The Role of Modelling and Scenario Development in Long-term Strategies
"It is difficult to make predictions, especially about the future"
A central question in the economics of managing climate change is, “How much will global decarbonisation cost?” This has spawned a profusion of economic models assessing likely future states of the world under different scenarios.
An economic model is essentially a simplified framework for describing the workings of the economy. This has the advantage of requiring internal consistency, for example, in ensuring that actions across sectors are sufficient to achieve economy-wide targets, and exerting the discipline of forcing the modeler to formally articulate assumptions and tease out relationships behind those assumptions.
Models are used for two main purposes. The first is conditional “as if” simulations (“Given what we know about the behavioural workings of the economy, how would the world change relative to some counterfactual if we assume a change in this or that variable?”). Such model simulations have produced valuable insights, highlighting risks and uncertainties and helping society formulate, examine, and understand interactive relationships.
But when it comes to the low-carbon transition, models are much less effective at their second common use: providing forecasts. This is because when making long-term forecasts, very little can be taken as given.1 Most sensible modellers offer the caveat that projections should not to be interpreted as forecasts, but the problem runs deeper than making clear, specific assumptions. The problem is that the very things that are most interesting when it comes to making predictions decades ahead are also those that are hardest to model. As a result, most often, they are simply not modelled and consequently “models” tell us little about the true costs and benefits of long-run policies to promote renewable technologies and encourage resource efficiency.
For example, economists understand the concept of learning by doing. This says that as you deploy a new technology and experiment with it, you learn how to engineer, fabricate, fit, and use it better. Deployment and experience drive innovation, making goods cheaper and more productive. Many models now explicitly incorporate this process.2 Yet it’s usually not the learning that models get wrong, it’s the doing.
Figure 1. New renewable capacity forecasts, excluding hydropower
Electrical capacity (GW)
Source: Matthieu Metayer, Christian Breyer and Hans-Josef Fell, 2015. ‘The projections for the future and quality in the past of the World Energy Outlook for solar PV and other renewable energy technologies‘. Energy Watch Group.
The International Energy Agency (IEA) has some of the most knowledgeable and experienced energy economists in the world. More than anyone, they understand how energy technologies evolve, and yet they systematically underestimate the deployment of renewables (Figure 1 compares IEA forecasts with actual outturns) and correspondingly overestimate the costs.
What they failed to predict were broader social responses, how some governments subsidised technologies when expensive, such that when prices fell and looked like they were falling further, others would deploy them. They did not predict that by now, renewables would be the biggest source of annual investment in energy generation the world over, outpacing coal, oil, gas, nuclear and hydro combined.
The price of solar photovoltaic (PV) modules has fallen 44% in the two years to the end of August 2017 and by 83% since 20103, a period over which the price of wind turbines dropped 35%.4 Even the cost of offshore wind energy is down around 46% since 2012.5 In places both are already at grid parity and competitive with coal, oil and gas without subsidy, while land-based wind costs have outcompeted gas and coal for many years in large parts of the world.
Conventional models: Precisely wrong
The tendency to underestimate the speed of major technology transitions is not surprising. It stems from economists’ adopting standard modelling approaches conventionally used for costing small changes a few years ahead. Such models are structured around fixed behavioural parameters and exogenous assumptions about key technology costs, economic behaviour, tastes and preferences in the future. These are then used to generate whole-economy cost forecasts associated with deploying technologies and processes in key sectors such as energy, transport, buildings and land use sufficient to meet a GHG emissions target.6 Virtually nothing in these economic models provides insights into how these assumptions will evolve. Models simply presuppose the things we want to know and invariably miss the dynamics of technological revolutions and social feedbacks, as well as the role of new networks, changing expectations, and social norms in driving these changes. The further out the forecast, the larger the uncertainties and the chance that structural breaks push the economy onto new paths driven by new technologies, institutions and behaviours.
This makes model projections at best illustrative, especially when trying to forecast the impact of nonmarginal impulses such as climate change impacts or the transformation of the global energy system. Predictions and forecasts are essentially only as good as the priors of the forecaster who makes them.
A good illustration of these shortcomings is provided by the Intergovernmental Panel on Climate Change (IPCC). Its expert analysis has helped improve understanding and guide policymakers globally. The IPCC is the world’s most respected and authoritative body on these matters. The Summary for Policymakers by IPCC Working Group III reported modelling results for meeting a 2°C stabilisation target as costing an average loss in global consumption of 2.9 percent to 11.4 percent a year by 2100.7 The decimal places should ring alarm bells—economists struggle to predict GDP two years out to one decimal place, let alone over 80 years, a period during which baseline GDP will likely have grown between 300 and 900 percent.
But it’s not just the spurious precision that’s the problem. More worrying are the conclusions. They state that after eight decades of clear policy signals’ deploying a uniform global carbon price, pumping trillions of dollars into investment and innovation in new technologies, employing the world’s brightest and best innovators and most creative entrepreneurs to extract energy for free from the sun, the wind, and the sea and store and use it efficiently, it will still be more productive in 2100 in every scenario to live in a society that employs labour across the world to dig up fossil fuels—oil, coal and gas—in ever-more remote locations, put these fuels on ships and railways and into pipelines, transport them halfway across the world, and burn them using essentially nineteenth-century technologies, losing up to half that energy in transmission or in inefficient combustion engines. Even allowing for the difficulty of squeezing carbon out of land, industry and buildings, this conclusion is highly implausible. The evidence suggests a postcarbon society could be cleaner, quieter, safer, more technologically advanced and more prosperous.
