly model, due to data constraints, we only use the
idiosyncratic risk36.
irj,t, is the change in short-term interest rates that proxy for the change in the
monetary policy stance. igj,t is the interest rate gap that measures the difference
between the policy rate and a benchmark rate. It is the variable of main interest
36Derivations of these measures will be discussed later. See also the appendix 3.1 and Table 9 for
summary statistics of variables used.
157
that captures the bank risk-taking channel, specifically the phenomenon of search-for-
yields as discussed in Section 3.2. Since a drop in interest may not necessarily imply
excessive low rates, a benchmark would provide a measure for how low is actually
low. If the bank risk-taking channel is operational, we expect  to be negative,
meaning higher interest rates (above some benchmark) reduce bank risk-taking or,
conversely, lower interest rates increase bank risk-taking. We follow Altunbas et al.
(2014) in using this indicator as a measure of the bank risk-taking channel37. In
addition, as per Altunbas et al. (2014), we use both the change in interest rates and
the interest rate gap in the estimations. The latter is a sharper measure of bank
risk-taking channel as it better captures the phenomenon of search-for-yield, while
the former is a coarse measure that also incorporates the financial accelerator effect
or the balance sheet channel. Specifically, any change in interest rates will affect
borrower’s collateral value (the balance sheet effect), and in turn the riskiness of
bank loans portfolio38.
The Taylor benchmark rates are consistently higher than the natural rates (or
more accurately the trend rates) in all economies. The interest rates in most of
the economies, although higher than the rates in developed countries, are still low
against the background of their trend output growth rates over the past several
years. This reinforces BIS (2012) which argues that policy rates in emerging market
economies, particularly in Asia, appeared unusually accommodative by the end of
2011. Following the global financial crisis of 2007, there was a downward movement
in the interest rates in line with the near zero-interest rate policy in the developed
world with some signs of reversal post 2010. Output and inflation also followed a
37Delis and Kouretas (2011) use levels of interest rates in their main estimations, and changes
in interest rates as a robustness check. They acknowledge that the changes in interest rates are a
measure of the bank risk-taking channel.
38It is important to distinguish the risk-taking channel from more widely understood channels
relating to credit market, namely, the financial accelerator or the balance sheet and bank lending
channels, which together constitute the broad credit channel. The risk-taking channel goes beyond
the change in the net worth of both lenders and borrowers (which is incorporated in the broad
credit channel). It focuses on “the impact of changes in policy rates on either risk perceptions or
risk-tolerance and hence on the degree of risk in the portfolios, on the pricing of assets, and on the
price and non-price terms of the extension of funding” (Borio and Zhu 2012). In other words, it
accounts for the amount of uncertainty a lender is willing to hold in its portfolio.
158
downward trend, most notably between 2008 and 2009, followed by a quick recovery
thereafter. This is also reflected in the rising Taylor benchmark rates post 2009.
grj,t is the real GDP growth rate of country j. ycj,t measures the slope of the
yield curve of country j. Both these variables are included to capture the effects of
the expectations channel of the transmission mechanism. As a priori, the coefficient
of gr is expected to be negative—better economic conditions raise the profitability
of a larger number of projects in accordance with the expected net present value,
which reduces the overall credit risk to banks (Kashyap et al. 1993, Altunbas et al.
2014). Similarly, the coefficient of yc should be negative—the steeper the slope of
the yield curve, the higher the bank profits (Viale et al. 2009; Entrop et al. 2012)
and the lower the bank risk.
cgj,t is bank credit to GDP of country j (Delis and Kouretas 2011; Mannasoo
and Mayes 2009). Previous studies have found mixed signs. For example, Gonzalez-
Hermosillo et al. (1997), Mannasoo and Mayes (2009), and Delis and Kouretas (2011)
find that financial deepening increases the vulnerability of the banking sector—a
positive relationship. This lends support to evidence from the global financial crisis
particularly in the developed economies where excessive finance was found to have
harmed the economy (Taylor 2009; Diamond and Rajan 2009; Turner 2012). On the
other hand, other studies consider financial deepening as a sign of economic maturity
and therefore expect a negative coefficient (Hutchison and Mc-Dill 1999).
erj,t measures exchange rate volatility of country j to account for the exchange
rate channel. As the countries studied in this paper are mostly small and open
economies, they tend to be susceptible to volatile capital flows. Hence, the more
volatile the economic environment, the more volatile is the exchange rate and the
higher is the bank risk.
smj,t measures changes in the broad stock market index of country j that captures
the wealth channel. We expect the coefficient to be negative. An increase in the asset
prices would increase the collateral value, thereby reducing bank risk.
szi,j,t size; , capitalization; cpi,j,t1 and pri,j,t1, profitability; are the bank specific
variables of bank i in country j included to control for the bank lending channel effects.
159
The bank lending channel focuses on the financial frictions derived from the balance
sheet situation of banks39. Based on empirical evidence, the signs on these variables
are expected to be negative as bigger, more capitalized, highly liquid and more
profitable banks would be as less risky. Bank specific characteristics, except size,
which is considered as a predetermined variable, are lagged to deal with endogeneity
(Delis and Kouretas 2011).
csj,t, capital stringency index; spj,t, supervisory power index; and mdj,t, market
discipline index; represent the regulatory indexes of country j (Barth et al. 2013)40.
They are included as per Delis and Kouretas (2011) to avoid the problem of omitted
variable bias, that is, to account for the bank regulatory environment. The higher the
value of each index, the greater the capital stringency, the stronger the supervisory
power, and the greater the market disclosure faced by banks. As such, we expect
these indices to have a negative impact on bank risk.
cei,t is individual country effects to account for unobserved cross-country differ-
ences. Time dummies are also included which vary according to the frequency of the
data. This is to account for technological change that occurs over time, which may
influence bank risk-taking behavior.
Both equations (1) and (2) are a dynamic model that includes the lagged bank-risk
variable as an explanatory variable. This can be justified on several grounds (Delis
and Kouretas, 2011). First, the persistence may reflect the existence of intense
competition, which tends to alleviate bank risk-taking. Second, the importance
and prevalence of relationship-banking with customers means bank risk is likely to
persist. Third, given that bank risk is likely to follow business cycles, we may find a
persistent bank risk behavior in line with the dynamics of business cycle. Hence, if
39The bank lending channel contends that a monetary policy tightening can affect the supply
of bank loans due to a decline in their total reservable deposits. This can be offset by raising
non-reservable funding. However, the cost of non-reservable funding would be higher for smaller
and low-capitalized banks if the market perceives them as riskier. Similarly, illiquid banks will
be more adversely affected by a monetary tightening due to a lower possibility of being able to
liquidate their assets. See Bernanke and Blinder (1988), Kashyap and Stein (1995), Kishan and
Opiela (2000), among others, for a good analysis of the specific effects of the variables on the bank
lending channel.
40See the appendix for detailed definition of each index.
160
risk is persistent, a static model would be biased, and a dynamic model is preferred41.
The main difference between the annual model and the quarter model is the
inclusion of persistence or one-period lagged macroeconomic variables in the latter
(Altunbas et al. 2014). In addition, the quarterly model excludes bank profitability
because of the unavailability of quarterly profitability data. Treatment of stock
variables (bank specific characteristics and regulatory indices) remains unchanged in
the two specifications.
3.4 Data Sample and Estimation Issues
Existing work on risk-taking channel of monetary policy has focused exclusively on
the U.S and Europe. This paper in contrast aims to fill this gap in the empirical
literature by studying the bank risk pattern in Asia. Given the linkages of emerg-
ing market economies with the advanced economies, they continue to grapple with
spillovers of monetary policy from advanced to emerging economies. External fac-
tors, prominently concerns about exchange rate and capital flow volatility, therefore
play a significant role in these economies’ monetary policy conduct. Monetary pol-
icy may be eased in the absence of overheating in order to reduce the incentive for
further inflows. But if capital flows result in inflationary pressures and credit booms,
then the policy interest rate may need to be tightened to ensure stability of the
domestic economy. This is however seldom done due to the fear of causing further
capital inflows. According to the BIS (2012) monetary policy stance in emerging
market economies, particularly Asia, as measured by real (inflation-adjusted) policy
rates and Taylor rule rates has been very accommodative. The risk taking channel
of monetary policy posits that this overly accommodative monetary policy stance
can translate into bank risk taking. Since a meaningful study of interaction between
bank risk behavior and monetary policy would require the banking sector to be both
41The coefficient of the lagged bank risk may be viewed as the speed of convergence to equilibrium.
A statistically significant value of zero implies that bank risk is characterized by a high speed of
adjustment; while a value of one means that the adjustment is very slow. Values between 0 and 1
suggest that risk persists, but will eventually return to its mean.
161
sizable and relatively well developed, in this study we focus on a group of emerging
market economies in Asia. This restricts our sample to ten economies, namely, the
PRC; Hong Kong, China; India; Indonesia; the Republic of Korea; Malaysia; the
Philippines; Singapore; Taipei,China; and Thailand. The panel is unbalanced due
to different data availability for the chosen countries. The largest panel for annual
data runs from 2000 to 2011. For quarterly data, the longest period is from 2003Q1
to 2011Q4. There are no quarterly data available for Hong Kong, China and India,
therefore they are not included in the quarterly model.
The choice of measures accounting for banks’ risk is of particular importance for
our empirical analysis. Measuring risk is a complicated issue and there is no specific
proxy for bank risk-taking. A priori, bank risk refers to the amount of uncertainty a
lender is willing to hold in his portfolio. Previous literature have used accounting or
market based measures to quantify the extent of bank risk tolerance. In the absence of
a well established bank risk proxy, this paper uses four measures of bank risk, namely
the idiosyncratic bank risk, z-score, ratio of non-performing loans to total loans and
the standard deviation of return on assets42. These indicators are considered to reveal
different type of risk related information and reflect diverse aspects of risk-taking,
hence each has its own advantages and disadvantages as measures of bank-risk taking.
In other words, neither of them is more accurate or superior to another, but rather
complementary to each other in capturing the main dimension of bank risk.
The first bank risk measure used in this paper is the idiosyncratic bank risk,
br1i. This indicator is derived from stock market information. As the variable name
suggests, this indicator measures the risk specific to a bank after accounting for the
broad market effects by decomposing risk into idiosyncratic (individual) and systemic
(market wide) elements. Thus it allows us to ascertain whether monetary policy
influences each individual bank’s risk position on top of systemic considerations.
The computation of this measure is based on a simple Capital Asset Pricing Model
(CAPM). The CAPM model is based on the following equation: Ri,j,t = i,tRm,j,t +
"i,t where Ri,j,t are the daily stock market logarithmic abnormal returns from each
42See appendix 3.1 for detailed derivations of the bank risk variables and sources of data.
162
bank i in country j. Rm,j,t are the daily stock market abnormal returns from the
broad stock market index m from country j. The term "i,t is the bank specific
residual. Specific beta coefficients, i,t, are calculated for each bank in the sample by
running separate regressions on daily data for every year (annual model) and quarter
(quarterly model). The idiosyncratic bank risk, br1i is therefore given by:
br1i =
mX
t=1
"2i,t
m
where m is the number of daily trading days in a year (for the annual model) or
quarter (for the quarterly model). Daily individual bank return index and daily stock
market return index for each country used in the CAPM estimation are obtained from
Datastream.
The other three indicators are accounting based constructed using data obtained
from Bankscope. The second indicator is the ratio of non-performing loans to total
loans, br2i, which is taken as a proxy of credit risk in a bank’s portfolio. Since some
proportion of non-performing loans will result in losses for a bank, a higher ratio
indicates a higher level of bank risk. Unlike the other measures for bank risk such
as z-score or standard deviation of return on assets, which reflect insolvency risk,
this measure directly refers to credit risk. However, it is instructive to note that this
measure of bank risk is backward looking.
The third indicator which is also based on accounting information from banks’
balance sheet is the z-score, which is a universal measure of individual bank fragility.
The index measures bank soundness in terms of insolvency risk or distance to default
and is often used as a measure of financial soundness or risk-taking (De Nicolo et
al. (2003), Demirgüç-Kunt et al. (2006); Maechler et al. 2007, Angkinand and
Wihlborg (2008), Berger et al. (2009), Tabak et al. (2010), Delis et al. (2011)).
Z-index combines in a single measure the profitability, leverage and return volatility.
It is given by the ratio:
br3i,t =
rai,t + eki,t
(rai,t)
163
where rai,t is the return on asset and eki,t the ratio of equity capital to total assets
for bank i and (rai,t) is the standard deviation of return on assets of bank i. In
essence, the z-score measures the lower bound in which expected returns need to drop
in order to exhaust the bank’s equity. In other words, it represents the probability
of a negative shock to profits that forces bank to default (Yeyati and Micco, 2003).
While z-index increases with higher profitability and capitalization levels, it decreases
with unstable earnings captured by the standard deviation of return on assets. So
a higher level of z-score corresponds to a greater distance to equity depletion and
therefore greater bank stability.
The data to compute z-score index, namely, the return on assets and equity to
total asset ratio is available from Bankscope. We use a three-year rolling time window
for calculating standard deviation of returns on assets (rai,t), following Cihak et al.
(2009). Furthermore, given that z-score is highly skewed, we use natural logarithm
of z-index, which is normally distributed, following Leaven and Levine (2009) and
Özsuca and Akbostancı (2016).
Finally, standard deviation of asset returns, br4i, which measures the volatility
of return on bank assets is used as the fourth proxy for banks’ risk exposure. Again,
we use a three-year rolling time window to compute the standard deviation.
