Economists Can Change Minds with Data

This post is the fourth about discussions at the 2021 Women in Economics Symposium, which focused on ways that economists make a difference in the world. The first two posts, Inspiring Young Women to Pursue Economics and How Economists Make a Difference, gave overviews of the event’s goals, topics and speakers, and the third, Using Economics to Solve Puzzles, highlighted the keynote speech of economist Lisa D. Cook.

How can economists make a difference in the world? One way is by changing minds with data, said Abigail Wozniak at the Women in Economics Symposium earlier this year.

Wozniak, who is the director of the Minneapolis Fed’s Opportunity & Inclusive Growth Institute and a senior research economist, began her keynote presentation by providing her definition of economics.

As she put it in her presentation slides, economics is “a practice of using theory, data, and observation to interpret human behavior… while allowing that one or more of these could be wrong.” During her remarks, Wozniak gave examples that she said illustrate the power data have to change minds.

“The end goal is not necessarily to always turn our assumptions and our understanding upside down,” she emphasized. “But it’s really to come back to that part of the practice of economics where we’re asking ourselves: ‘Is this correct, is this right, or is something I’m seeing or thinking misleading or incorrect?’”

That is why changing minds is important, she said.

“If we’re not changing minds at least sometimes, then we’re proceeding from some kind of basis of really, frankly, fantasy, because that would suggest that we all know all the answers already,” Wozniak said. “So, we have to have some of that turnover in what we think.”

Wozniak gave three examples of times when she and other economists changed minds with data. She focused on:

1. Declining geographic mobility
2. The impact drug testing has on employment
3. Origins of COVID-19 health disparities

Declining Geographic Mobility

Wozniak’s research on declining geographic mobility (PDF) in the U.S. changed some widely held views, she said. Geographic mobility refers to moving over a long distance—for example, to a different metropolitan area, a different state or even a different coast.

She noted that many people have the impression that migration and long-distance moves are common in the U.S., and some people were concerned that the 2007-09 housing crisis would reduce geographic mobility.

Coming out of the crisis in the late 2000s, many homeowners were underwater on their mortgages—meaning they owed more on their homes than they were worth. There was concern that these underwater mortgages would make it difficult for homeowners to relocate, which would hamper the economic recovery because people would be less willing to leave places that had been hard hit by the crisis and move to places that were starting to improve, Wozniak explained.

She and her co-authors decided to look into whether people were in fact moving less often in light of the housing crisis. They found that there were some downward blips in migration rates, but that long-distance moving had been declining pretty steadily since 1980.

Their findings were surprising, she said.

“It kind of confirmed that what people were really worried about was happening, but not in a way that was nearly as pronounced as they were expecting,” Wozniak said. “People were moving less at the trough of the housing crisis, but it really was much more a continuation of a very long-run pattern that had been happening for decades and that hadn’t really been widely discussed.”

Eventually, people were fairly persuaded by the data that people are moving less than they had in the past, she said.

The Impact Drug Testing Has on Employment

The second example Wozniak discussed was from her research on whether the expansion of employer drug screening was creating a barrier for those who were seeking employment—particularly for Black men. In this case, she said, the data changed her own mind.

She had done focus groups with people who were trying to move from unemployment to employment. She said that people in the groups frequently cited employer drug screening as a challenge in trying to get a job.

She investigated by looking at states that had encouraged employer drug testing and those that had limited employer use of drug testing, which she called “pro-testing” and “anti-testing” states, respectively.

As she viewed employment patterns in high-testing industries, she found differences between pro-testing and anti-testing states. The results surprised her because they were different from what she had expected based on her conversations with the focus groups.

In pro-testing states, she found that employment for Black workers actually rose relative to that of white workers after those states passed legislation encouraging the testing. She found the opposite in anti-testing states. In particular, while employment was declining for both groups, it declined more sharply for Black workers than it did for white workers after these states started restricting testing.

“The conclusion from this is that that impression that I had from conversations with folks was actually not borne out in the data, and in fact the opposite seemed to be true,” Wozniak said.

In exploring potential reasons for her findings, she said that the explanation she settled on in her paper was employer bias: Employers incorrectly assumed higher drug use among Black applicants. But she said that when testing was implemented, they were able to verify that the assumption was incorrect, which changed the employment patterns.

Origins of COVID-19 Health Disparities

For her third example, Wozniak highlighted work from other researchers, including one of her colleagues at the Opportunity & Inclusive Growth Institute, who looked at the origins of COVID-19 health disparities. In this case, data were used to change minds quickly and to change people’s perception of how the pandemic was unfolding, she said.

She noted that, early in the pandemic, Black and Latino Americans tested positive for COVID-19 at higher rates and also had higher mortality rates than white Americans. An idea took off in the media last summer that the virus was more lethal for Black and Latino Americans because of underlying health disparities, Wozniak said.

The researchers looked at other hypotheses that might help explain why infection rates were higher for Black and Hispanic neighborhoods in New York City, Wozniak noted. They used cell phone data to examine the living and working arrangements of large numbers of people—for example, how far the residents had to commute for work, and how densely populated their households and workplaces were.

The researchers found that a big part of the differences in COVID-19 positivity rates was due to factors other than race and ethnicity, Wozniak said.

“What that means is that all of these other things that go into the neighborhood differences are really the things explaining test positivity and not something about populations that are unrelated to work and commuting and living arrangements,” she explained.

She went on to say that as a result of this work and that of others in public health, we now hear much less about underlying health disparities being a big factor behind the infection disparities. “Instead, we’ve come to understand pretty quickly that economic circumstances are feeding into this much more strongly,” she said.