Tom Kemeny is an award-winning social scientist, with expertise in cities and economic development. In 2019, he won the Understanding Society Prize. For his work on local social networks, he won the 2016 Urban Land Institute Prize, for best paper in the Journal of Economic Geography. He is an author of The Rise and Fall of Urban Economies: Lessons from San Francisco and Los Angeles, published in 2015 by Stanford University Press. He is Senior Lecturer (Associate Professor) at Queen Mary, University of London, and a Visiting Fellow at LSE’s International Inequalities Institute. You can find him at Twitter at @KemenyThomas.
Michael Storper is Professor of Economic Geography at the London School of Economics, Professor of Economic Sociology at Sciences Po in Paris, and Distinguished Professor of Regional and International Development at the University of California, Los Angeles. For his work on regional development and economic geography, he has been recognized as a Corresponding Fellow of the British Academy; he has also received the Regional Studies Association‘s Sir Peter Hall Prize; and the Founder’s Medal from the Royal Geographical Society.
The problem of growing interregional inequality
Economic inequalities are widening. The richest individuals and largest corporations today enjoy a larger piece of the pie. But inequality is also feels territorial. Some cities and regions promise not just higher incomes but also improved life chances. In the U.S., our perception of expensive coastal ‘superstar’ metros contrast starkly against shrinking Rustbelt cities and struggling rural areas; in the UK, it manifests as the classical split between London and the north; in France, declining provincial areas are differentiated from Paris, Lyon and a few smaller metros. And it has become commonplace to argue that the significance of these gaps is not only economic – they are increasingly linked to the recent rise of polarization and populism that, in a wide range of countries over the last few years has disrupted politics.
Concerns about this kind of inequality find support in recent academic work. Using Gini coefficients, Figure 1 traces the evolution of interregional wage gaps in the U.S. at the level of Commuting Zones. It confirms a sense that things used to be different. Gaps between places diminished steadily over much of the mid-20th century. Then, around 1980, those gaps began to grow. This shift from convergence to divergence has been observed in the U.S. at other spatial scales (Ganong and Shoag, 2017, Giannone, 2017), as well as in several other developed economies, including Sweden (Enflo and Roses, 2015), Spain (Martinez-Galarraga et. al, 2015), the EU as a whole (Roses and Wolf, 2018), and certain developing countries, including Mexico (Aguilar-Retureta, 2016).
Figure 1. Evolution of interregional income inequality (-convergence), overall and by education, 1940–2017.
Note: Authors’ calculations. Based on year-specific Gini coefficients estimated using average estimated hourly wage and salary income for all in-sample workers in 722 1990-vintage Commuting Zones (CZs), and weighted by population. Incomes are adjusted for inflation to 2015 dollars using Bureau of Labor Statistics CPI. Source data are IPUMS public-use extracts of Decennial Censuses and the American Community Survey.
What is causing rising interregional inequality?
Interregional inequality now commands researchers’ attention. Among economists, much of the focus is on the locational choices of workers, especially those with college degrees. Figure 2 visualizes their locational patterns since 1980. On the horizontal axis are local shares in 1980 of workers with at least 4 years of college. On the vertical is subsequent growth in these shares between 1980 and 2017. The upward sloping relationship tells us that college-educated workers have sought out cities that started out with higher shares of college educated workers. The fact that larger circles lie on the left means that larger cities have been the biggest winners, both in absolute and relative terms. Agglomeration is cited as the cause – larger cities offer more chances for these workers to interact and in so doing make each other more productive (Moretti, 2004). And while this kind of concentration confers benefits on the winners, there are vicious cycles for places losing their pools of skilled labor.
While these patterns may be descriptively accurate, workers’ choices are not a sufficient explanation for the patterns we see in Figure 1. Crucially, to help explain the 1940 to 1980 period – one which saw dramatic growth and catch up – economists argue that workers of all kinds voted with their feet to satisfy preferences for warm, sunny winters and larger homes. But this begs the question: why did workers’ preferences suddenly change around 1980? And why was this alleged change concentrated mainly among those with a university education?
Figure 2. Initial shares of workers with 4+ years of college and 1980-2017 annual growth rates in the share of workers with 4+ years of college, U.S. Commuting Zones.
Note: Authors’ calculations. Circles represent 722 Commuting Zones, scaled according to 1980 population. Dashed line represents linear fit, not weighted by population. ‘College share’ on the y-axis is the share of the working population who have completed at least four years of university education. Source data are public-use (IPUMS) extracts of the Decennial Census and American Community Service. Full details of the data available upon request.
Minimal requirements for a theory of interregional inequality
What should a sufficient theory be able to explain? First, it needs to explain why, in the 1940 to 1980 period, people were spreading out across the landscape, and why this led to interregional convergence in incomes. Second, it must explain why mainly college-educated workers started concentrating around 1980, and how this led to rising interregional inequality. If the story is about workers’ preferences, we need to know what caused preferences to change all of a sudden. If some other cause is in play – housing costs or crime or the quality or geography of culture, we again need to provide an explanation for what deeper factors changed in 1980, and why.
We lack such a theory. And without it, how can we hope to understand and address rising interregional inequality?
We contend that labor demand, rather than supply, is the direct determinant of both convergence and divergence. Behind demand, the deep cause is major technological change of the kind people commonly describe as industrial revolutions. Industrial revolutions change the basis for the organization of the economy. But they also have a particular geography. An industrial revolution occurred around 1980, when long-percolating digital technologies began to transform our world. The geography of this third industrial revolution was initially strongly concentrated, creating new firms and industries in particular places that demanded highly-educated workers. These firms created their own markets and reaped rewards that were partly shared with their workers. This set off the divergence we see today.
Why were incomes converging in the 1940 to 1980 period? Consider that today’s frontier technology is tomorrow’s incumbent. The early period of Figure 1 represents what happens to the geography of an economy when a once-revolutionary set of technologies matures. In this particular case, we are talking about the ideas that powered the second industrial revolution, centered on electrification and machinery. With the maturation of these ideas came widespread diffusion of core ideas and also jobs, and therefore incomes.
Does this mean our current phase of divergence will end? We don’t pretend to be fortune tellers, but there are reasons to be pessimistic about a return to patterns of the 1940-1980 period. The technologies at the core of the third industrial revolution will surely mature and disperse. But while higher transport costs meant that, in the mid-20th century, jobs dispersed to the South and West of the country, in the 21st many are more likely to move abroad as they become more standardized. On top of that, newly concentrated market power among incumbents like Google, Amazon, Facebook, Apple, as well as the ecosystem of venture-related finance, mean that the next revolution could stay captured in existing superstar cities.
So why is this framework useful? There is widespread concern about the geography of inequality, and a resurgent desire to address it. One thing we lack is a clear set of tools to do so. But before we can start talking about policy, we need to understand the phenomenon itself. It will be impossible to successfully address inequality if we cannot understand its causes. If, as we believe, the main explanations today emphasizing the supply of affordable housing or particular worker preferences turn out to be intermediate outcomes rather than deep causes, then a crucial first step has been taken in the wrong direction. Our work aims to be a corrective, refocusing research and policy attention in the right direction.
 Figure 1 weights by each location’s population. If we do not do so, then the post-1980 period looks like the end of convergence, but not the start of a divergence phase. Yet, given the large disparities in the populations of Commuting Zones (from a low of under 1,000 to a high of over 15,000,000), we believe it makes sense to take size differences into account.
Are you currently involved with regional research, policy, and development, and want to elaborate your ideas in a different medium? The Regional Studies Association is now accepting articles for their online blog. For more information, contact the Blog Editor at RSABlog@regionalstudies.org.