Edward Glaeser on Cities
What this covers
Ed Glaeser of Harvard University sits down with Russ Roberts to trace why American cities rise and fall. The core argument centers on Detroit's twentieth-century trajectory: a city that organized itself around vertically-integrated mass production, which excelled at near-term productivity but proved catastrophic for long-term adaptation. Glaeser frames the city's decline not as inevitable but as a policy choice—betting on infrastructure and large firms rather than on human capital, education, and the conditions that allow small entrepreneurs to thrive and pivot when industries shift. The conversation moves from this historical anatomy into a comparative analysis of cities that did rebuild successfully—New York, Boston, Seattle, Silicon Valley—and identifies what separated them from Detroit's fate.
The discussion spans several distinct territories. Glaeser develops the case for Agglomeration / Idea Spillovers as the engine of urban success, anchoring it in patent citation studies and the persistence of geographic clusters despite long-distance communication. He then pivots to federal policy, arguing that Removing Anti-Urban Policy Biases—highway subsidies, mortgage interest deductions, zoning that privileges single-family houses—systematically drains cities in favor of sprawl. On Detroit specifically, Glaeser proposes Shrink to Greatness: consolidating city services into dense cores rather than maintaining infrastructure across depopulated terrain. Roberts introduces a counterpoint, questioning whether idea exchange is really the driver or whether small-firm density works mainly through lower search costs for hiring. Glaeser closes by distinguishing between what national policy should do—establish good baseline conditions, not pick winners—and what mayors must do: deliver schools, safety, and functional transit to make a city worth staying in.
Glaeser argues that cities thrive when they are dense networks of small firms and skilled, entrepreneurial people exchanging ideas, and that declining cities like Detroit fail by betting on large vertically-integrated firms and infrastructure rather than investing in human capital, education, and basic services.
- Detroit's decline traces to a large-firm industrial monoculture that produced middle managers rather than reinventive entrepreneurs
- Successful cities (New York, Boston, Seattle, Silicon Valley) rebuilt around information-intensive industries that reward face-to-face idea exchange
- Good city policy means schools, safety, and removing anti-urban biases (transport, housing, education zoning), not infrastructure spectacle
Cities succeed through geographic clustering that enables informal idea spillovers among skilled workers and entrepreneurs.
- The valuable idea exchange in creative work occurs not in formal seminars (which present ideas at a well-developed stage) but in ordinary informal conversations—walking down the hall, stopping by an office, having coffee with a colleague—while one is still figuring out the right topic and empirical strategy.
“seminars are ideas at a fairly well-developed stage. I think a lot of the sort of creative side occurs well before your paper is designed... That occurs in ordinary conversations, while you are walking down the hall”
- Cities are humanity's greatest invention because almost all of the greatest hits of human achievement—Athenian philosophy, Renaissance art, the mass-produced automobile—were collaborative and arose within clusters of genius where people learned from one another face-to-face, leveraging our best asset as a species: the ability to work with others.
“the life's blood of the city is enabling us to benefit from people around us... our best asset as a species is being able to work with others.”
- The irregular clustering of artistic genius in particular places and times (Florentine Renaissance art, French Impressionism) is explained by spillovers across people—one person's idea is played with and expanded by others—rather than by superior water, or even patronage like the Medicis, much of which postdates the artistic peak.
“That clustering is readily explainable because of the spillovers across people. Because one person comes up with an idea and then other people play with it.”
- Globalization increases the return to having a good idea because it can be sold across the whole planet, which paradoxically strengthens rather than dissolves geographic clusters—Hollywood remains tightly rooted even as American movies like Avatar sell in the poorest parts of India's cities.
“globalization increases the return to having a good idea, because you can sell it across the whole planet. That is nowhere more tangible than in Hollywood”
- New long-distance communication technology has not made face-to-face contact obsolete; the most information- and technology-intensive industry (Silicon Valley) is also the world's most famous geographic cluster, and Google—maximally connected by long-distance communication—deliberately builds the Googleplex with few walls to maximize face-to-face connection among its own employees.
