Fall 2018 Institute Conference

Wrap-up: The 2018 Fall Institute Conference

The Fall 2018 Institute Conference focused on regional shocks and regional divergence. Read a wrap-up and watch videos of the conference

Day 1: October 18, 2018

12:00 p.m. CT | October 18, 2018

Lunch Keynote | Esteban Rossi-Hansberg

1:30 p.m. CT | October 18, 2018
Esteban Rossi-Hansberg

Esteban Rossi-Hansberg of Princeton University gave the keynote address on U.S. trends in economic concentration at national and local levels. He and his coauthors have found a growing divergence: At the national level, markets have been concentrating for 25 years. Top producers in most industries have become larger, with greater market shares. But at the local level, paradoxically, markets are simultaneously growing less concentrated in most sectors, and probably more competitive. Due to transport costs and imperfect substitutability, most markets are local and product-specific. Rossi-Hansberg’s presentation documented these trends for different industry sectors, at several geographic levels—the most local levels were the least concentrated. Diverging trends are most prevalent in retail trade, least in manufacturing.

More significantly, his research suggests that the trends are connected: Top firms expand—and increase concentration at the national level—by adding establishments that decrease concentration at the local level. By estimating zip code level effects on the number of discount department stores before and after a Walmart opening, for example, they find that when Walmart enters the local market, the number of stores actually increases. The same is true for most industries: Big companies grow by building local establishments, decreasing local concentration. This effect endures for at least seven years after entry, their research shows.

Rossi-Hansberg noted that there are no credible explanations for how decreasing concentration would cause higher markups and profits, and empirical research has found mixed results. “In sum, product-market concentration does not seem to be a problem,” he concluded. Of course, large firms could cause other problems. Big companies could exert monopsony power (sole buyers of production inputs can influence purchase prices, including wages for workers) and might use wealth to exercise political power for regulatory advantages.

1:45 p.m. CT | October 18, 2018

Panel 1: Placed-based Policies

2:15 p.m. CT | October 18, 2018
Janet Currie

Kicking off the panel on Place-Based Policies, Janet Currie of Princeton spoke about the local economic and welfare consequences of hydraulic fracturing.

There is much and divergent public policy around polluting industries. Her study with three other co-authors found that, generally, fracking increased employment, income, and wages in different locations – mostly rural and lower-income areas. But fracking also had various costs, including increased pollution, congestion and spending on infrastructure and public safety because of an increase in crime. There was generally not an increase in spending on health care or education.

One key fact in her findings: infant health is adversely affected for those families living within 1.5 kilometers away from a fracking source.

2:30 p.m. CT | October 18, 2018

Panel 1: Place-based Policies

2:45 p.m. CT | October 18, 2018
Nathan Hendren

Nathan Hendren of Harvard University presented findings from the Opportunity Atlas, a national analysis seeking to answer: “Which neighborhoods in American offer children the best chance to rise out of poverty?” This effort has received wide media and academic attention.

The study’s objective was to measure average outcomes (such as earnings, incarceration rates, college graduation) of children who grew up in each neighborhood by demographic sub-group (race, gender, parental income). The researchers created a sample of 20.5 million Americans now in mid-30s, mapped back at the Census-tract level (70,000 geographic tracts of about 4,200 people) to where they lived until they were age 23. This enabled focus on any tract to measure outcomes for different population groups. He covered the data sources, sample definitions, methodological difficulties and efforts to overcome them.

Using tools on opportunityatlas.org Hendren showed examples in Minneapolis and Atlanta to demonstrate wide national variation. But outcomes also differ widely at much smaller geographic levels.

He showed which variables have positive impact, but noted “This is based on kids born years ago. Are these patterns stable over time?” Their research leads them to believe “there is considerable persistence.” But is variation driven by causal effects of place, or selection — better-off people moving to nicer places?

Previous research suggests it’s mostly driven by place: moving to a good location improves outcomes. Current research similarly suggests that the earlier a child moves, the better their outcome. Again, is that causal? Statistical refinements indicate that, yes, the effect appears to be caused by the relocation. Why don’t more families move to better neighborhoods? Affordable housing could be one obstacle. “If we relax barriers, will families move to better neighborhoods?” Hendren asked. Research currently underway in Seattle hopes to answer this question.