In terms of interest rate gaps, we use two measures as per Altunbas et al. (2014)
in calculating the deviation of policy rate from some benchmark levels43. The first
known as the Taylor gap (ig1) is the difference between the actual nominal interest
rates and that generated from a Taylor-type equation44. For data comprehensive-
ness, we use the 3-month money market or interbank rate from the International
Monetary Fund’s International Financial Statistics (IFS) as a proxy of the policy
rate. (The change in interest rates, irj, included in the specification is therefore the
first difference of this series). The second measure, for simplicity, called the natural
gap (ig2) is the difference between the real short-term interest rates and the natural
43See appendix 3.1 for detailed estimations of the interest rate gaps.
44The Taylor equation is estimated as ig1 = c+1.5(⇡⇡⇤)+0.5y , where ⇡ measures the inflation
rate, ⇡⇤ measures the average inflation rate from 2000Q1, y measures the output gap, and c is the
sum of the average inflation and real GDP growth since 2000Q1. See appendix 3.1 for more details.
164
interest rates calculated using the Hodrick-Prescott filter. The first measure is more
of an ad-hoc rule. And since not all central banks of the sampled countries follow
the Taylor type rule, we also rely on the estimates that generate natural rates, which
remove the long-run trend of the interest rate series.
Raw data for the macro variables—growth rate (gr); yield curve (yc); exchange
rate volatility (er); changes in stock market index (sm); and bank credit to GDP
(cg)—are obtained from the IFS, Datastream, Bloomberg, and CEIC, depending on
countries and availability. Growth rate is obtained directly from the IFS as period-on-
period change in the GDP volume index. Yield curve measures the difference between
the 10-year government bond yield and the 3-month treasury bill rate. Changes in
stock market index are the first differences in the natural log of the index. Exchange
rate volatility is calculated as the standard deviation of the first differences of the
daily natural log exchange rate. Bank credit to GDP is the ratio of bank lending to
the private sector over nominal GDP.
For the bank specific variables: size, szi, is the log of total assets (Kashyap and
Stein 1995, 2000); liquidity, lqi, is the liquidity to-total asset ratio; capitalization,
cpi, is the capital to asset ratio (Kishan and Opiela 2000; Van den Heuvel 2002).
In the annual model, profitability, pri, as measured by the return on equity is also
included. These are data obtained directly from Bankscope (for the annual model),
and Bloomberg (for quarterly model), except the variable, lq in the quarterly model,
where it is calculated based on the raw data obtained from Bloomberg.
Table 9 provides descriptive statistics for the variables used. Note the sample
shows a large variability as measured by bank size, which implies that the sample
includes both large and small banks that reduce the potential of selection bias. Table
10 provides the descriptive statistics of the data sorted by country. The average size of
bank (as measured by total assets) is largest in the PRC and lowest in the Philippines.
Overall, the distribution of average bank size is homogeneous reflecting that no single
country has a dominant role in driving the results. In general, the banks included in
the sample cover roughly the top 85% of the banking sector (publicly listed) in each
165
of the countries45. Table 11 presents correlation matrix of the variables included
in estimation. The four measures of bank risk are positively (and significantly)
correlated with each other. Furthermore the two measures of the stance of monetary
policy, ig1 and ig2, are also positively and significantly correlated to each other.
This paper uses the dynamic general method of moments (GMM) to estimate
model equations (Arellano and Bond 1991; Arellano and Bover 1995; Blundell and
Bond 1998). The dynamic modeling approach has been used in various studies, such
as, measuring economic growth convergence (Caselli, Esquivel, and Lefort 1996),
estimating labour demand (Blundell and Bond 1998), estimating the relationship
between financial intermediary development and economic growth (Beck, Levine,
and Loayza 2000), and more recently in the literature dealing with bank risk-taking
channel (Altunbas et al. 2014; Delis and Kouretas 2011).
One possible identification challenge of testing whether monetary policy affects
bank risk is that there could be, in principle, a two-way relationship between mon-
etary policy and bank risk. If financial stability is an important goal of monetary
policy, then interest rate setting is also affected by overall changes in financial fragility
and bank risk. Ioannidou et al. (2015) and Jimenez et al. (2014), using data for Bo-
livian and Spanish banks, respectively, find that bank risk-taking and interest rates
are endogenous in the countries. That said, for the countries studied in this paper,
the objective of monetary policy is generally focused on achieving stable prices or the
dual mandate of stable prices and sustainable growth. We are unaware that main-
taining low bank risk is an explicit monetary policy objective in any of the countries.
However, in accordance with theory and following earlier empirical work we treat
interest rates as endogenous in the estimation methodology described below to avoid
any simultaneity bias.
Another possible source of endogeneity is unobserved heterogeneity, whereby
there are factors unobservable to researchers that can affect both bank risk and
the explanatory variables. Econometrically, unobservable heterogeneity exists in the
baseline equation if E (⌘i|Xi,t) 6= 0, where ⌘i is the individual bank specific fixed ef-
45According to Datastream, the stock market data typically cover the top 85% of firms in each
industry group.
166
fect, and Xi,t is the vector of explanatory variables. Given the greater co-movements
of macroeconomic data, the problem of unobservable heterogeneity is likely going to
be an issue, which means that the OLS estimators are both biased and inconsistent.
To address this and given the availability of panel data, fixed-effects estimation is
the solution. Still, fixed-effects estimation is biased if past values of the dependent
variable are used as an explanatory variable—if there is dynamic endogeneity, where
past values of the dependent variable are correlated with the current values of other
explanatory variables46.
Therefore, to obtain consistent and unbiased estimates in the presence of unob-
served and dynamic endogeneity, a dynamic GMM panel estimator is used. This
method uses the dynamic endogeneity inherent in the explanatory variables, that is,
lags of endogenous variables as instruments. In the difference GMM estimator, lagged
levels of endogenous variables are used as instruments for the difference equation.
Arellano and Bover (1995) and Blundell and Bond (1998) show that the difference
GMM estimator can be improved by also including equations in levels in the esti-
mation procedure or what is known as the system GMM estimator. System GMM
estimator is an improvement over the difference estimator in two main ways. First,
Beck, Levine, and Loayza (2000) note that if the original model is conceptually in
levels, differencing may attenuate the signal-to-noise ratio and reduce the power of
statistical inferences. Second, Arellano and Bover (1995) suggest that variables in
levels may be weak instruments for first-differenced equations. Thus, in system GMM
estimation, difference equations are complimented with equations in levels, wherein
lagged levels of endogenous variables as well as lags of first-differenced variables are
used as instruments for difference and level equations, respectively.
In the estimations, besides the interest rate variables (ir and ig), we treat as
endogenous the following variables: yield curve (yc), lagged bank return on equity
(pr), lagged bank liquidity (lq), and lagged bank capitalization (cp). For annual
model, therefore, the second to fifth lags of these variables are used as instruments47.
46See Wooldridge (2001) for more details on the nature of the bias.
47The lags used refer to both the difference and levels equations. Given the limited number of
observations, especially in the annual model, we restrict the number of instruments to avoid falling
167
All the regulatory indices and bank size are treated as predetermined variables (Delis
and Kouretas 2011). This implies that banks already have information on the size
and regulatory environment before making their decisions. This suggests that the
first to fifth lags of these variables are included as instruments. For the quarterly
model, since first lag of endogenous variables is included in the estimation, we employ
as instruments third to fifth lags of endogenous variable and second to fifth lags of
predetermined variables.
The GMM estimator ensures efficiency and consistency, provided that the dy-
namic regression model is not subject to second-order serial correlation and that the
instruments used are valid. The p-value of AR(2) should be large accepting the null
hypothesis of no serial correlation of order two in first differences of the errors. This
is important as higher order autocorrelation would imply that lags of the depen-
dent variable is not actually endogenous and, hence bad instruments. Furthermore,
the validity of the instruments is checked by using Sargan test for overidentifying
restrictions.
3.5 Results
3.5.1 Annual Model
Table 12 and 13 present the results of the annual model with the two measures of
interest rate gaps, the Taylor gap (ig1) and the natural gap (ig2), respectively. The
different columns indicate the different bank risk variables: idiosyncratic bank risk
(br1); ratio of non-performing loans to total loans (br2); z-score (br3); and volatility
of return on assets (br4).
Several key results can be gleaned from Table 12. First, the significant lagged
dependent variables show that bank risk is highly persistent (first four rows). This
result also lends support to the choice of a dynamic model. Given that the coefficient
is less than one, this implies that the risk, while it persists, eventually returns to its
into the trap of “too many instruments” as pointed out by Roodman (2009).
168
equilibrium48.
Second, changes in short-term interest rates (ir) do not have a significant impact
on most measures of bank risk, except in the case of idiosyncratic bank risk (br1),
where the coefficient is positive. A positive coefficient is expected for the market
determined risk measure, as lower interest rate are supposed to decrease bank risk
on the outstanding loans.This is likely to capture the effects from the balance sheet
channel—lower interest rates increase the creditworthiness of borrowers, and via
valuation effects reduce the riskiness of existing banks’ loans. This is consistent with
the findings of Jimenez et al. (2014) and Altunbas et al. (2014) that lower rates
make loan repayment easier by reducing the interest burden of borrowers, which in
turn, makes loan less risky.
Third, the coefficient of the main variable of interest, the Taylor gap (ig1) is
significant and has the expected negative sign in all bank risk variables. (Since the
z-score as a measure of bank risk is interpreted as the opposite of other measures,
the correct a priori sign is positive). This means that higher interest rates above
the Taylor-rule benchmark decreases bank risk or, conversely, lower interest rates
below the benchmark increases bank risk—supporting the phenomenon that “too
low interest rates” can be harmful49. In terms of magnitude, if interest rates are
100-basis point lower than the benchmark, the short (1 year) and long term effect
on bank risk (in percent) as measured by different indicators is given below:
Bank risk variable Short term (%) Long term (%)
Idiosyncratic bank risk 0.04 0.07
Non-performing loans 0.07 0.2
Z-score 0.03 0.07
Volatility of return on assets 0.01 0.02
48We also included higher lags of the bank risk variables, but found no evidence of persistence
beyond the first year.
49Taken together with the preceding result, it is clear that the interest gap variable is a better
measure of risk-taking channel. Take for instance the current situation where interest rates in
emerging economies are not close zero-bound (unlike the case in many advanced economies); in this
environment, a fall in the interest rates does not necessarily lead to more risk-taking. Yet, if interest
rates are “too low” (lower than a benchmark), we find a significant evidence of more risk-taking by
banks which is robust to different measures of bank-risk.
169
The increase in bank risk in response to lax monetary policy is lower in the case
of Asia as compared to advanced economies like the US and euro area, for instance
Altunbas (2014) find that if the interest rate is 100 basis point below the value given
by taylor rule, the average probability for a bank to go into default increases by 0.6%
after a quarter and 0.8% in the long run. The estimates in our study however are
closer to findings in some other emerging market economies. For instance, Özsuca
and Akbostancı (2016) find that in Turkish banks, the average probability of loan
default increases by 0.09 percent after a quarter and 0.2 in the long run in response
to interest rate being 100 basis point below the benchmark rate.
Thus, the impact on bank risk varies depending on the risk measure utilized. As
mentioned earlier, a priori, no one measure is better than the other and therefore
these results should be viewed as complementary in understanding the impact of low
interest rates on bank risk taking.
Fourth, of the four bank specific variables, only profitability (pr) as measured
by return on equity is statistically significant and correctly negatively signed in two
measures of bank risk—idiosyncratic risk (br1) and volatility of the return on as-
sets (br4). This implies that more profitable banks have a lower riskiness measure.
On the other hand, both bank size (sz) and bank liquidity (lq) are appropriately
negatively signed in each case of idiosyncratic bank risk measure (br1) and ratio
of non-performing loans to total loans (br2), respectively. However non-performing
loans are positively related to bank size bank size (sz). One reason why large large
banks may undertake higher levels of risky assets is that they are more capable in
managing risk and have an easier access to external funds when needed.
Fifth, in terms of accounting for the expectations channel, the output growth
rate (gr) is statistically significant and correctly signed in all four measures of bank
risk. Better economic conditions increase the profitability or net present values of
projects, which reduce the overall credit risk of banks (Kashyap et al. 1993; Altunbas
et. al. 2014). This result is consistent with the findings of Gambacorta (2009)
and Altunbas et al. (2014). The steepness of the yield curve (yc), however, has a
significant negative impact only for idiosyncratic bank risk (br1).
170
Sixth, there is some evidence to suggest that financial deepening reduces bank
risk. Bank credit to GDP (cg) is statistically significant and appropriately signed in
the case of br3 and br4.
Seventh, exchange rate volatility and stock market performance does not seem
to effect bank risk taking in Asia. Both exchange rate volatility (er) and changes
in broad stock market index (sm) are statistically insignificant for all bank risk
measures.
Eighth, there is also some evidence to suggest that regulatory environment ame-
liorates bank risk-taking. Greater capital stringency is statistically significant and
reduces bank risk in the case of br1. Similarly, supervisory power has a negative
effect on non-performing loans, br2.
The validity and consistency of the results is confirmed by the tests for the second-
order serial correlation and instrument exogeneity (the Hansen test). The null hy-
pothesis of no second-order serial correlation cannot be rejected in each of the bank
risk specification. Similarly, the null of instrument exogeneity cannot be rejected
implying that the instruments used are valid. Time and country dummies are also
jointly significant in most specifications.
As robustness checks, we also use an alternate measure of interest rate gap based
on the natural real interest rates (Table 13). By and large, the results are consistent
with the findings in Table 12. There is strong evidence to support the specification of
a dynamic model—the lagged dependent variable is always statistically significant.
“Too low” interest rates (ig2) contribute to greater risk-taking in all measures of bank
risk (see table below), not lower interest rates per se (ir)50. If interest rates are 100
basis point lower than the benchmark, as measured by the natural rate, the short (1
year) and long term effect on bank risk (in percent) is:
50As in Table 12, changes in interest rates (ir) are only statistically significant when bank risk is
measured as idiosyncratic bank risk (br1).
