Oxford Review of Economic Policy , 2025, 41, 591–615 https://doi.org/10.1093/oxrep/graf036 Article Understanding and modelling structural economic change as a dynamic resource creation process—an application to low-carbon transitions Dimitri Zenghelis,* Hector Pollitt,** Jean-François Mercure,*** and Frank W. Geels**** ∗University of Cambridge, UK, email: dz320@cam.ac.uk ∗∗University of Cambridge, UK, email: hbp22@cam.ac.uk ∗∗∗University of Exeter, UK, email: J.Mercure@exeter.ac.uk ∗∗∗∗University of Manchester, UK, email: frank.geels@manchester.ac.uk This research was supported by The Productivity Institute, funded by the UK Economic and Social Research Council ( grant number ES/V002740/1) . Abstract The global economy is constantly undergoing large-scale structural change. Policy-makers face the task of deciding how technological change is directed, where limited public funds should be spent, and how to in- duce private investment. This paper provides an understanding of structural economic change as a dynamic process of resource creation, applying findings to the transition to a low carbon economy. The aim of this paper is to provide an explicit account of the mechanisms and processes that drive and steer innovation and adoption of new technologies, networks, and behaviours. These include strategic complementarities, expec- tation formation, and the role of multiple actors. We distinguish between structural change within a sector and knock-on cascades and spillovers across sectors, both of which play crucial roles in driving transformational change. Understanding these processes then allows us to develop the analytical tools to guide appropriate policy choices. With the support of a simple modelled illustration, we show that structural change cannot be analysed using a static optimization approach based on historic data. Economic models remain valuable in providing insights, but have fundament al limit ations when making predictions in the context of reinforcing feedbacks and increasing returns. Instead, they are best used to inform risks and steer policy in the direction suited to achieve strategic objectives. We find that a variety of models, complemented by a range of qualita- tive and non-modelling analytical approaches, with different strengths and weaknesses, can help to identify tipping points and allow scenario analysis to articulate risks and opportunities, thereby guiding policy choice. Keywords: economic modelling, macroeconomics, productivity, economics of innovation, structural change, policy, cli- mate policy, low-carbon investment, energy systems modelling. JEL codes: B41, Q43, E1, H2, Q55, Q48, C63, C73, O44, B22 I. Introduction This paper focuses on the role of macroeconomic modelling and supportive analytical tools in guiding policy-makers in making strategic decisions. It suggests that the entire class of current equilibrium-based models faces severe analytical constraints in providing useful economic analy- sis. It argues that the marginal, incrementalist views embodied in these models is no longer appro- priate for addressing the challenges of the twenty-first century ( Mercure et al., 2021 ; Peñasco et al., 2021 ; Sharpe and Lenton, 2021 ) . The paper reviews the current economic theories and modelling tools that are used for policy appraisal. It illustrates key shortcomings and provides examples of how new approaches, building on evolutionary and complexity economics literature, could © The Author( s) 2025. Published by Oxford University Press on behalf of The Oxford Review of Economic Policy Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( https://creativecommons.org/licenses/by/4.0/) , which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. https://doi.org/10.1093/oxrep/graf036 mailto:dz320@cam.ac.uk mailto:hbp22@cam.ac.uk mailto:J.Mercure@exeter.ac.uk mailto:frank.geels@manchester.ac.uk https://creativecommons.org/licenses/by/4.0/ 592 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels complement existing tools. However, in the face of deep uncertainty, it emphasizes that these tools are more usefully deployed in assessing risks and opportunities of policy strategies, rather than in creating scenarios of the distant future with potentially false precision. The starting point is to acknowledge that the modern economy is in a state of constant evo- lution. Companies develop new products and build new industries. They find ways to make ex- isting products better or more cheaply. Consumers choose the products they buy based on a set of evolving tastes. Policy-makers have multiple objectives, including driving innovation, shaping markets, promoting economic efficiency, and safeguarding competitiveness. Policy-makers also care about food and energy security, inflation, jobs, and skills. They seek to develop competi- tive industries and encourage urban development in more liveable and productive cities, while minimizing regional inequality. And they must do this while addressing the challenges of limited public finances associated with complex fiscal and financial dynamics, pre-existing infrastructure, security demands, and changing demographic pressures. Furthermore, the global economy is undergoing at least three revolutionary general purpose technology transformations, generating new competitive races for firms and economies. These are ( i) clean energy and electrification for the net-zero transition, ( ii) artificial intelligence ( AI) , big data, automation, and connectivity, and ( iii) genetic engineering. The clean energy transition replaces large operational costs ( e.g. mining or drilling for fuel) by moderate upfront capital costs ( e.g. solar panels, electric vehicles) and overall reduces energy input costs in most sectors ( Mercure and Pollitt, 2025 ) . AI will improve the management of systems and system change, particularly in cities, energy, transport, land, and water ( Stern and Romani, 2023 ) . Genetic engineering can help solve ageing-related disease burdens such as cancer and Alzheimer’s, and will improve food supplies. These transformations bring many opportunities, but also risks for social stability. For example, large proportions of jobs could be automated ( Frey and Osborne, 2017 ; PWC, 2017 ) with knock- on effects for other parts of the economy ( Ford, 2015 ) . A rapid clean-energy transition threatens many occupations and could strand the productive capital of entire regions whose economic output originates from outdated brown industries. Reducing age-related health burdens, in the context of low birth rates, may require the adoption of higher retirement ages. Unlike the other transformations, policy-makers are expected to drive the clean-energy tran- sition. However, many of the same questions are being asked across all three transformations: expected benefits and costs, in terms of potential opportunities crowded out, and new opportu- nities created by assessing full risk-adjusted returns to investment. Decision-makers face the task of deciding how technological change is directed, where limited public funds should be spent, how to induce private investment, how to secure reliable supply chains and benefit from trade, and how to re-tool and reskill workers in declining communities. Especially in the case of clean energy, policy-makers are already highly engaged. They are rolling out new infrastructure, reforming electricity markets, adopting new efficiency standards and reg- ulations, introducing renewable energy requirements, and supporting R&D and deployment, as well as relaxing planning laws to encourage development of clean infrastructure. These actions will have long-lasting effects. They will drive structural change. The merits of these policy actions must be evaluated holistically, in terms of systemic change. Flexibility and optionality in the face of future change, together with steering the path for devel- opment, may be more important than forecasting the future. Potential risks may be more relevant to decision-makers than mean expected outcomes. The challenge is substantial, but policy-makers can use economic tools and models as aids for understanding their policy options ( Hallegatte et al., 2024 ) . This, however, requires a set of modelling tools that captures the key system properties adequately. The following section describes the nature of structural change and why it poses a challenge to conventional economic modelling. Section III discusses how current models could be improved to account for structural change and section IV considers the wider assessment framework. Section V concludes. Understanding and modelling structural economic change 593 II. The nature of structural change ( i) The challenge to economics We define structural change as systemic transitions in particular sectors and industries, which in- volve the emergence and diffusion of new behaviours and technologies ( which are both shaped by institutions) , as well as the overthrow and decline of existing industries and institutions. The ef- fects are neither small nor local, and involve transformations in technologies, behaviours, and mar- kets that underlie any economy’s productive capacity. Structural change does not satisfy Samuel- son’s correspondence principle, meaning that the addition and subsequent removal of an input shock will not return the economy to its initial state. Structural change is irreversible and often self-reinforcing. It is closely linked to the concept of path dependency, which states that future outcomes are shaped by historic conditions ( Aghion et al., 2014 ) . What this describes is the presence of ‘tipping points’. A tipping point delineates the move of a system from one reinforcing characteristic to another when there are multiple equilibria. When a tipping point is reached, the positive feedbacks for an emerging innovation become so strong that inertia is overcome and a transition from one system to another becomes self-reinforcing and potentially irreversible. This traces a non-linear pathway to a new equilibrium ( Lenton, 2020 ) . The latter affords the possibility that the economy might be best thought of as being on a constant dynamic path, where new structures are created, and the economy never settles in an equilibrium ( Arthur, 1994 ) . In such circumstances, conventional approaches relying on a unique equilibrium as the basis of analysis will not be fit for purpose. Fundamentally, there are two kinds of problem in economics: � how to allocate efficiently existing resources and structures, � how to create effectively new resources and structures. Based on the presumed principle that human wants are infinite, but resources are finite, much of the field of economics is defined around the problem of resource allocation.1 In part this reflects the roots of economic theory at a time when scarcity of all items, including food, was a major societal threat ( Malthus, 1798 , ch. 1; Galbraith, 1998 , ch. 9) . The original versions of endoge- nous growth theory attempted to address resource creation, but excluded the possibility of tipping points and transformational change ( Aghion and Howitt, 1992 ; Romer, 1990 , 1994 ) . Standard neoclassical theory and New Keynesian theory are still dominated by resource allocation ( Stern, 2018 ) . Post-Keynesian theory adopts a more nuanced position based on underemployment of resources and accumulation of capital, but still focuses ultimately on resource allocation. Both environmental and ecological economics are also designed to assess questions of resource alloca- tion rather than creation.2 More recently, interest in resource creation economics has grown, not least due to the trans- formational societal changes described in section I . Increased use of industrial strategy has also demanded a form of economics that recognizes the strategy’s policy aims. Much of the theory has been derived from Schumpeter ( 1934) , from which the field of evolutionary economics was developed ( e.g. Nelson and Winter, 1982 , 2002 ) . Evolutionary economics has now largely been subsumed into the more recent field of complexity economics ( Anderson, 1972 ; Arthur, 1999 ; Mercure, 2022 ) , which links social and technological evolution with emergent macro-level prop- erties that form from micro-level interactions. Resource creation arises from increasing returns and learning by doing, and from network effects and cascades involving the interaction of multiple actors with strategic complementarities. For example, some of the most innovative and influential ideas emerge not from rigid plans, but from well-connected networks and environments which generate a cascade of new thinking and behaviours ( Cleevely, 2025 ) . At the same time, strategic complementarities arise when the returns to a particular investment depend on the investments other agents make. For example, the returns to investing in clean technologies increase when other agents are also investing in 1 See https://mises.org/mises-daily/four-hundred-years-dynamic-efficiency 2 Although Marx considered capital accumulation, detailed questions of resource creation were first addressed sys- tematically by the original institutional economics at the start of the twentieth century, building on Veblen ( 1898 ; 1899 a ,b ; 1900 ) . However, the field stagnated for multiple reasons ( Hodgson, 2004 ) and was eventually developed into an approach based heavily on resource allocation. https://mises.org/mises-daily/four-hundred-years-dynamic-efficiency 594 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels Figure 1: Two conceptions of public intervention, looking at marginal change or transformative change the same technologies or compatible networks, pushing down costs and opening new market opportunities. These features can give rise to multiple equilibria, and tipping points at which the economy will shift phase from one equilibrium to another. Anticipating and mapping the pathway to such tipping points is a core element of strategic policy frameworks. It is the move across tipping points from one equilibrium to another that constitutes structural change. Interventionist policy may be needed to push the economy over relevant tipping points ( Aghion et al., 2014 ; van der Ploeg and Venables, 2025 , this issue) . Tipping points do not necessarily arise in the process of resource creation, but they may do where, for