“all this new technology that enables long distance communication isn't making face-to-face contact obsolete. The clearest example I can think of is Silicon Valley itself.”
- Hard evidence for inter-firm idea spillovers comes from patent citation studies (Jaffe, Trajtenberg, and Henderson) showing patents are more likely to cite patents that are geographically close to them, even when they are not in the same company.
“we have the hard evidence of patent citation, which Jaffee, Trajtenberg, Jaffe and Henderson find are more likely to cite patents that are geographically close to them even when they are not in the same company”
Federal and local policies systematically subsidize sprawl and single-family housing while penalizing urban density and multifamily construction.
- Federal policy systematically stacks the deck against cities: highway spending massively increases the accessibility of far-flung areas (pulling people out) while transit aid like a Detroit people-mover does almost nothing, and pro-home-ownership policy (mortgage interest deduction, Fannie Mae/Freddie Mac) subsidizes owning, which favors single-family houses since over 85% of multi-family units are rented while the same share of single-family units are owner-occupied.
“highway spending really has a massive difference in the accessibility of far-flung area. There are huge differences created by running a highway down the area. So that really pulls people out. Whereas adding a people-mover to Detroit does virtually nothing.”
- Land-use regulation that began sensibly (19th-century fire risk limits) has metastasized into unnecessary, counterproductive regulation captured by interest groups—suburban NIMBYism and ever-expanding historic preservation districts (15% of lower Manhattan excluding Central Park) that act as no-building zones—which makes cities far more expensive than necessary and creates income uniformity among those who can afford the housing.
“many localities have let themselves be conquered by various interest groups in making it far too difficult to build... the net impact of these things is that they make cities far more expensive than they need to be.”
- The way local education systems tie school quality to residential location creates a strong, government-created incentive to suburbanize; a region-wide school-choice system (public or charter, anywhere in the region) would largely break those locational incentives, though it is politically infeasible.
“we've created such a strong schooling incentive for people to move out of those cities that have weak school systems.”
- Cities are better when they change and develop new space rather than being preserved as museum pieces; preservation should be capped at a fixed number of buildings rather than allowed to expand into ever-larger districts, and most barriers separating commercial and residential use should be removed since modern urban activity is fairly clean and free of major externalities.
“I don't much like the idea of cities being museum pieces... I think it's good that cities change and that they develop new space”
Detroit's decline stemmed from industrial structure and policy design rather than external economic forces alone, and recovery requires shrinking densely, charterizing schools, and improving basic services.
- The comeback of older cities since the 1970s was driven mostly by economic change, not city policy: globalization and new technology raised the returns to skill and the value of information-intensive industries that have always benefited from urban proximity—finance in New York (43% of Manhattan payroll in 2007), biotech and sciences in Boston, and software/retail entrepreneurship in Seattle (Amazon, Microsoft, Costco, Starbucks).
“I tend to think of this as being about what globalization and new techogies did to the returns to skill, the values of high end information-intensive industries that have always had a particular reason to be in urban areas.”
- Tom Holmes's work shows that differences in state policies toward unions and labor were remarkably powerful in explaining industrial growth after 1947, helping drive jobs from cities to lower-cost areas.
“The work of Tom Holmes at the U. of Minnesota shows how remarkable the difference is in state policies towards unions, labor, how powerful those policies were in explaining industrial growth after 1947.”
- Between 1998 and 2010, paid employees in Wayne County fell from 755,000 to about 570,000, manufacturing employment halved from roughly 120,000 to 60,000, while healthcare and social assistance—an industry traditionally driven by public sector spending—now accounts for 100,000 of the county's 568,000 jobs.
“Between 1998 and 2010 the number of paid employees in Wayne County declined from 755,000 to 570,000... the number of manufacturing employees... has literally declined by 50%, from about 120,000 to about 60,000.”
