Policy should be adaptive

Allowing automatic updates to policy implementation can nudge us towards better equilibria

I first really internalized the game theory concept of equilibrium when I started driving in Washington, DC during the summer of 2022. I’d just moved here from Seattle, where the streets are laid out in a grid and drivers are generally calm and accommodating. DC, by contrast, is laid out to look pretty on a map, and the drivers will do literally anything they think they can get away with to get to their destination ten seconds sooner.

Initially, I tried to maintain my unhurried style of driving, but came to see that this actually ramped up the aggression of people around me. If I stopped at a newly yellow light, the people behind me would scowl and tap their steering wheels impatiently. If I obeyed the DC-wide law that you can’t turn right at a red light, the honking would start — including from police. I’ve seen people refuse to merge when I make room for them, preferring instead to wait until I pass them.

DC drivers are in an equilibrium. They’re behaving optimally, given their beliefs about how other drivers act and about how unlikely any kind of traffic enforcement is here. I was frustrating their expectations and throwing everything out of equilibrium. Eventually, I realized that everything would go much smoother if I would just compromise a bit and drive more aggressively. And so I do, a little.

Of course, an equilibrium in which you assume that everyone around you is distracted and furious isn’t the only one that’s possible. Seattle is proof of that. But even in DC, we can see hints of a less harrowing state of affairs when there are traffic cameras around. Plenty of signs announce the cameras well before you’re in range of the cameras and, though the speed limits don’t change, everyone slows down dramatically.

If we wanted people to slow down in the current equilibrium, we’d simply have to blanket the city in cameras. I question, though, whether that would actually get us to sustainably calmer traffic patterns. Would pent-up aggression express itself in even wilder driving the moment anyone finds themselves unsurveilled? I’ve watched in my rearview mirror as the driver behind me clenched their jaw the whole time we were under a camera’s watchful eye, so I think that’s likely. Omnipresent cameras might just paper over the worst behavior without changing people’s beliefs about the equilibrium we were all in. This could ratchet up frustrations by forcing people to act in ways that they believe are suboptimal.

We could also, however, get people to slow down by promoting a new equilibrium. To shift towards this new equilibrium, we need to give people a common cause — some goal they can work towards constructively — and a new set of norms that they can operate under and enforce. We can do this by making policy responsive to people’s behaviors.

In the context of traffic, this might mean having traffic cameras on most lights, but with the explicit agreement that some fraction of them will be turned off if speeding drops below a certain level. The converse of this is that a greater number would be activated if the number of speeding citations exceeds some number. For example, we could create a rule that half of the currently activated speeding cameras will be turned off if the city catches fewer than 1,000 incidents of speeding within a month.

For this to really work, we’d need to have public displays around the city of how many speeding incidents there were. This would serve as a reminder of the goal that we’re all working towards and might even incentivize casual enforcement of traffic laws by other drivers. After all, nobody likes the stress of being monitored so closely — even me, and I usually don’t have a problem with driving the speed limit. As a bitter reminder of our collective failures, we’d also need highly visible indicators of which cameras were activated, such as with a conspicuous light on each camera. I suspect that, with clearly articulated rewards and punishments for collective behavior, we could adopt new norms that would lead to new beliefs about others’ behavior and, in turn, bring about a more peaceful equilibrium.

This type of adaptive policy could be generalized beyond traffic, of course. Taxes on unhealthy foods could be adjusted in near real time based on sales over the past month. The duration of the school year could be modulated based on student performance on standardized tests. Carbon taxes could be lowered in response to decreasing emissions (and raised if emissions increase).

What all of these have in common is that they’re not just trying to shift individual behavior, but rather collective behavior. Public policy largely forgets this mechanism for meeting its goals: we have a lot of incentives that are applied to individuals, but without a shared effort, it’s rare to arrive at a new equilibrium. This in turn means that our successes are more transitory than we might like.

One policy effort that’s created a new equilibrium is the campaign against smoking, and it achieved this through recruiting people into a shared cause to make smoking less acceptable. By convincing the public of the harms of secondhand smoke, people started advocating for ways to be separated from smokers. Smoking became stigmatized, and fewer people have taken it up. The result is that we live in an equilibrium where we expect people to abstain from smoking in most public places.

We can easily implement more adaptable policies, so why don’t we? My guess is that we don’t because it involves designing a very stark trade-off in terms of what level of deviance is acceptable. Our laws are nominally designed to accept zero violation. Speeding is always a violation of the law, even though our roads are planned to accommodate it. This tacitly means that all violations of the law are equal. We don’t experience deviance this way. Rather, we understand that scale matters a lot in the public’s experience of a violation of laws or norms. It’s one thing to pass by a smoker on the sidewalk now and then, but quite another to be regularly obliged to share confined spaces with large numbers of them.

Politicians might also balk at adaptive policies, since they doesn’t play well with the signaling games that they usually take part in. You can imagine, for example, politicians getting into a kind of feedback loop where everyone wants to be seen as tougher or more lenient on some type of deviance than everyone else. One failure mode of an adaptive policy is that everyone resents it and rewards lawmakers who revert to the status quo; another is that it tries to reach a complete absence of deviance, which is probably not optimal.

Despite their potential for failures, it’s worth investigating ways that we could create more responsive policies as a way of achieving sustainable behavioral equilibria. One way of starting to implement these would be to focus on areas where people have a strong self-interest in changing to a different equilibria. For instance, most people I talk to complain about the drivers here, and would welcome calmer, more orderly traffic. Regardless of the suitability of any specific goal for the use of an adaptive policy, though, I’m optimistic that explicitly thinking out and aiming for better equilibria could produce superior policy outcomes than incentivizing individuals.