People Adapt, Thwarting Government Plans

If you derive no other takeaway from the political response to the ongoing COVID-19 pandemic, let it be this: human behavior adapts in real-time to ever-changing expectations of the future. We make decisions based on imperfect bits and pieces of information available to us, and use them to mitigate the uncertainties that lie ahead.

You’ve probably done this many times in the last few months without even thinking about it, or at least not systematically. Do you wash your hands more frequently today than you did a month ago, or perhaps carry a bottle of hand sanitizer with you in the event that you must venture out in public? If so, you’ve adapted your behavior based on available information about the virus threat and reasonable expectations to reduce your own risk of contracting the disease. And more likely than not, you were already doing these things before waiting for any president, governor, or regulatory agency official to instruct you to do so.

In fact, evidence of risk-mitigating behavioral modification is all around us. Almost every single day, stores run through their entire supply of hand sanitizer, Lysol spray, and similar disinfectants because people are using these items at much higher rates. Most of us have incorporated social distancing when venturing into public by maintaining spaces between us, and many are now wearing facemasks as part of their routines for going out to the store for necessities. Some of these decisions may prove more effective at risk mitigation than others, and it is not my objective here to advise your own coping strategy. But these actions also share a common theme: they are all voluntary behavioral changes that have attained mass-adoption in very short order.

Behavioral changes also carry complex dimensions that are not as easy to observe: decisions to venture outside or stay put, decisions to stay open or close, decisions on how many people to allow into your business at a time, decisions to order something online or try to obtain it at the store. These too occur in real-time, and are happening all around us – often based on a personal calculus of needs and risks, acted on intuitively at the margin of the decision-making process itself.

In the last several weeks, the heavy hand of government has moved into many of these decision-making processes with alarming speed. Its more benign manifestations include public messaging to encourage the same hygienic and social distancing practices that most people were already doing on their own. 

On the other end are restrictions and highway checkpoints to dissuade travel, anti-gouging laws that prevent the price mechanism from rationing sought-after goods following a surge in demand, mandatory business closures based on what the state deems “non-essential” (with the definition often varying from state to state or even municipality to municipality), and even surveillance to catch, fine, and break up unapproved social interactions.

Most of these actions encounter unintended complications that work against their efficacy, even when a policy is adopted for well-intentioned reasons.

First, the state itself is usually a lagging indicator in crisis responses. It tries to project the exact opposite through a cultivated image of leadership and “taking charge” to defeat the crisis. But most of its decisions are actually reactive, and often lag several weeks behind even the most basic voluntary behavioral responses.

We see this in the responses to a pair of surveys about how people are coping with the pandemic, taken about a month apart between early February and early March 2020. In the February survey, about 40% indicated they were avoiding large social gatherings and 66% said they had increased hand-washing. When surveyors asked similar questions on March 10-12, large social gathering avoidance had increased to 68%, and handwashing increased to 88%.

This pattern appears to be well attested in other areas. Another early March survey reported similar levels of handwashing, in addition to 61% reporting that they had adopted social distancing. Another late-March survey using slightly different questions suggests these patterns have rapidly intensified, with 91% stating that they now avoid social events such as parties and 77% indicating that they would be uncomfortable eating out in a restaurant.

These attested behavioral changes are in no sense comprehensive – either as a reflection of how people have altered their habits, or as a risk-aversion strategy when taken alone. But they do offer a clear glimpse of a major shift in public behavior that had already taken place before the federal government issued its national COVID mitigation guidelines for social distancing and various stages of economic shutdown (March 16). They also predate most of the stringent lockdown and shelter-in-place orders that city and state governments began to impose during the same period.

Second, the timing of these shifts in behavior is also important, because they complicate many of the epidemiological models on which government officials are basing their decisions. One widely-publicized model released around March 17 projected 2.2 million deaths in the United States as its worst case scenario. 

While politicians and the press hyped this number into a sense of widespread alarm, few news reports mentioned a substantial caveat to the projection: it modeled a “do nothing” scenario that was already unrealistic for a population with rapidly changing behavioral modifications, as well as governments that had already started to impose lockdowns. 