- Detroit should 'shrink to greatness' by restricting where city services are available, charging the incremental cost of services to those outside compact core areas (or not providing some services there), pressuring people to relocate to denser areas, and selling off remaining space to private developers—combined with near-complete charterization of schools and borrowing successful policing tactics from cities like New York.
“I would do more to shrink to greatness, of trying to make the city's physical footprint somewhat smaller, use more private competition the school side, and do more in terms of borrowing from other cities on police side.”
- The hallmark of declining cities is funding structures and infrastructure far beyond actual demand—Detroit, where over 90% of central-city homes are valued well below construction cost, gained nothing from federally subsidized urban renewal housing or a monorail gliding over empty streets; what it needed was investment in schools and safety.
“The hallmark of declining cities is to have funded structures and infrastructure relative to the level of demand in that city.”
Small firms and entrepreneurial human capital drive urban reinvention better than large vertically-integrated firms that cannot pivot to new industries.
- Today's Silicon Valley giants like Google and Yahoo increasingly resemble Detroit's large firms: deeply concerned with exchanging ideas within their companies but much less likely to have employees connect with people in other companies—a contrast to the 1960s when many little firms talked to each other, as documented by Saxenian's comparison with Boston's walled-off Route 128 cluster.
“these are much more like Detroit's giants in the sense that they are very large firms. They are firms that are deeply concerned with the exchange of ideas within their companies but much less likely to have the people in those companies connect with people in other companies”
- Detroit's decline was seeded by its central innovation—the vast vertically-integrated factory—which is great for short-run productivity but bad for long-run reinvention, because such plants don't need or give to the city, can be relocated to lower-cost areas, and crowd out small entrepreneurs in favor of middle managers who can't pivot to new industries.
“The problem was the big idea was a vast, vertically integrated factory. And that's a great recipe for short run productivity, but a really bad recipe for long run reinvention.”
- It is built into the genes of every industry that it will eventually decline; the real question for a region's survival is whether it can reinvent itself, which depends on skills (human capital) and especially entrepreneurial human capital rather than dominance by a few large firms.
“I think it's built into the genes of every industry that they will eventually decline. The question is whether or not the region then reinvents itself.”
- Weak proxies for entrepreneurial talent (establishments per worker, share of employment in startups) are nonetheless strong predictors of urban regeneration: places with many little firms have done much better than places dominated by a few large firms, because small entrepreneurs are adept at switching industries when one declines while middle managers of giant firms are not.
“places that have lots of little firms have managed to do much better than places that were dominated by a few large firms, particularly if they are in a single industry”
Educational attainment predicts city recovery, but mayors cannot manufacture human capital; they must focus on schools, safety, and basic functionality.
- Telling a mayor that the path to city success is 'more college graduates' is only slightly more helpful than reminding them of the predictive power of January temperature—there are limited short-run levers, so the advice really pushes back toward basics: good schools, safe streets, decent commutes, since educated people want the same things.
“it's only slightly more helpful than reminding the mayors of the power of January temperature in predicting areas' success, which is an even more useless fact”
- Following Enrico Moretti's work, national policy to establish land grant colleges in particular places before 1940 appears to have had lasting effects, with those places enjoying a very good thirty-year run after 1980—though this is not a cost-benefit analysis and offers little a mayor can act on.
“the national policy to establish land grant colleges in particular places does appear to have had lasting effects”
- Policy should focus primarily on people in Detroit—ensuring its kids are well-educated and find a brighter future wherever that is, even if they leave—rather than on propping up the place itself; the federal government should set good policies and not pick winning or losing cities, just as it shouldn't prop up individual companies, while mayors retain a place-specific duty to make their city as functional as it can be.
“our primary focus is on people in Detroit, whether or not they choose to live in Detroit or elsewhere. So, the important thing is that those kids are well-educated and find a brighter future wherever they find a brighter future.”
