Constantine Yannelis: Who Has Been Stockpiling during the Crisis?

Constantine Yannelis is assistant professor of finance at Chicago Booth. This transcript is taken from an interview conducted April 2, 2020.

How has COVID-19 shaped consumer behavior?

I have a new study—with coauthors at Columbia, Northwestern, and the University of Southern Denmark—demonstrating that this crisis has had profound effects on consumer behavior. First, we saw a massive increase in consumer spending, from early until mid-March, that is consistent with stockpiling. We saw increased spending on groceries as well as air travel. Presumably, people were leaving urban centers or trying to quickly get back to their families, anticipating shelter-in-place orders.

Now, this changed drastically toward the middle and the end of March, when we saw a corresponding sharp decrease in consumer spending. This was quite severe in many categories. For example, restaurant spending collapsed sharply, as did spending on public transportation and air travel, and in a number of other categories. We find that this decrease in spending was much more severe in states and cities that issued shelter-in-place orders. A couple of other things actually increased—for example, spending on food-delivery services, consistent with this being driven by COVID-19 and the corresponding massive changes to individuals’ lifestyles.

We explored heterogeneity across groups of consumers by demographics, income, and political affiliation. Just reading the news, [you see that] there are all kinds of stories about people behaving in different ways. We observed this with younger people: there was a delayed response, consistent with them not taking things as seriously early on. This could also be due to differential risk exposure, as COVID-19 doesn’t hit young people as badly as it does old people, on average.

We also find that families with children were likelier to stockpile. We find little effect by income, with high- and low-income households behaving quite similarly in response to this crisis.

We also explored heterogeneity by predicted political affiliation—using geography and demographics to predict whether someone was likelier to be a Republican or a Democrat. For example, consider an older rich man in West Texas or a young, lower-income woman in the Bay Area. According to surveys, these people are likely to have different political affiliations. We use that to explore differences among political partisanship.

According to a Pew [Research Center] poll conducted in early March, twice as many Democrats as Republicans said that they were very concerned about the COVID-19 crisis. Yet, surprisingly, we find that Republicans were stockpiling more than Democrats in the early weeks of the crisis.

We do see some other small differences in behavior. For example, Republicans were likelier to spend at restaurants or to spend on retail, which could be consistent with them having differing beliefs [from Democrats] about the severity of the crisis, or the differential risk exposure for these groups. By and large, even though we observed differences in the magnitudes, we do find Republicans and Democrats behaving similarly, at least in terms of their household-spending portfolios.

In terms of whether this is a new normal, or whether this is just transitory behavior and people are substituting spending from one time period to another, it’s too early to tell. Of course, this crisis is new, so it’s too early to tell what the persistent effects will be.

How has real-time data helped us address the crisis?

There are tremendous opportunities for researchers given the new, high-frequency data that we didn’t have in the 2008-09 financial crisis. You can explore these questions in close to real time and inform the broader academic community and policy makers if things are changing.

One of the last things I did before self-isolation was to attend a play with my wife and one of my colleagues. We saw [former Chicago mayor] Rahm Emanuel there, and he said, “A crisis is a terrible thing to waste.” As terrible as this crisis is, we’re going to learn a lot about how the world works. Pandemics have been with mankind since the beginning of recorded history—think of the Spanish flu, the bubonic plague, or Justinian’s plague in the Eastern Roman Empire. This is the first time in history that we actually have high-frequency data on how household spending changes during a pandemic.

The silver lining to this is that we can learn a lot about the world quickly in a way that is going to be useful for policy makers. Hopefully we won’t end up Monday-morning quarterbacking, as was very much the response to the last financial crisis [in 2008–09], because we just have much more available data. Researchers can study the world and how it works, and then policy makers can make informed decisions about how to respond to this new threat.

What characterizes this crisis is tremendous uncertainty. There’s just a lot of information that we don’t know. For example, we’re not sure what the case fatality rate is for this virus. If you look across countries, there are very different numbers, which has to do with differences in how countries record deaths, administer tests, and so on and so forth.

Beyond not knowing the true damage that this virus will cause from a health perspective in the short run, we don’t know much about the medium and the long runs. It’s possible that treatments will be developed, or vaccines. On the other hand, perhaps the virus will mutate and become even worse than it is, and perhaps there will be long-lasting effects.

One factor that has been missing is thinking, in a formal economic framework, about how the government should react in a situation with a great deal of uncertainty. There’s been a lot of work by my colleague [University of Chicago’s and Chicago Booth’s] Lars Peter Hansen and others about how policy should be made in cases where there’s a lot of uncertainty about future damage. This is called “robust control,” and an example of the application of this is in climate change. We know that temperatures are rising, and we know that this is caused by human activity, but there’s a lot of model uncertainty. We don’t know (1) by how much temperatures will rise and (2) what the impact on humans and the economy will be.

To summarize this literature, which is formal and mathematical, one of the big findings is that in cases of uncertainly, policy makers should adopt a cautious approach and try to minimize the maximum damage that can be done. The recommendation for policy makers in cases where there’s a lot of uncertainty is that we should be very cautious. One thing that we’ve seen in the media is the opposite conclusion from what economic theory would tell us. A lot of people have been saying, “Oh, we don’t know how bad this virus is; therefore, we should do nothing or proceed as normal.” No, that’s not what theory would tell us to do if we don’t know what’s going to happen. If we have a lot of uncertainty in a situation—it’s true, we don’t know how bad this virus is—we should be more, not less cautious.

There’s certainly a risk of people overreacting and taking too-severe responses, but the basic takeaway is that in a situation where we don’t know what’s happening, where we don’t know the true extent of potential damages, that’s a situation where we want to proceed cautiously, and a basic guiding principle should be to avoid severe damage—for example, a large portion of the population dying, if it turns out that this virus has a high fatality rate or that there’s a significant potential for it to mutate and turn into a more virulent and deadly form.

An overarching question is: How should the government respond in a situation where there’s uncertainty? Essentially, how should governments model the world? That’s one question I want to explore. Another question is: Looking at the longer-term effects of how households responded and the persistent effects of this crisis, what led households to behave in different ways? Not everybody stockpiled. People responded at different times, and why did one household respond differently than another? Was this due to differences in beliefs, or was this due to differences in potential risk exposure? I also want to explore how access to liquidity can save companies and jobs.