3:15 p.m. CT | October 18, 2018

When Work Disappears: Manufacturing Decline and the Falling Marriage Market Value of Young Men

David Autor

David Autor of MIT explained the consequences of joblessness in poor communities on the formation of families and the availability of “marriageable” men. In 2014, 3/4ths of young manufacturing workers were men, compared to 2/3rds in 1970. Manufacturing jobs offer high earnings to low education men, partially due to higher wages, but also from long and stable hours. However, shocks resulting from trade policies in the 1990s and 2000s reduced the relative employment and earnings for young men. Autor also assessed and explained the downstream effects on marriage, fertility, and childhood poverty. Autor found that trade shocks have differential effects dependent on gender; men generally see larger negative effects in rates of employment and their earnings; additionally, men see an increase in idleness and mortality. For women, the effects aren’t quite as dire. Fertility is generally cyclical, that is, when the economy is stronger, the birth rate goes up. While an adverse shock to male earnings reduces the birth rate, an adverse shock to female earnings increases the birth rate. Additionally, a decrease in women’s earnings does not raise rates of childhood poverty; the data shows that declines in women’s earnings often lead to marriage and therefore, a dual income household.

4:00 p.m. CT | October 18, 2018

Panel 2: Land Use and Regulations

4:30 p.m. CT | October 18, 2018

The Effects of Rent Control Expansions on Tenants, Landlords and Inequality

Rebecca Diamond

Rebecca Diamond of Stanford University discussed rent control as a solution to affordable housing issues, using a 1994 law change in San Francisco to analyze address-level migration data for residents. This change provided a natural experiment of rent control expansion that affected all small multi-family structures built prior to 1980 in San Francisco. Tenants in these buildings, relative to those living in buildings not affected the rent control law, were 13-20% more likely to remain at their address in the medium to long term, with black and Hispanic renters most likely to remain (relative to white renters). Landlords of rent-controlled buildings tend to redevelop their buildings by razing or by converting to condominiums, thereby decreasing the rent-controlled supply of housing. Ultimately, Diamond found that rent control laws can prevent displacement, especially among minorities, but rent control policies tend to indirectly reduce the supply of affordable housing. Renters benefit in the short term, but are hurt the long run due to changing housing supplies and the possibility of being priced out of one’s neighborhood.

5:00 p.m. CT | October 18, 2018

Land-Use Restrictions and U.S. Macroeconomic Performance

Kyle Herkenhoff

Kyle Herkenhoff of the University of Minnesota spoke about the impact of land-use regulations on U.S. economic performance. He first noted that regional resource reallocation has long been a central feature of the American economy, with population shifting first from the East to the Midwest and Great Plains, then from rural areas to cities, and from the mid- to late-1990s, to California; by 1990, California was home to 12 percent of all Americans. This sequential reallocation reflected changing opportunities and amenities—people moved from locations that were less productive and desirable to areas that were more so. Recently, however, this trend has slowed substantially, with the interstate migration rate declining 40 percent from previous levels. And the slowdown coincides with slower economic growth, increasing housing prices and declining income convergence among U.S. states. From 1950 to 2014, employment in Rust Belt states declined dramatically as a share of national employment. Shares in California, Texas, the Pacific and Mountain West, and the South have climbed.

His research with coauthors links these three trends to tighter land-use regulations (such as zoning and environmental laws), noting that these regulations have progressively tightened, especially in highly productive states like New York and California, where housing prices are very high. If, hypothetically, urban land-use regulations were lowered nationally from current levels to 1950 levels, GDP would increase by more than $130 billion per year. In further experiments, they determine that rolling back land regulations from 2014 to 1980 levels would increase GDP and productivity about 6 percent by reducing housing and production costs, and by encouraging people to move to high productivity states. New York, California and mid-Atlantic states would gain population; the South and Rust Belt would shrink.

5:30 p.m. CT | October 18, 2018
Morris Kleiner

The last presenter on the panel was Morris Kleiner of the University of Minnesota. Much of Kleiner’s research focuses on occupational licensing practices and their effects as regulations on labor markets, and his talk served as a summary of some of this work. He began with a chart comparing the decline in unionization among American workers since the 1950s along with a concomitant rise in the share of the workforce required by regulations to have a license to work in their chose professions. This teed up why licensing is an important topic: it has become a more important factor in labor markets than unions or the minimum wage, both of which have been heavily studied by economists. Further, these regulations vary greatly from state to state, providing a rich data terrain for study. Kleiner then walked through the results of several of his recent studies on the issue. First, a series of papers with Alan Krueger and other co-authors has analyzed the influence of licensing on increasing wages for licensed professions. Meanwhile, other work has documented the rise in prices for services in licensed professions, however, the effect on the quality of those services is less clear.

Of particular interest to many conference participants was a paper Kleiner co-authored with Janna E. Johnson documenting the influence of licensing practices on a phenomenon that has been heavily studied recently—the recent decline in U.S. interstate migration. Kleiner and Johnson’s study compared migration rates across professions and found that migration was lower for professions that were subject to more stringent licensing requirements. This effect appears larger for professions that have greater state-specific licensing policies, suggesting that licensing policies have played a role in reducing interstate migration by raising the costs of moving. Finally, Kleiner discussed new work with Evan Soltas on the overall effects of state licensing policies. Licensing increases wages and hours for individual workers in licensed occupations, but reduces total employment and hours worked (by raising barriers to entry), so what is the net effect? Kleiner’s results suggest there is a net loss in welfare from these policies.

6:15 p.m. CT | October 18, 2018

Dinner Keynote | Amy Finkelstein

7:45 p.m. CT | October 18, 2018

Geographic Variation in Health and Healthcare

Amy Finkelstein

Amy Finkelstein of MIT discussed her research on healthcare spending and outcomes, and began by asking why there is virtually no correlation between healthcare spending per capita and mortality. One explanation is that people are different—they have different experiences and preferences (what Finkelstein calls the “demand factor”) and another is that places are different (“supply factor”). These two possible yet different reasons intuitively have different implications. The literature in this space has found that healthcare differences are related to preferences and healthcare spending is related to place. The theoretical assumption Finkelstein operated from is that place has an effect on the person; the goal, then, is to isolate these place-specific treatment effects and to use them to conduct a counterfactual analysis. Finkelstein examined people who moved from one healthcare spending area to another to identify the impact of a person’s place on their healthcare outcomes. The project’s data set comes from Medicare claims data that shows when people move from one area to another. Finkelstein uses opioid abuse as a case study to better understand healthcare outcomes when people move; the data suggests that when opioid abusers move to a county with a 20% higher rate of opioid abuse, the rate of abuse increases by 6 percent. For healthcare, life expectancy, and rates of opioid abuse, place matters.


Day 2: October 19, 2018

8:30 a.m. CT | October 19, 2018

Panel 3: Regional Economic Activity: Theory & Data

9:00 a.m. CT | October 19, 2018
Elisa Giannone

In the day’s first panel, Elisa Giannone of Penn State University kicked things off by presenting research on the change in the wage premium for high skilled workers across cities. The motivating facts were that prior to 1980, wages were converging across U.S. cities (cities with lower wages saw faster wage growth while higher-wage cities saw slower wage growth), but this convergence stopped after 1980; the correlation between wage levels and growth disappeared. But this pattern is different depending on skill levels—regional wage convergence among workers without a college degree has continued since 1980, but not for workers with a college degree.

It’s tempting to draw a connection to several other well-studied phenomena: The increasing dispersion of wages between high-income cities like San Francisco and lower income cities, the increasing premium to education and training due to “skill-biased technical change” (SBTC), the tendency of workers and with similar types of skills and industries that hire them to locate in the same place (spatial agglomeration), and the long-run decline in migration across U.S. cities. Giannone presented a spatial economic model that can account for SBTC and geographic agglomeration along with other factors like migration and differences in housing and other amenities. Using data to estimate the model and then run some counterfactual “experiments,” the model suggests that SBTC and spatial agglomeration are responsible for most of the decline in wage convergence across cites, with housing and migration responsible for less. Interestingly, the model also goes some way in accounting for the increased wage dispersion between cities and reduced migration, though it wasn’t intended to explain those phenomena. Giannone then presented some preliminary results of new research that also accounts for these patterns between cities in other countries than the U.S.

9:30 a.m. CT | October 19, 2018

Unpacking complexity in workplace skills

Morgan Frank

Morgan Frank of MIT presented his research on the differential impact of artificial intelligence (AI) and how job skills are clustered in geographical locations. AI’s impact varies based on how much of the job requires cognitive skills; that is, some workers are augmented by technology while others are replaced, and relatedly, small cities face a greater impact from AI than do big cities. Frank’s research finds that there exists a high polarization between types of workplace skills. By using networking models to cluster work skills based on how similar they are to each other and each other’s neighbors, Frank found there are two basic skill types: sensory/physical skills and social/cognitive skills. Occupations that rely on social/cognitive skills tend to make higher and higher wages, relative to jobs that require more physical skills. The data also show that particular skills are connected to spatial mobility, both in terms of permanent migration and frequency of business air travel. In essence, different types of skills can demonstrate how easily or not it is to move between job titles and how that affects a person’s spatial mobility. Using big data to model the network of relatedness between job skill categories helps to explain the distribution of job types in cities, which ultimately accounts for the disparities between city economies.

9:50 a.m. CT | October 19, 2018

Panel 3: Regional Economic Activity: Theory & Data

Left to right: Elisa Giannone, Morgan Frank, Simeon Alder, Katarína Borovičková

10:00 a.m. CT | October 19, 2018
Simeon Alder

The last presenter on the panel was Simeon Alder from the University of Wisconsin-Madison, who discussed joint work with Lee Ohanian and David Lagakos on the role of labor market conflict on the economic decline of the “Rust Belt” region of the Great Lakes and the Ohio River Valley. He began by documenting four facts of the region’s economic history: The Rust Belt had a large share of national economic activity after World War II which declined slowly and steadily, wages in the region were substantially higher at the beginning of this period, relations between labor and management were more prone to conflict (strikes and lock-outs), and the industries in the Rust Belt saw weaker productivity growth over this period. Further, some data suggest this decline was not simply due to a general decline in manufacturing or in certain industries which happened to be concentrated in the Rust Belt but appears more specific to manufacturing in the Rust Belt itself.

This suggests a connection to research by Jim Schmitz and others about the relationship between work rules and other policies that restrict competition with reduced innovation and productivity growth. Alder then sketched a model in which there are two geographic regions—the Rust Belt and the rest of the country—where the labor market in the latter is non-competitive while it is competitive in the latter (the model also has two sectors, a manufacturing and a services sector). When running the model, calibrated to match some of the other empirical parameters such as the change in productivity between foreign and U.S. manufacturing and the wage premium in the Rust Belt, it also reproduces the decline in the Rust Belt’s employment share. Alder said this suggests labor conflict is an important part of the story in the decline of the Rust Belt.

11:15 a.m. CT | October 19, 2018
Oren Cass

The Manhattan Institute’s Oren Cass began the conference’s final panel by discussing his forthcoming book The Once and Future Worker. His primary contention was that consumer welfare—measuring people’s well-being by their incomes and what they can buy with them—is an insufficient metric for understanding the experiences of many people and communities left behind by structural economic transitions over the last generation. Understanding people’s welfare as workers, he argued, is just as important. His presentation focused the economic performance of regions that he termed prosperous (mostly big cities) versus those that are distressed (primarily rural and small towns), and the much-discussed impact on the American political climate. In particular, he criticized the view that our current political conflict is primarily between those who favor openness and economic dynamism on the one hand and a closed, static worldview on the other.

Cass then turned to the conventional policy solutions that arise from the consumer welfare focus—geographic mobility (moving from underperforming regions to successful ones), retraining workers, and transfer payments. Government payments as a share of local income have dramatically increased in distressed areas, and Cass argued that this was an illustration of the insufficiency of looking only at consumer welfare. Under that view, the fact that the overall economy was growing and that the winners were compensating the losers could be viewed as a success story. But most people wouldn’t see it that way, and for good reason—meaningful work and the capacity for self-sufficiency are more important to people than simply their value in providing an income. He also argued that the emphasis on automation as a driver of this transition was misplaced, and that trade was a more important factor, echoing David Autor. He closed by briefly discussing policy implications of his view, such as trade restrictions and emphasizing broader workforce goals in education than college attainment.

10:45 a.m. CT | October 19, 2018

Panel 4: Social Consequences of Regional Disparities

11:45 a.m. CT | October 19, 2018
Yannis Ioannides

Yannis Ioannides’ (Tufts University) talk had a deceptively simple title, “Social Interactions Do Matter; Two examples.” His presentation was a technical analysis in two examples of how economies are affected by social interactions, culture, values and institutions. “Social interactions are ubiquitous and central to the urban economy.” His first example was how adoption of Information and Communication Technologies (ICT) affects city-size distribution. (A nation’s city population sizes usually follow an intriguing pattern, the rank-size rule, which says that if a country’s cities are ranked top-to-bottom by population, the second largest city will have half the population of the largest, city No. 3 will have a third the population of No. 1, and so on.) Ioannides’s research finds that as nations adopt ICT, this distribution becomes less dispersed, increasing decentralization—people grow farther apart. (If social networks grow, it’s through technology, not face-to-face.)

The second example was focused on the interaction of “non-economic” factors such as culture, politics, history and institutions (e.g., enforcement of property rights) and economic growth. His is a theory of corruption and rent-seeking, given history, culture and institutions “which operate, critically, through social interactions.” This theory explains corruption and rent-seeking as outcomes, but also as influences on long-term economic performance. He supplemented his theoretical model with empirical analyses, regressions measuring the effect on per capita GDP of corruption, institutions, human capital and culture. Results largely confirm theoretical predictions: lagged corruption and collectivistic culture (rather than individualistic) have negative impact on growth; institutions and human capital have positive impact.

