Expect Above Average Temperatures: Identifying the Economic Impacts of Climate Change
AbstractOne of the most glaring gaps in the economic understanding of climate change centers on the economic
costs of future climate change. A recent literature has attempted to empirically estimate these costs by
using within-unit variation in weather to identify the costs of being exposed to above-average
temperatures. These panel settings have made great progress towards identifying the implications of
weather shocks for outcomes such as GDP, farm values, crop yields, labor productivity, health, air
conditioning use, and crime. What can we learn from this literature?
Climate change is a permanent, anticipated change in the distribution of weather. I formally analyze the
effects of climate change on payoffs and control variables within a setting that captures the direct
effects of today’s altered weather, the effect of past expectations of today’s altered weather on past
long-term investments, and the effect of today’s expectations of future altered weather on today’s longterm
investments. I show that expectations matter for the costs of climate change, except in particular
special cases that may be unlikely to apply to contexts of interest. I show that identifying the costs of
climate change requires identifying the causal effect of weather shocks on dependent variables and also
the causal effect of weather forecasts on dependent variables. Further, I show that standard panel
regressions obtain systematically biased estimates of the causal effect of weather due to forecasts’
presence as unobserved covariates. I offer constructive suggestions for how to overcome this bias.
Finally, I relate my analysis to arguments from the literature. The standard appeal to the envelope
theorem to justify the informativeness of weather regressions ignores the presence of past controls in
today’s payoff functions and also ignores that many dependent variables of interest are themselves
controls, not objectives.