This post is inspired by a set of similar referee comments I’ve gotten over the last several months. The frequency and consistency of the nature of the comment made me think it would be worth reflecting on here.
Before I get to the comment itself, here is a brief bit of background to provide necessary context. Much empirical climate economics is interested in understanding the impact of climate shocks on different economically relevant outcomes. The central goal of this work is identifying a dose-response function between a climate shock (e.g. high temperatures) and an outcome (e.g. mortality). In theory these dose response functions can then be used to estimate damages from future climate change.
However, it is clear these dose response functions, estimated on past extreme weather events, are unlikely to reflect future responses to climate shocks because those future responses are likely to include some level of adaptation. One attempt to address this problem that has become increasingly common is to estimate the dose response function across various measures of heterogeneity in the data. This is done as an attempt to measure the scope for adaptation to moderate the dose response function. The argument is that some units in the data might be more adapted than others and so estimating the dose response function separately in these two groups can yield an estimate of what an adaptation inclusive dose response function might look like.
By far the two most common measures of heterogeneity that have been examined empirically are (a) past exposure to the particular shock and (b) incomes/wealth. The idea is that places with more past exposure to a given shock are more likely to invest in adaptive approaches than places with less exposure (e.g. central AC penetration is greater in the Southern US than in the Northern US). Similarly, places with higher income/wealth have the ability to invest more in adaptive responses than those with lower income/wealth.
With this background, the comments I received were both with respect to this kind of heterogeneity analysis. Both referees expressed surprise that, in different settings, I and co-authors found that the adverse effects of heat declined less as income increased than as other markers of adaptation (in one case past exposure, in the other housing age) increased. The surprise, in my assessment, is driven by the intuition that income growth is a particularly effective marker of adaptation to climate change.
For a long time, this was also my intuition and is, I think, fairly common. Despite this, empirical examinations of the question pretty consistently find that this intuition is not exactly correct.
Aside from my work that motivated the referee comments, the difference in heat’s impact on mortality in hot vs. cold places seems to be larger than that between rich and poor locations. Other work finds similar impacts for education. In both cases, the estimated effects are smaller in hot places than cold places. They are also generally smaller in rich places than poor places. But the reduction in impacts is larger moving from cold to hot than poor to rich. Also worth mentioning is work that shows increasing incomes does not appear to permanently mitigate the heat-crime relationship.
What explains the gap between the common intuition that income is more predictive of adaptation than other markers and much of the evidence? I think there are a few possibilities. One is that the intuition that income is a better proxy is pervasive because it is easier to conceptualize mechanisms that would justify it. It may be easier to imagine how increasing incomes can lead to adaptation than think of how adaptation might occur just based on exposure without changing incomes. That lends itself to the intuition that income matters more.
A second possibility is that these differences have been empirically mismeasured. Many of the estimates of the difference in effect sizes between hot and cold vs. rich and poor places are not convincingly different from one another. Confidence intervals on the estimated effects allow for the possibility that they are quite similar. This is compounded by the fact that nearly all of this analysis is cross-sectional heterogeneity and so potentially confounded. It is also true that there are many more cold, rich places than hot, rich places and conversely many more poor, hot places than poor, rich places.
The third possibility is the one I find most interesting, that is that income may simply be a poor proxy for adaptation compared to past exposure. These heterogeneity analyses are done in the name of capturing the scope for adaptation to reduce damages. But neither approach – using income or past experience – directly measures adaptation. They measure things we think are good proxies for adaptation. Saying that you believe effects should decline more with increases in income than increases in past exposure is in effect saying you believe that income is a better proxy for adaptation than past exposure. That is, people with higher incomes will have done more to adapt than those with more past experience. It’s easy to imagine how high income could facilitate adaptation and so I think this is a natural intuition.
But the empirical evidence suggests that income may not be a good proxy for adaptation, or at least not as good a proxy as past exposure. Why? Answering that question is important because it can inform how policy is designed to increase adaptation. It could be the case that income is not as highly correlated with adaptation as one might expect because high income places have not had the exposure to shocks to motivate adaptive investments. If that is the case, then increasing incomes in places that frequently experience shocks should lead to increased adaptation as the motivation to invest exists already and the increase in incomes provides the ability.
On the other hand, it could be the case that the most effective investments in adaptation are not within the scope of individuals to make, even with higher incomes, but require collective action. That would indicate that increasing income alone is unlikely to lead to large increases in adaptation on its own.
The reality of what facilitates adaptation likely lies somewhere between these two extremes. It is also probable that the extent to which income or experience is a better marker of adaptation is context and outcome dependent.
But better understanding why past experience with climate shocks seems to be better correlated with reduced impacts than high incomes will help guide decisions about how to encourage future adaptation. This is also a call for better measuring and assessing the effectiveness of adaptation directly, rather than relying on proxies like income or experience.
Join the Conversation