Abstraction to distraction
The reason for these consistently gloomy results is the requirement that models tend towards a preordained steady-state equilibrium. Indeed, most models focus almost exclusively on narrow distortionary costs associated with a carbon prices, just because they can. Yet the real world is what economists call endogenous—that is, subject to systemic changes that originate within the system and determine the structure of the equilibrium itself. Large, nonmarginal transformations, such as the low-carbon transition, are hard to estimate empirically and even harder to model because of the nonlinear dynamics. Hard-to-predict events have persistent effects. Who could have deterministically modelled the opening up of world trade in the second half of the last century, the diffusion of internet mobility and social media, or the attacks of September 11, 2001?
Changes in direction can become mutually reinforcing, leading to tipping points and multiple equilibria where initial choices are amplified through new networks and path dependencies, especially where these are subject to lock-in or irreversibility.8 Investing in renewable technologies pushes their price down, making further investment increasingly attractive relative to conventional technologies, where the gains from additional learning or scaling are smaller. New engineers learn how to cheaply install, connect and repair the technology (one reason why solar PV is considerably cheaper in Germany than in the United States). Planning institutions are updated and new networks are built or transfigured. New industrial interests emerge and lobby for supportive policies. Consumers change behaviour and demand new infrastructure and technologies (such as networked efficiency, recycling and pedestrianisation). Very quickly, an economy can switch from one technology network to another as it becomes more attractive than the incumbent.
Take cities. Once built, they are hard to change retrospectively, as infrastructures become locked in. Behaviours and institutions then follow. Citizens in sprawling cities tend to lobby for fuel subsidies and highway lanes, while those in dense cities call for public transport, congestion charging and cycling lanes. The decisions on urban connectivity made by planners in China, India and elsewhere will go a long way toward determining the efficiency and resource security of their economies. They also create sizable new markets, which stimulate innovators and investors across the world. Yet these processes are nowhere to be found in economic models.
The full interaction of an endogenous system is fiendishly complex to replicate and any error spreads through the model like a runaway virus. Meteorological model ensembles make consistent and accurate global forecasts up to two weeks ahead, but then start to diverge widely because of the infamous “butterfly wing” effect. The same is true with economic models. The IPCC forecast for 2100 is likely to be about as accurate as a forecast for rain showers over Paris at 9:30 a.m. in six months’ time. Far too many important causal links in the chain between now and then will have been missed.
This is why modelling requires abstraction, because not all variables can be included and not all causal processes simulated. But abstracting is fine until you abstract from the key properties of the system and then purport to forecast that system as whole. Endogeneity is a feature of human development. Thomas Malthus famously took the structure of the global economy as given, such that the world would run low on resources in the face of a growing population.9 In fact, it turned out that every extra human mouth was born with a brain. And human innovation allowed agricultural yields to skyrocket and industrialisation to provide an unprecedented array of consumer possibilities. These endogenous dynamics are knowable as processes (“Scarcity breeds innovation”) but not as outcomes (“Therefore we predict the plough”).
The problems faced by economic forecasters tasked with examining the impact of large, transformative change, such as transitioning to a resource-efficient global economy over longer periods, are immense. Technology costs, economic structures, and tastes and preferences are not just hard to predict, they depend on actions taken by decision-makers today.10
Expecting more from expectations
Expectations are crucial to determining the cost and effectiveness of managing change. Social psychologists have long understood that solving coordination problems requires building expectations into your models and generating “common knowledge.”11 Likes and dislikes, social norms and technology costs are not fixed, they can all be influenced. Crucially, all these factors feed off each other endogenously, changing the dynamic in a way that lowers costs.
Agents base their decisions on how they anticipate others will act. A mayor, politician, business leader, or consumer is unlikely to invest early in deploying or developing renewable energy and energy efficiency if they believe nobody else is going to invest. They will expect the technology costs to be high, the financing to be niche, and the markets to remain limited and immature. They will invest, however, if they think everyone else will do likewise in anticipation of falling technology costs, cheap financing, and growing and profitable new market opportunities. This becomes a self-fulfilling expectation as increased investment begets lower costs through learning by doing.