171
Bank risk variable Short run (%) long run (%)
Idiosyncratic bank risk 0.03 0.05
Non-performing loans 0.11 0.33
Z-score 0.04 0.08
Volatility of return on assets 0.02 0.04
Bank profitability (pr) seems to be a more important bank specific variable that
affects bank risk-taking compared to liquidity and size. Likewise, growth rate (gr)
rather than slope of the yield curve (yc) is a more important determinant of eco-
nomic activity or indirectly profitability of projects that affects credit risk of banks
and hence bank risk-taking. Similarly, regulatory environment, particularly capital
stringency and supervisory power appears to be significant in some risk specifica-
tion. One result that differs slightly between the two measures of interest rate gap,
is the more mixed evidence of financial deepening in reducing bank risk using the
natural gap measure. Finally, the specification tests on the absence of second-order
serial correlation and instrument exogeneity are passed, implying that the parameter
estimates are consistent and valid.
Further robustness checks are employed to ensure that the model does not suffer
from multi-collinearity problem. There are two interest rate variables included in
the model, namely ir and ig. Table 14 and 15 present results of the model with
only ig i.e. the main variable of interest. The coefficient is still significant and of
expected negative sign. Furthermore, the magnitudes of the effect are also similar to
the baseline models confirming that the two variables capture separate effects. The
impact of other macroeconomic and bank specific controls are also similar as in the
baseline model.
The sample used in the paper consists of countries where interest rates are not
the main policy instrument and in some cases where interest rates are not necessarily
market determined. This may affect the computation of interest rate gap variable.
Furthermore, China is a peculiar case wherein the credit boom following the gov-
ernment stimulus efforts in the aftermath of the global financial crisis may distort
any relation between interest rates and bank risk taking. Therefore, table 16 and
172
17 present results without China, Hong Kong and Singapore (where interest rates
are not the main policy instruments), to further assess the robustness of the find-
ings. The results from the models are qualitatively unchanged. In particular, the
coefficient on both the taylor rate and natural rate gap is significant and appropri-
ately negatively signed suggesting that inclusion of these countries, especially China,
does not bias the findings. Furthermore, quantitatively the effect of low interest rate
(below a benchmark) on bank risk taking is similar to that reported above.
3.5.2 Quarterly Model
The effects of short-term interest rate on bank’s risk tolerance can be very brisk, and
hence the use of annual data to study the risk-taking channel has been contested
by Altunbas et al. (2014). Furthermore, Jimenez et al. (2014) find interest rates
could affect bank risk in two opposing ways. In the short term, low interest rates,
by reducing the probability of default on existing loans, leads to a fall in the risk of
bank’s portfolio. Over time, however, banks start to undertake more risky operations
and lending to projects with a higher probability of default. To further investigate
this relationship and to distinguish between any likely short- and longer-term effects
that might be missing in the annual specification, this paper also estimates a model
with quarterly data. In the quarterly model, idiosyncratic risk is the only bank risk
used because of lack of data availability.
Indeed, the quarterly model provides more nuanced results (Table 18, Column 1).
The contemporaneous Taylor gap (ig1) is statistically significant and positive, mean-
ing if interest rates are higher than benchmark, bank risk increases, or conversely, if
interest rates are lower than benchmark, bank risk decreases. This runs contrary to
what is expected of an operating risk-taking channel. That said, lagged ig1 has the
a priori sign that supports the risk-taking channel (as in the annual model). In fact,
the combined magnitude of the contemporaneous and lagged ig1 is still negative,
implying overall the risk-taking channel is still operating and the effects of “too low”
interest rates is likely to dominate. The long run impact of low interest rate on bank
risk is 0.05 and 0.1 percent respectively in the quarterly model.
173
The coefficient of exchange rate volatility, which captures market risk, is positive
and significant, implying that countries with more volatility in the exchange rate tend
to have a more risky banking sector. The other result that is statistically significant
is the adverse impact of financial deepening on bank risk—the contemporaneous
coefficient of credit to GDP is positive. Again as a robustness check, the other
measure of interest rate gap (ig2) is used (Column 2). The results are likewise
similar to that of the Taylor gap.
3.6 Conclusion
Recent studies have found a significant negative relationship between interest rates
and bank risk-taking in the US and the euro area. This paper explores the impact of
low interest rates on bank risk-taking in selected Asian economies with a relatively
developed banking sector. Using a recent line of empirical and theoretical literature,
and a panel dataset of publicly listed banks, we find evidence of an operational
bank risk-taking channel. In particular, we find that when interest rates are “too
low”—lower than a benchmark—bank risk increases. This result is robust to other
factors, which might have influenced banks’ risk-taking behavior, namely, general
economic conditions, bank specific characteristics, country specific market risk, and
the regulatory environment.
The results of this paper point to three main policy considerations. First, the
conduct of monetary policy, in particular, the setting of interest rates does not seem
neutral to the financial stability goal. It is therefore encouraging to see work done to
incorporate financial stability considerations into macroeconomic models, especially
in the context of monetary policy formulation (Agur and Demertzis 2010; Curdia
and Woodford 2011). Second, more importantly, this suggests a greater role of
macroprudential policy focusing on possible excesses in finance and its related areas,
even during periods of benign inflation rate when interest rates are usually kept low.
In particular, the capital stringency requirement seems to pose a limiting check on
bank risk. Finally, as the Asian economies are not well insulated from the loose
174
monetary policy of the advanced economies, it is important for policy makers to
allow the exchange rates to move more flexibly, especially in level terms. There
will be periods when policy makers—by allowing interest rates to fall too low—not
only fuel the risk-taking channel, but may also interfere with the pursuit of domestic
stabilization goals.
175
Table 13 – Summary Statistics of the Variables used in Annual Regression, 2000-
2011
Variable No ofobs. Mean
Std. De-
viation Min Max
Idiosyncratic risk, br1 824 0.1240 0.7250 0.00004 12.2978
Non-performing loans to total loans, br2 768 0.0533 0.0563 0.0013 0.2904
Z-score, br3 734 3.7791 1.2864 -2.0381 7.8179
Volatility of returns on asset, br4 731 0.4093 0.8872 0.0116 11.0302
Changes in short-term interest rate, ir 826 -0.2056 2.2732 -13.2634 7.4900
Taylor gap, ig1 826 -5.7417 4.8519 -23.5854 10.5615
Natural gap, ig2 826 0.1140 2.3175 -7.9591 7.6715
Growth rate, gr 813 5.7603 3.2587 -7.2370 14.7630
Yield curve, yc 805 1.8120 1.2050 -0.6431 6.7590
Credit to GDP, cg 800 0.7519 0.3990 0.1816 2.0212
Exchange rate volatility, er 826 0.0014 0.0011 0.0000 0.0075
Change in stock market index, sm 826 0.0347 0.1239 -0.3915 0.4770
Bank size, sz 815 10.2053 1.3598 6.8954 14.7142
Profitability, pr 818 12.8695 10.7146 -70.4210 44.8120
Liquidity, lq 824 22.2159 11.3602 2.1370 80.7420
Bank capitalization, cp 790 8.1472 3.1576 0.1600 23.7900
Supervisor index, sp 826 11.5387 1.3327 7.0000 14.0000
Market discipline index, md 826 6.8535 0.8408 5.0000 9.0000
Capital stringency index, cs 826 5.2143 1.5386 3.0000 8.0000
Note: See the appendix for details of each variable. The variables in index are index numbers, while idiosyncratic
risk, z-score, volatility of returns on asset, and exchange rate volatility are unit-free. All the other variables are
in percent, except bank size which is the natural log of US million.
Source: Authors’ calculations.
176
Table 14 – Descriptive Statistics by Country (mean values), 2000 - 2011
Variable PRC HK IND INO KOR MAL PHI SIN TAP THA
br1 0.0047 0.0094 0.0094 0.2234 0.0413 0.0185 0.0628 0.0103 0.0592 0.2667
br2 0.0150 0.0129 0.0313 0.0581 0.0164 0.0852 0.0990 0.0409 0.0124 0.1068
br3 4.3097 4.2498 4.2620 2.9952 3.0652 4.0856 4.1526 4.2024 3.4799 2.8399
br4 0.1104 0.2009 0.1420 0.8701 0.4187 0.3845 0.2858 0.1872 0.2983 1.0297
ir 0.0192 -0.4222 0.1823 -1.3363 -0.1549 -0.0311 -0.4441 -0.1432 -0.0017 0.0584
ig1 -11.6494 -4.0297 -8.7797 -5.7888 -4.1938 -4.5033 -2.4541 -6.9155 -3.0482 -4.3433
ig2 -0.1012 -0.1711 0.7403 -0.6128 0.0874 0.0458 0.3676 0.0186 -0.0615 0.1700
gr 10.9097 4.5150 7.7555 5.4174 4.2668 5.0121 4.7129 5.9411 4.2113 3.9494
yc 1.5117 2.2253 1.3069 1.9677 1.1235 1.4977 3.5072 1.7385 0.8975 2.2589
cg 1.1847 1.5315 0.4452 0.2441 0.9406 1.1073 0.3097 1.0013 0.5378 0.9858
er 0.0004 0.0001 0.0015 0.0027 0.0029 0.0009 0.0017 0.0014 0.0012 0.0014
sm 0.0355 0.0199 0.0009 0.0775 0.0411 0.0282 0.0744 0.0153 0.0203 0.0336
sz 12.0988 10.7471 10.1456 9.2576 11.6287 9.8637 8.5764 11.5259 10.7194 9.6430
pr 18.4509 16.4763 17.2230 17.8226 10.8787 11.0267 9.9985 10.4788 6.3182 6.6352
lq 27.3887 29.8162 10.2021 28.6074 13.8684 29.7413 29.9321 29.9521 14.9926 16.4083
cp 5.5872 8.6296 6.4593 10.5523 6.0612 8.7753 11.4927 9.6669 6.5575 7.7838
sp 10.3377 9.8261 12.2109 12.5652 10.3654 11.3448 11.3636 12.5000 13.1233 10.7273
md 6.4545 7.0000 6.7891 6.3913 7.6154 7.3362 7.0000 7.8333 6.8767 6.0455
cs 3.4416 4.6522 7.0000 5.8913 4.3462 4.0345 5.1515 5.8333 5.2192 5.2500
Number of
banks
14 4 16 9 5 10 9 3 8 8
Weight of
country i
16.28 4.65 18.60 10.47 5.81 11.63 10.47 3.49 9.30 9.30
Note: PRC=the People’s Republic of China; HKG=Hong Kong, China; IND=India; INO=Indonesia; KOR=the Republic of
Korea; MAL=Malaysia; PHI=the Philippines; SIN=Singapore; TAP=Taipei,China; and THA=Thailand. br1=idiosyncratic
risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility of returns on asset, ir =changes in short-term
interest rate, ig1=Taylor gap, ig2=natural gap, gr =growth rate, yc =yield curve, cg =credit to GDP, er =exchange rate
volatility, sm=change in stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. Number of banks=number of banks
included from each country. Weight of country i = number of banks (country i)/total number of banks.
Source: Authors’ calculations.
177
T
ab
le
15
–
C
or
re
la
tio
n
M
at
rix
br
1
br
2
br
3
br
4
ir
ig
1
ig
2
cg
gr
yc
sm
sz
lq
cp
pr
er
sp
m
d
cs
br
1
1.
00
0
br
2
0.
11
6
1.
00
0
br
3
-0
.2
11
-0
.3
19
1.
00
0
br
4
0.
12
9
0.
31
4
-0
.6
59
1.
00
0
ir
-0
.0
25
-0
.0
98
0.
13
9
0.
02
4
1.
00
0
ig
1
0.
02
9
0.
28
8
-0
.1
52
0.
10
3
0.
01
5
1.
00
0
ig
2
-0
.0
14
-0
.0
26
0.
05
2
-0
.0
20
0.
30
8
0.
62
0
1.
00
0
cg
0.
00
4
-0
.0
32
0.
08
2
-0
.0
82
0.
03
7
-0
.0
67
-0
.0
11
1.
00
0
gr
-0
.1
36
-0
.1
65
0.
14
5
-0
.1
22
0.
18
2
-0
.3
81
-0
.0
46
-0
.0
32
1.
00
0
yc
0.
06
7
0.
29
7
-0
.0
63
0.
13
1
-0
.1
15
0.
01
0
-0
.2
22
-0
.1
24
-0
.0
85
1.
00
0
sm
-0
.0
32
0.
02
9
0.
06
3
-0
.0
52
0.
08
9
-0
.0
43
-0
.0
94
-0
.1
10
0.
32
4
-0
.0
10
1.
00
0
sz
-0
.1
36
-0
.4
83
0.
19
1
-0
.2
24
0.
10
2
-0
.3
31
0.
00
4
0.
44
5
0.
19
3
-0
.3
25
-0
.0
61
1.
00
0
lq
-0
.0
43
0.
20
1
0.
08
3
0.
01
9
-0
.1
16
0.
10
5
-0
.0
52
0.
19
6
-0
.0
53
0.
19
7
0.
08
6
-0
.0
92
1.
00
0
cp
-0
.0
63
0.
25
8
0.
20
1
-0
.0
24
-0
.0
68
0.
21
9
-0
.0
11
-0
.1
77
-0
.1
68
0.
29
5
0.
10
7
-0
.4
39
0.
40
2
1.
00
0
pr
-0
.0
91
-0
.3
09
0.
29
7
-0
.1
81
0.
01
0
-0
.2
17
-0
.0
25
-0
.0
55
0.
18
1
-0
.0
69
0.
10
2
0.
17
1
0.
04
8
0.
00
0
1.
00
0
er
0.
03
5
-0
.0
27
-0
.2
33
0.
17
5
-0
.0
99
-0
.0
53
-0
.0
98
-0
.4
07
-0
.2
02
0.
06
2
-0
.1
28
-0
.1
18
-0
.1
05
0.
06
6
-0
.0
52
1.
00
0
sp
-0
.0
78
-0
.2
09
0.