example, lower costs attract demand, and rising demand causes costs to decline further. They may also arise in situations of network externalities. By contrast, they are explicitly ruled out in optimal resource allocation models that neglect resource creation. Understanding structural change therefore requires a rejection of the notion that the economy is guaranteed to find the one single efficient equilibrium, provided we solve all market failures. There are, in fact, too many possible outcomes to be able to count them ( Kauffman, 2000 ) . Analysing any scenario that may involve tipping points must therefore abandon the efficient static equilibrium and allow for a resource creation perspective. This perspective extends the dis- tinction between static efficiency and dynamic efficiency, noting that what may seem statically inefficient turns out to be dynamically efficient, because of resources that are created. This dis- tinction lies at the epistemological core of the two types of economic problem. The notion is captured by the phrase ‘kickstarting the clean innovation machine’. Here, ambi- tious early policy efforts, designed to divert resources to less productive sectors ( such as renew- ables in the early days of development) may initially tie up factors of production. However, they also generate economies of scale and increasing returns, allowing new sectors to outcompete the incumbents without the need for further policy intervention ( Aghion et al., 2014 , p. 6) . The old system is thus replaced by a new, superior one ( Figure 1 ) . The implication is critical for policy-making ( illustrated in Figure 1 ) . In the resource allocation perspective, price mechanisms are used to correct for market failures and allow the economy to find its efficient frontier. But intervention is politically costly, especially if it is needed indefinitely. In contrast, putting the economy onto a new, innovative trajectory does not require indefinite intervention, but it may require a front-loaded effort before self-propelling change is induced. In that perspective, upfront investment is critical in mainstreaming new technologies, infrastructures, or systems. This is precisely what occurred with solar and wind energy, and electric vehicles ( EVs) , where early policy support has crowded in private investment, brought costs down, and created a new range of equilibria for energy and mobility systems ( Nemet, 2019 ; Grubb et al., 2021 b ; RMI, 2023 a ,b ) . The essential feature of this kind of endogenous growth is that knowledge increases the produc- tive potential of factor inputs so that traditional diminishing returns are overcome. Where once adding more machines per employee might have been expected to lead to a less-than-proportional Understanding and modelling structural economic change 595 Figure 2: Cost of energy Source : Way et al. , ( 2022) . increase in output, now the gains from learning, experimentation, and innovation derived from using more machines allow unit costs to decline and whole-economy output to expand more than proportionately. This has been observed at times of rapid technological change since the eighteenth century ( Freeman and Louçă, 2001 ; Mercure, 2022 ) . It is crucial to understanding the resource creation dynamics in a technology transition. Allocative efficiency is still important for policy that leads only to marginal ( i.e. non- transformational) changes that are unlikely to cross tipping points ( Mercure et al., 2021 ) . How- ever, for problems of transformational change, policy-makers must ask what new markets, tech- niques, and behaviours could be created or induced by their choice of policy design. In this context, a focus on dynamic efficiency and effectiveness is more appropriate. ( ii) The clean energy transition as an example of dynamic structural change Fossil fuel production and use technologies are mature; the petrochemical and internal combus- tion engine revolution occurred in the early 1920s. Therefore, most of the innovation gains have already been exploited ( Farmer and Lafond, 2016 ) . There have been technological breakthroughs in extraction ( such as fracking) , distribution, refining, and burning, but the cost of fossil fuel based energy has not fallen substantially over the last century ( Figure 2 ) . More importantly, fossil fuel systems operate on a cost basis and, being extractive, require a substantial and ongoing supply of costly labour. Expansion is thereby subject to diminishing returns to scale and increasingly expensive and complex resource extraction systems are needed. Accessing more fossil fuels on the margin means tapping ever more remote sources and deploying more exotic extraction pro- cesses, such as tar sands, shale oil, and deep offshore oil. It also utilizes expensive resources, most notably labour, in systematically and repeatedly transporting, refining, and burning the fuels nec- essary to supply uninterrupted global energy. The operational supply chain for utilizing the fossil fuel system is both expensive and subject to increasing costs on the margin. A key feature of the creation economy is that many modular, replicable, and scalable new tech- nologies ( such as solar PV and batteries) are characterized by powerful economies of scale, both in discovery and production as businesses fabricate and distribute things more efficiently, thereby dramatically lowering costs ( Geels et al., 2021 ) . Not all new zero-carbon technology turns out to have lower costs than a carbon-intensive alternative, but those that fall into this category generally will, especially when integrated into a vastly expanded electrified energy system. The remainder of this section explores some of the ways in which the cost reductions for clean technologies have been realized. ( iii) Reinforcing technological feedbacks Powerful reinforcing technological feedbacks can alter significantly the state of the world. As such, it is not possible to anticipate the full extent of the economic changes resulting from a 596 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels transition. However, these processes are unlikely to happen of their own accord and often need intervention to provide a trigger. Two key types of self-propelling feedbacks matter for growth and transition: ( i) the innovation-diffusion reinforcing cycle, which can be further split into learning effects, economies of scale, and network effects; and ( ii) the transformation of the ‘design space’ of available inputs for making new technologies. The first feedback indicates that investing in and demand for new technology induces cost reductions, which induces investment in and demand for more of the same technology. The sec- ond feedback suggests that new low-cost technologies or inputs also change the ‘design space’, the universe of inputs and resources within which entrepreneurs operate, opening possibilities for new product development that were not possible before ( Stankiewicz, 2000 ) . This is an important source of irreversibility. For example, the AI revolution at affordable costs is changing fundamen- tally how businesses design their products and services, and there is no way back to the pre-AI economy. These processes involve complementarity effects that may cross sectors ( see below) . For exam- ple, solar power increases incentives to develop batteries and smart energy management systems, which become cheaper, thereby encouraging solar power and EV purchases. At the same time, the incumbent technological system has an innate advantage with incumbent firms possessing extensive supply networks, deep pockets, and enhanced lobbying power ( Geels, 2025 ) as well as high network effects and switching costs embedded in physical infrastructure and technologies ( Hidalgo et al., 2007 ; Aghion et al., 2014 ; Klitkou et al., 2015 ; Seto et al., 2016 ) . The interaction between economies of scale to production and discovery causes tipping points, where the drive to the new optimum eventually overcomes the inertia related to the old system. Transformational change therefore may take a period of time, before a ‘critical mass’ is reached, but when it happens the process may be quick. ( iv) Agglomeration, clustering, and early-mover advantage Real-world transitions do not unfold in an abstract space, but involve competitive struggles be- tween firms and countries. Creation economics and the presence of increasing returns provides incentives for a geographical race to develop new markets. This growing realization is becoming a key driver of cost reducing reinforcing feedbacks. Countries that successfully invest early can corner future green product markets. Moreover, a firm’s choice over whether to innovate ‘clean’ or ‘dirty’ products is influenced by the practice of the countries where its inventors and researchers are located and the technology and supply adjacencies of existing sectors ( Aghion et al., 2016 ; Hidalgo et al., 2007 ) . Therefore, investing early in knowledge clusters and supply networks will determine medium- and long-term outcomes. Even economies that are too small to influence global technology trends must account for endogenous responses and increasing returns to scale. Evidence shows that countries that suc- cessfully invest early in low carbon, resource efficient capabilities have greater success in diversi- fying into future green product markets ( Hidalgo et al., 2007 ; Mealy and Hepburn, 2020 ; Mealy and Teytelboym, 2022 ) . Increasing returns also generate spatial clustering and agglomeration at the spatial level. This imparts a competitive advantage to moving early to develop new clusters and supply lines. Because early movers gain at the expense of laggards, this may be a zero-sum game. China’s early strategic investment in EVs, batteries, and solar PV has allowed Chinese firms to corner the market in fast-growing global sectors. Clean sectors accounted for 40% of China’s GDP growth in 2023 ( Carbon Brief, 2024 ) . Other countries are waking up to the dangers of falling behind. The Inflation Reduction Act in the US channelled hundreds of billions of dollars into clean sectors. European car makers ( who were slow to adopt battery technologies) are lobbying for import tariffs in a vain effort to stem the inevitable tide of change. ( v) Induced innovation applies to institutions and human behaviour Structural change relates not only to innovation in technologies, but also to innovation in institutions and behaviour ( Geels and Turnheim, 2022 ) . There are associated socio-political Understanding and modelling structural economic change 597 effects: as new technologies become cheaper, consumers start to favour them and politi- cians begin to back them. This triggers corresponding change in political institutions, indus- try lobbying power, consumer tastes, and social norms that amplify the effects of any policy intervention. Consequently, a technological transition relates to wider systems change concerning syn- chronous complementarities across a range of key assets. The assets might be produced, human, intangible, or natural. They include infrastructures, complementary technologies, training and skills, institutions, markets, and business models. Inertia, innovation, and path dependency characterize behaviour over periods of large-scale structural change ( Barnes et al., 2004 ; David, 1994 , Aghion et al., 2014 ) . Societies build on what they have, unless and until given reason not to by virtue of the rush to shift to a superior system. Consequently, to model processes of change, we cannot assume a unique equilibrium based on a combination of optimization and fixed ‘behavioural’ coefficients. The economy emerges from its behaviours and technologies as ‘an expression’ of choices, expectations, and investments made along the way. Choices and expectations change the character of the economy. ( vi) Shifts in expectations and risk Shifts in expectations are a specific example of changes in social institutions but are important enough to consider separately. Expectation formation plays a critical role in the presence of in- creasing returns to scale, strategic complementarities, and interactions between actors, since what is expected to happen can help push the economy across the tipping points ( van der Ploeg and Venables, 2005 , this issue) . For example, the assumption that reducing greenhouse gas emissions is costly ( often by model design) can itself be costly. The challenge is that this narrative, to the extent that it is believed and forms expectations, can prove self-fulfilling. Conventional analyses based on allocative efficiency routinely lead to policy delay ( Peñasco et al., 2021 ; Grubb et al., 2021 b ) . Delayed investment postpones the reinforcing feedbacks between deployment and cost-reductions ( van der Meijden and Smulders, 2017 ) . This then creates a vicious rather than virtuous circle involving inertia and locking in to dynamically inferior incumbent technologies. By contrast, a technology breakthrough that eventually reduces energy costs will generate a more favourable political environment to support the technology and limit the alternatives. It is also likely to generate consumer tastes favouring the new technology. For example, if your neighbour has a solar PV system on their roof that cuts their bills, you are more likely to purchase such a unit and vote for a party that subsidizes its adoption. Social norms, politics, and other institutions tend to move in lockstep. New lobby groups will emerge, disrupting the incumbents. Consumer tastes and preferences will adapt. The clean energy transition comes with innovation in ideas, technologies, business models, and changes in preferences that reinforce one another and trigger a virtuous circle with sustainable reinforcing dynamics. Feedback loops become important drivers of the transition ( see Table 1 ) and drive expectations of its scale. The result is that the transition, once under way, is often both faster than expected and unstop- pable, catching many people by surprise, rendering cost estimates for mitigation grossly over- stated, and delivering cheaper and more efficient energy ( Mercure et al., 2025 a ) . Policy plays a crucial role in guiding behaviour and anchoring shared expectations for effective coordina- tion and collective action. For example, announcing a date for ending internal combustion en- gine ( ICE) car sales, while enforcing interim industry mandates ( e.g. 50% sales by 2030) , can crowd in investment by lowering the risk premium under the certainty of a market for EVs in the future. Expectations formation is thus a key part of the reinforcing feedback process. Economists using tools based on allocative efficiency not only get the future wrong, they make it wrong, because of the power attached by decision-makers to often biased forecasts ( Krugman, 1991 ) . Narratives can create resources by changing behaviour and inducing innovation ( Akerlof and Shiller, 2009 ; Shiller, 2017 ) . This reinforces the finding that economic models are at their best when they communicate ‘insights not numbers’ ( Peace and Weyant, 2008 ) . 598 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels Table 1: Factors causing increasing returns and reinforcing feedbacks that models should consider Factor Brief description Reinforcing technological feedbacks Learning effects With deployment, lessons are learned on how to manufacture, distribute, install, run, and maintain equipment more efficiently. This can be represented in models by using ‘learning curves’. Economies of scale in production and distribution Unit costs also fall because of scale effects accrued from larger production and distribution networks, especially if high fixed costs can be shared across wider production. These effects may also be included using learning curves. Network and coordination effects Combinatorial technologies and network and coordination effects mean there are greater advantages of moving in tandem with others, with spillovers, shared benefits, and more potential to generate new ideas. Diffusion-based model approaches can capture some of these effects. The ‘design space’ effect The deployment of new products with rapidly declining costs opens up new possibilities for entrepreneurs to design cost-competitive products that were not possible or competitive previously ( Stankiewicz, 2000 ) . Institutions and human behaviour Social and institutional feedbacks Social norms provide a shared understanding of acceptable actions and are sustained through social interactions ( Ostrom, 2000 ) , and form the basis of legislative action ( Posner, 1997 ) . They evolve in tandem with technology. Relaxing assumptions about optimizing behaviour only partially addresses this issue; diffusion-based modelling goes further. Consumer tastes Consumers routinely influence each other, leading to crowd effects, and changing consumption patterns ( Mercure, 2018 ) . Changing consumer tastes are critical in the diffusion of new technologies and diffusion-based modelling captures some of these effects ( Knobloch and Mercure, 2016 ; Mercure et al., 2021 ) . Expectations and risk Risk perceptions Changed perceptions of risks also raise the cost of capital for investment in incumbent technologies which risk being stranded, towards new technologies which are eventually seen as less risky and more productive. Cross-sectoral structural change Reinforcing technological feedbacks In the same ways as described above, technological effects can spill over between sectors. Conventional macroeconomic models can capture some of these effects ( for example, through prices) , but some are too uncertain to model accurately. Institutions and human behaviour Institutional behavioural patterns are likely to cross over between sectors quickly, as ideas are identified and imitated. Shifts in expectations and risk Risk perceptions may also change quickly across sectors. For example, financial instability usually draws in sectors not directly affected by initial shocks. ( vii) Cross-sectoral structural change Finally, there is also the need to distinguish between: 1. transitions and structural changes within particular sectors, which is where product inno- vation dynamics unfold, and 2. knock-on effects, spillovers, cascade effects between sectors ( which may be more important than the sectoral transitions for their macroeconomic effects) . There are several mechanisms in which innovation in one sector can create structural change in another sector. For example, falling prices of IT equipment led to rapid adoption, new products, and new business models across the whole economy. Components used in IT equipment also started to appear in other products such as cars, with increased use leading to economies of scale and further price reductions in the IT sector. In some cases, company structures have been altered to look like those in the IT industry. This generates a reinforcing feedback which drives rapid and hard-to-predict change. Falling prices of IT equipment lead to rapid adoption of new products. The rising demand for these products causes increases in the demand for IT equipment encouraging further economies of scale and yet further falls in prices of IT equipment. Understanding and modelling structural economic change 599 Looking forward, the widespread use of batteries in EVs and intelligent AI systems to manage real time demand responsiveness will likely lead to further cost reduction and technical improve- ments that, in turn, can increase the use of batteries in homes ( to store self-generated electricity) , in grids ( to provide back-up capacity) , and even in heavy trucks, trains, ships, and short-haul airplanes ( RMI, 2023 b ) . Green hydrogen and other zero-carbon technologies could follow the same path. It is possible that two-thirds of current global emissions could have cost-effective low carbon substitutes by 2030 ( Systemiq, 2021 ) . The previous ‘great waves’ of innovation are defined as the development of general purpose technologies that had transformational impacts across all economic sectors ( Freeman and Perez, 1988 ; Freeman and Louca, 2001 ; Perez, 2010 ) . Examples include steel, oil, and modern com- munications. Even if the development of these technologies could have been predicted, it would not have been possible to predict the full range of impacts across the economy at the time. We currently face the same challenge with clean energy, AI, and genetic engineering. III. Modelling structural change Having assessed the purpose of an intervention and its focus, an assessment can be made on the ap- propriate method of policy analysis. In this section, we focus on when to apply static optimization- based approaches and when fundamental uncertainty associated with path-dependent structural processes makes precise expected future costs and benefits unknowable. This distinction reflects the nature of ( often cumulative) uncertainty in the real world. Forecasts can only be made as contingent on a set of uncertain assumptions which illuminate risks and contingent probabilities. When resource creation is involved, there are simply too many ways policy can, positively or neg- atively, influence the direction of economic development to be able to rank them all and identify an optimal course of action. The number of possible pathways increases combinatorially with each successive innovative development, and parsing through these futures may take the whole of humanity longer than the duration of human history, let alone a single computer ( Mercure and Pollitt, 2025 , p. 42; Mercure, 2022 , p. 151) . Therefore, when we talk about ‘modelling structural change’ it is important to be clear what we are not saying. We are not merely advocating better modelling of sectors and explaining what is required to do it well. In principle, one can model any scenario one can logically envisage, alongside the processes underlying it, but if economies of scale, strategic complementarities, and non-linear network effects dominate, then the impact of initial conditions and impulses, together with subsequent errors, will tend to cumulate and reinforce each other, leading to non-linearly diverging pathways and outcomes ( Mercure, 2022 ) . One simply cannot therefore expect an accurate forecast in the way that one can with a sin- gle equilibrium in a marginal change situation. There are many excellent models that formally articulate dynamic and tipping point relationships ( see Van der Ploeg and Venables, 2025 , this issue) . But the value of these models lies not in that they might get the future right ( Zenghelis, 2019 ) . What they do very well is show how sensitive the future might be to particular types of early decisions. As Paul Romer put it: ‘Instead of suggesting that we can relax because policy choices don’t matter, it suggests to the contrary that policy choices are even more important than traditional theory suggests.’ 3 The challenge in assessing policies with resource creation potential is therefore considerable. Most current models deny the possibility for resource creation and therefore may produce answers to the wrong question for policy-makers. However, by drawing on evolutionary and complexity theory, the modelling community is now developing tools that could give important insights. A critical question thus concerns the degree to which increasing returns in technology deploy- ment are transformational and whether they can lead to self-propelling transitions. This varies by sector of energy use and emissions but, in the clean energy transition, solar energy and EVs are cases in point. Model development and analysis are needed to determine whether shifts to technologies like solar and EVs facilitate each other, to the extent that an economy-wide self-propelling transition 3 See Romer’s commentary on conditional optimism, October 2018: https://paulromer.net/conditional-optimism- technology- and- climate/ https://paulromer.net/conditional-optimism-technology-and-climate/ 600 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels could occur away from traditional combustion and traditional industries towards a green and smart, digitalized and intelligent global economy. As noted previously, past industrial revolutions invariably involved constellations of technologies cooperating in creating new economic sectors with high returns ( Freeman and Perez, 1988 ; Freeman and Louca, 2001 ; Perez, 2010 ) . Typically, the emergence of new sectors ( e.g. IT) changes the relationships between the existing ones as they increase their productivity, For example, digitalization can increase the productivity of agriculture, AI can increase the productivity of retail, and EVs, with their very low marginal costs of operation, can substantially reduce operating costs of logistics and distribution. These are questions that the modelling com- munity is not yet grappling with. However, tools to study this are beginning to exist in complexity and evolutionary economics ( Balint et al., 2017 ; Hepburn et al., 2025 , this issue) . ( i) Modelling structural change Models are used for policy analysis in lieu of carrying out macro-level experiments. The im- portance of computer models for policy analysis has grown substantially ( Süsser et al., 2021 ) and models are now used to support policy-making in most developed countries. Models are also used to develop policy narratives ( Beck, 2017 ) , including a range of different narratives of the impacts of AI on jobs ( Frey and Osborne, 2017 ) and the persistent narrative that climate policy inevitably leads to economic costs ( Mercure et al., 2016 ; Pollitt et al., 2024 ) . Models can be thought of as for- malized instruments for testing possible courses of action ( Ellenbeck and Lilliestam, 2019 ) . They provide a consistent simplifying framework for analysis and comparison and impose a discipline on the modeller to articulate assumptions and parameters. The need for new analytical models follows from improvements to economic theory. Equi- librium economic models based on the standard theory are designed to study marginal change occurring on an otherwise fixed economic structure. This premise makes the weighing of costs ver- sus benefits, against a counterfactual ( do nothing) scenario, very straightforward. In doing so, they often understate the benefits and overstate the costs, largely because the costs ( e.g. investment) are better known than the benefits ( e.g. productivity growth benefits of innovation induced by the policy change) . For example, ex-post analysis of the costs of environmental policies tend to find that they are significantly cheaper than ex-ante analyses suggest, because models fail to cap- ture the broad range of innovations in technologies, behaviours, and institutions that may occur because of strong and coherent policies ( Grubb et al. 2021 a , 2023 ; Mercure et al., 2016 , 2021 ) . This issue can also be expressed by saying that current models focus on resource allocation but ignore resource creation ( e.g. Dixon and Jorgensen, 2012 ) . They therefore face severe con- straints that limit their usefulness in the key questions relevant to policy-makers. By assuming a fixed economic structure ( technologies, productive capital, skills) , they cannot provide an ade- quate representation of any structural change beyond the marginal. Models struggle to represent institutions, innovation, and human behaviour in a realistic way. The creation economics view is an extension of the theory of endogenous technical change and is supported by early work on transitions of socio-technical regimes, including the multi- level perspective that has a long history in characterizing social transformations ( Geels, 2002 ; Dechezleprêtre et al., 2022 ) . The approach analyses the contextual societal factors that enable or impede the diffusion of innovations from niches to dominance. Models struggle with such reinforcing feedbacks because the multiple equilibria they imply makes them unstable. Nevertheless, understanding the process of change is increasingly seen as an essential part of informing decisions regarding current key economic issues. Sharpe ( 2023) describes ‘the problem of allocation is that of how to divide the pie, the problem of creation is that of how to make the pie in the first place’. A key aspect of market creation and the early exploitation of S-shaped deployment curves implies that looking backward at deployment and price trends in the early period running up to the inflection point affords a poor guide to future developments ( Figure 3 ) . The relative shares of the old and new products appear slow to change initially; for example, the