Thursday 29 August 2019

Dirty Density: Air Quality and the Density of American Cities


Air pollution tends to be worst in large cities and their urban cores. As a result, it is urban air pollution that makes the headlines when the media report on pollution and its effects (see, for example, this, this and this).  Since it is mainly an urban problem, air pollution exposure is shaped by urban planning and policy. In particular, it can be affected by population density, the defining feature of urbanization that distinguishes cities from smaller towns and villages.
In a recent paper we study how air pollution - as measured by fine particulate matter (PM2.5) concentration - is shaped by cities’ population density in the United States. In particular, we want to find out whether residents in dense urban areas are exposed to worse air quality. For this purpose, we use new data on satellite-derived measures of PM2.5 concentration at a fine spatial scale and demographic information from the US census.
Intuitively, our analysis is conducted by comparing cities of different densities and their average concentration of PM2.5. The resulting scatter plot is provided in Figure 1 and serves to illustrate our main result: denser cities tend to have worse air quality. But this naïve approach cannot give us a definitive (causal) answer on whether population density affects air pollution. Why? Because many confounding factors can bias the associated estimates. For example, people decide on where to locate based on various factors including local amenities and employment opportunities. Given that many productive activities (e.g. factories) generate pollution; if people move into areas close to these activities, a naïve estimation which ignores these confounders will overstate the true effect of density on pollution. 

Figure 1 – PM2.5 Concentration and Population Density

Note: The vertical axis represents PM2.5 average residential exposure (in μg /m3), as obtained from the satellite-derived measures. The horizontal axis represents the natural logarithm of population density. The points represent 933 CBSAs (metro and micropolitan areas). The black line is estimated by OLS using the underlying data.

To deal with this and other endogeneity issues, we use data on the geological characteristics at and around US cities as instrumental variables for density. Intuitively, we use these variables to generate variation in density that is not shaped by pollution or other confounders.[1] Our instrumental variable estimates confirm the message in Figure 1: denser locations are associated with higher concentrations of PM2.5. How much higher? Quite a bit. According to our estimates, doubling density – say, increasing the density in Houston to match that of Chicago – increases PM2.5 concentrations by 0.73 μg/m3, which is roughly 10% of the average pollution across cities. Using well-established dose-response functions that map pollution concentration to mortality rates, in conjunction with official mortality costs estimates from the US Environmental Protection Agency, we find that doubling densities would lead to annual mortality costs of as much as USD 630 per capita.
So denser cities lead to higher pollutant concentration. But was it not the case that denser cities are also greener? A prolific strand of research has emphasized the environmental advantages of denser cities (see Kahn and Walsh 2015 for a review). In dense cities, households enjoy shorter commutes when driving to work (Duranton and Turner 2017) and may even switch to other transport modes when these are available (Cervero and Guerra 2011). In a world in which a significant amount of emissions is generated by transport – especially driving – this observation has led to the conclusion that denser cities are also greener (see Glaeser and Kahn 2010). But our results indicate that this should not be interpreted as meaning that high density leads to better air quality. Yes, denser cities are associated with lower emissions, and this is important for reducing global greenhouse gas concentrations. However, if we are concerned about local air quality, having lower emissions does not suffice. Even if emissions are low in denser cities, it is the concentration of pollutants that determines local pollution levels. Our paper shows that pollution exposure is higher in denser cities, making the environmental quality in these cities lower than in other locations.[2]
What should be done?
For decades now, the compact city urban planning approach has been promoting urban densification as a way to contain sprawl, reduce car use and promote some of the beneficial agglomeration forces normally associated with density. There are many things to enjoy about compact cities, including shorter commutes and better access to commercial and recreational activities. Yet our results indicate that the purposed environmental advantages of compactness may be limited to reductions in global pollutants. When it comes to our lungs and hearts, denser cities create a more polluted, harmful environment. Urban planners should take note of these trade-offs when designing the cities of the future.

References:
Cervero, R., & Guerra, E. (2011). Urban densities and transit: A multi-dimensional perspective. Institute of Transportation Studies, University of California, Berkeley.
Combes, P. P., Duranton, G., Gobillon, L., & Roux, S. (2010). Estimating Agglomeration Economies with History, Geology, and Worker Effects. Agglomeration Economics, 15.
Combes, P. P., & Gobillon, L. (2015). The empirics of agglomeration economies. In HHandbook of regional and urban economics (Vol. 5, pp. 247-348). Elsevier.
Duranton, G., & Turner, M. A. (2018). Urban form and driving: Evidence from US cities. Journal of Urban Economics, 108, 170-191.
Glaeser, E. L., & Kahn, M. E. (2010). The greenness of cities: carbon dioxide emissions and urban development. Journal of Urban Economics, 67(3), 404-418.
Kahn, M. E., & Walsh, R. (2015). Cities and the Environment. In Handbook of regional and urban economics (Vol. 5, pp. 405-465). Elsevier.




[1] This strategy was initially developed in the agglomeration literature trying to estimate the productive advantages of cities (see Combes et al. 2010, Combes and Gobillon 2015). In our case, we use it to study one of the congestion forces that constrain city growth.
[2] Complementary results in the paper show that the observed increase in PM2.5 is not driven by a different sectoral composition of production in larger cities or by differences in total city population. See details in the paper.