- In most American cities the failure of public services like schools is not about the sheer dollars spent but how those dollars are spent; the best case for spending more is usually that it's the only way to buy off interest groups needed to enact reform.
“It's not the sheer dollars involved in most of American cities, it's the way those dollars are being spent.”
- The share of an area's population with a college degree (measured as of 1940, 1960, or 1970) is a very good predictor of whether northeastern and northwestern metropolitan areas have been able to turn themselves around.
“The year, the share or the fraction in the metropolitan area with a college degree as of 1940 or 1960 or 1970 has been a very good predictor of whether, particularly northeastern or northwestern metropolitan areas, have been able to turn themselves around.”
Infrastructure and user fees should replace general taxation; if projects cannot be user-funded, that signals they lack genuine value.
- Almost all infrastructure can and should be paid for by its users rather than general tax revenues—a crumbling bridge or an airport should be funded by tolls and fees on those who use it, not distant taxpayers; if a project can't be funded by user fees, that is itself reason for skepticism about its value, a principle Adam Smith stated 240 years ago as the best defense against white elephant projects.
“almost all infrastructure can be paid for by users rather than the general tax revenues. I just cannot possibly see why we think that the crumbling bridge is a job for taxpayers far away rather than the users of that bridge.”
- Cities genuinely need government because density produces negative externalities: unclean water historically required public intervention (the Croton Aqueduct turned cities from killing fields, where a boy born in New York in 1900 was expected to live 7 years less than the national average), and a public street grid creates enormous holdup problems that necessitate public congestion management like congestion charging.
“cities do need, because there are these negative externalities associated with density, they actually do need government.”
Local governments should not run aggressive social safety nets or education systems because redistribution and service delivery drive away high-human-capital residents and entrepreneurs.
- Localities should be wary of running their own social safety nets because of the ease of exit for companies and richer people; aggressive local redistribution (as Detroit pursued after 1970) drives away high-human-capital residents.
“there are lots of reasons why we should be wary of localities trying to run their own social safety nets, the ease of exit for companies and for richer people”
- The main reason cities don't aggressively privatize education, services, and land to revitalize is politics, not the difficulty of the ideas: for a mayor to admit he can't manage part of the city feels like an admission of failure, and interested parties in education and policing wield enormous power to preclude large-scale change.
“I think the main reason why you don't see more of this is indeed politics. And for a mayor to admit that... he can't manage part of the city is an admission of failure”
- The more troubled a school system is, the stronger the case for going almost entirely private; Detroit should contemplate moving nearly completely out of the public school business toward charter and voucher-like systems, since competition naturally produces delivery of economically valuable skills.
“the more troubled a school system is, the more the case is for going almost entirely private. Which I think would not be a crazy thing for Detroit to contemplate.”
City governance is pragmatically focused on service delivery rather than ideology, showing little partisan difference in outcomes.
- 120 years ago Detroit looked like one of the most entrepreneurial places on the planet with an automotive genius on every street corner, and 60 years ago it was among the most productive places on the planet.
“If you look back 120 years ago or so, Detroit looked like one of the most entrepreneurial places on the planet. It seemed as if there was an automotive genius on every street corner.”
- Because cities are involved in very tangible outputs, they tend to be less ideological and more driven by delivering basic services; regression-discontinuity evidence (Ferreira and Gyourko) comparing cities where a Republican or Democrat barely won shows little partisan difference in outcomes—'there's no Democratic or Republican way to clear out the trash.'
“what they find is very little difference in the center between the two parties. Which goes back to the old line that there's no Democratic or Republican way to clear out the trash.”
- Rather than subsidizing artificial tech clusters, cities should make land-use regulation near universities friendly to building relevant space—not zoning it off for single-family detached houses—so that the documented spillovers from universities can actually be captured.
“you want to make sure that your land use regulations or spaces close to those universities are relatively friendly for building new relevant space. So it's not an issue of subsidies for these areas, but it's an issue of making sure you haven't zoned it off for single family detached houses”