left-hand side of the figure shows that the new product gains less than 5% market share in the first six periods. If this process is interpreted as a linear response to policy signals, change will always be slow. Understanding and modelling structural economic change 601 Figure 3: S-shaped deployment of network technology Source : Grubb ( 2018) . However, the right-hand side of the figure shows that the relative growth rates are maintained throughout the early part of the period, leading to near-exponential growth in the new product, and pushing the old product out of the market at a similar rate. Using a linear approximation for such a non-linear process belies the degree to which experts failed, over the last decade, to predict the speed of the transition to renewables and EVs. The adoption of batteries, EVs, and photovoltaics maps neatly onto such an illustrative representation. ( ii) Modelling costs means modelling innovation As discussed in section II , structural change is often ( although not always) driven by innovation. When modelling innovation, it is important to make the distinction between product and process innovation. Product innovation involves the design of new types of product that may stimulate consumer demand. Process innovation involves improving the efficiency of producing existing products ( Utterback and Abernathy, 1975 ) . AI will lead to substantial process innovation ( e.g. reducing labour costs) but will likely also lead to an extensive range of new products. Product innovation makes it impossible to predict the net economic impacts of AI because of the underlying fundamental uncertainty ( Kauffman, 2000 ) . In contrast, the clean energy transition is to some extent easier to model because many of the required technologies exist already and the effects of process innovation are highly predictable ( Farmer and Lafond, 2016 ) . When combined, the two types of innovation lead to the creation of new markets, institutions, behaviours, and social norms ( Zenghelis et al., 2024 ) . The scale of the innovation inevitably leads to knowledge spillovers into other sectors ( Dechezleprêtre et al., 2017 ) , potentially generating a further round of structural change. For example, advances in clean lab processes necessary for computer chip fabrication have also played a role in the evolution of silicon solar wafers used in photovoltaic technologies, and vice versa. Standard Computable General Equilibrium ( CGE) , Dynamic Stochastic General Equilibrium ( DSGE) models, and the first-generation climate Integrated Assessment Models ( IAMs) such as DICE ( Nordhaus, 1992 , 2016 ) do not usually include any representation of technology beyond an implicit formation in price elasticities. Productivity growth is exogenously determined, and therefore innovation is pre-written and independent of policy. Grey productivity growth happens ‘for free’, while green growth is assumed costly. These models therefore abstract technology from the problem, missing the defining nature of the transition. Current IAMs ( and energy system 602 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels models) normally include a wide array of technologies; however, total factor productivity ( TFP) nevertheless remains exogenous in most cases ( Dellink et al., 2017 ) . Thus, current equilibrium-based models do not adequately capture the drivers of innova- tion and technological change in their relationship with economic growth. The solar revolution ( Nemet, 2019 ) in particular has alerted modellers to the issue of endogenous or induced inno- vation and some models have attempted to incorporate innovation. However, it is reported that technical difficulties mean that even in these models the option for endogenous technology is of- ten disabled ( Grubb et al., 2023 ) . This is not unexpected. Incorporating processes with increasing returns in resource allocation algorithms causes ambiguities and multiple solutions in an opti- mization framework, a reflection that in the real world technological change is path-dependent. Excluding endogenous process innovation makes the models inadequate for assessing climate policy or economic growth, given that they do not model the key mechanisms involved. Current climate-economy models also suffer from the paradox that, while TFP growth is as- sumed to originate from general innovation taking place across sectors for free, benefiting eco- nomic growth, low-carbon innovation, for its part, is assumed economically costly and does not benefit the economy. Predictions from these models are therefore biased towards incumbent and existing technologies. Much of the discussion about the use of methane as a ‘bridging fuel’ results from mis-representing the rapid cost declines of solar and wind energy in these models. Properly accounting for path dependencies in models would show that early intervention in the innovation system and leadership in market creation could be desirable, even under high discounting assumptions. Identifying potential systemic tipping points can allow targeted policies to have outsized effects, such that climate goals can be met with smaller policy interventions. Farmer et al. ( 2019) focused on how to identify, model, and trigger sensitive intervention points to transition rapidly to a post-carbon society. These sensitive intervention points ( SIPs) can exist in technological, psychological, socio-political, or economic domains. To summarize, escaping lock-in to high-carbon technologies could save substantial resources in the long run ( Way et al., 2022 ) , but models without endogenous technological progress miss this relationship. Inadequate modelling of innovation has effectively worked against policies that could accelerate the clean energy transition and benefit the economy ( Stern, 2022 ; Stern et al., 2022 ; Zenghelis et al., 2024 ) . Research into tipping points and potential cascade effects could help to yield a list of ‘watch fors’ and early warning signals to guide policy choice in strategic decision-making. ( iii) Alternatives to conventional modelling The need for an alternative to equilibrium-based optimization modelling has been noted for some time ( Trutnevyte, 2016 ; Trutnevyte et al., 2019 ) , even by current modellers ( Peng et al., 2021 ) . The limitations of econometrics in periods of rapid change and technological advances are also acknowledged ( Galbraith, 1975 ; Keynes, 1938 , 1939 , p. 347; Godley, 1996 ; Lucas, 1976 ; Peters, 2019 ; Stern et al., 2022 ) . The problem is one of historical specificity ( Hodgson, 2001 ) ; how can information from one context be used to inform another? Generally speaking, modelling changing behaviour, habits, and technology adoption becomes easier as each becomes established, which attracts more and more users, and becomes the new nor- mal. As the poet Antonio Machado said: ‘Traveller, there is no path. The path is made by walking.’ In other words, new technological futures always start off more difficult than the socio-technical systems already in place, almost by definition; but their adoption can generate substantial societal returns. This basic fact is generally not recognized in models, largely because the number of pos- sible futures is too large for current methods based on equilibrium and optimization ( Mercure, 2022 ) . Current equilibrium-based models assume that: 1. resources and all of their possible allocations are always fully known and used in the best possible way; 2. resources endowments and comparative advantage are fixed and not created by the eco- nomic process; Understanding and modelling structural economic change 603 3. general innovation underpins growth and happens for free on its own in the background, but green innovation is costly and sacrifices resources. The key shortcomings of these models could be overcome if we instead assumed that: 1. the amounts of resources are uncertain, but the economy has ample reserve capacity; 2. the economy innovates and invests to create resources and capacity according to what it anticipates it will need; 3. all innovation requires investment, of which the outcome is uncertain, but when done in sectors with pervasive positive economic impact, it often increases the productivity of the economic process as a whole, although it may concurrently cause declines in output and investment in sectors becoming obsolete. The current model shortcomings can effectively be traced to a focus on optimization-based ap- proaches to allocate fixed resources; which ultimately is required to generate equilibrium results. Common assumptions with little empirical basis, such as perfect knowledge, fully flexible mar- kets, and rational behaviour/expectations may sometimes be adopted for the practical purpose that an optimization algorithm can be used to solve the problem posed. However, given that there is no analytical solution to optimization problems about ‘resource formation’ and path depen- dency, and because we want to understand change over time ( not the state of the economy at a fixed point in time) , we need models that are simulating, not optimizing. This means models that answer ‘what if’ questions empirically without constraining results to equilibrium outcomes. A move toward simulation-based approaches would also allow models to incorporate positive- reinforcing dynamics, including increasing returns to scale ( as most businesses report in reality) and, critically, innovation, technological evolution, and resource creation. Table 1 provides ex- amples of such effects that models should cover, based on the discussion in section II . Model categories that accommodate this dynamic description include system dynamics ( SD) , agent-based approaches ( ABMs) , and potentially dynamic macroeconometric input–output mod- els. All these approaches accept the possibilities of fundamental uncertainty, heterogeneous pop- ulations, and bounded rationality. These modelling approaches are only now being applied to the clean energy transition ( Pollitt, 2019 ) , but in the examples where they have been applied, we can already see that they give qual- itatively different insights to those from equilibrium, optimization-based models ( e.g. Mercure et al., 2019 ) . Two critical insights stand out in relation to the clean energy transition: first, that there need not be net costs to reducing greenhouse gas emissions and second that carbon pricing as a sole instrument may not be the most effective approach for emissions reductions ( Mercure et al., 2016 ; Rosenbloom et al., 2020 ; Pollitt et al., 2024 ) . Moreover, to the extent that pricing and policy effort matter, their optimal dynamic application may be relative, targeted early on at sectors with large innovation potential. Rather than applying a uniform carbon price to pick off the cheapest abatement options first and then ramping up, a disproportionate policy effort should be applied to bringing key sectors to competitiveness up front. It is important to differentiate between conceptual models that are used to demonstrate system properties and empirical models that are used for real-world policy evaluation ( Holland, 1995 , p. 156; Romanowska et al., 2021 , p. 236) . Because of their scale and computing requirements, ABMs are likely to be more suitable as conceptual models. For example, Nelson and Winter ( 1982) demonstrated the innovation process at firm level in such a model. Conceptual models may have a role in policy analysis, but empirical models are needed most. Empirical models must have broad coverage ( global, whole economies) with substantial sec- toral disaggregation to capture properly the technology effects. Currently, the most widely used simulation-based tools that do this are E3ME ( Mercure et al., 2025 b ) and GINFORS ( Lutz et al., 2010 ) models. The GEM-E3-FIT CGE model ( Fragkiadakis et al., 2020 ) has adapted several assumptions to bring it more into line with the theories of structural change. ( iv) A demonstration of modelling structural change The Future Technology Transformations ( FTT) modelling tool was designed to simulate technol- ogy decision-making within individual sectors ( Mercure, 2012 ) . Here we focus on the model for 604 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels Figure 4: Simulation results for CO2 emissions in the power and chemicals sectors ( GtCO2 ) the power sector ( Mercure et al., 2014 ) . In the model, whenever new power generation capacity is required ( either to replace old capacity or to expand total capacity) investors choose between a large number of generation technologies.4 The model assesses investor preferences by compar- ing levelized costs between technology options that are subject to learning effects. Distributions of these costs indicate local variabilities as well as the heterogeneous character of investors. The diffusion of technology follows a set of coupled non-linear differential equations, sometimes re- ferred to as ‘Lotka-Volterra’ or ‘replicator dynamics’ equations. These equations represent the potential for larger or better-established industries to capture market share and influence investor preferences. The FTT model is coupled to the E3ME macroeconometric model ( Cambridge Econometrics, 2022 ; Mercure et al., 2025 b ) . E3ME is a non-equilibrium, post-Keynesian model that allows for the possibility of spare economic capacity. However, E3ME’s conventional macroeconometric specification based on elasticities does not allow for hysteresis effects; for sectors not covered by the FTT model ( e.g. chemicals) , the modelling approach is thus more basic. An illustrative scenario is assessed in which a global carbon tax of $80/tCO2 is introduced globally for a decade over the period 2030–40. In all other years there is no carbon tax. Figure 4 shows global emissions in the scenario compared to a baseline case with no carbon tax. Results are presented for the power sector, modelled using the evolutionary FTT approach. Results are also shown for the chemicals sector, which uses the conventional elasticity-based econometric approach in the E3ME model. Results for the power sector, show that half of the reduction in emissions is sustained by 2060, even though climate policy is abandoned after 2040 ( and renewables hit capacity constraints in some countries) . The policy, while it was in action, has altered the path