While the 2.2 million person body count grabbed the headlines, the more realistic scenarios accounting for both behavioral and policy modifications were buried in the model’s accompanying academic study and received little attention until weeks later.

These and similar complications hint at a larger problem facing forecasters, both in epidemiology and across the physical and social sciences. Behavioral modification is constant and ongoing, making it exceedingly difficult to accurately predict. While some epidemiological models attempt to predict the effects of a narrow range of voluntary and politically imposed behavioral changes (albeit with debatable accuracy), it’s much more difficult to account for unintended consequences.

Yet if the last few weeks have taught us anything, unintended consequences abound – and not always in ways that are optimal for virus containment. Consider how hundreds of colleges adopted a sudden and sometimes haphazardly executed decision to shut their doors in early March. 

The policy intended to preempt an outbreak from taking place on campus, where close-quarter dorms posed a potential risk for rapid transmission. But its hasty enactment, often without much thought given to the implementation, thrust hundreds of thousands of college students into finding new residences or making last-minute travel arrangements to get home, often to the other side of the country. This approach successfully mitigated the risk of on-campus infection, but also gave little consideration to the risks of transmission and exposure due to increased travel.

We saw similar unintended consequences at airports amid the adoption of restrictions on international travel in mid-March. While these measures were intended to slow the transmission of the disease, they also had the unintended effect of corralling thousands of arriving passengers in close quarters for hours on end as they waited to clear customs and newly imposed safety screenings – conditions that are ripe for disease transmission.

Another unintended consequence may be seen in the more recent policy of some states – Rhode Island and Florida come to mind – to impose police checkpoints at their borders to discourage interstate travel and transmission. Let’s consider how the adoption of such policies could backfire. 

Suppose you live in a region that could be constrained by a checkpoint in the near future, but you also have another residence or family or some form of home-away-from-home that isn’t  as restricted. Before the checkpoint policy is imposed, there’s a decent chance you’d be willing to ride it out at home by sheltering in place. With checkpoints looming however, some people may start to weigh the risk of being stuck there for a long time, of being cut off from family and loved ones in other states, or of having to deal with increasingly draconian police enforcement in their own home towns even though it’s well past the point of diminishing returns for disease-mitigation.

The unavoidable effect is that some people who were previously sheltering in place in one region may now make the decision to relocate to another before stricter travel regulations are imposed. And they will do so as a way to mitigate the uncertainty of an indefinite lockdown, even if by travelling they now increase the risk of transmission. In short, the announcement of a policy intended to restrict travel may have the offsetting effect of inducing people to travel to avoid being trapped in a single place after it goes into effect.

While the state’s natural impulse in such situations is to crack down even harder on travel, that too comes with unintended consequences. Should we accept, for example, the prospect of law enforcement checkpoints that must now differentiate between permissible “essential” trips to the grocery store and forbidden travel across state lines from a hotbed region to a comparatively isolated one? Considering that such enforcement will also increase social interactions between anyone who passes a checkpoint and the police, it may even become a vehicle for disease transmission itself. The more draconian the lockdown, the more opportunities that disease will spread as a result of the enforcement actions to impose that lockdown.

At the root of these and other unintended consequences is the recurring problem of information uncertainty. This happens at the policy level where politicians imagine they can design and seamlessly implement a quick fix to an exceedingly complex system that also defies our capacity to predict and forecast. It also happens in our daily lives, where the personal decisions we make to mitigate the risk of the disease incorporate individualized knowledge of our own circumstances that no collective entity or state actor could ever come close to obtaining.

When information is scarce, costly, and difficult to obtain, the risk of unintended consequences increases dramatically. Let that serve as a caution against aggressive and sweeping policy decisions. The chances are high that they are not adequately taking into account individual behavioral changes that are already well underway, let alone additional responses provoked by the policy itself. Given the epistemic constraints in which we are operating, such policies may prove costly, ineffectual and at attaining the results they seek, or worse, backfire.

Related Articles – Crisis, Policy, Regulation