Dulani Chunga moved from a safe, quiet but poor village in Malawi to Blantyre, the prime business city, in the hopes of changing his destiny. He was drawn to the city by stories of streetlights, the opportunity to make money, and the chance to send his children to school. He lives in Ndirande, an immense slum with squalid conditions. While his income is higher than what it used to be in his village, it is barely enough to feed his family of four—food and shelter cost a lot more in Blantyre.
The fortunes of many Dulanis are stuck in a low-development trap of developing-country cities. Yet, evidence increasingly highlights the productivity advantages of living and working in dense cities, particularly in the developing world. While productivity benefits of density, measured as the elasticity of wages with respect to density, are significant for developed country cities—0.043 in the United States and 0.03 in France—some recent estimates for developing countries are multiples higher: 0.19 in China, 0.12 in India, and 0.17 in Africa. What do these estimates tell us? A 10 percent increase in density increases productivity by 1.7 percent in Africa compared to 0.4 percent in the United States. These estimates appear implausible if we take Dulani’s experience into perspective. More broadly, 54 percent of Africa’s urban population lives in slums and 38 percent in South Asia.
How do we reconcile these elasticity estimates with reality? In a recent working paper, we examine more than 1,200 estimates of urban productivity from 70 studies covering 33 countries from 1973 to 2020. In addition, we constructed new estimates to show how urban costs, with respect to crime, congestion, and pollution, changed with density. For this, we collected data from hundreds of cities around the world, including several in developing countries.
A quick look suggests high agglomeration economies in developing countries
A casual glimpse at productivity estimates measured through wage premiums shows that these are on average nearly 5 points higher in developing countries (Figure 1).
Figure 1. People in developing countries appear to benefit more by living in cities
Note: This figure computes unweighted average wage elasticity estimates for each country using individual worker data for the non-services sector (reflecting either manufacturing or the entire economy). This comprises two-thirds of the developing country estimates. It reflects 271 raw elasticity estimates (144 in non-high-income countries), aggregated across different studies with different methodologies.
A broader examination tells a different story
A meta-analysis is a technique that helps explain differences in estimates across studies based on their attributes, including methodology, time period of the study, and so on. For example, studies estimating agglomeration benefits using nominal wages or labor productivity have elasticities that are 6.3 and 4.3 percentage points higher than those using total factor productivity (TFP). This suggests that part of the wage premium is driven by higher capital intensity, perhaps a result of thicker capital markets, in urban areas, rather than efficiency or spillovers per se.
Some studies also control for the fact that skilled workers are attracted to dense cities that make them productive. These studies include human capital controls such as an individual’s education, lowering agglomeration gains. Finally, econometric analysis that employs panel fixed effects, thereby controlling for the selection of better workers or firms, lowers estimates by about 1.8 percentage points. Once such study-level idiosyncrasies are accounted for, elasticity estimates for developing countries are only 1 percentage point higher than those for developed countries (Figure 2).
Figure 2. Agglomeration premiums on labor productivity nearly disappear after controlling for urban costs
Note: The figure uses a rope-ladder representation of a subset of the estimated coefficients from the meta-analysis model. The meta-analysis probes into the factors—methodological, data-related, controls—that influence agglomeration elasticity estimates. The methodology for meta-analysis minimizes the Bayesian Information Criterion. Using the standard errors of the coefficients, it also plots the 90 percent confidence intervals, where standard errors are clustered at the study level. Similar estimated coefficients are obtained by model selection using the Akaike Information Criterion or Bayesian Model Averaging methods.
Agglomeration premia should reflect the higher cost of working in cities (such as higher housing costs or time lost in transport) or compensation for urban disamenities such as pollution and crime. But most empirical work does not factor in these costs. Our meta-analysis shows that studies controlling for urban costs would estimate net agglomeration benefits to be 0.1 percent for high-income countries and 1 percent for non-high-income countries when using labor productivity as the outcome. These results are in line with French and Colombian data, suggesting that the net benefits from city size are close to being flat.
Our estimates on the extent of urban disamenities with respect to pollution, congestion, and crime suggests that urban disamenities are higher in developing countries. For the average city density, in high-income countries 19-30 percent fewer hours are spent in traffic congestion, pollution is 16-28 percent lower, and the homicide rate is around four times lower. In particular, the elasticity of the homicide rate is positive and very high (24 percent) in developing countries and negative (56 percent) in developed countries. This suggests that if urban costs pertaining to crime are accounted for, the magnitude of net agglomeration elasticity in developing countries would be smaller or even negative.
Are cities in developing countries different?
The findings from our systematic meta-analysis and estimates on cost elasticities support Dulani’s view on the ground. People in developing countries are concentrating—but not because they are attaining the productivity benefits of urbanizing. Developing-country cities are generally dense but not productive, and they are crime-ridden and polluted to boot. This evolution is consistent with what has been called “premature urbanization.”
In light of these findings, can we really hope that the migration of people from villages to dysfunctional cities will pull them out of poverty? The experiences of China and South Korea suggest that cities become productive when urbanization is accompanied by industrial dynamism and broader structural transformation of the national economy. Developing countries ought to focus on removing distortions that limit structural transformation that creates the impetus for spatial transformation. It is only then that cities will attain economic density, achieve higher productivity, and live up to the hopes of many more Dulanis to come.
This article originally appeared in the Future Development blog and was written by Arti Grover, Somik V. Lall, and Jonathan Timmis