of development. In contrast, in the chemicals sector, the limitations of using a standard econometric approach become apparent. When the carbon tax is removed, emissions revert quickly to baseline levels as if the historic legacy of any induced innovation was wiped out. This means in practice that a pricing policy would need to be in place indefinitely to reduce emissions permanently. This result stems from the implicit assumption that policy action induces temporary behavioural change, but the economic structure remains the same. The evolutionary FTT model illustrates two connected reasons for the persistence of the policy effect. The carbon tax allows solar and wind power to gain market share and achieve critical mass and, relatedly, learning effects accelerate cost reductions in solar and wind power, which cannot be reversed. The model results highlight the need to consider technology-induced structural change explic- itly in modelling exercises. Relying on econometric and other coefficient-based approaches can- not address questions of structural change that we would expect to see in an economic transition. This simple illustration offers a lower bound for the dynamic effect, as it covers only structural 4 The model includes 22 technologies: nuclear, oil generators, coal plants, gas turbines, hydro, ocean, solar PV, solar CSP, wind onshore, wind offshore, geothermal, and a few variants. Understanding and modelling structural economic change 605 changes in the power sector. It does not cover the full dynamic knock-on effects of spillovers, cascades, and network effects between sectors, nor the wider socio-behavioural responses which, if included, would be expected to widen the disparity between the two approaches. ( v) Recognizing the uncertainty in our models There is uncertainty at each stage in the modelling process, including the data and definitions used, forward-looking assumptions, behavioural assumptions, and parameter estimation. Current programmes like the Energy Modelling Forum attempt to reduce uncertainty in climate-economy models by comparing ensembles of runs from different models ( e.g. Harmsen et al., 2021 ; van Vuuren et al., 2020 ) . However, the models participating in the comparisons are based on similar data and optimization structures, and have generally collectively avoided the issues of innovation and resource formation discussed above. As a result, the comparisons have demonstrated well where the modelling frontier lies with optimization models and, correspondingly, the degree to which their results naturally flow from the above assumptions over resource use ( Mercure et al., 2019 ) . However, the methodology in comparison exercises often self-selects because of the way they are designed. For example, the strict output requirements for submitting scenarios to the IPCC often disqualify models that stray from the standard ( Peters et al., 2023 ) . Moving forward, a more pluralist approach to modelling is required for all questions that re- late to structural change and its inherent uncertainty ( Hepburn et al., 2025 , this issue) . As further progress is made to improve and enhance the modelling frameworks, models will be able to cap- ture better the potential economic benefits of structural change, alongside the costs that are well identified already. Even so, no single model will ever tell the full story of economic transition and structural change, largely because there are so many ways it could be achieved. For example, there are many ways that models could adhere to both empirical observation and theories of structural change by altering unobservable behavioural relationships. Given the strong possibility of positive- reinforcing feedbacks, a small change in a key parameter can lead to different outcomes where the difference grows over modelling time spans, in contrast with general equilibrium modelling which usually returns to a pre-defined path after disturbances. The outcomes themselves matter less than what this tells us about the process of delivering them and the possible scenarios they point to. Economic modelling is a critical part of articulating assumptions and enhancing understand- ing of systemic relationships. However, models’ role in prediction diminishes when it comes to structural change. This suggests that a variety of models should be complemented by a range of qualitative and non-modelling analytical approaches, with different strengths and weaknesses, designed to illuminate possible pathways and identify potential tipping points ( Hepburn et al., 2025 , this issue) . The key is to gain an understanding of the mechanisms and processes behind structural change, even if one cannot accurately predict them ( except on the margin) . Prediction matters only in an illustrative and contingent way: ‘if we do policy X, then the result will be something like Y’, and not in an absolute way, ‘Y is going to happen’. This is already recognized in some formal policy formation processes ( e.g. European Commission, 2015 ) , in which models do not constitute a comprehensive assessment. The next section discusses the role of modelling in policy formation. IV. Using models to guide policy for bringing about structural change Models play an important role in the making of policy. The purpose of this section is to inves- tigate the role of models in bringing about structural change where this has clear, anticipated economic and social benefits. When there are multiple equilibria and tipping points, one cannot proceed in the manner used by economists seeking to answer questions relating to marginal per- turbations within a unique equilibrium set-up. Instead, in these circumstances, the focus needs to shift from predictions based on analysis of incremental change to understanding the process of change, through illustrative scenario analysis. The purpose of scenario analysis is to describe how, and in what ways, the economy might look once it has crossed a particular tipping point 606 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels ( recognizing that in a path-dependent process driven by reinforcing feedbacks, the likelihood of any specific scenario occurring is near zero) , and then to ask how we can steer the economy to increase the likelihood that we get to such a tipping point. Section III showed that models which satisfy the Samuelson correspondence principle will never entertain the possibility of a change in the underlying structure of the model, nor articulate where tipping points exist and what happens when one crosses them. They are therefore inadequate to address questions of structural change. However, we discussed in section II how it is impossible for a model to know all future outcomes in an evolving economy. The key insights required from economists are therefore those which describe key risks and opportunities associated with that evolutionary process. These findings suggest that a ‘market shaping’ approach may be the more appropriate rationale for policy where structural change is a potential outcome ( Kattel et al., 2018 ; Sharpe and Lenton, 2021 ; Mercure et al., 2021 ) . This requires a shift in our thinking about how to take decisions. Previous work has suggested moving from cost–benefit analysis to risk–opportunity analysis ( see below) , where costs and benefits are assumed to be unknowable with any great degree of precision because of the multitude of potential outcomes. As these futures cannot be enumerated exhaus- tively, one cannot form reliable probability distributions, and therefore cannot calculate expected values without admitting such huge uncertainties that results can become meaningless ( Keynes, 1921 ; Knight, 1921 ; Shackle, 1949 , 1952 ; Weitzman, 2009 ; Mercure et al., 2021 ; Mercure, 2022 ) . One must embrace the fact that the outcomes involve unquantifiable risks and opportunities. The EU’s analysis of the Energy Efficiency Directive in 2018 provided an example of the ben- efits of exploratory modelling. A simulation model suggested that energy efficiency measures could benefit GDP, whereas an equilibrium model estimated costs. Discussion about model as- sumptions highlighted the importance of informing equipment producers of the need to develop increased production capacity to meet higher future demands. Although informing companies may have seemed an obvious policy strategy, it had not been covered in previous discussions. Furthermore, neither model on its own would have identified the need for clear communication, because the simulation model neglected supply constraints and the optimization model assumed perfect knowledge. ( i) Scenario analysis can tackle evolutionary change Analysing risks and opportunities related to evolutionary economic transformation processes re- quires thought-through scenario approaches. Given that models can anticipate process innova- tion but not predict what new goods and services could be developed, there is a limit to what is knowable about the future. Due to this limit to knowability, identifying an optimal policy is not possible. Our best chance then lies with scenario analysis to draw the contours of what we consider reasonably encompasses the worst and best case scenarios. Below, we give an overview of productive approaches to tackling non-linear feedbacks enumerated in Table 1 . Learning and diffusion effects Learning curves and diffusion dynamics, when coupled dynamically together, give rise to tipping points, where more diffusion leads to more cost reductions which lead to more diffusion and so on. Calibrating carefully to empirical data, one can identify the moment in time of cost-parity between new technologies and incumbents ( if it arises at all) . We can also identify points in time at which this movement becomes self-propelling. We can then add into the mix different types of policy instruments, to see if cost parity arises at all and, if it does, whether it is sooner or later. Identifying a tipping point is never completely unambiguous, but exploring scenario space is the way to the conditions that could give rise to one. Network and coordination effects Technologies rarely evolve in a vacuum, they typically co-evolve in constellations, for example railways and steam engines, or petrochemicals and the internal combustion engine ( Freeman and Perez, 1988 ) . As the costs and deployment of one technology reach a tipping point, those of its companions are drawn towards tipping as well. Cascading positive tipping points have been Understanding and modelling structural economic change 607 discussed ( Sharpe and Lenton, 2021 ) . In climate-relevant technologies, battery deployment for vehicles has the potential to revolutionize the power sector, household production and use of electricity, and other modes of transport. The design space effect This effect is where input costs for designing new products changes and creates new possibilities. In principle, it is not possible to predict what entrepreneurs, geeks, and scientists may invent, because doing so would amount to doing the invention ourselves. However, aggregate economic models can have sector output that responds to input prices, and they can also, in principle, have dynamically changing input coefficients. For example, if solar energy becomes cheaper than coal, electricity may replace coal as an input for many industries, requiring changing the input– output coefficients in a multi-sectoral macro-model. All equilibrium models by their definition keep input–output coefficients constant but alter prices to determine capital allocations. Mod- elling the evolving design space would require moving to dynamically evolving economic struc- tures. Doing this could become relatively uncertain, but could give a sense of how upstream in- dustries ( e.g. basic materials, agriculture, mining) react to transformations occurring downstream ( e.g. EVs replacing conventional ones, and solar replacing coal) . Social and institutional feedbacks and consumer tastes Consumers are not born with innate knowledge of all the markets and technologies they will be using over their lifetimes; they learn this from peers over the course of their lives. This leads to ex- plosive feedbacks, commonly called fads and fashions, where agents learn by interacting with their social groups. Social norms constrain behaviours in a similar way, where dominant behaviours naturally emerge ( Young, 2009 ) . This phenomenon, comparable to the diffusion of diseases in populations, can also be modelled using non-linear systems dynamics; but the uncertainty gener- ated by the non-linear processes call for a scenario approach, similar to how predictions of the Covid pandemic were made. Risk perceptions and shifts in expectations Risk is a subjective quantity, and its estimation typically takes into account a subjective set of factors while it excludes others. For example, climate transition-related risks include, in some reports, risks of stranded fossil fuel assets that could arise in a situation of rapid re-pricing of those assets. Expectations involve some degree of consensus in the beliefs of groups of agents in certain specific futures. Shifts in expectations therefore involve those agents changing their beliefs towards a new, different consensus. Modelling this naturally calls for understanding the differ- ences between different sets of plausible quantified futures that we could reasonably propose as agent expectations. From quantified differences can be drawn, for instance, financial risk assess- ments, such as on the future value of fossil fuels and the impact of those risks on the economy and broader financial system ( Mercure et al., 2018 ; Semieniuk et al., 2022 ) . ( ii) Robust Decision Making Robust Decision Making ( RDM) is a framework for making decisions under conditions of un- certainty, including uncertainty relating to structural change. Rather than searching for optimal outcomes, it seeks solutions that produce acceptable outcomes under a range of possible futures ( Guivarch et al., 2017 ) and aim to identify trigger points ( including tipping points) at which pol- icy change may be required ( Workman et al., 2021 ) . The bridging approach advocated in Geels et al., ( 2016) provides a similar framework. These approaches develop scenarios of possible future outcomes that account for the fact that risks are not