06
3
-0
.0
67
-0
.0
63
-0
.0
19
-0
.0
45
-0
.5
29
-0
.0
16
-0
.1
92
0.
04
8
-0
.1
05
-0
.1
34
0.
03
2
0.
03
2
0.
17
2
1.
00
0
m
d
-0
.1
67
-0
.1
28
0.
23
8
-0
.2
13
0.
14
4
0.
11
7
0.
09
2
0.
11
8
-0
.0
39
-0
.1
23
0.
08
4
0.
16
8
0.
07
8
0.
11
0
0.
00
8
-0
.1
02
-0
.0
35
1.
00
0
cs
0.
03
6
-0
.2
55
0.
15
1
-0
.1
15
0.
07
4
-0
.0
76
0.
11
9
-0
.5
00
0.
00
7
-0
.1
13
-0
.0
08
-0
.0
77
-0
.2
64
-0
.0
02
0.
12
2
0.
22
1
0.
42
8
0.
06
6
1.
00
0
178
Table 16 – Results: Annual Model with Taylor Interest Rate Gap (ig1)
and Different Bank Risk Measures
Variables Idiosyncratic
risk, br1
Non-
performing
loans, br2
Z-score, br3 Volatility of
Return on Assets,
br4
L.br1 0.386***
(0.0398)
L.br2 0.659***
(0.0749)
L.br3 0.546***
(0.0694)
L.br4 0.506***
(0.0610)
ir 0.0436** 0.000732 -0.000613 -0.0178
(0.0187) (0.000497) (0.0353) (0.0191)
ig1 -0.0221** -0.000847** 0.0305** -0.0114*
(0.0106) (0.000359) (0.0141) (0.00644)
cg 0.434 0.0193 1.221*** -0.481**
(0.347) (0.0123) (0.455) (0.213)
gr -0.0255** -0.000853** 0.0270** -0.0111*
(0.0117) (0.000334) (0.0131) (0.00629)
yc -0.0874* -0.000779 0.0820 0.0228
(0.0490) (0.00209) (0.0765) (0.0326)
sm 0.00450 -0.0115 -0.337 -0.218
(0.187) (0.00916) (0.411) (0.155)
sz -0.600* 0.0135** -0.0532 -0.0490
(0.312) (0.00637) (0.339) (0.188)
L.lq 0.0125 -0.000624** -0.00492 -0.000533
(0.00965) (0.000281) (0.0112) (0.00677)
L.cp -0.0440 7.01e-05 0.00556 -0.00617
(0.0318) (0.000911) (0.0492) (0.0234)
L.pr -0.0330*** -9.25e-05 0.0153 -0.0152*
(0.00410) (0.000281) (0.0114) (0.00840)
er -78.62 1.829 -33.70 -22.59
(47.86) (1.325) (78.33) (32.24)
sp -0.0688 -0.00334*** 0.0705 -0.0145
(0.0481) (0.000905) (0.0558) (0.0218)
md -0.112 0.00171 -0.0808 0.00921
(0.0783) (0.00162) (0.0643) (0.0465)
cs -0.0511* -0.000319 0.0429 -0.000106
(0.0285) (0.00123) (0.0606) (0.0362)
Observations 684 646 624 623
F-statistic 72.86*** 262.25*** 37.58*** 53.11***
AR(2) 0.581 0.636 0.918 0.150
Hansen 0.220 0.266 0.0920 0.267
Wald test for country 1.38 3.46*** 3.56*** 3.15***
and time effects
Note: br1=idiosyncratic risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility
of returns on asset, ir =changes in short-term interest rate, ig1=Taylor gap, ig2=natural gap, gr
=growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate volatility, sm=change in
stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. L refers to
one-period lag. F-stat and its p-value is the goodness of fit of the regression, AR(2) is the test
for second-order serial correlation and Hansen is the test for over-identifying restrictions. Wald
test is the test for joint significance of country and time dummies. Robust standard errors are in
parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
179
Table 17 – Results: Annual Model with Natural Interest Rate Gap
(ig2) and Different Bank Risk Measures
Variables Idiosyncratic
risk, br1
Non-
performing
loans, br2
Z-score, br3 Volatility of
Return on Assets,
br4
L.br1 0.414***
(0.0291)
L.br2 0.649***
(0.0791)
L.br3 0.512***
(0.0731)
L.br4 0.518***
(0.0659)
ir 0.0284* 0.000376 -0.00477 -0.00988
(0.0151) (0.000440) (0.0282) (0.0137)
ig2 -0.0258** -0.00114** 0.0393** -0.0211**
(0.0108) (0.000568) (0.0194) (0.0104)
cg 0.537* 0.0270** 1.188** -0.382**
(0.313) (0.0122) (0.492) (0.186)
gr -0.0209** -0.000774** 0.0300** -0.0120*
(0.0101) (0.000336) (0.0133) (0.00641)
yc -0.0721* -5.84e-05 0.0433 0.0121
(0.0386) (0.00194) (0.0664) (0.0343)
sm -0.0111 -0.0118 -0.269 -0.221
(0.197) (0.00878) (0.405) (0.166)
sz -0.509** 0.0108* -0.434 -0.0867
(0.249) (0.00616) (0.320) (0.257)
L.lq 0.0133 -0.000555** -0.00727 0.00113
(0.0128) (0.000270) (0.0119) (0.00703)
L.cp -0.0353 -9.65e-05 -0.00865 -0.0101
(0.0259) (0.000889) (0.0572) (0.0247)
L.pr -0.0303*** -0.000193 0.0143 -0.0151*
(0.00400) (0.000269) (0.0108) (0.00787)
er -72.48* 1.763 -64.65 -30.18
(41.82) (1.278) (82.78) (35.72)
sp -0.0816 -0.00365*** 0.0608 -0.0241
(0.0498) (0.000984) (0.0579) (0.0260)
md -0.0931 0.00243 -0.0764 0.00857
(0.0731) (0.00159) (0.0634) (0.0432)
cs -0.0505** -0.00129 0.0451 -0.0109
(0.0239) (0.00128) (0.0597) (0.0414)
Observations 684 646 624 623
F-statistic 75.92*** 265.02*** 18.23*** 63.08***
AR(2) 0.400 0.696 0.992 0.156
Hansen 0.204 0.227 0.107 0.168
Wald test for country 1.68* 3.31*** 1.85** 1.98**
and time effects
Note: br1=idiosyncratic risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility
of returns on asset, ir =changes in short-term interest rate, ig1=Taylor gap, ig2=natural gap, gr
=growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate volatility, sm=change in
stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. L refers to
one-period lag. F-stat and its p-value is the goodness of fit of the regression, AR(2) is the test
for second-order serial correlation and Hansen is the test for over-identifying restrictions. Wald
test is the test for joint significance of country and time dummies. Robust standard errors are in
parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
180
Table 18 – Results: Annual Model with Taylor Interest Rate Gap (ig1)
- without ir
Variables Idiosyncratic
risk, br1
Non-
performing
loans, br2
Z-score, br3 Volatility of
Return on Assets,
br4
L.br1 0.382***
(0.0421)
L.br2 0.676***
(0.0711)
L.br3 0.538***
(0.0653)
L.br4 0.490***
(0.0671)
ig1 -0.0171* -0.000798** 0.0298** -0.0128*
(0.00973) (0.000371) (0.0139) (0.00660)
cg 0.488 0.0192 1.258*** -0.464**
(0.349) (0.0125) (0.459) (0.195)
gr -0.0202* -0.000756** 0.0267** -0.0135**
(0.0109) (0.000336) (0.0133) (0.00623)
yc -0.0501 -0.000104 0.0833 0.00927
(0.0337) (0.00197) (0.0739) (0.0294)
sm -0.00998 -0.0120 -0.315 -0.202
(0.179) (0.00909) (0.412) (0.154)
sz -0.619* 0.0147** -0.0788 -0.0760
(0.332) (0.00640) (0.332) (0.166)
L.lq 0.0104 -0.000624** -0.00939 0.00295
(0.0112) (0.000281) (0.00992) (0.00615)
L.cp -0.0561* 0.000215 -0.00187 -0.00212
(0.0309) (0.000913) (0.0476) (0.0226)
L.pr -0.0320*** -2.46e-05 0.0141 -0.0200***
(0.00431) (0.000272) (0.0100) (0.00675)
er -67.05 2.104 -46.99 -29.87
(45.50) (1.312) (80.22) (32.55)
sp -0.0798* -0.00339*** 0.0651 -0.0102
(0.0475) (0.000901) (0.0546) (0.0202)
md -0.0740 0.00214 -0.0747 -0.00872
(0.0696) (0.00155) (0.0630) (0.0354)
cs -0.0541* -0.000168 0.0387 -0.00525
(0.0296) (0.00127) (0.0605) (0.0295)
Observations 684 647 625 624
F-statistic 62.55*** 250.74*** 38.78*** 53.28***
AR(2) 0.472 0.577 0.522 0.147
Hansen 0.593 0.341 0.142 0.233
Wald test for country 1.64* 3.67*** 3.45*** 2.96***
and time effects
Note: br1=idiosyncratic risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility
of returns on asset, ir =changes in short-term interest rate, ig1=Taylor gap, ig2=natural gap, gr
=growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate volatility, sm=change in
stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. L refers to
one-period lag. F-stat and its p-value is the goodness of fit of the regression, AR(2) is the test
for second-order serial correlation and Hansen is the test for over-identifying restrictions. Wald
test is the test for joint significance of country and time dummies. Robust standard errors are in
parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
181
Table 19 – Results: Annual Model with Natural Interest Rate Gap
(ig2) - without ir
Variables Idiosyncratic
Bank risk,
br1
Non-
performing
loans, br2
Z-score, br3 Volatility of
Return on Assets,
br4
L.br1 0.376***
(0.0395)
L.br2 0.666***
(0.0750)
L.br3 0.506***
(0.0734)
L.br4 0.504***
(0.0709)
ig2 -0.0275** -0.00105* 0.0381* -0.0245**
(0.0135) (0.000577) (0.0194) (0.0110)
cg 0.570 0.0263** 1.249** -0.351*
(0.352) (0.0121) (0.520) (0.180)
gr -0.0200* -0.000704** 0.0293** -0.0143**
(0.0107) (0.000341) (0.0132) (0.00670)
yc -0.0561 0.000347 0.0390 0.00279
(0.0362) (0.00187) (0.0671) (0.0326)
sm -0.00875 -0.0119 -0.240 -0.211
(0.191) (0.00872) (0.406) (0.167)
sz -0.619* 0.0119* -0.498 -0.112
(0.352) (0.00605) (0.338) (0.242)
L.lq 0.0113 -0.000555** -0.0120 0.00484
(0.0114) (0.000269) (0.0103) (0.00662)
L.cp -0.0581* -2.75e-05 -0.0197 -0.00639
(0.0314) (0.000889) (0.0547) (0.0234)
L.pr -0.0315*** -8.97e-05 0.0121 -0.0202***
(0.00438) (0.000253) (0.00975) (0.00689)
er -65.51 1.963 -84.27 -34.00
(44.89) (1.188) (83.60) (35.70)
sp -0.0835* -0.00363*** 0.0539 -0.0212
(0.0496) (0.000967) (0.0559) (0.0227)
md -0.0690 0.00257 -0.0740 -0.00149
(0.0710) (0.00155) (0.0649) (0.0368)
cs -0.0670* -0.00112 0.0384 -0.0192
(0.0359) (0.00131) (0.0592) (0.0362)
Observations 684 647 625 624
F-statistic 55.01*** 253.31*** 18.56*** 59.93***
AR(2) 0.386 0.612 0.592 0.150
Hansen 0.518 0.217 0.156 0.229
Wald test for country 1.78** 3.50*** 2.32*** 2.77***
and time effects
Note: br1=idiosyncratic risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility
of returns on asset, ir =changes in short-term interest rate, ig1=Taylor gap, ig2=natural gap, gr
=growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate volatility, sm=change in
stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. L refers to
one-period lag. F-stat and its p-value is the goodness of fit of the regression, AR(2) is the test
for second-order serial correlation and Hansen is the test for over-identifying restrictions. Wald
test is the test for joint significance of country and time dummies. Robust standard errors are in
parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
182
Table 20 – Results: Annual Model with Taylor Interest Rate Gap (ig1)
- restricted sample
Variables Idiosyncratic
risk, br1
Non-
performing
loans, br2
Z-score, br3 Volatility of
Return on Assets,
br4
L.br1 0.389***
(0.0405)
L.br2 0.743***
(0.144)
L.br3 0.610***
(0.0634)
L.br4 0.562***
(0.112)
ir 0.0473** 0.00117* 0.0102 -0.0223
(0.0184) (0.000646) (0.0327) (0.0188)
ig1 -0.0121* -0.000887** 0.0263** -0.0106*
(0.00640) (0.000445) (0.0130) (0.00587)
cg 1.169 0.0877*** 1.721* -1.781**
(0.846) (0.0326) (0.972) (0.715)
gr -0.0332** -0.000727** 0.0333* -0.0115
(0.0145) (0.000338) (0.0172) (0.0103)
yc -0.0820 -0.000245 0.0863 0.0429
(0.0520) (0.00263) (0.0754) (0.0309)
sm -0.0631 -0.0299 -0.258 -0.224
(0.178) (0.0191) (0.416) (0.186)
sz -0.235 -0.00352 -0.185 -0.216
(0.296) (0.00575) (0.285) (0.212)
L.lq 0.0138 -0.000713** -0.0170 -0.00556
(0.0147) (0.000353) (0.0121) (0.00651)
L.cp -0.0104 -0.00191** -0.00120 0.00301
(0.0290) (0.000932) (0.0470) (0.0232)
L.pr -0.0304*** -0.000273* 0.00307 -0.00590*
(0.00572) (0.000159) (0.00554) (0.00317)
er -109.9 3.951* -111.6 7.185
(68.39) (2.035) (87.38) (31.30)
sp -0.0776 -0.00183 0.0586 -0.0310
(0.0499) (0.00185) (0.0617) (0.0271)
md -0.186* 0.00257 -0.104 0.0899
(0.102) (0.00256) (0.0897) (0.0833)
cs -0.0229 0.000719 0.0145 -0.0146
(0.0325) (0.00140) (0.0804) (0.0541)
Observations 568 552 511 516
F-statistic 196.57*** 147.99*** 33.53*** 40.83***
AR(2) 0.392 0.347 0.923 0.265
Hansen 0.419 0.283 0.515 0.229
Wald test for country 2.60*** 3.54*** 2.20** 1.86**
and time effects
Note: br1=idiosyncratic risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility
of returns on asset, ir =changes in short-term interest rate, ig1=Taylor gap, ig2=natural gap, gr
=growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate volatility, sm=change in
stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. L refers to
one-period lag. F-stat and its p-value is the goodness of fit of the regression, AR(2) is the test
for second-order serial correlation and Hansen is the test for over-identifying restrictions. Wald
test is the test for joint significance of country and time dummies. Robust standard errors are in
parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
183
Table 21 – Results: Annual Model with Natural Interest Rate Gap
(ig2) - restricted sample
Variables Idiosyncratic
risk, br1
Non-
performing
loans, br2
Z-score, br3 Volatility of
Return on Assets,
br4
L.br1 0.384***
(0.0360)
L.br2 0.719***
(0.142)
L.br3 0.596***
(0.0689)
L.br4 0.558***
(0.116)
ir 0.0388** 0.000816 0.00584 -0.0237
(0.0185) (0.000617) (0.0313) (0.0172)
ig2 -0.0241* -0.00143** 0.0443** -0.0144*
(0.0143) (0.000703) (0.0221) (0.00839)
cg 1.278 0.108*** 1.606 -1.662**
(0.902) (0.0320) (1.043) (0.627)
gr -0.0347** -0.000604* 0.0356** -0.0104
(0.0157) (0.000320) (0.0176) (0.0108)
yc -0.0875* -0.000943 0.123 0.0322
(0.0521) (0.00314) (0.0784) (0.0214)
sm -0.00359 -0.0235 -0.285 -0.201
(0.186) (0.0177) (0.431) (0.175)
sz -0.186 -0.00260 -0.222 -0.197
(0.319) (0.00758) (0.321) (0.266)
L.lq 0.0152 -0.000430 -0.0197 -0.00381
(0.0147) (0.000343) (0.0146) (0.00699)
L.cp -0.0162 -0.00375*** -0.00954 0.000230
(0.0321) (0.00109) (0.0482) (0.0277)
L.pr -0.0305*** -0.000196 0.00309 -0.00577*
(0.00560) (0.000175) (0.00586) (0.00328)
er -94.89 3.445* -119.8 10.41
(70.65) (1.904) (88.18) (34.30)
sp -0.0842 -0.00286* 0.0706 -0.0361
(0.0519) (0.00159) (0.0668) (0.0296)
md -0.174* 0.00401 -0.0930 0.0856
(0.101) (0.00280) (0.106) (0.0831)
cs -0.0247 9.65e-05 0.0331 -0.0176
(0.0349) (0.00196) (0.0830) (0.0577)
Observations 568 552 511 516
F-statistic 134.73*** 112.32*** 29.85*** 43.78***
AR(2) 0.372 0.281 0.955 0.257
Hansen 0.448 0.212 0.524 0.240
Wald test for country 3.89*** 4.79*** 2.37*** 2.07**
and time effects
Note: br1=idiosyncratic risk, br2=non-performing loans to total loans, br3=z-score, br4=volatility
of returns on asset, ir =changes in short-term interest rate, ig1=Taylor gap, ig2=natural gap, gr
=growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate volatility, sm=change in
stock market index, sz =bank size, pr =profitability, lq =liquidity, cp =bank capitalization, sp
=supervisor index, md =market discipline index, and cs =capital stringency index. L refers to
one-period lag. F-stat and its p-value is the goodness of fit of the regression, AR(2) is the test
for second-order serial correlation and Hansen is the test for over-identifying restrictions. Wald
test is the test for joint significance of country and time dummies. Robust standard errors are in
parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
184
Table 22 – Results: Quarterly Model with Idiosyncratic Bank Risk
(br1)
Variable Equation 1 Equation2
br1(t-1) 0.895*** 0.907***
(0.1420) (0.1440)
ir -0.0705 -0.0473
(0.0888) (0.0750)
ir(t-1) -0.0145 -0.016
(0.0535) (0.0527)
ig1 0.0158*
(0.0090)
ig1(t-1) -0.0209**
(0.0095)
ig2 0.0233*
(0.0138)
ig2(t-1) -0.0367*
(0.0185)
gr 0.0245 0.024
(0.0215) (0.0210)
gr(t-1) -0.0288 -0.029
(0.0191) (0.0193)
yc -0.212 -0.213
(0.1700) (0.1750)
yc(t-1) 0.185 0.184
(0.1190) (0.1200)
cg 0.0779* 0.0771*
(0.0423) (0.0410)
cg(t-1) -0.0651 -0.0641
(0.0436) (0.0469)
er 46.76** 47.92***
(17.96) (17.70)
er(t-1) -16.78 -14.89
(22.43) (24.41)
sm -0.312 -0.331
(0.2710) (0.2910)
sm(t-1) 0.0615 0.0414
(0.1460) (0.1570)
sz -0.0339 -0.0476
(0.0564) (0.0641)
cp(t-1) 0.00157 0.00221
(0.0026) (0.0026)
lq(t-1) 0.00394 0.00346
(0.0049) (0.0049)
sp -0.0316 -0.0303
(0.0227) (0.0220)
185
md 0.0271 0.0304
(0.0299) (0.0311)
cs -0.0643 -0.0623
(0.0510) (0.0497)
Observations 1,616 1,615
F-statistics 397.59 *** 409.10***
AR(2) 0.276 0.276
Hansen 0.438 0.36
Wald test for Country 1.79* 1.81*
and seasonal effects
Note: br1=idiosyncratic risk, ir =changes in short-term interest rate, ig1=Taylor gap,
ig2=natural gap, gr =growth rate, yc =yield curve, cg=credit to GDP, er =exchange rate
volatility, sm=change in stock market index, sz =bank size, lq =liquidity, cp =bank
capitalization, sp =supervisor index, md =market discipline index, and cs =capital stringency
index. L refers to one-period lag. F-stat and its p-value is the goodness of fit of the regression,
AR(2) is the test for second-order serial correlation and Hansen is the test for over-identifying
restrictions. Wald test is the test for joint significance of country and time dummies. Robust
standard errors are in parentheses. *** p<0.01, ** p<0.05, and * p<0.1.
Source: Authors’ estimations.
186
Appendix 3.1: Data, Sources and Transformation
Variable Data, Source and Period Transformation/Remarks
Bank risk
Idiosyncratic
risk, br1
Daily total return index of bank i;
index number. Source: Datastream
Daily total market return index of
country j, index number. Source:
Datastream, TOTMK...(RI).
Period (annual model):* 2000 to
2011, except the People’s Republic
of China (PRC) from 2004; and
India and Taipei,China from 2002.
Idiosyncratic risk is calculated as:
br1i =
Pm
t=1
"2i,t
m ,
where m is the number of trading
days in a year or a quarter
depending on the model used, and
is the residual of the following
estimated capital asset pricing
model:
ri,j,t  ¯ri,J = i,t(rm,j,t  ¯rm,J) + "i,t
where ri,j,t is the daily return of
stock price index of bank i at time
t, ¯ri,J is the average historical daily
return of stock price index of bank
i, rm,j,t is the daily return of stock
market price index of country j,
and ¯rm,J is the average historical
daily return of the stock market
price index of country j. Note, all
daily returns are first transformed
into natural log before estimation.
Non-
performing
loans to total
loans, br2 (%)
Non-performing loans to total
loans, %. Source: Consolidated
bank balance sheet, universal bank
model, Bankscope.
Period: 2000 to 2011, except the
PRC from 2004; the Republic of
Korea from 2001; and India and
Taipei,China from 2002.
187
Z-score, br3 Equity capital to total assets of
bank i, %. Source: Consolidated
bank balance sheet, universal bank
model, Bankscope.
Return on assets of bank i , %.
Source: Consolidated bank balance
sheet, universal bank model,
Bankscope.
The z-score is calculated as:
br3i,t = zi,t =
rai,t+eki,t
(rai,t)
where rai,t is the return on asset
and eki,t the ratio of equity capital
to total assets for bank i.
Period: 2000 to 2011, except the
PRC and Taipei,China from 2004;
and India from 2003.
(rai,t) is the standard deviation of
rai,t computed using data for
period t, t 1 and t 2, that is, a
three-year rolling window.
Volatility of
return on
assets, br4
Return on assets of bank i , %.
Source: Consolidated bank balance
sheet, universal bank model,
Bankscope.
Period: 2000 to 2011, except the
PRC from 2004; Indonesia,
Malaysia, the Philippines, Thailand
from 2001; and India and
Taipei,China from 2002.
Standard deviation of the return on
assets, computed over a three year
rolling window (over time period t,
t 1 and t 2).
Interest rate variables
188
Change in
short-term
interest rate,
irt, (%)
3-month money market rate or
3-month interbank deposit rate;
end of period, %. Source: IFS,
60B..ZF
For India: Call money rate (%)
Source: Reserve Bank of India
For Taipei,China: Policy rate:
discount rate (%). Source: CEIC,
45972101 (WMAD)
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Change in short-term interest rate
is simply: irt = it  it1, where i is
the 3-month money market or
interbank rate.
For the annual model, it is an
average of monthly rates.
Taylor rule
gap, ig1 (%)
3-month money market rate/3-
month interbank deposit rate; end
of period, %. Source: IFS, 60B..ZF
Annual CPI % change (quarterly
frequency, %.
Source: IFS, 64..XZF
GDP volume index, 2005=100.
Source: IFS, 99BVPZF
For Taipei,China: Inflation:
Consumer price index, year-on-year
% change. Source: CEIC,
249101701 (WIGEBAAA)
Quarterly Model: ig1 is computed
as the difference between the actual
nominal interest rate at end-period
and that generated by a Taylor-rule
as follows:
ig1 = c+ 1.5(⇡  ⇡⇤) + 0.5y, where
⇡ is the quarter-on-quarter inflation
rate, and y is the measure of output
gap created using HP filter (with a
smoothing parameter  = 1600).
The constant c is defined as the
sum of the average inflation and
real GDP growth since 2000Q1.
⇡⇤is computed as the average level
of the inflation rate since 2000Q1.
189
Gross domestic product, 2006p,
national currency. Source: CEIC,
261788101 (WARKAA)
GDP deflator, index number.
Source: CEIC, 261820301
(WARRAA)
GDP volume index = Nominal
GDP/GDP deflator (adjusted to
ensure that 2005 is base year).
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Annual Model: ig1 for the annual
model is calculated as the average
of quarterly ig1.
Natural rate
gap, ig2 (%)
3-month money market rate/3
month interbank deposit rate; end
of period, %. Source: IFS, 60B..ZF.
Annual CPI % change, quarterly
frequency, %.
Source: IFS, 64..XZF.
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Quarterly Model: Real interest rate
is computed using the Fisher
equation: rt = it  ⇡t where rt is
the real interest rate, it the nominal
interest rate and ⇡t is annual
inflation rate (quarter–on-quarter).
ig2 is computed as the difference
between the above real interest
rates and the natural rate
calculated using the
Hodrick-Prescott filter (with a
smoothing parameter  = 1600).
Annual model: ig2 for the annual
model is calculated as the average
of quarterly ig2.
190
Real GDP
growth rate,
gr (%)
Quarter: GDP volume index,
quarter-on- quarter % change. (%)
Source: IFS, 99BPXZF.
Annual: GDP volume index,
year-on-year % change. (%)
Source: IFS, 99BPXZF
For Taipei,China: Real GDP, y-o-y
% change, quarterly series.
Source: CEIC, 261981801
(WARKAB).
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
191
Yield curve, yc
(%)
Malaysia, the Philippines,
Thailand, Singapore: 3 month
T-bill rate, end-period, %.
Source: IFS, 60C..ZF
The People’s Republic of China:
Central Bank Bill, middle-rate,
end-period, %. Source:
Datastream, CHBNK3M.
Hong Kong, China: Exchange fund
bills (3 months), end-period, %.
Source: Datastream, HKEFB3M.
India: 91-day T-bill rate,
end-period, %. Source:
Datastream, INTB91D.
Indonesia: SBI 90-day middle rate,
end-period, %.
Source: Datastream, IDSB90.
The Republic of Korea: Yield on
CD-91-day, end-period, %.
Source: Datastream, KOCD91D.
Taipei, China: T-bill rate 91-day,
end-period, %.
Source: CEIC, 45984701
(WMPGA).
India, the Republic of Korea,
Malaysia, the Philippines,
Singapore, Thailand: 10-year
government bond yield, end-period,
%. Source: IFS, 61. . . ZF.
The People’s Republic of China:
10-year government bond yield,
end-period, %. Source: Bloomberg,
GCNY10YR.
Hong Kong, China: HK Exchange
Fund note, 10-year, end-period, %.
Source: Datastream, HKEFN10.
Quarterly Model: yc is the
difference between the 10 year
government bond yield and the
3-month T-bill rate.
Annual Model: The average of
quarterly yield curve is used as a
measure of the steepness of yield
curve for the annual model.
192
Indonesia: 10-year government
bond yield, end-period, %.
Source: Bloomberg, GIDN10YR.
Taipei,China: Government bond
yield, 10-year, end-period, %.
Source: CEIC, 45985701
(WMUAE).
Period (annual model): From 2000
to 2011, except the PRC from 2005;
India and Taipei,China from 2002;
and Indonesia from 2003.
193
Bank credit to
GDP ratio, cg
(%)
For all countries: Bank’s claim on
private sector, national currency,
billion.
Source: IFS, 22D..ZF.
GDP, national currency, billion.
Source: IFS, 99B..ZF
The PRC: GDP, national currency,
billion. Source: Datastream,
CHGDP...A.
Taipei,China: Bank Credit: Loans
and investments of financial
institutions, national currency,
billion.
Source: Datastream,
TWBANKLPA.
GDP, national currency, billion.
Source: Datastream, TWGDP...B.
Period (annual model): 2000 to
2011, except the PRC from 2004;
India and Taipei,China from 2002;
and Indonesia, Malaysia,
Philippines and Thailand from
2001.
Ratio of bank credit to GDP using
quarterly credit and quarterly
GDP, and annual credit and annual
GDP for the quarterly and annual
model, respectively.
194
Exchange rate
volatility, er
Daily nominal exchange rate, daily
average (USD/national currency).
Source: Bloomberg
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
er is calculated as the standard
deviation of the first difference in
the natural log of the daily
exchange rate: er = lnet  lnet1,
where e is the daily exchange rate.
Quarterly exchange rate volatility
is therefore calculated based on the
daily rate in each quarter; and
annual volatility, daily rate in each
year.
Change in
broad stock
market index,
sm
Share price: end period, index.
Source: IFS, 62.EPZF
Taipei,China: Equity Market
Index: TAIEX Capitalization
Weighted, end-month, index
number.
Source: CEIC, 46107601 (WZJA).
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Calculated as (annual/quarterly)
changes in the natural log of the
broad stock market index.
Bank specific variables
195
Total assets
(log), sz
(Millions
USD)
Annual:
Total Assets, USD million.
Source: Consolidated bank balance
sheets, universal bank model,
Bankscope.
Quarter: Total Assets, USD
million. Source: Consolidated bank
balance sheets, Bloomberg.
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
In cases where accounting standard
had changed from local Generally
Accepted Accounting Principles
(GAAP) to International Financial
Reporting Standards (IFRS),
information from both C* (local
GAAP) and C2 (IFRS) accounts
was used to obtain historical series.
Profitability
(return on
equity), pr
(%)
Annual: Return on Average Equity
(ROAE), %.
Source: Consolidated bank balance
sheets, universal bank model,
Bankscope.
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
As above
196
Liquidity (%
of total
assets), lq (%)
Annual:
Liquid assets/short-term funding,
% Source: Consolidated bank
balance sheets, universal bank
model, Bankscope.
Quarter, USD million: Cash and
bank balance. Interbank assets.
Short-term investment. Total
Assets.
Source: Bloomberg
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
For annual data, when accounting
standards had changed from local
GAAP to IFRS, information from
both C* (local GAAP) and C2
(IFRS) accounts was used to obtain
historical series.
For quarterly model: Liquidity =
(Cash and bank balance +
interbank assets + short term
investments)/Total Assets
Capitalization
(% of total
assets), cp (%)
Equity capital as a percentage of
total assets
Annual: Equity/total assets, %.
Source: Consolidated bank balance
sheets, universal bank model,
Bankscope.
Quarter: Total capital ratio. (%)
Source: Bloomberg
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Capitalization is the ratio of equity
capital to total assets.
For annual data, when accounting
standards had changed from local
GAAP to IFRS, information from
both C* (local GAAP) and C2
(IFRS) accounts was used to obtain
historical series.
For quarterly data, total capital
ratio is the ratio of total capital to
total assets.
Regulatory Indexes
197
Supervisory
Power, sp
Source: Barth et al. (2013)
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Supervisory power, reflects the
powers invested with the
supervisory agencies to take specific
actions against bank management
and directors, shareholders, and
bank auditors. The supervisory
power index can take values
between 0 and 14, with higher
values denoting higher supervisory
power. This variable is determined
by the answers to14 survey
questions. This index like the other
two regulatory indexes is an annual
series.
To obtain the quarterly series, the
value for each year is used as the
value of each quarter in that
particular year.
Market
Discipline, md
Source: Barth et al. (2013)
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Market discipline reflects the
degree of transparency and
disclosure of accurate information
to the public that the banks are
forced to undertake. This index can
take values between 1 and 7
(depending on answers to 7 survey
questions), again with higher values
denoting higher market discipline.
198
Capital
Stringency, cs
Source: Barth et. al. (2013)
Period (annual model): 2000 to
2011, except the PRC from 2004;
and India and Taipei,China from
2002.
Capital stringency shows the extent
of both initial and overall capital
stringency. Initial capital
stringency refers to the constraints
on the sources of funds counted as
regulatory capital, as well as
whether the regulatory or
supervisory authorities verify these
sources. Overall capital stringency
indicates whether or not provisions
are made for market risk and value
losses while calculating the
regulatory capital. Theoretically,
the capital stringency index can
take values between 0 and 8
(depending on answers to 8 survey
questions), with higher values
indicating more stringent capital
requirements.
Note: * All quarterly data for the PRC, Indonesia, Malaysia, Philippines, the Republic of Korea,
and Singapore are from 2004:1 to 2011:4, except the yield curve of the PRC from 2005:1. All
quarterly data from Thailand and Taipei,China are from 2003:1 to 2011:4.
199
References
[1] V. Acharya and M. Richardson (eds). 2009. Restoring Financial Stability: How to Repair a
Failed System. Wiley Finance.
[2] T. Adrian and A. Ashcraft. 2012. Shadow Banking Regulation. Annual Review of Financial
Economics, Annual Reviews, 4(1), 99-140.
[3] T. Adrian and H.S. Shin. 2011. Financial Intermediaries and Monetary Economics. In
B.M.Freedman and M. Woodford, eds. Handbook of Monetary Economics, Volume 3A. Ox-
ford, UK: Elsevier.
[4] T. Adrian and H.S. Shin. 2010. Liquidity and Leverage. Journal of Financial Intermediation,
19, 418-37.
[5] T. Adrian and H.S. Shin. 2009. Money, Liquidity, and Monetary Policy. American Economic
Review, 99(2), 600-605.
[6] T. Adrian and H.S. Shin. 2009. The shadow banking system: Implications for financial
reform. Federal Reserve Bank of New York. Staff Report no. 382.
[7] T. Adrian and H.S. Shin. 2008. Financial intermediaries, financial stability and monetary
policy. In Maintaining stability in a changing financial system. Federal Reserve Bank of
Kansas City Economic Policy Symposium, 287 - 334.
[8] P.R. Agenor, K. Alper, and L. Pareira da Silva. 2012. Capital requirements and business
cycles with credit market imperfections. Journal of Macroeconomics, 31, 687-705.
[9] I. Agur and M. Demertzis. 2010. Monetary Policy and Excessive Bank Risk-Taking. Discus-
sion Paper. No. 2010-30S. Tilburg: Center for Economic Research, Tilburg University.
[10] G.A. Akerlof and R.J. Shiller. 2009. Animal Spirits: How Human Psychology Drives the
Economy, and Why It Matters for Global Capitalism. Princeton University Press.
[11] R. Albuquerque, M. Eichenbaum and S. Rebelo. 2012. Valuation Risk and Asset Pricing.
NBER Working Papers 18617.
[12] F. Allen, E. Carletti, and R. Marquez. 2009. Stakeholder Capitalism, Corporate Governance
and Firm Value. Economics Working Papers ECO2009/10, European University Institute.
[13] F. Allen and D. Gale. 2000. Comparing Financial Systems, Cambridge, MIT Press.
200
[14] Y. Altunbas, L. Gambacorta, and D. Marqués Ibañez. 2014. Does Monetary Policy Affect
Bank Risk? International Journal of Central Banking, 14/1, 95 - 135.
[15] Y. Altunbas, L. Gambacorta, and D. Marqués Ibañez. 2010. Bank Risk and Monetary Policy.
Journal of Financial Stability, 6, 121-129.
[16] B. W. Ambrose, M. Lacour-Little and A. B. Sanders. 2005. Does regulatory capital arbitrage,
reputation, or asymmetric information drive securitization? Journal of Financial Services
Research, 28(1) 2/3, 113-133.
[17] I. Angeloni, F. Ester and M. Lo Duca. 2015. Monetary policy and risk taking. Journal of
Economic Dynamics and Control, 52(C), 285-307.
[18] A. Angkinand, and C. Wihlborg. 2008. Deposit Insurance, Risk-Taking and Banking Crises:
Is there a Risk-Minimizing Level of Deposit Insurance Coverage? Working Paper, Center for
Law, Economics and Financial Institutions, Copenhagen Business School.
[19] M. Arellano and S. Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo
Evidence and an Application to Employment Equations. Review of Economic Studies, 58(2),
277–297.
[20] M. Arellano and O. Bover. 1995. Another Look at the Instrumental Variables Estimation of
Error- Components Models. Journal of Econometrics, 68, 29-51.
[21] J. Ayuso, D. Pérez and J. Saurina. 2004. Are capital buffers procyclical? Evidence from
Spanish panel data. Journal of Financial Intermediation, 13, 249-264.
[22] Bank for International Settlements. 2012. 82nd Annual Report. June.
[23] Bank for International Settlements. 2010. 80th Annual Report. June.
[24] R.B. Barsky, L.H. Christopher, and M.S. Kimball. 2007. Sticky-Price Models and Durable
Goods. American Economic review, 97(3), 984–98.
[25] J.R. Barth, G. Caprio, and R. Levine. 2013. Bank Regulation and Supervision in 180 coun-
tries from 1999 to 2011. Journal of Financial Economic Policy, 5(2), 111-219.
[26] T.H. Beck, R. Levine, and N. Loayza. 2000. Financial Intermediation and Growth: Causality
and Causes. Journal of Monetary Economics, 46, 31-77.
[27] G. Bekaert, M. Hoerova and M. Lo Duca. 2013. Risk, uncertainty and monetary policy.
Journal of Monetary Economics, 60(7), 771-788.
201
[28] A.N. Berger. 2003. The economic effects of technological progress: Evidence from the banking
industry. Journal of Money, Credit and Banking, 35, 141–176.
[29] A.N. Berger, L.F. Klapper, and R.Turk-Ariss. 2009. Bank Competition and Financial Sta-
bility. Journal of Financial Services Research, 35(2), 99–118.
[30] A. N. Berger and G. F. Udell. 1993. Securitization, Risk, and the Liquidity Problem in
Banking, in Michael Klausner and Lawrence White, eds.: Structural Change in Banking
(Irwin, Homewood: IL).
[31] B. Bernanke and A. Blinder. 1988. Credit, Money, and Aggregate Demand. American Eco-
nomic Review, 78(2), 435-439.
[32] B. Bernanke and M. Gertler. 1989. Agency Costs, Net Worth and Business Fluctuations.
American Economic Review, 79, 14-31.
[33] B. Bernanke and M. Gertler. 1995. Inside the Black Box: The Credit Channel of Monetary
Policy Transmission. Journal of Economic Perspectives, 4, 27-48.
[34] B. Bernanke, M. Gertler and S. Gilchrist. 1996. The Financial Accelerator and the Flight to
Quality. The Review of Economics and Statistics, 78, 1-15.
[35] B. Bernanke, M. Gertler, and S. Gilchrist. 1999. The financial accelerator in a quantitative
business cycle framework. Handbook of Macroeconomics, volume 1c: 1341–93. Elsevier.
[36] M. Blair. 2010. Financial Innovation, Leverage, Bubbles and the Distribution of Income.
Review of Banking and Financial Law, 30, 225-311.
[37] R. Blundell and S. Bond. 1998. Initial Conditions and Moment Restrictions in Dynamic
Panel Data Models. Journal of Econometrics, 87(2), 115–143.
[38] C. Borio and R. McCauley. 1996. The Economics of Recent Bond Market Yield Volatility.
Economic Papers. No. 45. Basel: Bank for International Settlements.
[39] C. Borio and H. Zhu. 2012. Capital Regulation, Risk-Taking, and Monetary Policy: A Miss-
ing Link in the Transmission Mechanism? Journal of Financial Stability, 8(4), 236-251.
[40] J. Boyd and E. Prescott. 1986. Financial Intermediary-Coalitions. Journal of Economic The-
ory, Elsevier, vol. 38(2), 211-232.
[41] J. Boyd and G. De Nicoló. 2005. The Theory of Bank Risk-Taking and Competition Revisited.
Journal of Finance, 60(3), 1329-1343.
202
[42] C. Calomiris. 1995. Financial Fragility: Issues and Policy Implications. Journal of Financial
Services Research, 9, 241-257.
[43] C. Calomiris and J. R. Mason. 2004. Credit Card Securitization and Regulatory Arbitrage.
Journal of Financial Services Research, 26(1), 5-27.
[44] M. I. Cambon and R. Losada. 2014. Competition and structure of the mutual fund industry
in Spain: the role of credit institutions. Spanish Review of Financial Economics, forthcoming.
[45] T. Campbell and W. A. Kracaw. 1980. Information Production, Market Signaling, and the
Theory of Financial Intermediation. Journal of Finance, 35, 863- 882.
[46] J. Campbell and G. Mankiw. 1989. Consumption, Income, and Interest Rates: Reinterpreting
the Time Series Evidence in O.J. Blanchard and S. Fischer, eds., NBER macroeconomics
annual. Vol. 4. Cambridge, MA: MIT Press, 185–216.
[47] J. Cao and G. Illing. 2015. “Interest Rate Trap”, Or: Why Does the Central Bank Keep
the Policy Rate Too Low for Too Long Time? Scandinavian Journal of Economics, 117(4),
1256-1280.
[48] C. Carlstrom and T. Fuerst. 1997. Agency Costs, Net Worth, and Business Fluctuations: A
Computable General Equilibrium Analysis. American Economic Review, 87, 893-910.
