Skip to content

An Air Pollution Paradox?

Should pollution abatement investment be relatively low in places where pollution levels are high?

The evidence is mounting that what you can’t see can kill you. What you can’t see are the tiny particles in the air you’re breathing (unless you are reading this blog through an electron microscope). Really tiny particulate matter measuring less than 2.5 micrometers (about 1/30th the diameter of a human hair) is the most pernicious because it can work its way deep into your lungs and circulatory system. Research has linked exposure to these particles, also known as PM2.5, to a range of outcomes you don’t want to mess with, including asthma, acute bronchitis, lung cancer, stroke, and heart disease.

Concentrations of PM2.5 are increasingly being monitored around the world. More monitors and better data are helping us develop a clearer picture of who’s being exposed to how much. As the evidence piles up, we’re coming to appreciate what a big problem these little particles create. And for those of us who think about the economics of air pollution, we’re discovering some important ways in which PM2.5 seems to contradict our textbook understanding.

The Current State of the Air

The map below summarizes PM2.5 concentrations across the globe. The colors code these concentrations relative to the World Health Organization’s recommended limit (an annual average concentration of 10 µg/m3). It’s estimated that over 90% of people on earth are living in areas that exceed this health-based guideline.WHOmapSource: State of Global Air 2017

Not a pretty picture. But before you get too down about the state of the air, consider this good news. We are seeing reductions in PM2.5 exposure in the U.S. and other developed countries (thank you pollution regulation!). If you want to see what’s been happening to the air quality in your neighborhood, this website makes it easy to plot PM2.5 trends by zip code (US locations only).

That’s the good news. The bad news is that this trend is not universal. Earlier this summer, the Lancet medical journal published the most comprehensive assessment of global trends in air pollution exposure and associated health impacts I have seen. This report did not make anyone’s list of 2017 feel-good summer beach reads. But if you care about air quality it’s worth a look.

For starters, the report estimates population-weighted average concentrations of PM2·5 for each country. The global average has increased from 40 µg/m³ (in 1990) to 44 µg/m³ (in 2015). The variation across countries is really striking. In several countries (such as Canada, Australia, Sweden), population-weighted annual average concentrations fall below 8.0 µg/m³.  But in some of the worlds most populated countries,  these average concentrations are terribly high. Examples include Bangladesh (89.4 µg/m³), India (74.3 µg/m³), and China (58.4 µg/m³).

These numbers are all the more troubling when you consider that many people in these countries breathe below-average air. Josh Apte has done some truly amazing work measuring the quality of the air people actually breathe in cities like Delhi and Beijing. In India’s capital, for example, annual average PM2.5 concentrations exceeds 150 µg/m³. That’s more than 10 times the level experienced by the average American. Wow.


Economics of Air Pollution

The second part of the Lancet study estimates the effects of PM2.5 exposure on health and mortality. The results are devastating. Premature deaths attributed to PM2.5 pollution are estimated at 4.2 million in 2015 (more than half of these are in India and China).

One caveat is that disentangling these health impacts of PM2.5 exposure is hard because air pollution levels are correlated with all sorts of things that determine health outcomes (like industrial activity, climate, income, other pollutants). So researchers have to find ways to separate the effects of PM2.5 exposure from the effects of other stressors.

This gets complicated quick. So, let’s retreat for a moment to the relatively simple world we find in economics textbooks. When economists think about the relationship between pollution concentrations and health impacts, we often think in terms of damage functions. I must have drawn these pictures a hundred times for my undergraduates:


The graph on the left plots the total health damages (external costs not reflected in market prices) caused by pollution over a range of pollution levels. The slope/steepness of this line at any given pollution level tells us how damages improve with an incremental reduction in pollution. The graph on the right plots these marginal damage changes (on a different scale). The key is that marginal damages (and the marginal benefits from cleaning up pollution) are typically assumed to be higher at higher pollution levels.

These stylized relationships help us think about where we should invest in reducing emissions. The economists’ prescription is to focus pollution abatement efforts where the returns on investment are highest. If abatement costs are similar across locations, this prescription sends abatement efforts to places where air quality is at its worst and people are most at risk. This makes good economic (and equitable) sense.

PM2.5 Curve Ball

There are a growing number of careful and well-executed studies investigating the health impacts of PM2.5. These have mostly been conducted in wealthier countries with relatively low pollution levels. To carry out global assessments of health impacts, researchers need to extrapolate the results of these studies over the full range of pollution levels that people are exposed to. Loosely speaking, researchers collect existing studies, connect the dots between them, and then extrapolate to pollution levels not covered by existing research.

The picture below summarizes the current thinking about the global relationship between the per capita mortality rate and PM2.5 concentrations:

ApteSource: this important paper by Joshua Apte, Julian Marshall, and colleagues.

The dashed line is analogous to my total damage function above. It shows how the mortality rate increases with PM2.5 exposure. The shape of this curve is rather surprising to those of us who feel like we can draw these damage functions in our sleep.

The slope of this line is decreasing as pollution concentrations increase, suggesting that the marginal damage function slopes down. In other words, an incremental reduction in PM2.5 delivers greater reductions in per capita mortality rates in places where pollution levels are relatively low (e.g. the United States). In places located on the flatter part of the curve (e.g. India, China), the effect of an incremental reduction in pollution exposure is relatively small.