independent, but correlated and prone to cascades. They are participatory in nature, drawing on both qualitative and quantitative inputs. The results from models may feed into the process ( Workman et al., 2021 ) but the models must allow for the possibility of substantive struc- tural change and not constrain results to equilibrium outcomes based on current structures. Sce- nario analysis and a market-shaping approach to policy better enable decision-makers to make 608 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels robust and risk-reducing proposals for policy in response to the key tipping points articulated in Table 1 in section III . An example in the clean energy transition would be to find strategies that are effective both in a scenario in which the costs of clean technologies remain high and the policy environment is hostile, and in another scenario in which clean energy costs fall sharply and decarbonization policy becomes ambitious. The two risks related to costs and policy are not independent, but are mutually correlated and reinforcing; one is more likely to come with the other. The literature on complexity economics has provided similar guidance for policy-makers op- erating under conditions of deep uncertainty ( Probst and Bassi, 2014 ) . Notably, the trade-off between optimality and resilience must be acknowledged ( Mercure et al., 2021 ; Hynes et al., 2022 ) and the benefits of diversity must be recognized ( Meadows, 2008 , pp. 59–160) . It must be possible to update policy on a continuous basis ( Dembo, 2021 ; Kirman, 2016 , p. 77) and trialling policy ahead of full implementation may be necessary ( Wilson, 2016 ) . ( iii) Risk–opportunity analysis When assessing the risks and opportunities of a particular investment decision at a time of struc- tural change, analysing the drivers of innovation and the reinforcing feedbacks that enable sys- temic transformation may offer better guidance to decision-makers than analysing cost estimates based on simple static analysis. Uncertainty is deep, and thus we are generally better able to steer the direction of economic transformation than to predict exactly where the destination of travel is. This approach has been called risk–opportunity analysis ( ROA) ( Grubb et al., 2021 b , 2023 ; Mercure et al., 2021 ) . A clear understanding of the changing landscape of risks and opportunities is a prerequisite for applying appropriate analytical approaches. ROA is a general framework that seeks to guide decision-making by embracing risk and uncertainty. It attempts to identify the most important risks and opportunities generated by policy action, even if this list cannot be exhaustive. For exam- ple, the net-zero transition generates at least as many new occupations and business opportunities as it eliminates. It also seeks to assess the resilience of the economy through various scenarios of transformative structural change. The goal is not necessarily to minimize uncertainty, but rather, to work with it. ROA combines the ideas of cost–benefit analysis ( CBA) with risk assessment. CBA is a specific case of ROA where the risks are known deterministically. Policy action can generate benefits but could also at the same time reduce resilience by increasing vulnerability to shocks. Policy action can also deliberately increase uncertainty to create opportunities for business investment, such as by supporting disruptive innovation. ROA thus allows analysts to bring together, under one broad decision-making framework, re- lated but distinct policy-making functions that may be classified between strategy, regulation, and accounting. Strategy involves determining where we want to go, while regulation concerns the management of existing systems to maintain their behaviour within acceptable bounds, and ac- counting analyses the allocation of existing resources. These domains of policy-making are tradi- tionally managed by independent institutions ( e.g. executive, central bank, finance and economics ministry) , but may need to work together to manage processes of structural change effectively. For example, a range of different decarbonization policies could achieve our stated goal of net-zero emissions rapidly and cost-effectively, but they could also make the financial system vulnerable to shocks ( Bolton et al., 2020 ) . Current debates over the cost-effectiveness of the net-zero transition and financial risk management are taking place independently. However, policy-making in those domains critically depends on each other. ROA offers a suitable frame- work to determine how they fit together ( Mercure et al., 2021 ) . ROA approaches remain un- derdeveloped and underemployed, and further development of applied techniques is urgently required. ( iv) Complementary analytical approaches Section III laid out the difficulties that models face when assessing structural transformation. Typical failures include excessive focus on sectors that were important historically and excessive Understanding and modelling structural economic change 609 weight put on costs that are known with greater confidence, at the expense of uncertain future benefits. However, different types of models can be used to understand processes, articulate risks, and inform choices. These approaches range from technological engineering, historical accounts of structural transitions, geographical contagion models, economic theory, systems understanding and mapping, and the formation of expectations from social psychology ( see below) . Policy-makers must understand the important dynamics in economies of scale in production and discovery, increasing returns, complementary systems technologies, social norms, and strate- gic complementarities and expectation formation. A selective array of clearly specified, well tar- geted, and properly understood models can guide more efficient investments in the right tech- nologies ( Aghion et al., 2016 ; Stern and Valero, 2021 ) . These types of more diverse approaches— blending the quantitative and the qualitative—often lead to a greater emphasis on good risk man- agement, applying the precautionary principle to robust responses which retain optionality, rather than seeking ‘optimal’ policies produced by conventional cost–benefit analysis ( Peng et al., 2021 ) . It should be clear, therefore, that we need to realign our expectations of what models can and cannot do when it comes to understanding and steering structural change. At best economic models, together with insights from other disciplines ( which can range from history to geography and social psychology) , can help to explain and compare possible processes and outcomes, thereby illuminating possibilities which matter to the decision-maker. What they will not do, given the nature of the problem, is allow us to predict what will happen ( for reasons outlined in section II ) . Unlike unique equilibrium models, whose value can be assessed by virtue of the accuracy of their forecasts, which in turn depend on the structural properties of the model, the value of mod- els investigating structural change is their insights in illuminating risk or opportunity. Forecast accuracy is inappropriate as a benchmark for assessing value of models with tipping points and reinforcing feedbacks. For example, some simple toy models are clearly unrealistic and wrong, but their insights can still be hugely valuable ( Hepburn et al., 2025 , this issue) . This is why we recommend the deployment of a broad range of models with different strengths and weaknesses. The key is to gain an understanding of the mechanisms and processes behind a structural tran- sition. Past changes in policies, institutions, and networks have resulted from complex systemic effects and spillovers. These cannot be picked up in one overall formal model. V. Conclusions The low carbon transition is well under way and, in lockstep with revolutions in AI and genetic engineering, is now unstoppable. The scale of structural change is so large, and replete with a mix of uncertainty and technological advances, that it cannot be analysed using a static opti- mization approach based on historic data. Explicit account must be taken of the processes that drive and steer innovation and adoption of new networks, including strategic complementarities, expectation formation, and the role of multiple actors. Policy-makers are responding to, and proactively managing, this change. The merits of their actions need to be evaluated holistically, in terms of systemic change, evolutionary economics, and the drivers of innovation to create resources. Strategic visioning and scenario planning by eco- nomic decision-makers is required, based on a conceptual framework and analytical toolkit that can inform policy-makers of the dynamic risks and opportunities associated with transformative change. Analysing these implications requires an interdisciplinary conceptual approach that addresses the specificities of non-marginal structural change, including technological discontinuities, sys- temic transformation, multi-actor processes, and uncertainty. It requires replacing cost–benefit analysis with analysis of risk and opportunity: a process that has already prompted financial markets to alter their asset pricing and investment decisions. Policy action to boost economic prosperity in these conditions must focus on dynamic market shaping, rather than static market failures. In line with the theories of technical discontinuities and transitions, governments seek- ing to boost productivity sustainably must strategically design and steer, rather than passively forecast, the future. This inevitably entails a role for what is currently described as ‘industrial policy’. 610 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels The task for decision-makers in the current macroeconomic environment of rapidly mounting uncertainty and structural change is therefore to maximize opportunities and minimize risk. For economists, this means adopting a more flexible and dynamic approach to modelling future sce- narios and the drivers of innovation pathways. The choice to decarbonize an economy rapidly is a fundamental strategic investment decision, centred around the management of very large risks. It cannot be construed as a marginal project appropriate for types of narrow cost–benefit analysis that are commonly in use. Policy-makers face a choice among very different growth paths with hugely different consequences. While there is still an allocation problem, this problem is dynamic and expressed in terms of assets rather than factor inputs. Investing in a range of assets ( i.e. physical, human, and natural capital) that is future-proofed and unlikely to become ‘stranded’ will support economic develop- ment. This means recognizing and understanding forward dynamics. When examining structural transitions, conventional economics has largely got things the wrong way round. Instead of treating technology as an add-on to existing economic models, the focus might be better placed on applying economics to modelling technology and social de- velopment. This might be the most powerful lesson taken from Malthus. The challenge is acute in the case of AI, because it is impossible to predict the direction of technological advances even though AI will likely introduce advances to all economic sectors. Policy flexibility will thus be key in supporting structural change that benefits all of society. For the net-zero transition, the immediate role for government is clearer because the main technolo- gies are already known. Governments must steer private investment and reinforce expectations of tangible returns from the net-zero transition. Clear and credible policy is necessary to align expectations on the direction of change and ensure policy effectiveness until the transition gains momentum, and economic costs are minimized. Being knowledge and technology driven, net-zero carbon investments will have greater scope for learning, innovation, and cost reduction than cur- rent alternative technologies and are therefore likely to generate a higher economic rate of return, stronger growth, and higher GDP in the long run. Positioning the economy to benefit from this innovation requires a considered and fit-for-purpose analytical toolkit. References Aghion , P., and Howitt, P. ( 1992) , ‘A Model of Growth through Creative Destruction’, Econometrica , 60( 2) , 323–51. https://doi.org/10.2307/2951599 — Hepburn, C., Teytelboym, A., and Zenghelis, D. ( 2014) , ‘Path Dependence, Innovation and the Eco- nomics of Climate Change’. In: R. Fouquet ( ed.) Handbook on Green Growth , Edward Elgar Publish- ing, https://doi.org/10.4337/9781788110686.00011 . https://www.lse.ac.uk/granthaminstitute/wp-content/ uploads/2014/11/Aghion_et_al_policy_paper_Nov20141.pdf — Dechezleprêtre, A., Hémous, D., Martin, R., and Van Reenen, J. ( 2016) , ‘Carbon Taxes, Path Dependency, and Directed Technical Change: Evidence from the Auto Industry’, Journal of Political Economy , 124( 1) , 1–51, https://doi.org/10.1086/684581 . Akerlof , A., and Shiller, R. J. ( 2009) , Animal Spirits: How Human Psychology Drives the Economy, and Why It Matters for Global Capitalism , Princeton, NJ: Princeton University Press. Anderson , P. W. ( 1972) , ‘More is Different’, Science , 177( 4047) , 393–96. https://doi.org/10.1126/science.177. 4047.393 Arthur , W. B. ( 1994) , Increasing Returns and Path Dependence in the Economy , Ann Arbor, MI: University of Michigan Press. https://doi.org/10.3998/mpub.10029 — ( 1999) , ‘Complexity and the Economy’, Science , 284( 5411) , 107–109. https://doi.org/10.1126/science.284. 5411.107 Balint , T., Lamperti, F., Mandel, A., Napoletano, M., and Roventini, A. ( 2017) , ‘Complexity and the Economics of Climate Change: A Survey and a Look Forward’, Ecological Economics , 138, 252–65. https://doi.org/10. 1016/j.ecolecon.2017.03.032 Barnes , W., Gartland, M., and Stack, M. ( 2004) , ‘Old Habits Die Hard: Path Dependency and Behavioral Lock- in’, Journal of Economic Issues , 38( 2) , 371–77. https://doi.org/10.1080/00213624.2004.11506696 Beck , M. ( 2018) , ‘Telling Stories with Models and Making Policy with Stories: An Exploration’, Climate Policy , 18( 7) , 928–41. https://doi.org/10.1080/14693062.2017.1404439 Bolton , P., Despres, M., Pereira Da Silva, L. A., Samama, F., and Svartman, R. ( 2020) , The Green Swan: Central Banking and Financial Stability in the Age of Climate Change , Paris: Banque de France. https://doi.org/10.2307/2951599 https://doi.org/10.4337/9781788110686.00011 https://www.lse.ac.uk/granthaminstitute/wp-content/uploads/2014/11/Aghion_et_al_policy_paper_Nov20141.pdf https://doi.org/10.1086/684581 https://doi.org/10.1126/science.177.4047.393 https://doi.org/10.3998/mpub.10029 https://doi.org/10.1126/science.284.5411.107 https://doi.org/10.1016/j.ecolecon.2017.03.032 https://doi.org/10.1080/00213624.2004.11506696 https://doi.org/10.1080/14693062.2017.1404439 Understanding and modelling structural economic change 611 Cambridge Econometrics ( 2022) , ‘The E3ME Manual’, www.e3me.com . Carbon Brief ( 2024) , ‘Analysis: Clean Energy was Top Driver of China’s Economic Growth in 2023’, https: //www.carbonbrief.org/analysis- clean- energy- was- top- driver- of- chinas- economic- growth- in- 2023 / Cleevely , D. ( 2025) , Serendipity: It Doesn’t Happen by Accident , C&P, ISBN 9781068439216. David , P. A. ( 1994) , ‘Why are Institutions the “Carriers of History”? Path Dependence and the Evolution of Conventions, Organizations and Institutions’, Structural Change and Economic Dynamics , 5( 2) , 205–20. https://doi.org/10.1016/0954- 349X(94)90002- 7 Dechezleprêtre , A., Martin, R., and Mohnen, M. ( 2017) , ‘Knowledge Spillovers from Clean and Dirty Technolo- gies’, Working Paper 135, https://www.lse.ac.uk/granthaminstitute/wp-content/uploads/2013/10/Working- Paper- 135- Dechezlepretre- et- al_updateOct2017.pdf — Fabre, A., Kruse, T. et al. ( 2022) , ‘Fighting Climate Change: International Attitudes towards Climate Policies’, Centre for Climate Change Economics and Policy Working Paper 409, Grantham Research Institute on Cli- mate Change and the Environment Working Paper 384 , London: London School of Economics and Political Science. Dellink , R., Chateau, J., Lanzi, E., and Magné, B. ( 2017) , ‘Long-term Economic Growth Projections in the Shared Socioeconomic Pathways’, Global Environmental Change , 42, 200–14. https://doi.org/10.1016/j.gloenvcha. 2015.06.004 Dembo , R. S. ( 2021) , Risk Taking … in an Uncertain World , Bloomington, IN: Archway Publishing. Dixon , P. B., and Jorgensen, D. ( 2012) , ‘Handbook of Computable General Equilibrium Modelling’, Amsterdam: North-Holland. Ellenbeck , S., and Lilliestam, J. ( 2019) , ‘How Modelers Construct Energy Costs: Discursive Elements in Energy System and Integrated Assessment Models’, Energy Research & Social Science , 47, 69–77. https://doi.org/10. 1016/j.erss.2018.08.021 European Commission ( 2015) , ‘Better Regulation Guidelines’, European Commission document SWD 111 final , http://ec.europa.eu/smart-regulation/guidelines/docs/swd_br_guidelines_en.pdf Farmer , J. D., and Lafond, F. ( 2016) , ‘How Predictable is Technological Progress?’, Research Policy , 45( 3) , 647– 65. https://doi.org/10.1016/j.respol.2015.11.001 — Hepburn, C., Ives, M. C., Hale, T., Wetzer, T., Mealy, P., Rafaty, R., Srivastav, S., and Way, R. ( 2019) , ‘Sensitive Intervention Points in the Post-carbon Transition’, Science , 364( 6436) , 132–34. Ford , M. ( 2015) , The Rise of the Robots: Technology and the Threat of Mass Unemployment , London: Oneworld Publications. Fragkiadakis , K., Fragkos, P., and Paroussos, L. ( 2020) , ‘Low-carbon R&D Can Boost EU Growth and Compet- itiveness’, Energies , 5236, 13( 19) , Article ID: 5236. https://doi.org/10.3390/en13195236 Freeman , C., and Louçă, F. ( 2001) , As Time Goes By: From the Industrial Revolutions to the Information Rev- olution , Oxford: Oxford University Press — Perez, C. ( 1988) , ‘Structural Crisis of Adjustment, Business Cycles and Investment Behaviour’. In: G. Dosi, C. Freeman, R. Nelson, G. Silverberg, and L. Soete ( eds) , Technical Change and Economic Theory , London: Pinter, 38–66. Frey , C. B., and Osborne, M. A. ( 2017) , ‘The Future of Employment: How Susceptible are Jobs to Computerisa- tion?’, Technological Forecasting and Social Change , 114, 254–80. https://doi.org/10.1016/j.techfore.2016. 08.019 Galbraith , J. K. ( 1975) , Money: Whence it Came, Where it Went , Princeton, NJ: Princeton University Press, reprinted 2017. — ( 1998) , The Affluent Society , 5th edn, London: Penguin. Geels , F. W. ( 2002) , ‘Technological Transitions as Evolutionary Reconfiguration Processes: A Multi-level Perspec- tive and a Case-study’, Research Policy , 31( 8–9) , 1257–74. https://doi.org/10.1016/S0048- 7333(02)00062- 8 — ( 2025) , ‘Escaping and Creating Lock-ins in Accelerating Low-carbon Transitions: Conceptual Reflections and Empirical Insights from Electricity and Auto-mobility Transitions in Europe, the United States, and China’, Cambridge Journal of Regions, Economy and Society , forthcoming. https://doi.org/10.1093/cjres/rsaf011 — Turnheim, B. ( 2022) , The Great Reconfiguration: A Socio-technical Analysis of Low-carbon Transitions in UK Electricity, Heat, and Mobility Systems , Cambridge: Cambridge University Press. — Berkhout, F., and Van Vuuren, D. ( 2016) , ‘Bridging Analytical Approaches for Low-carbon Transitions’, Na- ture Climate Change , 6( 6) , 576–83. — Pinkse, J., and Zenghelis, D. ( 2021) , ‘Productivity Opportunities and Risks in a Transformative, Low-carbon and Digital Age’, Working Paper 9, The Productivity Institute. Godley , W. ( 1996) , ‘Money, Finance and National Income Determination: An Integrated Approach’, Working Paper 167, Levy Economics Institute of Bard College. http://www.e3me.com https://www.carbonbrief.org/analysis-clean-energy-was-top-driver-of-chinas-economic-growth-in-2023 https://doi.org/10.1016/0954-349X(94)90002-7 https://www.lse.ac.uk/granthaminstitute/wp-content/uploads/2013/10/Working-Paper-135-Dechezlepretre-et-al_updateOct2017.pdf https://doi.org/10.1016/j.gloenvcha.2015.06.004 https://doi.org/10.1016/j.erss.2018.08.021 http://ec.europa.eu/smart-regulation/guidelines/docs/swd_br_guidelines_en.pdf https://doi.org/10.1016/j.respol.2015.11.001 https://doi.org/10.3390/en13195236 https://doi.org/10.1016/j.techfore.2016.08.019 https://doi.org/10.1016/S0048-7333(02)00062-8 https://doi.org/10.1093/cjres/rsaf011 612 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels Grubb , M. ( 2018) , ‘Conditional Optimism: Economic Perspectives on Deep Decarbonization’, Institute for New Economic Thinking [INET] at the Oxford Martin School, University of Oxford, https://www.ineteconomics. org/perspectives/blog/growth- with- decarbonization- is- not- an- oxymoron — Wieners, C., and Yang, P. ( 2021 a) , ‘Modeling Myths: On DICE and Dynamic Realism in Integrated Assessment Models of Climate Change Mitigation’, WIREs Climate Change , 12( 3) , e698. — Drummond, P., Mercure, J.-F., Hepburn, C. et al. ( 2021 b) , ‘The New Economics of Innovation and Transition: Evaluating Opportunities and Risks’, Technical Report, Economics of Innovation and System Transition ( EEIST) , www.eeist.co.uk . — Sharpe, S., Agarwal, V., Anadon, L. D., Barbook-Johnson, P., Chaudhury, S., Farmer, D., Han, Z., Jones, A., Kapur, V., Kelkar, U., Lamperti, F., Lenton, T. M., Melehk, Y., Mercure, J.-F., Nijsse, F., Pasqualino, R., Peñasco, C., Pereira, M., and Zhu, S. ( 2023) , Economics of Energy Innovation and System Transition: Synthesis Report , https://eeist.co.uk/wp-content/uploads/EEIST-Synthesis-Report-2023IIB094.pdf Guivarch , C., Lempert, R., and Trutnevyte, E. ( 2017) , ‘Scenario Techniques for Energy and Environmental Re- search: An Overview of Recent Developments to Broaden the Capacity to Deal with Complexity and Uncer- tainty’, Environmental Modelling & Software , 97, 201–10. https://doi.org/10.1016/j.envsoft.2017.07.017 Hallegatte , S., McIsaac, F., Dudu, H., Jooste, C., and Knudsen, C. ( 2024) , ‘Macroeconomic Implications of a Transition to Net Zero Emissions’ Peterson Institute for International Economics Working Paper, ( 24–6) . https://doi.org/10.2139/ssrn.4767989 Harmsen , M., Kriegler, E., van Vuuren, D. P., van der Wijst, K.-I., Luderer, G., Cui, R., Dessens, R., Drouet, L., Emmerling, J., and Morris, J. F. ( 2021) , ‘Integrated Assessment Model Diagnostics: Key Indicators and Model Evolution’, Environmental Research Letters , 16( 5) , 054046. https://doi.org/10.1088/1748-9326/ abf964 Hepburn , C., Ives, M., Loni, S., Mealy, P., Barbrook-Johnson, P., Farmer, J. D., Stern, N., and Stiglitz, J. ( 2025) , ‘Economic Models and Frameworks to Guide Climate Policy’, Oxford Review of Economic Policy , 41( 2) , 616–52. Hidalgo , C. A., Klinger, B., Baraba´si, A.-L., and Hausmann, R. ( 2007) , ‘The Product Space Conditions the De- velopment of Nations’, Science , 317( 5837) , 482–87, https://doi.org/10.1126/science.1144581 Hodgson , G. M. ( 2001) , How Economics Forgot History: The Problem of Historical Specificity in Social Science , London: Routledge. https://doi.org/10.4324/9780203519813 — ( 2004) , ‘The Evolution of Institutional Economics: Agency, Structure and Darwinism in American institution- alism’, London: Routledge. Holland , J. H. ( 1995) , Hidden Order: How Adaptation Builds Complexity , New York: Basic Books. Hynes , W., Trump, B. D., Kirman, A., Haldane, A., and Linkov, I. ( 2022) , ‘Systemic Resilience in Economics’, Nature Physics , 18, 381–84. https://doi.org/10.1038/s41567- 022- 01581- 4 Kattel , R., Mazzucato, M., Ryan-Collins, J., and Sharpe, S. ( 2018) , ‘The Economics of Change: Policy and Ap- praisal for Missions, Market Shaping and Public Purpose’, Policy Report, WP 2018-06 , Institute for Innova- tion and Public Purpose. Kauffman , S. A. ( 2000) , Investigations , New York: Oxford University Press. https://doi.org/10.1093/oso/ 9780195121049.001.0001 Keynes , J. M. ( 1921) , A Treatise on Probability , London: Macmillan. — ( 1938) , ‘Untitled’, letter to R. F. Harrod, reproduced in R. Skidelsky ( ed., 2015) , John Maynard Keynes: The Essential Keynes , London: Penguin Classics, 280–81. — ( 1939) , ‘The League of Nations: Professor Tinbergen’s Method’, The Economic Journal , 49( 195) , 558–77. https://doi.org/10.1093/ej/49.195.558 Kirman , A. ( 2016) , ‘Complexity and Economic Policy: A Paradigm Shift or a Change in Perspective? A Review Essay on David Colander and Roland Kupers’s Complexity and the Art of Public Policy ’, Journal of Economic Literature , 54( 2) , 534–72. https://doi.org/10.1257/jel.54.2.534 Klitkou , A., Bolwig, S., Hansen, T., and Wessberg, N. ( 2015) , ‘The Role of Lock-in Mechanisms in Transition Processes: The Case of Energy for Road Transport’, Environmental Innovation and Societal Transitions , 16, 22–37. https://doi.org/10.1016/j.eist.2015.07.005 Knight , F. H. ( 1921) , Risk, Uncertainty and Profit , Boston, MA: Houghton Mifflin. Knobloch , F., and Mercure, J.-F. ( 2016) , ‘The Behavioural Aspect of Green Technology Investments: A General Positive Model in the Context of Heterogeneous Agents’, Environmental Innovation and Societal Transitions , 21, 39–55. https://doi.org/10.1016/j.eist.2016.03.002 Krugman , P. ( 1991) , ‘History Versus Expectations’, The Quarterly Journal of Economics , 106( 2) , 651–67. https: //doi.org/10.2307/2937950 Lenton , T. M. ( 2020) , ‘Tipping Positive Change’, Philosophical Transactions of the Royal Society B: Biological Sciences , 375( 1794) , 20190123. https://doi.org/10.1098/rstb.2019.0123 . Lucas , R. ( 1976) , ‘Econometric Policy Evaluation: A Critique’. In: K. Brunner, and A. Meltzer ( eds.) The Phillips Curve and Labor Markets, Carnegie–Rochester Conference Series on Public Policy , American Elsevier, 19–46. https://www.ineteconomics.org/perspectives/blog/growth-with-decarbonization-is-not-an-oxymoron http://www.eeist.co.uk https://eeist.co.uk/wp-content/uploads/EEIST-Synthesis-Report-2023IIB094.pdf https://doi.org/10.1016/j.envsoft.2017.07.017 https://doi.org/10.2139/ssrn.4767989 https://doi.org/10.1088/1748-9326/abf964 https://doi.org/10.1126/science.1144581 https://doi.org/10.4324/9780203519813 https://doi.org/10.1038/s41567-022-01581-4 https://doi.org/10.1093/oso/9780195121049.001.0001 https://doi.org/10.1093/ej/49.195.558 https://doi.org/10.1257/jel.54.2.534 https://doi.org/10.1016/j.eist.2015.07.005 https://doi.org/10.1016/j.eist.2016.03.002 https://doi.org/10.2307/2937950 https://doi.org/10.1098/rstb.2019.0123 Understanding and modelling structural economic change 613 Lutz , C., Meyer, B., and Wolter, M. I. ( 2010) , ‘The Global Multisector/Multicountry 3E–Model GINFORS. A De- scription of the Model and a Baseline Forecast for Global Energy Demand and CO2-Emissions’, International Journal of Global Environmental Issues , 10( 1–2) , 25–45. https://doi.org/10.1504/IJGENVI.2010.030567 Malthus , T. ( 1798) , An Essay on the Principle of Population: Volume 1 , reproduced and edited by P. James ( 1989) , Cambridge: Cambridge University Press. Meadows , D. H. ( 2008) , Thinking in Systems: A Primer , White River Junction, VT, and London: Chelsea Green. Mealy , P., and Hepburn, C. ( 2020) , Transformational Change: Parallels for Addressing Climate and Development Goals , Edward Elgar Publishing eBooks, https://doi.org/10.4337/9780857939067.00025 — Teytelboym, A. ( 2022) , ‘Economic Complexity and the Green Economy’, Research Policy , 51( 8) , 103948, https://doi.org/10.1016/j.respol.2020.103948 Mercure , J.