We do not know how much climate change will cost us. Yet that cost is the ‘externality’ missing from fossil fuel prices (Gollier 2021), which needs to be incorporated by either taxing or capping carbon emissions (as in the EU Emission Trading System). Plausible assumptions about that cost could justify carbon prices anywhere from a few dollars to hundreds of dollars per tonne. The problem is that measuring the cost of climate change is inherently difficult because climate change is a novel experiment being run in real time on a global scale.
How could we back out the cost of future climate change from data available today? You have surely noticed that different places have different climates. So you might compare outcomes between cooler place A (say, Germany) and warmer place B (say, Italy) to learn about the effects of climate change. The problem is that any two places differ in many ways, and these differences often correlate with their climates. Within the EU, myriad differences in institutions, histories, land use patterns, infrastructure, and culture arise as we move from south to north, so that it is impossible to firmly attribute any particular difference in social or economic outcomes to more southerly places being warmer. This method can estimate either net benefits or net costs from climate change for US agriculture, depending on how one accounts for differences in irrigation between different places (Mendelsohn et al. 1994, Schlenker et al. 2005).
What else could we do? You have surely noticed that the same place can have very different weather at different times. Rather than rely on comparisons across space, you might instead compare times when place A was warmer to times when the same place A was colder. These changes in weather are as good as random, thus far beyond the control of society. Over the last decade, economists have shown that weather matters for a host of social and economic outcomes (Dell et al. 2014, Carleton and Hsiang 2016). Once they estimate the effects of hotter days, many economists have begun extrapolating to the effects of climate change by using scientific projections of how weather will change over the century. In US agriculture, this approach forecasts losses of up to 56% by the end of the century (Deschênes and Greenstone 2007, Fisher et al. 2012).
The drawback of this approach is that random weather events differ from climate change in important ways – climate change makes previously unusual weather events occur over and over, we know it will alter future weather, and it affects all locations’ weather at once (on the final point, see Cruz and Rossi-Hansberg 2021). Troublingly, we don’t even know whether extrapolating from the effects of weather overestimates or underestimates the cost of climate change. For instance, farmers may adapt better to climate change than to short-run weather events because they will have a chance to install irrigation and adjust their planting decisions, in which case actual costs will be less than estimated. But it could also be true that farmers take actions in response to short-run weather events that they could not sustain year after year (as with using more groundwater), in which case actual costs will be greater than estimated.
If economists are going to be able to offer clear guidance about the appropriate ambition of climate change policy, we need firmer damage estimates – or at least to know whether data-driven estimates are upper or lower bounds. In a recent paper (Lemoine 2021), I develop new methods of estimating damages that account for how climate change makes agents live with altered weather over and over, and leads agents to expect altered weather in the future. I ground my methods in a dynamic model of decision-making in which agents can take actions to protect themselves against current and future weather events. These actions can be based on knowledge of the weather outside their doors (‘ex-post adaptation’), as when farmers irrigate in response to hot weather. They can also be based on forecasts of coming weather (‘ex-ante adaptation’), as when farmers adjust acreage planted in anticipation of hot weather.
Within the world of my model, I derive the true effect of changing the climate (in symbols, not numbers). This is the great unknown in climate change economics, the target many aim at by extrapolating from estimated effects of weather but which is never observed directly in the real world. Because we usually never see the target, we have no idea how good our aim is. But I here do know the target. So I can assess the accuracy of using changes in weather to get at changes in climate and can try to improve our aim.
Optimistically, it turns out that conventional approaches can be successful in some special cases. Pessimistically, these cases do appear rather special. In particular, they require that agents’ decisions not be linked over time, essentially reducing the dynamic environment to a static one (as in Hsiang 2016, Deryugina and Hsiang 2017). As soon as agents interact with a capital stock (such as irrigation or air conditioning infrastructure) or a natural resource stock (such as groundwater), their decisions depend both on past decisions and on projected future decisions – and thus on past weather and on projected future weather. These dependencies matter for the true effect of climate change but cannot be fully captured by conventional methods that extrapolate from effects of short-run weather events. The accuracy of current methods may not be great.
