We hear about the pernicious effects of climate change constantly. There is a broad consensus among scientists, politicians, journalists, and others that raising awareness about the negative effects of climate change is useful (Weder di Mauro 2021). For instance, it can strengthen support for mitigation policies or preemptive action to avoid the largest damages. Although warning people about the future impact of climate change might seem desirable, little is known about the impact of anticipation on the local and aggregate effects of climate change. In a recent paper (Bilal and Rossi-Hansberg 2023), we propose a novel quantitative framework to evaluate the economic impact of climate change on all US counties. Because our framework features forward-looking migration and investment adaptation responses, agents anticipate climate change and react immediately to future expected changes in climate. We use this framework to gauge the local and aggregate importance of anticipation responses.
To many readers, it might seem surprising that economists do not know more about the local role of anticipatory adaptation responses to climate change. The reason is simple: dynamic general equilibrium models with aggregate climate shocks and many locations are hard to analyse and compute. Consider a person living in Florida, who expects storms and heatwaves to intensify severely close to their home over the next few decades. Should they move elsewhere? And if so, where? Clearly, they need to consider how climate will affect other counties. Furthermore, they need to predict how other people and investors will anticipate these changes and react to them themselves. Naturally, all these reactions affect local economies since they determine how much capital is invested and how many people live there. Our Florida resident faces this complex problem, as do millions of others that live all across the US. Their reactions determine how climate change ultimately impacts all regions. Technically, we are faced with a heterogenous agent dynamic economy with aggregate shocks, where the number of state variables is the stock of capital and population size in each county. For the US economy and its 3,143 counties, this implies more than 6,000 state variables. Solving the transition of this economy in response to climate change has proven intractable so far. 1
To make progress, we leverage the solution method in Bilal (2023) of a ‘Master equation’ state-space approach to this type of economy. This consists of an analytical first-order (and potentially higher-order) representation of the directional derivatives of policy functions around the steady state. The result is a solution to this complex problem that allows us to solve the evolution of this economy in seconds. With this methodology, we are able not only to compute counterfactual exercises but also to structurally estimate the migration and investment elasticities that govern the responses of people and investments to changes in the economic environment.
An important part of the quantification of economic climate models is to estimate structural ‘damage functions’, namely, to determine how changes in temperatures affect local fundamentals in the model. In our case, we want to study how heatwaves and storms affect local productivity, amenities, and capital depreciation rates. Our starting point is a reduced-form estimation of the effect of heatwaves and storms on local outcomes using an event study design. We use data since 2000 at the county level and estimate the impulse response from these events on population, income per capita, and investments. For storms, for example, we estimate that, in coastal counties, a once-in-50-years storm leads to a 5% decline in county population, a 2.5% decline in income per capita, and a large increase in investment for the first few years that reverts back after about six years. Figure 1 shows our estimates for population and investment. We find no significant effects in inland counties. We also estimate the impact of heatwaves. We find that a once-in-20-years heatwave reduces population and income by 2.5%, although only temporarily, and has a small negative effect on investment in counties with average temperatures above the median. We do not find significant effects in colder counties.