-F. ( 2012) , ‘FTT:Power:A Global Model of the Power Sector with Induced Technological Change and Natural Resource Depletion’, Energy Policy , 48, 799–811. https://doi.org/10.1016/j.enpol.2012.06.025 — ( 2018) , ‘Fashion, Fads and the Popularity of Choices: Micro-foundations for Diffusion Consumer Theory’, Structural Change and Economic Dynamics , 46, 194–207. https://doi.org/10.1016/j.strueco.2018.06.001 — ( 2022) , Complexity Economics for Environmental Governance , Cambridge: Cambridge University Press. https://doi.org/10.1017/9781108553650 — Pollitt, H. ( 2025) , Reframing Climate Policy: Innovation and the Economics of Resource Creation , Cambridge: Cambridge University Press ( in press) . — — Geels, F. W., and Zenghelis, D. ( 2025 b) , ‘The Effects of Low-carbon Transitions on Labour Productivity: Analysing UK Electricity, Heat, and Mobility with a Techno-economic Simulation Model’, Climate Policy , 1–18. https://doi.org/10.1080/14693062.2025.2522836 — Lam, A., Buxton, J., Boulton, C., and Lenton, T. ( 2025 a) , ‘Evidence of a Cascading Positive Tipping Point towards Electric Vehicles’, forthcoming in Nature Communications . — Pollitt, H., Bassi, A. M., Viñuales, J. E., and Edwards, N. R. ( 2016) , ‘Modelling Complex Systems of Het- erogeneous Agents to Better Design Sustainability Transitions Policy’, Global Environmental Change , 37, 102–15. — Knobloch, F., Pollitt, H., Paroussos, L., Scrieciu, S. S., and Lewney, R. ( 2019) , ‘Modelling Innovation and the Macroeconomics of Low-carbon Transitions: Theory, Perspectives and Practical Use’, Climate Policy , 19( 8) . — Pollitt, H., Chewpreecha, U., Salas, P., Foley, A. M., Holden, P. B., and Edwards, N. R. ( 2014) , ‘The Dynamics of Technology Diffusion and the Impacts of Climate Policy Instruments in the Decarbonisation of the Global Electricity Sector’, Energy Policy , 73, 686–700. — — Viñuales, J. E., Edwards, N. R., Holden, P. B., Chewpreecha, U., Salas, P., Sognnaes, I., Lam, A., and Knobloch, F. ( 2018) , ‘Macroeconomic Impact of Stranded Fossil Fuel Assets’, Nature Climate Change , 8( 7) , 588–93. — Sharpe, S., Viñuales, J. E., Ives, M., Grubb, M., Lam, A., Drummond, P., Pollitt, H., Knobloch, F., and Ni- jsse, F.J.M.M. ( 2021) , ‘Risk-opportunity Analysis for Transformative Policy Design and Appraisal’, Global Environmental Change , 70, 102359. Nelson , R. R., and Winter, S. G. ( 1982) , An Evolutionary Theory of Economic Change , Cambridge, MA, and London: Harvard University Press. — — ( 2002) , ‘Evolutionary Theorizing in Economics’, Journal of Economic Perspectives , 16( 2) , 23–46. https: //doi.org/10.1257/0895330027247 Nemet , G. F. ( 2019) , ‘ How Solar Energy Became Cheap: A Model for Low-carbon Innovation ’, Abingdon: Rout- ledge. https://doi.org/10.4324/9780367136604 Nordhaus , W. D. ( 1992) , ‘An Optimal Transition Path for Controlling Greenhouse Gases’, Science , 258( 5086) , 1315–19. https://doi.org/10.1126/science.258.5086.1315 — ( 2016) , ‘Revisiting the Social Cost of Carbon’, Proceedings of the National Academy of Sciences , 114( 7) , 1518–23. https://doi.org/10.1073/pnas.1609244114 Ostrom , E. ( 2000) , ‘Collective Action and the Evolution of Social Norms’, Journal of Economic Perspectives , 14, 137–58. https://doi.org/10.1257/jep.14.3.137 Peace , J., and Weyant, J. ( 2008) , Insights not Numbers: The Appropriate Use of Economic Models , White Paper, April, Washington, DC: Pew Center on Global Climate Change. Peñasco , C., Anadón, L. D., and Verdolini, E. ( 2021) , ‘Systematic Review of the Outcomes and Trade-offs of Ten Types of Decarbonization Policy Instruments’, Nature Climate Change , 11, 257–65. Peng , W., Iyer, G., Bosetti, V., Chaturvedi, V., and Edmonds, J. ( 2021) , ‘Climate Policy Models Need to Get Real about People—Here’s How’, Nature , 594, 174–76. https://doi.org/10.1038/d41586- 021- 01500- 2 Perez , C. ( 2010) , ‘Technological Revolutions and Techno-economic Paradigms’, Cambridge Journal of Eco- nomics , 34, 185–202. https://doi.org/10.1093/cje/bep051 Peters , G. P., Al Khourdajie, A., Sognnaes, I., and Sanderson, B. M. ( 2023) , ‘AR6 Scenarios Database: An Assess- ment of Current Practices and Future Recommendations’, NPJ Climate Action , 2( 1) , 31. https://doi.org/10. 1038/s44168- 023- 00050- 9 https://doi.org/10.1504/IJGENVI.2010.030567 https://doi.org/10.4337/9780857939067.00025 https://doi.org/10.1016/j.respol.2020.103948 https://doi.org/10.1016/j.enpol.2012.06.025 https://doi.org/10.1016/j.strueco.2018.06.001 https://doi.org/10.1017/9781108553650 https://doi.org/10.1080/14693062.2025.2522836 https://doi.org/10.1257/0895330027247 https://doi.org/10.4324/9780367136604 https://doi.org/10.1126/science.258.5086.1315 https://doi.org/10.1073/pnas.1609244114 https://doi.org/10.1257/jep.14.3.137 https://doi.org/10.1038/d41586-021-01500-2 https://doi.org/10.1093/cje/bep051 https://doi.org/10.1038/s44168-023-00050-9 614 Dimitri Zenghelis, Hector Pollitt, Jean-François Mercure, and Frank W.Geels Peters , O. ( 2019) , ‘The Ergodicity Problem in Economics’, Nature Physics , 15, 1216–21. https://doi.org/10.1038/ s41567- 019- 0732- 0 Pollitt , H. ( 2019) , ‘The Contribution of Post-Keynesian Economics to Climate Policy and Meeting Global De- carbonisation Targets’, Post-Keynesian Economic Society annual conference , see http://www.postkeynesian. net/downloads/events/Pollitt_2019.pdf — Mercure, J.-F., Barker, T. S., Salas, P., and Scrieciu, S. ( 2024) , ‘The Role of the IPCC in Assessing Actionable Evidence for Climate Policymaking’, NPJ Climate Action , 3, article 11. Posner , R. A. ( 1997) , ‘Social Norms and the Law: An Economic Approach’, The American Economic Review , 87( 2) , 365–69. Probst , G., and Bassi, A. M. ( 2014) , Tackling Complexity: A Systemic Approach for Decision Makers , Sheffield: Greenleaf Publishing. PWC ( 2017) , ‘Will Robots Steal Our Jobs? The Potential Impact of Automation on the UK and other Major Economies’, ch. 4 of UK Economic Outlook, March 2017 , see http://www.pwc.co.uk/economic-services/ ukeo/pwcukeo- section- 4- automation- march- 2017- v2.pdf RMI ( 2023 a) , ‘X-change Cars: The End of the ICE Age’, Rocky Mountains Institute, Colorado, United States ( by K. Bond, S. Butler-Sloss, D. Walter, H. Benham, E. J. Klock-McCook, D. Mullaney, Y. Numata, L. Speelman, C. Stranger, and N. Topping) . — ( 2023 b) , ‘X-change Batteries: The Battery Domino Effect’, Rocky Mountains Institute, Colorado, United States ( by D. Walter, K. Bond, S. Butler-Sloss, L. Speelman, Y. Numata, W. Atkinson) . Romanowska , I., Wren, C. D., and Crabtree, S. A. ( 2021) , Agent-based Modelling for Archaeology: Simulating the Complexity of Societies , Santa Fe, NM: Santa Fe Institutional Press. Romer , P. M. ( 1990) , ‘Endogenous Technical Change’, Journal of Political Economy , 98( 5) , S71–S102. https: //doi.org/10.1086/261725 — ( 1994) , ‘The Origins of Endogenous Growth’, Journal of Economic Perspectives , 8( 1) , 3–22.0. https://doi.org/ 10.1257/jep.8.1.3 Rosenbloom , D., Markard, J., Geels, F. W., and Fuenfschilling, L. ( 2020) , ‘Why Carbon Pricing is not Sufficient— and how a “Sustainability Transition Policy”can Help Mitigate Climate Change’, Proceedings of the National Academy of Sciences , 117( 16) , 8664–68. https://doi.org/10.1073/pnas.2004093117 Schumpeter , J. A. ( 1934) , The Theory of Economic Development , 2nd edn, Cambridge, MA: Harvard University Press. Semieniuk , G., Holden, P. B., Mercure, J.-F., Salas, P., Pollitt, H., Jobson, K., Vercoulen, P., Chewpreecha, U., Edwards, N. R., and Viñuales, J. E. ( 2022) , ‘Stranded Fossil-Fuel Assets Translate into Major Losses for Investors in Advanced Economies’, Nature Climate Change , 12( 6) , 532–38. https://doi.org/10.1038/s41558- 022- 01356- y Seto , K. C., Davis, S. J., Mitchell, R. B., Stokes, E. C., Unruh, G., and Urge-Vorsatz, D. ( 2016) , ‘Carbon Lock- in: Types, Causes, and Policy Implications’, Annual Review of Environment and Resources , 41, 425–52. https://doi.org/10.1146/annurev- environ- 110615- 085934 Shackle , G.L.S. ( 1949) , ‘A Non-additive Measure of Uncertainty’, The Review of Economic Studies , 17( 1) , 70– 74. https://doi.org/10.2307/2295790 — ( 1952) , Expectations in Economics , Cambridge: Cambridge University Press. Sharpe , S. ( 2023) , Five Times Faster: Rethinking the Science, Economics, and Diplomacy of Climate Change , Cambridge: Cambridge University Press. https://doi.org/10.1017/9781009326506 — Lenton, T. M. ( 2021) , ‘Upward-scaling Tipping Cascades to Meet Climate Goals: Plausible Grounds for Hope’, Climate Policy , 21( 4) , 421–33. Shiller , R. J. ( 2017) , ‘Narrative Economics’, American Economic Review , 107( 4) , 967–1004. https://doi.org/10. 1257/aer.107.4.967 Stankiewicz , R. ( 2000) , ‘The Concept of “Design Space‘, ch. 17, In: J. Ziman ( ed.) , Technological Innovation as an Evolutionary Process , Cambridge: Cambridge University Press, 234–47. Stern , N. ( 2018) , ‘Public Economics as if Time Matters: Climate Change and the Dynamics of Policy’, LSE Working Paper , see https://eprints.lse.ac.uk/100895/3/Public_economics_as_if_time_matters.pdf — ( 2022) , ‘A Time for Action on Climate Change and a Time for Change in Economics’, The Economic Journal , 132( 644) , 1259–89. https://doi.org/10.1093/ej/ueac005 — Romani, M. ( 2023) , ‘ The Global Growth Story of the 21st Century: Driven by Investment and Innovation in Green Technologies and Artificial Intelligence ’, London: Grantham Research Institute on Climate Change and the Environment, London School of Economics and Political Science. — Valero, A. ( 2021) , ‘Innovation, Growth and the Transition to Net-zero Emissions’, Research Policy , 50( 9) , 104293, https://doi.org/10.1016/j.respol.2021.104293 — Stiglitz, J., and Taylor, C. ( 2022) , ‘The Economics of Immense Risk, Urgent Action and Radical Change: Towards New Approaches to the Economics of Climate Change’, Journal of Economic Methodology , 29( 3) , 181–216. https://doi.org/10.1038/s41567-019-0732-0 http://www.postkeynesian.net/downloads/events/Pollitt_2019.pdf http://www.pwc.co.uk/economic-services/ukeo/pwcukeo-section-4-automation-march-2017-v2.pdf https://doi.org/10.1086/261725 https://doi.org/10.1257/jep.8.1.3 https://doi.org/10.1073/pnas.2004093117 https://doi.org/10.1038/s41558-022-01356-y https://doi.org/10.1146/annurev-environ-110615-085934 https://doi.org/10.2307/2295790 https://doi.org/10.1017/9781009326506 https://doi.org/10.1257/aer.107.4.967 https://eprints.lse.ac.uk/100895/3/Public_economics_as_if_time_matters.pdf https://doi.org/10.1093/ej/ueac005 https://doi.org/10.1016/j.respol.2021.104293 Understanding and modelling structural economic change 615 Süsser , D., Ceglarz, A., Gasching, H., Stavrakas, V., Flamos, A., Giannakidis, G., and Lilliestam, J. ( 2021) , ‘Model- based Policymaking or Policy-based Modelling? How Energy Models and Energy Policy Interact’, Energy Research & Social Science , 75, 101984. Systemiq ( 2021) , ‘The Paris Effect—COP26 Edition’, https://www.systemiq.earth/wp-content/uploads/2021/11/ The- Paris- Effect- COP26- edition- SYSTEMIQ.pdf Trutnevyte , E. ( 2016) , ‘Does Cost-optimization Approximate the Real-world Energy Transition’, Energy , 106, 182–93. https://doi.org/10.1016/j.energy.2016.03.038 — Hirt, L. F., Bauer, N., Cherp, A., Hawkes, A., Edelenbosch, O. Y., Pedde, S., and van Vuuren, D. P. ( 2019) , ‘Societal Transformations in Models for Energy and Climate Policy: The Ambitious Next Step’, One Earth , 1( 4) , 423–33. Utterback , J. M., and Abernathy, W. J. ( 1975) , ‘A Dynamic Model of Process and Product Innovation’, OMEGA , 3( 6) , 639–56. https://doi.org/10.1016/0305- 0483(75)90068- 7 Van der Meijden , G. C., and Smulders, S. ( 2017) , ‘Carbon Lock-in: The Role of Expectations’, International Economic Review , 58( 4) , 1371–415. https://doi.org/10.1111/iere.12255 van der Ploeg , F., and Venables, A. J. ( 2025) , ‘Green Transitions: Complementarities, Multiple Equilibria, and Tipping Points’, Oxford Review of Economic Policy , 41( 2) , 377–94. van Vuuren , D. P., van der Wijst, K.-I., Marsman, S., van den Berg, M., and Hof, A. F. ( 2020) , ‘The Costs of Achieving Climate Targets and the Sources of Uncertainty’, Nature Climate Change , 10, 329–34. https://doi. org/10.1038/s41558- 020- 0732- 1 Veblen , T. ( 1898) , ‘Why is Economics not an Evolutionary Science?’, The Quarterly Journal of Economics , 12( 4) , 373–97. https://doi.org/10.2307/1882952 — ( 1899 a) , ‘The Preconceptions of Economic Science I’, The Quarterly Journal of Economics , 13( 2) , 121–50. https://doi.org/10.2307/1882197 — ( 1899 b) , ‘The Preconceptions of Economic Science II’, The Quarterly Journal of Economics , 13( 4) , 396–426. https://doi.org/10.2307/1883645 . — ( 1900) , ‘The Preconceptions of Economic Science III’, The Quarterly Journal of Economics , 14( 2) , 240–69. https://doi.org/10.2307/1883770 Way , R., Ives, M. C., Mealy, P., and Farmer, J. D. ( 2022) , ‘Empirically Grounded Technology Forecasts and the Energy Transition’, Joule , 6( 9) , 2057–82. https://doi.org/10.1016/j.joule.2022.08.009 Weitzman , M. L. ( 2009) , ‘On Modeling and Interpreting the Economics of Catastrophic Climate Change’, Review of Economics and Statistics , 91( 1) , 1–19. https://doi.org/10.1162/rest.91.1.1 Wilson , D. S. ( 2016) , ‘Two Meanings of Complex Adaptive Systems’. In: D. S. Wilson, and A Kirman ( eds.) , Complexity and Evolution , pp. 31–46. Cambridge, MA: MIT Press. Workman , M., Darch, G., Dooley, K., Lomax, G., Maltby, J., and Pollitt, H. ( 2021) , ‘Climate Policy Decision Making in Contexts of Deep Uncertainty—from Optimisation to Robustness’, Environmental Science & Policy , 120, 127–37. https://doi.org/10.1016/j.envsci.2021.03.002 Young , H. P. ( 2009) , ‘Innovation Diffusion in Heterogeneous Populations: Contagion, Social Influence, and Social Learning’, American Economic Review , 99( 5) , 1899–924. https://doi.org/10.1257/aer.99.5.1899 Zenghelis , D. A. ( 2019) , ‘ Nobel for Economics 2018—A Question of Imbalance ’, Bennett Institute for Public Policy, https://www.bennettinstitute.cam.ac.uk/blog/nobel- economics- 2018- question- imbalance/( Date ac- cessed 16 August 2023) . — Serin, E., Stern, N., Valero, A., van Reenen, J., and Ward, B. ( 2024) , ‘Boosting Growth and Productivity in the United Kingdom through Investments in the Sustainable Economy’, CEP Reports, Centre for Economic Performance, LSE , https://EconPapers.repec.org/RePEc:cep:cepsps:43 . © The Author( s) 2025. Published by Oxford University Press on behalf of The Oxford Review of Economic Policy Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( https://creativecommons.org/licenses/by/4.0/) , which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. https://www.systemiq.earth/wp-content/uploads/2021/11/The-Paris-Effect-COP26-edition-SYSTEMIQ.pdf https://doi.org/10.1016/j.energy.2016.03.038 https://doi.org/10.1016/0305-0483(75)90068-7 https://doi.org/10.1111/iere.12255 https://doi.org/10.1038/s41558-020-0732-1 https://doi.org/10.2307/1882952 https://doi.org/10.2307/1882197 https://doi.org/10.2307/1883645 https://doi.org/10.2307/1883770 https://doi.org/10.1016/j.joule.2022.08.009 https://doi.org/10.1162/rest.91.1.1 https://doi.org/10.1016/j.envsci.2021.03.002 https://doi.org/10.1257/aer.99.5.1899 https://www.bennettinstitute.cam.ac.uk/blog/nobel-economics-2018-question-imbalance/ https://EconPapers.repec.org/RePEc:cep:cepsps:43 https://creativecommons.org/licenses/by/4.0/ I. Introduction II. The nature of structural change III. Modelling structural change IV. Using models to guide policy for bringing about structural change V. Conclusions References