The Reopening Party
Why does reopening trigger a surprisingly strong response by people compared to the caution they displayed before lockdowns?
Welcome to Plugging the Gap (my email newsletter mainly about Covid-19 and its economics). My goal is for several posts a week explaining economic research and the economic approach to understanding the pandemic. (In case you don’t know me, I’m an economist and professor at the University of Toronto. I have written lots of books including most recently on Covid-19. You can follow me on twitter (@joshgans) or subscribe to this email newsletter at the link below).
In yesterday’s post, I argued that people respond to infection risk by changing their behaviour; something that alters the path of a pandemic but also a government’s ability to control that behaviour. Today, I look at research that shows that people seemed to respond to reopening by doing more in a way that is surprising given that they curtailed activity prior to any government lockdowns. This suggests that government relaxation of restrictions may play a larger role in reopening and the potential for more infections than did government restrictions in the first place.
Here is a key graph from a new paper from economists Edward Glaeser, Ginger Zhe Jin, Benjamin Leyden, and Michael Luca.
These graphs show the timing of government lockdowns and government reopening (the two red vertical lines in each graph) alongside measures of activity (for SafeGraph visits to points of interests and restaurants) and Yelp (for restaurant visits and orders). If you look at what happened at the time of lockdowns, people had already stopped going out (especially to restaurants) before governments told those restaurants to shutdown. In other words, people were already responding to pandemic risk prior to governments enforcing that response.
By contrast, when governments reopened and lifted those restrictions, people did not wait and see what would happen before venturing out. Instead, significant numbers of people bolted for the door. In other words, there was an asymmetry in the reactions to government policy. For closing down, it did little. But, by the time things could open up people were ready and waiting to go out. You might have thought that government policy was either irrelevant (people would pay attention to risk and make up their own minds) or relevant (the government would convey information in their policy to constrain people who weren’t paying attention). The data here shows a mixed response. The question is why?
The trusted government theory
The paper presents a theory to explain this pattern. I’ve called it the trusted government theory but the authors call it a signalling theory. The theory starts with a particular premise. In order to understand how much risk there is in going out, for consumers, they need to actually go out and meet people. For the government, they obtain information by continual monitoring. This means that, prior to a lockdown, consumers are able to go out and so they have pretty much the same information as government. Add a little lag to the government’s ability to lockdown (perhaps due to political risk aversion) and consumers will take matters into their own hands and reduce their activity prior to lockdown.
On the other side of the lockdown, things are different. In this case, people have been sitting at home and so don’t really know how risky things are. But the government has this information. Thus, when the government actually reopens, this signals to consumers that things are relatively safe and so they go out themselves. Also, they continue to go out as they confirm the government’s information was accurate. Of course, this all relies on the government actually having information that consumers don’t have and consumers trusting that information. If that wasn’t the case, consumers would not rush out. In other words, the fact that they did suggests that they somewhat trust the government.
What does the data say about this theory? The authors point to the fact that in states which reopened and there was more Covid-19 prevalence, the increased activity was attenuated suggesting that people did look at things other than government to assess risk. But the other thing the researchers examine is GOP vote share in a particular state. Other work has established that people who voted Republican in the US tended to believe that Covid-19 was less serious than those who voted Democrat. This means that in states with more Republicans there will be more people who are less worried about Covid-19 and so happier to rush out when restrictions were lifted. This, however, is not quite what they found. Instead, they found that higher GOP vote share states were more likely to have an asymmetric response to closing and reopening. In other words, people were more likely to trust governments in reopening plans than they were in their closing plans. This seems surprising as the GOP voters have long been thought to be less rather than more trusting in government. The bottom line here is that the behaviour seems strange but then again 2020 is strange.
Another odd thing that occurs is summarised by this graph:
This graph shows what was happening with regard to Covid-19 infection rates just before reopening (date 0 in each graph). Notice that, in some states, infections were falling (the blue line) and there was more activity overall (the green dashed line) while in other states infections were rising (the orange dotted line) while there was less activity overall (the red line). In other words, it didn’t matter if governments had a good reason to reopen (cases were falling) or no reason to reopen (cases were rising); in either situation, consumers went out. The authors interpret this as consistent with their signalling model but I have to say that I am unconvinced. Case counts are pretty public information. People understood that going into lockdown why would they, en mass, no longer understand that coming out. It is true, they appear to but still, it seems like a puzzle to me.
The trust government/signalling theory posited is interesting but I think it is more convoluted than what might actually be the case. Let me speculate a little regarding some alternative theories.
‘Let me out of here!’: The LMOOH theory is that people were all very worried about becoming infected prior to the lockdown but after several months cooped up, as soon as there were restaurants to go to, out they went. In other words, the lockdown reflected people’s preferences according to the traditional theory of diminishing marginal utility — that is, deprive people of something for long enough and they might over-compensate when the opportunity presents itself. It is the reason why college students seem to act a little crazy when they leave home and why people on Survivor are happy to bid $300 for a burger and fries. With great frustration comes great consumption.
It’s Safer now than Then: The frustration theory is all very well but, in actuality, consumption did not rise to more than prior than the lockdown. This graph from The Economist (which does not properly account for reopening timing) shows that for Open Table reservation data.
But what might account for the behaviour change is an improvement in safety. In many jurisdictions, restaurants could open but had to provide social distancing options like outdoor seating (even at the curbside) or lower density. Mask-wearing might also make going out safer but not really to restaurants. The point here is that the world learned something from March to May and that those things led to safer practices. In this scenario, people naturally felt that things were less risky going out of the lockdown than going in. Add to that, the fact that lower density at restaurants may make reservations more scarce and there could be a little rush to get out while the going is good.
To really explore this theory, you would have to dig deeper into data on risk and also more nuanced government regulations and how these might have differed across states. I’ll keep a lookout for that type of study.
In summary, it remains a puzzle how people respond or don’t respond to government interventions that restrict activity. It could be that government policy changes represent strong signals for certain transitions but it could also be that lockdowns change important other behaviours which interact with what consumers want to do. We can add all of this to the ever-increasing list of challenges of predicting pretty much anything with regard to Covid-19.