A Pollution Paradox?

It’s important to note that the shape of this mortality rate-pollution concentration curve is still being debated. I’m no epidemiologist, but the estimated relationship above seems plausibly consistent with some kind of saturation response. At higher levels of exposure, it could be that much of the health damage has already been done and the marginal damage is relatively less. That said, there are remarkably few studies of this critical relationship in the most polluted parts of the world. One thing I take away from this is that more research is urgently needed.

Let’s assume for now that the estimated concentration-mortality relationship above is approximately correct. Does the shape of this curve create a tension between the economists’ prescription to maximize returns on investment and the moral imperative to focus efforts on places where people are most at risk?

In many of the world’s most polluted places, I think the answer is likely to be no. Some of the most polluted places on earth are very densely populated. A relatively small reduction in per capita mortality can have a big impact (in terms of the number of deaths avoided) in places where a lot of people live. Another consideration is abatement costs. The marginal costs of abatement are likely to be relatively low in many of the areas where PM2.5 concentrations are highest. In other words, there is low hanging pollution abatement fruit (e.g introducing cleaner cars, cleaner factories) to be found in places located on the flatter portion of the curve.

Even in places where the answer could be yes, there is, of course, a balance to be struck between efficiency (i.e. getting the biggest health improvement bang for our pollution mitigation buck) and equity. The research suggests that hundreds of thousands of people exposed to very high levels of PM2.5 are dying prematurely each year from what amounts to a preventable cause. We must do more to reduce this risk. With each incremental improvement in air quality, the next step becomes more valuable.


11 thoughts on “An Air Pollution Paradox? Leave a comment

  1. Thanks for the great article.
    What I’m confused by is at the low end (say 0-10ug), where the function is flat then steep… this seems a bit weird!
    I looked at reference [12] from the paper about the counterfactual exposure. It might make more sense to allow either: 1) the curve to go negative (maybe somehow giving completely pure air to someone is good for them?!) or 2) maybe rerun the bayesian analysis with a sigmoid-like function? Maybe there’s a more gentle increase in the effect of pollution.


  2. Excellent article. Thank you.
    At Paharpur Business Centre, Nehru Place, New Delhi, India, we have been testing PM 1’s in the air for the past two years using a Dust Track, on a daily basis at around 11 am every day. The PM 1’s are around 70-90% of the PM 2.5 readings. The question is where do these come from? My guess is that they come from CNG used by Buses, automobiles and three wheelers in Delhi. These PM 1 particles cannot be seen by the eye unlike the black some of the past from Diesel!.
    another issue is the high level of Chlorine in the air. It is around 14ug /m3 against WHO cut off of 20ug/m3 – coming from burning of plastic and unknown materials in incinerators in Delhi?
    The good news is that people ( politicians and decision makers ) may study the Harvard study – http://www.the and try and reduce the CO2 levels in their own dwelling area to achieve higher cognitive ability of up to +299%. This will get them to understand the benefits of clean and fresh air and make a change.
    We should be able to file a petition in the Supreme Court for the right to clean air.
    Best wishes
    Kamal Meattle

  3. Hi Meredith.

    Great article. You asked “Does the shape of this curve create a tension between the economists’ prescription to maximize returns on investment and the moral imperative to focus efforts on places where people are most at risk?” and concluded that “In many of the world’s most polluted places, I think the answer is likely to be no. Some of the most polluted places on earth are very densely populated. A relatively small reduction in per capita mortality can have a big impact (in terms of the number of deaths avoided) in places where a lot of people live.”

    We who developed the risk functions used to make the estimates published in The Lancet were also concerned about the policy implications of their non-linearity, and published a paper on this issue a few years ago. We reached the same conclusion as you did.

  4. “The marginal costs of abatement are likely to be relatively low in many of the areas where PM2.5 concentrations are highest.” Maybe, but we need to know more about the sources of air pollution. Your map suggests a rough correlation between PM2.5 and non-anthropogenic causes, e.g. Saharan dust in Mauritania, Mali, and Niger. According to Time, PM2.5 in Beijing reached “a staggering 630” a few months ago, largely blown in from the Gobi Desert.

  5. Great article, got me thinking…Wonder if the curveball is a result of poorer performance in certain countries across the broader range of the “social determinants of health”? Totally hypothetical example, maybe Country A performs better across the social determinants of health (education, support systems, employment, list goes on..) so PM reductions have a larger impact on health outcomes because that country enjoys a relatively strong performance across a pretty broad set of ‘healthfulness factors’ and PM exposure is one of the last remaining problems, whereas Country B performs poorly across the social determinants of health and therefore PM reductions don’t have the same beneficial impact, because the citizens of Country B are still subject to environmental and social factors that leave them at higher risk.

    • Causality and correlation?
      Perhaps in a country like Japan, where people wear masks out of social ‘pressure’ to not spread disease, the reduction in PM would not have had such a great affect on health. Where knowledge about health improves [even as PM does not decline].
      Another potential factor: in China and India it could be that the health data are for the ‘wealthy’ who have access to healthcare and manage their exposure to PM.

      Lets remember that there are many many [so many] factors that affect health that one should not to look at it in ‘isolation’.

%d bloggers like this: