Inequality and location, location, location

Rebecca Diamond’s research touches on some of today’s most pressing issues: Gender pay disparities. Immigrants and innovation. The long-term impacts of rent control.

“I’m getting a lot of emails from reporters about that one right now,” the labor economist said.

Diamond, M.A. ’11, Ph.D. ’13, joined the faculty this fall as the inaugural Martin Feldstein Professor of Economics.

Feldstein, who died in 2019, was a public finance expert, adviser to presidents of both political parties, and president emeritus of the influential National Bureau of Economic Research. He taught at Harvard for five decades and presided for 18 years over the “Ec 10” introductory course.

Known for innovative uses of data and advanced econometric modeling, Diamond honors Feldstein’s legacy by lending empirical rigor to research with real-world impact.

In an interview with the Gazette, edited for length and clarity, Diamond recalled how the economics department shaped her studies and unique research approach.

---

As a graduate student, your advisers included Edward Glaeser, Lawrence Katz, and Ariel Pakes. How did they shape your path?

I learned empirical methodology in Ariel’s industrial organization course, while Larry’s labor course was much more topic-based. We had to write a term paper for Larry’s class, and he had given a lecture on spatial differences in local labor markets.

Meanwhile, I learned some techniques in Ariel’s class that I thought would be useful for thinking about how people decide where to live and how to quantify the value of different characteristics of cities.

I put those two methodologies together, and that was the beginning of … everything.

It turned into your 2016 American Economic Review paper on U.S. workers’ location choices. Can you offer a summary?

That paper was trying to understand the combination of two key facts.

One, from 1980 to 2000 the college wage premium went way up, with college-educated workers earning a lot more than those without bachelor’s degrees. At the same time, we saw an increase in what is called “spatial segregation,” where college-educated workers chose to settle in different cities than those without college degrees.

College-educated workers were specifically living in cities that were increasingly high-wage but also had increasingly high housing costs. On the one hand, you might think these housing costs eroded the consumption value of their higher wages.

On the other, why are college-educated workers choosing to live in these expensive places? Maybe some of the city’s desirable amenities were compensating them for the high housing costs. And if that’s the case, the increase in wage inequality could actually understate the extent of the total increase in inequality.

I found that a lot of the increase in spatial inequality was driven by changes in labor demand.

But I also found a snowball effect. When you get a higher share of college-educated workers in an area, you get more variety in restaurants and salons. You get lower crime, you get better schools, you get better air quality. That furthers segregation because college-educated workers are willing to pay more for better amenities.

Can you say more about the method you used to study these phenomena?

The part of my model that quantifies how people decide which city to live in, and how they trade off different characteristics — say, higher wages with higher housing costs — came from one of Ariel’s methods for estimating how people buy cars. Why do they buy a Ford over a BMW?

There are a lot of parallels. With cars, the model includes how much people value miles per gallon, how much people value air conditioning. With cities, my model included how much people value low crime, how much do people value air quality or school quality.

What do you see as the throughline connecting your work?

A lot of what I focus on is the role of geographic location and interactions with the housing or labor market that lead to heterogenous welfare effects for different groups of people. And then there are random papers that pick up what seems like important questions.

Like your 2021 paper on the gender pay gap for Uber drivers?

My co-author Paul Oyer, one of my colleagues at Stanford, came into my office one day and said, “I think I can get Uber data. What’s the coolest thing we could study?”

As a labor economic, I saw a unique setting for studying the gender wage gap. You know there can’t be discrimination in how the company is setting pay, because they use a formula.

What did you end up finding?

Despite the formulaic pay, there was still a 7 percent gender pay gap. Women tended to drive fewer hours per week than men, and that mechanically created large disparities.

We could estimate a learning curve separately for men and women and show they were identical. It’s not like women were learning slower. It’s just that they had spent less time driving. They were not as far up the learning curve.

And we were able to estimate the returns of learning. As you drive more, you get better at knowing peak pricing hours. You get better at knowing which neighborhoods to pick up in.

Part of it also had to do with the neighborhoods men and the women lived in. People usually pick up close to home. And for whatever reason, female Uber drivers tended to live in neighborhoods with slightly lower demand.

Men also drive faster, and Uber drivers get paid per mile.

You’ve studied another topic we’re hearing a lot about right now: rent control.

My co-authors and I looked at an expansion of rent control in San Francisco in the mid ’90s. A law, passed in 1979, gave rent control to all properties built as of that year with five units or more. In 1994, that was expanded to include buildings with two to four units. But because they were adding on to the old law, it was only buildings with two to four units built prior to ’79.

That created this nice treatment control. We could look at buildings with two to four units built in the ’70s relative to other two-to-four-unit buildings in the same neighborhood that were built in the ’80s. We could track the outcomes for tenants who lived in the properties when rent control was implemented. We could also track what happened to the properties over time.

What did you learn?

It does provide stability for those lucky enough to live in a building that suddenly gets rent control. We found they stayed in their apartments longer; they stayed in San Francisco longer than they would have. This was especially true for racial minorities.

But owners found ways to switch their properties to the owner-occupied market, where they can still command full market value. It just took a long time. It’s pretty challenging for owners in San Francisco to evict their rent-control tenants; they usually need to wait for them to move out on their own accord.

So over the long run, the supply of rent-controlled housing dropped 25 percent. So those who were renters when the law passed benefitted, but it hurt the next generation. They had fewer properties to choose from — and their rents were much higher.

Can you talk about your working paper concerning immigrants and innovation?

My co-authors and I used some clever data linkages to observe, for the universe of U.S. patents since 1990, which inventors are immigrants versus U.S.-born.

Something like 10 percent of the population in our dataset identified as immigrants. But they comprised 16 percent of inventors, and their share of patents was 23 percent. The average immigrant inventor is producing way more patents than the average American inventor.

Because most patents are not written by one person, just like with academic papers, we were also curious about externalities through collaboration.

To study this, you would ideally randomly force people to write a patent together and then study how that affects their subsequent careers. But we were able to approximate that by looking at untimely, early deaths of inventors and measuring the subsequent productivity of their collaborators.

We can do this separately for when the dying inventor is an immigrant versus when the dying inventor is U.S.-born. Both show substantial collaboration externalities, but they’re about twice as big for the immigrant inventors.

Putting that all together, we find that 32 percent of all innovation can be attributed to immigrants once you account for the fact that some patents by those born in the U.S. can be indirectly attributed to these collaborations.