[49] C. Carroll and W.Dunn. 1997. Unemployment Expectations, Jumping (S,s) Triggers, and
Household Balance Sheets. NBER Chapters, in: NBER Macroeconomics Annual 1997, 12,
165-230.
[50] F. Caselli, G. Esquivel and F. Lefort. 1996. Reopening the Convergence Debate: A New
Look at Cross-Country Growth Empirics. Journal of Economic Growth, 1(3), 363-89.
[51] S.G. Cecchetti, J. Gyntelberg and M. Hollanders. 2009. Central Counterparties for Over-
the-Counter Derivatives. BIS Quarterly Review, September.
[52] N. Chen. 2001. Bank Net Worth, Asset Prices, and Economic Activity. Journal of Monetary
Economics 48, 415–436.
[53] I. Christensen and A. Dib. 2008. The Financial Accelerator in an Estimated New-Keynesian
Model. Review of Economic Dynamics, 11, 155–78.
[54] L.J. Christiano, M. Eichenbaum and S. Rebelo. 2011. When is the Government Spending
Multiplier Large? Journal of Political Economy, 119(1), 78-121.
203
[55] L.J. Christiano, M. Eichenbaum and C.L. Evans. 2005. Nominal Rigidities and the Dynamic
Effects of a Shock to Monetary Policy. Journal of Political Economy, 113(1), 1-45.
[56] L.J. Christiano, R. Motto, and M. Rostagno. 2014. Risk Shocks. American Economic Review,
104(1), 27-65.
[57] L.J. Christiano, R. Motto, and M. Rostagno. 2008. Shocks, structures or monetary policies?
The Euro Area and US after 2001. Journal of Economic Dynamics and Control, 32(8), 2476-
2506.
[58] L.J. Christiano, R. Motto, and M. Rostagno. 2003. The Great Depression and the Friedman-
Schwartz hypothesis. Federal Reserve Bank of Cleveland, Proceedings, 1119-1215.
[59] M. Cihak, A. Maechler, K. Schaeck, and S. Stolz. 2009. Who Disciplines Bank Managers?
IMF Working Paper WP/09/272.
[60] S. Claessens, Z. Pozsar, L. Ratnovski and M. Singh. 2012. Shadow Banking: Economics and
Policy. IMF Staff Discussion Note. SDN 12/12.
[61] D. Clement. 2010. Interview with Gary Gorton. The Region - Federal Reserve Bank of
Minneapolis, December.
[62] L. Clerc, A. Derviz, C. Mendicino, S. Moyen, K. Nikolov, L. Stracca, J. Suarez and A.
P. Vardoulakis. 2015. Capital Regulation in a Macroeconomic Model with Three Layers of
Default. International Journal of Central Banking, 11(3), 9-63.
[63] T. Cordella and E.L. Yeyati. 2002. Financial Opening, Deposit Insurance, and Risk in a
Model of Banking Competition. European Economic Review, 42, 471-485.
[64] F. Covas and S. Fujita. 2010. Procyclicality of Capital Requirements in a General Equilibrium
Model of Liquidity Dependence. International Journal of Central Banking, 6(4), 137-173.
[65] V. Curdia and M. Woodford. 2011. The Central Bank Balance Sheet as an Instrument of
Monetary Policy. Journal of Monetary Economics, 58(1), 54-79.
[66] V. Curdia and M. Woodford. 2010. Credit Spreads and Monetary Policy. Journal of Money,
Credit and Banking, 42, 3-35.
[67] M.A. Davis and J. Heathcote. 2005. Housing and the Business Cycle. International Economic
Review, 46(3), 751–84.
204
[68] F. De Graeve. 2008. The external finance premium and the macroeconomy: US post-WWII
evidence. Journal of Economic Dynamics and Control, Elsevier, 32(11), 3415-3440.
[69] M. Del Negro, G. Eggertsson, A. Ferrero, and N. Kiyotaki. 2011. The Great Escape? A
quantitative evaluation of the Fed’s liquidity facilities. Staff Report no. 520.
[70] P. M. DeMarzo and D. Duffie. 1999. A Liquidity-Based Model of Security Design, Econo-
metrica, 67, 65-99.
[71] P. M. DeMarzo. 2003. Portfolio Liquidation and Security Design with Private Information,
Stanford University working paper.
[72] P. M. DeMarzo. 2005. The pooling and tranching of securities: A model of informed inter-
mediation. Review of Financial Studies, 18(1), 1–35.
[73] A. Demirgüç-Kunt, E. Detragiache, and T.Tressel. 2006. Banking on the Principles: Compli-
ance with the Basel Core Principles and Bank Soundness. IMF Working Paper WP/06/242.
[74] W. J. den Haan and V. Sterk. 2010. The myth of financial innovation and the great moder-
ation. Economic Journal, 121, 707–739.
[75] F. De Fiore and H. Uhlig. 2011. Bank Finance versus Bond Finance. Journal of Money,
Credit and Banking, 43(7), 1399-1421.
[76] G. De Nicolo, P. Honohan, and A. Ize. 2003. Dollarization of the Banking System: Good or
Bad. IMF Working Paper, WP/03/146.
[77] G. De Nicolò, G. Dell’Ariccia, L. Luc, and F. Valencia. 2010. Monetary Policy and Bank
Risk- Taking. Staff Position Note. No. 2010/09. International Monetary Fund.
[78] M. Delis and G. Kouretas. 2011. Interest Rates and Bank Risk Taking. Journal of Banking
and Finance, 35, 840-855.
[79] G. Dell Ariccia and R. Marquez. 2006. Lending Booms and Lending Standards. Journal of
Finance, 61(5), 2511-2546.
[80] D. Diamond and P. Dybvig. 1983. Bank Runs, Deposit Insurance, and Liquidity. Journal of
Political Economy 91, 401-419.
[81] D.W. Diamond and R.G. Rajan. 2012. Illiquid Banks, Financial Stability, and Interest Rate
Policy. Journal of Political Economy, 120(3), 552 - 591.
205
[82] D.W. Diamond and R.G. Rajan. 2009. The Credit Crisis: Conjectures about Causes and
Remedies. American Economic Review: Papers and Proceedings, 99(2), 606-610.
[83] A. Dib. 2010. Capital Requirement and Financial Frictions in Banking: Macroeconomic
Implications. Working Papers 10-26, Bank of Canada.
[84] I. Drumond, M. Martins and F. Verona. 2013. (Un)anticipated Monetary policy shocks in
a DSGE model with a shadow banking system. International Journal of Central Banking,
9(3), 73-117.
[85] S. Dubecq, B. Mojon, and X. Ragot. 2015. Risk Shifting with Fuzzy Capital Constraints.
International Journal of Central Banking, 11(1), 71-101.
[86] D. Duffie. 2008. Innovations in credit risk transfer: Implications for financial stability. BIS
Working paper 255.
[87] S. Eickmeier and B. Hofmann. 2013. Monetary Policy, Housing Booms, And Financial
(Im)Balances. Macroeconomic Dynamics, Cambridge University Press, 17(04), 830-860.
[88] K. Ellis, R. Michaely, and M. O’Hara. 2000. When the Underwriter Is the Market Maker:
An Examination of Trading in the IPO Aftermarket. Journal of Finance 55 (3), 1039–74.
[89] O. Entrop, C. Memmel, B. Ruprecht and M. Wilkens. 2012. Determinants of bank inter-
est margins: Impact of maturity transformation. No 17/2012, Discussion Papers, Deutsche
Bundesbank, Research Centre.
[90] E. Faia. 2010. Credit risk transfers and the macroeconomy. ECB Working Paper 1256.
[91] L. H. Fang. 2005. Investment Bank Reputation and the Price and Quality of Underwriting
Services. Journal of Finance 60 (6), 2729–61.
[92] G. Favara and J. Imbs. 2015. Credit Supply and the Price of Housing. American Economic
Review, 105(3), 958-992.
[93] F. Ferrante. 2015. A Model of Endogenous Loan Quality and the Collapse of the Shadow
Banking System. Finance and Economics Discussion Series 2015-021.
[94] Financial Crisis Inquiry Commission. 2010. The Shadow Banking and the Financial Crisis.
Staff report. FCIC.
[95] J. D. M. Fisher. 2007. Why Does Household Investment Lead Business Investment over the
Business Cycle? Journal of Political Economy, 115(1), 141–68.
206
[96] I. Fisher. 1933. The Debt Deflation Theory of Great Depressions. Econometrica, 1, 337-357.
[97] FSB (Financial Stability Board). 2011. Shadow Banking: Strengthening Oversight and Reg-
ulation. Recommendations of the Financial Stability Board (Basel: Bank for International
Settlements).
[98] V. Gabriel, P. Levine, J. Pearlman and B. Yang. 2010. An Estimated DSGE Model of the
Indian Economy. NIPE WP29/2010.
[99] P. Gaggl and M.T. Valderrama. 2010. Does a Low Interest Rate Environment Affect Risk
Taking in Austria? Monetary Policy and the Economy, 4, 32-48.
[100] J. Gali and P. Rabanal. 2004. Technology Shocks and Aggregate Fluctuations: How Well
Does the RBC Model Fit Postwar U.S. Data? NBER Macroeconomics Annual 2004, 225-288.
[101] L. Gambacorta. 2009. Monetary Policy and the Risk-Taking Channel. BIS Quarterly Review.
Bank for International Settlements. December.
[102] A. Gavazza. 2011. Demand spillovers and market outcomes in the mutual fund industry.
RAND Journal of Economics, RAND Corporation, vol. 42(4), pages 776-804.
[103] J. Geanakoplos. 2010. The leverage Cycle. NBER Macroeconomic Annual 2009, 24, 1-65.
[104] J. Geanakoplos and A. Fostel. 2008. Leverage Cycles and the Anxious Economy. American
Economic Review, 98, 1211-1244.
[105] N. Gennaioli, A. Shleifer, and R. Vishny. 2013. A model of shadow banking. The Journal of
Finance, 68, 1331 - 1363.
[106] A. Gerali, S. Neri, L. Sessa, and F. M. Signoretti. 2010. Credit and banking in a DSGE
model of the Euro Area. Journal of Money, Credit and Banking, 42(6), 107–141.
[107] M. Gertler. 2010. Macroeconomics in the wake of the financial crisis. Journal of Money,
Credit and Banking, 42(6), 217–219.
[108] M. Gertler and P. Karadi. 2011. A model of unconventional monetary policy. Journal of
Monetary Economics, 58, 17–34.
[109] M. Gertler and N. Kiyotaki. 2011. Financial intermediation and credit policy in business cycle
analysis. Handbook of Monetary Economics, volume 3A, chapter 11, 547–599. Elsevier.
207
[110] M. Gertler, N. Kiyotaki, and A. Queralto. 2012. Financial crises, bank risk exposure and
government financial policy. Journal of Monetary Economics, 59, S17-S34.
[111] F. Giavazzi and A. Giovannini. 2010. Central Banks and the Financial System. Working
Paper. No. 16228. Cambridge, MA: National Bureau of Economic Research.
[112] S. Gilchrist, V. Yankov and E. Zakrajsek. 2009. Credit Market Shocks and Economic Fluctu-
ations: Evidence from Corporate Bond and Stock Returns. Journal of Monetary Economics,
56(4), 471-493.
[113] C.A.E. Goodhart, A.K. Kashyap, D.P. Tsomocos, and A.P. Vardoulakis. 2012. Financial
regulation in general equilibrium. NBER Working Paper 17909.
[114] G. Gorton and J. G. Haubrich. 1990. The Loan Sales Market, in George Kaufman, ed.:
Research in Financial Services 2, 85-135.
[115] G. Gorton and A. Kahn. 1993. The Design of Bank Loan Contracts, Collateral, and Rene-
gotiation. NBER Working Papers 4273.
[116] G. Gorton and A. Metrick. 2010. Regulating the Shadow Banking System. Brookings Papers
on Economic Activity, 261–97.
[117] G. Gorton and A. Metrick. 2012. Securitized Banking and the Run on Repo. Journal of
Financial Economics, 104(3), 425-51.
[118] G. Gorton and G. G. Pennacchi. 1995. Banks and loan sales. Marketing nonmarketable
assets. Journal of Monetary Economics, 35, 389-411.
[119] S. I. Greenbaum. 1986. Securitization, Asset Quality and Regulatory Reform. BRC working
paper no. 147, Northweatern University, Evanston, IL.
[120] S. I. Greenbaum and Anjan V. Thakor. 1987. Bank Funding Modes: Securitization versus
Deposits, Journal of Banking and Finance 11, 379-401.
[121] R. Greenwood, S. Hanson, and J. Stein. 2012. A Comparative-Advantage Approach to Gov-
ernment Debt Maturity. Harvard Business School Working Paper 11-035.
[122] J. Greenwood, Z. Hercowitz and P. Krussel. 1997. Long-Run Implications of Investment-
Specific Technological Change. American Economic Review, 87(3), 342-362.
[123] B. Greenwald and J. Stiglitz. 1993. Financial Market Imperfections and Business Cycles.
The Quaterly Journal of Economics, 108, 77-114.
208
[124] S. Haltenhof, S. Lee and V. Stebunovs. 2014. The Credit Crunch and Fall in Employment
During the Great Recession. Journal of Economic Dynamics and Control, 43, 31-57.
[125] F.A. Hayek. 1939. Profits, Interest, and Investment. London: Routledge, Kegan Paul.
[126] V. Heideken. 2009. How Important are Financial Frictions in the United States and the Euro
Area?. Scandinavian Journal of Economics, 111(3), 567-596.
[127] A. C. Hess and C. W. Smith Jr. 1988. Elements of Mortgage Securitization. Journal of Real
Estate Finance and Economics 1, 331-346.