So what can we do? Is there any hope for a data-driven approach to estimating the cost of climate change? I develop a new type of approach that should improve our aim. We need to estimate effects not just of contemporary weather but also of lagged weather and forecasts of weather. Once we have done so, the theory tells us how to manipulate and combine the results in a way that mimics the effects of the climate. With this method, we apportion the effects of climate between the direct effects of altered weather, ex-post adaptation to living with altered weather over and over, and ex-ante adaptation to expecting to live with altered weather in the future. And we can sign the remaining gap between the new estimates and the true effect of climate change, so that we now know whether we have estimated a lower bound or an upper bound on long-run climate damages.
I apply my new method to the debate around impacts on agriculture in the eastern half of the US. Updating Deschênes and Greenstone (2007) and Fisher et al. (2012), I show that conventional methods would project losses of around 42% by the end of the century, driven by a greater number of extremely hot days. In contrast, my new methods estimate that climate change will eliminate profits from the average acre of current farmland. As before, increases in extremely hot days are important and bad, but now having more days with common heat also appears harmful. The change is driven by ex-post adaptation. I find that farmers’ ex-post adaptation to common heat provides short-run benefits. However, I also find that the direct effects of common heat are harmful and that the short-run benefits of ex-post adaptation reflect tradeoffs with long-run costs, such as from depleted groundwater or reduced soil quality. Surprisingly, adaptation now actually appears to increase the cost of climate change in the long run.
My work suggests a re-examination of the enormous literature that extrapolates from consequences of weather to consequences of climate change. It offers a roadmap for how to construct estimates that are grounded in economic theory and which could ultimately generate more credible carbon price recommendations. In the debate about the effects on US agriculture, these new estimates generate much more pessimistic results and alter common intuition about adaptation reducing the cost of climate change. The future health of the sector depends on expanding farmland in cooler regions and on the development of new crop varieties that can thrive in a warmer world.
Carleton, T A, and S M Hsiang (2016), “Social and Economic Impacts of Climate”, Science 353(6304).
Cruz, J-L, and E Rossi-Hansberg (2021), “Unequal gains: Assessing the aggregate and spatial economic impact of global warming”, VoxEU.org 2 March.
Dell, M, B F Jones, and B A Olken (2014), “What Do We Learn from the Weather? The New Climate-Economy Literature”, Journal of Economic Literature 52(3): 740–98.
Deryugina, T, and S Hsiang (2017), “The Marginal Product of Climate”, NBER Working Paper 24072.
Deschênes, O, and M Greenstone (2007), “The Economic Impacts of Climate Change: Evidence from Agricultural Output and Random Fluctuations in Weather”, American Economic Review 97(1): 354-85.
Fisher, A C, W M Hanemann, M J Roberts, and W Schlenker (2012), “The Economic Impacts of Climate Change: Evidence from Agricultural Output and Random Fluctuations in Weather: Comment”, American Economic Review102(7): 3749–60.
Gollier, C (2021), “Efficient carbon pricing under uncertainty”, VoxEU.org, 6 April.
Hsiang, S M (2016), “Climate Econometrics”, Annual Review of Resource Economics 8: 43-75.
Lemoine, D (2021), “Estimating the Consequences of Climate Change from Variation in Weather”, CEPR Discussion Paper no. 16194.
Mendelsohn, R, W D Nordhaus, and D Shaw (1994), “The Impact of Global Warming on Agriculture: A Ricardian Analysis”, American Economic Review 84(4): 753-71.
Schlenker, W, W M Hanemann, and A C Fisher (2005), “Will U.S. Agriculture Really Benefit from Global Warming? Accounting for Irrigation in the Hedonic Approach”, American Economic Review 95(1): 395-406.