To give a practical example, some 10 years ago, the ill-fated Copenhagen conference failed to achieve an ambitious climate deal because the dominant narrative was based on burden-sharing and sacrifice. This was informed by early model predictions that the costs of action would be prohibitive. This bred distrust. Few wanted to invest in costly emissions reductions if their competitors didn’t, encouraging free-riding off the efforts of others. The result was a race to the bottom and delayed global action. By contrast, the Paris Agreement of 2015 focused on voluntary contributions based on opportunity and self-interest.12 Self-interest breeds cooperation. Taking advantage of perceived opportunities, rather than sacrificing your citizens for the global good, has led to a comprehensive agreement. By influencing expectations, poorly specified models not only get the future wrong, they make the future wrong. To the extent that they are believed, they become part of the problem.13
Conclusion and policy recommendations
Economic models were never designed to serve as estimates of the long-term impacts of things like reducing carbon emissions. By missing the dynamics of technological revolutions, the feedbacks that new networks propagate and the role of expectations, they generate a structural bias towards maintaining the current high-carbon path and overestimating the net costs of a transition to a low-carbon economy. This postpones policy action and short-circuits the more pertinent questions of how structural change can be brought about in a transparent, market-friendly manner, one that promotes competition and growth and limits the scope for rent-seeking by vested interests.
There is an urgent need to shift from long-term deterministic predictions to a coherent theory behind long-run processes of systemic change. This requires the development of alternatives to static optimisation and welfare maximisation approaches, deploying instead insights from history, spatial geography, planning and social psychology. Quantitative models might better inform policy choices under uncertainty by helping manage risk and direct change, through applying dynamic choice criteria including real options theory, no regrets analysis, robust decision-making and agent-based modelling. For example, a new programme at Oxford University seeks to overcome the limitations of conventional modelling by focusing instead on insights from complex systems to identify sensitive intervention points where policymakers can deliver impact at scale and accelerate the achievement of global net-zero emissions.14
Adopting such approaches would yield immediate and material benefits, cost-effectively accelerating the transition to a postcarbon society. Governments would be encouraged to set clear, transparent, and credible frameworks to steer expectations to avoid locking into unsustainable and vulnerable physical, human and knowledge capital.
The lesson here is that models are not the whole story, but they are essential tools to help tell the story. In a sufficiently complex path-dependent world, where timely choices have long-term consequences, it might make more sense to drive change through directing and designing the future than to try erroneously to predict it.
1 MIT economist Robert Pindyck claims models tell us “very little” about climate change and how we solve it (Pindyck, R. S. 2013. Climate Change Policy: What Do the Models Tell Us? Journal of Economic Literature, 51 (3), 860–72. Lord Nick Stern suggests models “may be profoundly misleading on the issues of great significance,” (Stern, N. 2013. “The Structure of Economic Modeling of the Potential Impacts of Climate Change: Grafting Gross Underestimation of Risk onto Already Narrow Science Models.” Journal of Economic Literature 51 (3), 838–59. while Frank Ackerman and Elizabeth Stanton suggest models are “not remotely consistent with the recent research on climate impacts” (Ackerman, F., & Stanton, E. A. 2013. Climate economics: the state of the art. Routledge. )
2 See, e.g., the World Induced Technical Change Hybrid (WITCH) model, http://www.witchmodel.org/.
3 Bloomberg New Energy Finance, Solar Spot Price Index, December 2017, https://www.bnef.com/core/insights/17357
4 Bloomberg New Energy Finance, 2H 2017 Wind Turbine Price Index, September 2017, https://www.bnef.com/core/insights/17017
5 Bloomberg New Energy Finance, Skeptical About Climate, Clean Energy Skeptics, November 2017, https://about.bnef.com/blog/mccrone-skeptical-about-climate-clean-energy-skeptics/
6 Benjamin H. Mitra-Kahn, “Debunking the Myths of Computable General Equilibrium Models,” Schwartz Center for Economic Policy Analysis and Department of Economics, New School for Social Research, Working Paper 01-2008 (March 2008), http://www.economicpolicyresearch.org/images/docs/research/economic_growth/SCEPA%20Working%20Paper%202008-1_Kahn.pdf
7 IPCC, “Summary for Policymakers (IPCC AR5, Working Group III),” 2014; see Table SPM.2.
8 Philippe Aghion, Cameron Hepburn, Alexander Teytelboym, and Dimitri Zenghelis. “Path Dependence, Innovation and the Economics of Climate Change,” Centre for Climate Change Economics and Policy, Grantham Research Institute on Climate Change and the Environment, Policy Paper, November 2014, http://2014.newclimateeconomy.report/wp-content/uploads/2014/11/Path-dependence-and-econ-of-change.pdf.
9 Thomas R. Malthus, An Essay on the Principle of Population (London: Joseph Johnson, 1798).
10 Robert M. Solow, “Perspectives on Growth Theory,” Journal of Economic Perspectives 8(1) (1994): 45–54.
11 K. Thomas, O.S. Haque, S. Pinker, and P. DeScioli, “The Psychology of Coordination and Common Knowledge,” Journal of Personality and Social Psychology 107 (2014): 657–76.
12 A. Averchenkova, N. Stern, and D. Zenghelis, “Taming the Beasts of ‘Burden-Sharing’: An Analysis of Equitable 2030 Mitigation Pledges,” Grantham Research Institute and Centre for Climate Change Economics and Policy, London School of Economics, 2014.
13 Paul Krugman, “History versus Expectations,” Quarterly Journal of Economics 106(2) (1991): 651–67.
14 Oxford Martin School, “Programmes: Post-carbon Transition,” www.oxfordmartin.ox.ac.uk/research/programmes/post-carbon, accessed December 1, 2017.