Heat, air pollution, and global temperature shocks

Developing countries face a host of environmental challenges, but two of the most significant are air pollution and climate change. Wealthy countries face both as well, but the highest levels of air pollution, the fastest growth in pollution, and the least climate-resilient populations are all in low- and middle-income countries (LMICs). Some policies address both problems at once (reducing coal-fired power generation, for example). But win-win solutions do not always exist.

These trade-offs have led some to call for a greater focus on air pollution in LMICs, at the expense of climate, based on the view that air pollution is the more pressing problem. The argument is that air pollution causes substantial mortality and economic costs in the here and now, while climate change is a slow-moving challenge whose most damaging effects will arrive at the end of the century.

The case for air pollution first

Based on the historical evidence, this was a reasonable argument. Estimates of the mortality impacts of heat, which appear to make up a large part of global mortality from climate change, are more than an order of magnitude smaller through 2050 than current estimates of mortality from air pollution. Back-of-envelope calculation from their end-of-century results puts mortality from heat in 2050 somewhere between 0.03 and 0.04 deaths per 1,000 people per year, assuming warming through 2050 follows a moderate (RCP4.5 to RCP6.0) path.

Mortality from air pollution today is somewhere between 0.8 and 1.2 deaths per 1,000 people per year. In the most severely affected countries, such as India, the rate may be closer to 1.6 per 1,000. Even assuming substantial improvements in air quality by 2050, reasonable projections imply that mortality from air pollution would still be about five times the impact of heat in 2050. By 2100, with continued progress on air pollution, the impact of heat based on these historical estimates is likely to be comparable to, or slightly larger than, the impact of air pollution.

So the claim that air pollution deserves relatively more attention when resources are scarce seems broadly consistent with our existing understanding of the relative contributions of heat and air pollution to mortality. Climate change has other consequences that might motivate action. But so does air pollution, and excess mortality is one of the largest welfare costs of both. 

Why local estimates may understate climate mortality

New work on the mortality impacts of climate change calls all of the above into question. One challenge with much of the existing work on heat exposure, and climate change more generally, is that effects are estimated from highly localized weather shocks. Those estimates may not account for what happens when temperatures rise in other locations at the same time. But that is exactly what climate change does: temperatures increase, on average, in many places at once. One (imprecise) way to think about this is that much of the existing empirical literature estimates a partial equilibrium effect, while the impact of climate change is a general equilibrium question. The authors of existing work have tried a variety of careful approaches to go from partial to general effects, but it is a difficult problem.

To get around this challenge when estimating the GDP impact of climate change Adrien Bilal and Diego Känzig use an empirical approach that relies on global deviations from expected temperature to approximate the impact of large-scale warming. The basic identification argument is that these deviations from forecast are driven by natural variability in the climate system – El Niño cycles, solar cycles, and volcanic eruptions – that is exogenous to local economic and health conditions. We’re currently going through a particularly strong El Niño, which is pushing global temperatures up.

What global temperature shocks say about mortality

They’ve now applied that approach to mortality. They estimate that a permanent 1°C increase in global temperature raises annual mortality by 2.5 deaths per 1,000 people, compared to 0.1 per 1,000 from local heat exposure alone, an estimate between one and two orders of magnitude larger than previous work. Their estimates imply that by 2050, mortality from warming will be greater than present mortality from air pollution, and substantially greater than mortality from air pollution in 2050 with expected improvements in air quality. The impacts in Africa, Asia, and Latin America will be substantially larger than the global average.

Based on these numbers, the case for prioritizing air pollution today is much less clear. Their results also suggest, because of the way their baseline is constructed, that there is already substantial mortality from warming today, regardless of whether that warming is due to climate change. If so, actions to reduce the future damages of climate change will likely have immediate benefits, which further strengthens the case for climate adaptation.

Where the extra deaths come from

What generates the large impact on mortality that they document? In short, the second-order effects of warming. They decompose their estimate by channel and find a direct effect of heat exposure that is similar to Carleton et al. But because their approach gets closer to a general equilibrium estimate, it also captures a variety of second-order impacts.

Consider the income channel, which is where most of their effect comes from. In an earlier paper, Bilal and Känzig estimate that a permanent 1°C rise in global temperature lowers world GDP per capita by 22% in the long run. Lower global income in turn raises mortality. Combining that estimate with a separate estimate of how world income affects death rates, they show that the decline in global GDP explains most of the mortality response to warming. A localized weather shock does not capture this channel because it does not have the same impact on global GDP. The advantage of the new approach is that it can capture those kinds of channels.

The caveats

This paper has empirical challenges. Relative to local shock based approach, the sample here is small because it relies on a single global time series. That also means it does not include year fixed effects. The authors control for observable global shocks (recessions, commodity prices, epidemics, and conflicts), but any unobserved global shock that happens to coincide with a temperature shock is not averaged out. If such shocks raise mortality for reasons unrelated to temperature, they will contaminate the estimated effect. But this approach is a useful counterpoint to the localized approaches that have driven much of the literature to this point.

What this does and does not change

This is not an argument for prioritizing climate adaptation at the expense of work on air pollution on the basis of a single study. Air pollution remains a significant source of mortality in the short term. It may still be the largest single environmental source of mortality in many LMICs, even with these new results. And the reductions in pollution expected by 2050, which drive the lower projected mortality at that point, will only happen if pollution receives attention and resources today.

But these new results are important both because they suggest the mortality burden of warming may be far greater than previously estimated and because they show, again, how much depends on using approaches that account for equilibrium impacts when estimating the consequences of climate change. 

Source : World Bank