[128] B. Gonzalez-Hermosillo, C. Pazarbasioglu and R. Billings. 1997. Determinants of Banking
System Fragility: A Case Study of Mexico. Staff Papers, International Monetary Fund, Vol.
44, 295 - 314.
[129] B. Hobijn and F. Ravenna. 2010. Loan securitization and the monetary transmission mech-
anism. Working Paper.
[130] B. Holmstrom and J. Tirole. 1997. Financial Intermediation, Loanable Funds and the Real
Sector. The Quarterly Journal of Economics, 112, 663-691.
[131] M. Hutchison and K, McDill. 1999. Are All Banking Crisis Alike? The Japanese Experience
in International Comparison. Journal of the Japanese and International Economies, 13(3),
155-180.
[132] M. Horvath. 2000. Sectoral Shocks and Aggregate Fluctuations. Journal of Monetary Eco-
nomics, 45(1), 69–106.
[133] M. Hume and A. Sentance. 2009. The Global Credit Boom: Challenges for Macroeconomics
and Policy. Journal of International Money and Finance, 28(8), 1426-1461.
[134] M. Iacoviello. 2005. House Prices, Borrowing Constraints and Monetary Policy in the Busi-
ness Cycle. American Economic Review, 95, 739-764.
[135] M. Iacoviello and S.Neri. 2010. Housing Market Spillovers: Evidence from an Estimated
DSGE Model. American Economic Journal, 125-164.
[136] V. Ioannidou, S. Ongena, and J.L. Peydrò. 2015. Monetary Policy, Risk-Taking, and Pricing:
Evidence from a Quasi-Natural Experiment. Review of Finance, 19(1), 95-144.
[137] International Monetary Fund. 2009. Global Financial Stability Report: Navigating the Fi-
nancial Challenges Ahead. International Monetary Fund, Washington D.C.
209
[138] D.M. Jaffee and T. Russell. 1976. Imperfect Information, Uncertainty, and Credit Rationing.
The Quarterly Journal of Economics, 90(4), 651-66.
[139] T. Jappelli and M. Pagano. 1989. Consumption and Capital Market Imperfections: An
International Comparison. American Economic Review, 79(5), 1088–1105.
[140] U. J. Jermann and V. Quadrini. 2012. Macroeconomic effects of financial shocks. American
Economic Review, 102(1), 238–271.
[141] G. Jiménez, S. Ongena, J.L. Peydrò, and J. Saurina. 2014. Hazardous Times for Monetary
Policy: What Do Twenty-Three Million Bank Loans Say About the Effects of Monetary
Policy on Credit Risk-Taking? Econometrica, Econometric Society, 82(2), 463-505.
[142] A. Justiniano, G. Primiceri and A. Tambalotti. 2013. Household leveraging and deleveraging.
Federal Reserve Bank of New York, Staff Reports 602.
[143] A. Justiniano, G. Primiceri and A. Tambalotti. 2011. Investment Shocks and the Relative
Price of Investment. Review of Economic Dynamics, 14(1), 101-121.
[144] A. Justiniano, G. Primiceri and A. Tambalotti. 2010. Investment Shocks and Business Cycles.
Journal of Monetary Economics, 57(2), 132-145.
[145] A. K. Kashyap, J.C. Stein and D.W. Wilcox. 1993. Monetary Policy and Credit Conditions:
Evidence from the Composition of External Finance. American Economic Review, 83(1),
78-98.
[146] A.K. Kashyap and J.C. Stein. 1995. The impact of monetary policy on bank balance sheets,
Carnegie-Rochester Conference Series on Public Policy, 42(1), 151-195.
[147] A.K. Kashyap and J.C. Stein. 2000. What Do a Million Observations on Banks Say About
the Transmission of Monetary Policy? American Economic Review, 90 (3), 407-428.
[148] M. C. Keeley. 1990. Deposit Insurance, Risk, and Market Power in Banking. American
Economic Review, 80(5), 1183–200.
[149] C. Kindleberger. 1978. Manias, Panics, and Crashes. New York: Basic Books.
[150] R.P. Kishan and T.P. Opiela. 2000. Bank Size, Bank Capital, and the Bank Lending Channel.
Journal of Money, Credit, and Banking, 32 (1), 121-141.
[151] N. Kiyotaki and J. Moore. 2005. Financial deepening. Journal of the European Economic
Association, 3(2/3, Papers and Proceedings), 701–713.
210
[152] N. Kiyotaki and J. Moore. 1997. Credit Cycles. Journal of Political Economy, 105, 211-248.
[153] T. Kobayashi. 2008. Incomplete Interest Rate Pass-Through and Optimal Monetary Policy.
International Journal of Central Banking 4(3), 77-118.
[154] L. Laeven and R. Levine. 2009. Bank Governance, Regulation and Risk Taking. Journal of
Financial Economics, 93(2), 259-275.
[155] A. T. Levin , F. M. Natalucci and E. Zakrajsek. 2004. The magnitude and cyclical behavior
of financial market frictions. Finance and Economics Discussion Series 2004-70, Board of
Governors of the Federal Reserve System.
[156] G. Liu and N.E. Seeiso. 2011. Business Cycle and Bank Capital Regulation: Basel II Pro-
cyclicality. Working Paper No. 221, University of Stellenbosch.
[157] Z. Liu, P. Wang and T. Zha. 2010. Do credit constraints amplify macroeconomic fluctuations.
Federal Reserve Bank of Atlanta, Working Paper 2010-01.
[158] M. Lopez, F. Tenjo, and H. Zarate. 2011. The Risk-Taking Channel and Monetary Trans-
mission Mechanism in Colombia. Ensayos sobre Política Económica, Banco de la Republica
de Colombia, 29(64), 211-234.
[159] A. Maddaloni and J.L. Peydrò. 2011. Bank Risk-Taking, Securitization, Supervision, and
Low Interest Rates: Evidence from Lending Standards. Review of Financial Studies. 24(6),
2121-2165.
[160] A. Maechler, S. Mitra, and D. Worrell. 2007. Decomposing Financial Risks and Vulnerabili-
ties in Eastern Europe. Working Paper. No. 248. International Monetary Fund.
[161] K. Mannasoo and D. Mayes. 2009. Explaining Bank Distress in Eastern European Transition
Economies. Journal of Banking and Finance, 33 (2), 244- 253.
[162] Martin-Oliver and J. Saurina. 2007. Why do Banks Securitize Assets? XV Spanish Finance
Forum Conference Proceedings.
[163] B. Markovic. 2006. Bank Capital Channels in the Monetary Transmission Mechanism. Bank
of England Working Paper No. 313.
[164] P. McCulley. 2007. Teton Reflections. PIMCO Global Central Bank Focus, Au-
gust/September.
211
[165] R. McKinnon. 2013. The Unloved Dollar Standard: From Bretton Woods to the Rise of the
[People’s Republic of] China. New York: Oxford University Press.
[166] R. McKinnon and G. Scnabl. 2004. The East Asian Dollar Standard, Fear of Floating, and
Original Sin. Review of Development Economics, 8 (3), 331–360.
[167] R. Meeks, B. Nelson, and P. Alessandri. 2014. Shadow Banks and Macroeconomic Instability.
Bank of England Working Papers 487.
[168] C.A. Meh and K. Moran. 2010. The Role of Bank Capital in the Propagation of Shocks.
Journal of Economic Dynamics and Control, 34, 555–576.
[169] H. Mehran and A. Thakor. 2011. Bank Capital and Value in the Cross-Section. Review of
Financial Studies, Society for Financial Studies, 24(4), 1019-1067.
[170] A. Mian and A. Sufi. 2012. What explains high unemployment? The aggregate demand
channel. NBER Working Papers 17830
[171] A. Mian and A. Sufi. 2011. House Prices, Home Equity-Based Borrowing, and the US House-
hold Leverage Crisis. American Economic Review, 101(5), 2132-56.
[172] A. Mian and A. Sufi. 2009. The Consequences of Mortgage Credit Expansion: Evidence from
the 2007 Mortgage Default Crisis. The Quarterly Journal of Economics, 124, 1449-1496.
[173] H. P. Minsky. 1992. The Financial Instability Hypothesis. Economics Working Paper Archive
wp_74, Levy Economics Institute.
[174] T.D. Nadauld and M.S. Weisbach. 2012. Did securitization affect the cost of corporate debt?
Journal of Financial Economics, 105(2), 332-352.
[175] E. A. Özsuca and E. Akbostanci. 2016. An Empirical Analysis of the Risk Taking Channel
of Monetary Policy in Turkey. Emerging Markets Finance and Trade, 52(3), 589-609.
[176] T. Paligorova and J.A.C. Santos. 2017. Monetary Policy and Bank Risk-Taking: Evidence
from the Corporate Loan Market. Journal of Financial Intermediation, 30, 35-49.
[177] C. Parlatore. 2016. Fragility in Money Market Funds: Sponsor Support and Regulation.
Journal of Financial Economics, 121(3), 595-623.
[178] E. Perotti. 2010. Systemic liquidity risk and bankruptcy exceptions. CEPR Policy Insight
No. 52.
212
[179] G. G. Pennacchi. 1988. Loan sales and the cost of bank capital. The Journal of Finance, Vol.
XLIII, No. 2, 375-396.
[180] J. Poschmann. 2012. The shadow banking system: survey and typological framework. Work-
ing paper on global financial markets no. 27.
[181] Z. Pozsar. 2011. Institutional cash pools and the Triffin dilemma of the U.S. banking system.
IMF Working Paper 11/190.
[182] Z. Pozsar. 2008. The Rise and Fall of the Shadow Banking System. Moody’s Economy.com
– Regional Financial Review, July, at www.economy.com/sbs.
[183] Z. Pozsar, T. Adrian, A. Ashcraft, and H. Boesky. 2013. Shadow banking. Economic Policy
Review, Federal Reserve Bank of New York, issue Dec, 1-16.
[184] R. G. Rajan. 2006. Has Finance Made the World Riskier? European Financial Management,
12(4), 499–533.
[185] D. Roodman. 2009. How to Do Xtabond2: An Introduction to “Difference” and “System”
GMM in Stata”. Stata Journal, 9(1), 86-136.
[186] R.J. Rosen. 2009. Too much right can make a wrong: Setting the stage for the financial
crisis. Financial Institutions and Market, Palgrave Macmillan US, 37-69.
[187] J. J. Rotemberg and M. Woodford. 1997. An Optimization Based Econometric Framework
for the Evaluation of Monetary Policy. NBER Macroeconomics Annual, 12, 297-346.
[188] G.W. Schwert. 1989. Why Does Stock Market Volatility Change Over Time? Journal of
Finance, 44 (5), 1115-1153.
[189] H.S. Shin. 2009. Securitisation and financial stability. Economic Journal, 119, 309–332.
[190] F. Smets and R. Wouters. 2007. Shocks and Frictions in U.S business cycles: A bayesian
DSGE approach. American Economic Review, 97(3), 586-606.
[191] F. Smets and R. Wouters. 2003. An Estimated Dynamic Stochastic General Equilibrium
Model of the Euro Area. Journal of European Economic Association. 1(5), 1123-1175.
[192] D. Solomon. 2010. Firm Leverage, Household Leverage and the Business Cycle. MPRA Paper
26504.
213
[193] J.C. Stein. 2010. Securitization, Shadow Banking, and Financial Fragility. Daedalus 139(4),
41 -51.
[194] J.E. Stiglitz and A. Weiss. 1981. Credit Rationing in Markets with Imperfect Information.
American Economic Review, 71, 393-410.
[195] B. Tabak, M. Laiz and B D Cajueiro. 2010. Financial Stability and Monetary Policy – the
Case of Brazil. Banco Central Do Brazil Working Paper Series 217.
[196] J. Taylor. 2012. Monetary Policy and the Next Crisis. Wall Street Journal. 4 July.
[197] J. Taylor. 2009. The Financial Crisis and the Policy Responses: An Empirical Analysis
of What Went Wrong. Working Papers. No. 14631. Cambridge, MA: National Bureau of
Economic Research.
[198] J.B. Taylor, ed. 1999. Monetary Policy Rules. NBER Book Series Studies in Business Cycles.
Chicago: University of Chicago Press.
[199] H. Thomas. 2001. Effects of asset securitization on seller claimants. Journal of Financial
Intermediation, 10, 306-330.
[200] J. Tirole. 2011. Illiquidity and All Its Friends. Journal of Economic Literature, 49, 287–325.
[201] A. Turner. 2012. Economics After the Crisis: Objectives and Means. London: MIT Press.
[202] S. J. Van den Heuvel. 2002. Does Bank Capital Matter for Monetary Transmission?, FRBNY
Economic Policy Review, 8(1), 259-265.
[203] A. Viale, J. W. Kolari, D.R. Fraser. 2009. Common Risk Factors in Bank Stocks. Journal of
Banking and Finance, 33 (3), 464-472.
[204] F. Verona, M.M.F. Martins and I. Drumond. 2013. (Un)anticipated Monetary Policy in a
DSGE Model with a Shadow Banking System. International Journal of Central Banking,
9(3), 78-124.
[205] S. Williamson. 1987. Financial Intermediation, Business Failures and Real Business Cycles.
Journal of Political Economy, 95, 1196-1216.
[206] M. Woodford. 2010. Financial intermediation and macroeconomic analysis. Journal of Eco-
nomic Perspectives, 24(4), 21–44.
214
[207] E. L. Yeyati and A. Micco. 2003. Concentration and Foreign Penetration in Latin American
Banking Sectors: Impact on Competition and Risk. IDB Working Paper 499.
[208] L. Zhang. 2009. Bank Capital Regulation, Lending Channel, and Business Cycles. Deutsche
Bundesbank Discussion Paper Series 1: Economic Studies 2009-33.
[209] S.P. Zeldes. 1989. Consumption and Liquidity Constraints: An Empirical Investigation. Jour-
nal of Political Economy, 97(2), 305–46.
215