Further empirical analysis estimated the impact on individual institutional growth of corruption, per capital GDP, and human capital. Here, theory implies that previous levels of corruption (“norms”) lead to weaker institutions without necessarily reducing per capita GDP. Results confirm: “Bad norms tend to reproduce themselves.” Therefore, change is difficult: “Getting to be Denmark is not exactly easy.” For institutions as a whole, regressions further confirmed that lagged corruption and collectivism had a negative impact though lagged per capita GDP and human capital were largely positive.

Overall conclusion: Social interactions do indeed matter—for city-size distribution, institutions, corruption, rent-seeking and economic growth.

12:15 p.m. CT | October 19, 2018
Abigail Wozniak

In her presentation, “Coming and Going: Increasing Geographic Mobility at College Entrance and Exit,” Notre Dame’s Abigail Wozniak argued that college attendance is less likely for those who live far from campuses. That leads to disparities between those who live close to and those who live at a distance from post-secondary institutions. It shows that skill levels in a local economy are linked to the proximity of a post-secondary institution. And, because there is general decline in worker mobility, this further leads to disparities between locations themselves as those with college degrees concentrate in larger labor markets.

Geography is a determinant in individual economic opportunity, she stated, and, to the point, the data is stark about college attendance and college proximity. For example, the majority of public four-year college students attend an institution within 50 miles of home, and around 70 percent attend within 100 miles of home. And that’s significantly higher than students’ attendance patterns decades ago.

Similarly, there’s less geographic migration for college graduates. These newly educated workers stay close to their higher education location.

It all adds up to challenges to encourage potential students who live far from colleges to attend college at all and, further, to incentivize those who leave college to migrate for work.

She offers two proposals. She proposes increasing the Federal Pell Grant Program up to $5,000 per student to help those who don’t have access to a local college to overcome the costs of relocating for college. She estimates that would cover the costs of college living for a year.

She also proposes that college grads receive payback deferments to the Federal Student Loan (FSL) program when relocating as they start their careers.

In her view, there would be a substantial return on investment as it will increase two-year and four-year college goers in the labor market, and it will increase income opportunity for individual college grads and even for those who don’t complete their degrees.

The full details of Wozniak’s proposal have been previously published by the Hamilton Project.

1:00 p.m. CT | October 19, 2018

Lunch Keynote: Edward Glaeser

2:15 p.m. CT | October 19, 2018
Edward Glaeser

Edward Glaeser gave the closing keynote, a description and tentative prescription for what he called “The Geography of Joblessness,” in joint research with Ben Austin and Lawrence Summers, also of Harvard University. He began with evidence of long-term decline in U.S. employment and labor-force participation rates for working-age men. About 5 million fewer men are in the labor force now than in 1950. This is among the most worrisome failures of American society, said Glaeser.

Joblessness is not uniform. It’s concentrated among unmarried, less-educated men of recent cohorts. Geographic patterns of joblessness and economic growth fall generally in three regions: Coastal states (19 of 48) have done very well in GDP growth, Eastern heartland states (12) not on coasts have done poorly and Western heartland states (17), have recently climbed from worst-off to best. In joblessness, the Eastern heartland is again worst-off and Western best, though employment growth since the late 1970s has been best in the Coastal states.

What should be done to address joblessness? Should regions doing poorly receive government assistance? Economists have been skeptical about such “place-based” government strategies, believing they keep people in dysfunctional economies, artificially increase rents to the disadvantage of the poor, and create pockets of high unemployment and low human capital. Finally, infrastructure spending can lead to “monumental waste.”

Still, there are arguments for such policies. Place-based spending can generate externalities with broad benefit and provide social insurance for low-income people. Glaeser is skeptical about these. But more plausibly, place-based policies make sense if applied flexibly. The marginal impact of an employment subsidy is probably higher in West Virginia than in Seattle. Indeed, increases in labor demand seem to have more employment impact where joblessness is historically high, suggesting that pro-employment policies used in such areas could be most effective there.

This work is suggestive, not definitive. Again, the key point is that place-based policies should adapt to local conditions, not be applied uniformly. Social insurance programs, employment subsidies, educational interventions and regulatory reforms should all be “place-specific.” What would he not encourage? “Infrastructure and wholesale attempts to move economic activity.”