Much of the current epidemiological modelling involves inputs "from the ground up", using detailed data on contact patterns, geographical dispersion, travel and transport information and time use surveys. However, they do not directly attempt to model individual behaviour. In this respect, epidemiology is akin to the type of aggregate macroeconomic models widely used in the past.
This insight is important, because many economists might be tempted to simply take the epidemiologists' models off the shelf, to treat as the "biological block" of their economic models and assume that the aggregate predictions of the disease run as expected. But the kind of criticism of macroeconomic models for their past failure is valid also for economic-epidemiological models. It is important to introduce “microfoundations” into them to make useful predictions based on how the epidemiology and economics interact.
To see how introducing behaviour and decision making into models of epidemics can help us better understand the problems we face, this column focuses on the policies surrounding social distancing. This will both highlight the possible pitfalls of ignoring individuals’ behaviour but also help us critically evaluate some of the recent policies of the UK government and the scientific advice on which these policies were based.
When designing policies, ignoring spontaneous social distancing creates two potential pitfalls, that may seem contradictory. On the one hand, it may boost alarmist and misleading worst-case scenarios that are communicated to the public to prepare the ground for direct state intervention such as lockdowns. At the same time, it may make public health planners and decision makers complacent, so they fail to act sufficiently decisively when they must. And, perhaps surprisingly, it can do both at the same time.
What is the worst-case scenario?
The first question the UK government and its scientific advisors asked was, "What will the consequences be if we do nothing?" The response is contained in the well-known Imperial College report, based on the work of Ferguson et al. (2020). Their research considers a benchmark "laissez-faire" scenario with no intervention. The report is initially cautious, stating:
"[...] it is highly likely that there would be significant spontaneous changes in population behaviour even in the absence of government-mandated interventions" (p. 3) and that "In the (unlikely) absence of any control measures or spontaneous changes in individual behaviour, we would expect a peak in mortality (daily deaths) to occur after approximately 3 months" (p. 6).
But they go on to make statements such as: "In total, in an unmitigated epidemic, we would predict approximately 510,000 deaths in GB and 2.2 million in the U,S” (p. 7). The Imperial team themselves seem at times to use this do-nothing benchmark as an actual prediction.
The problem with this prediction of half million deaths is that it is based on the assumption that people will not in fact act spontaneously to mitigate the risks of unprotected exposure to infection. And while the authors of the report correctly spell out that this benchmark scenario is unlikely to be relevant, much of the subsequent messaging and media coverage made no mention of these caveats. On the contrary, the severity of this worst-case scenario based purely on the non-behavioural epidemiological forecast remains in the policy debate as the de facto default outcome in the absence of no decisive policy intervention. But it bears repeating that this scenario assumes that people will do nothing to protect themselves, an assumption that is untenable.
Advise, and they will follow?
This leads to the second pitfall, namely the belief that if directed to do so, people will voluntarily follow the government's recommendations. But will they? We again need to think about behaviour and incentives. Returning to the Imperial report, having outlined what might happen in their (unlikely) worst case scenario, they try several policy experiments to see which of these most effectively curbs the epidemic and flattens the curve. As an example, they consider the effect of a policy intervention titled Social distancing of entire population (Table 2, p. 6):
"All households reduce contact outside household, school or workplace by 75%. School contact rates unchanged, workplace contact rates reduced by 25%. Household contact rates assumed to increase by 25%."
How do we know it is a realistic assumption that households would reduce contact outside the home by 75% if the Prime Minister were to ask people to stay home? Again, we need to carefully consider people’s behaviour.
As it became increasingly clear that widespread restrictions of movement in the public space would become necessary to combat the disease, the government hesitated to impose robust measures to do so, relying instead on recommendations that it hoped people would follow. The Prime Minister spoke of us "[living in] a mature and grown-up and liberal democracy where people understand very clearly the advice that is given to them." This approach also evidences a lack of behavioural considerations. As is now well understood, when it comes to social distancing, there are positive externalities. This means that when people engage in social distancing, everybody else benefits. Unfortunately, positive externalities are typically under-supplied and social distancing is no exception. While government ministers were imploring its citizens to follow its advice and to stay at home, the media reported widespread instances of people gathering in large crowds in parks and socialising in pubs. So telling people how they should behave, without compulsion, might not achieve its aims. The government has of course since realised that more robust policies were necessary and made clear that it will enforce social distancing if needed.
How would economics have informed these policies?
In my work (Toxvaerd, 2020), I use a classical economic-epidemiological framework to revisit the idea of social distancing. A simple framework is useful for highlighting the points I make here about incentives and behaviour. In the model, each person trades off the costs and benefits of social distancing for themselves, without any regard for others. The figure below illustrates the outcomes. In a purely non-behavioural model, the epidemic would follow the now-famous curve, indicated by a dotted line. This is akin to the progression of the "unmitigated epidemic" that forms the do-nothing benchmark in the Imperial report. But in this behavioural model, individuals can instead choose to socially distance themselves, and they start doing so when the risks of infection become too high. The result is the curve in solid black. Initially, there is little social distancing and so the disease follows the benchmark dotted line, but at some point, spontaneous social distancing really kicks in and completely alters the course of the epidemic by flattening the curve. This flattening has nothing to do with government restrictions or with the capacity of hospitals or the health care sector; it is the result of spontaneous, uncoordinated social distancing by people themselves, acting rationally in their own best interests.
So the economic model including behavioural responses yields sharply different results compared to the non-behavioural model. Importantly, it allows us to think about how different characteristics of the disease, beyond biologically-determined infection and recovery rates, influences how people behave. For example, what would happen if the symptoms of infection become more unpleasant? The biological model cannot tell us anything about this, as unpleasantness plays no role in how the biology of the disease works. But in the economic model, we can easily do "what if" experiments to see how this changes behaviour. The result is seen in the blue curve, which shows that the more severe the disease is, the more individuals choose social distancing. This flattens the curve even more by spreading infection out over time. In other words, social distancing also makes the epidemic last longer.
Policymakers need to ensure not only that the economic analyses they use are properly rooted in individuals’ reactions, but also that the epidemiological and economic modelling communities work together to integrate our expertise and help manage the current and future epidemics. There is a lot of work to be done.
Imperial College COVID-19 Response Team (2020): Impact of Non-Pharmaceutical Interventions (NPIs) to Reduce COVID19 Mortality and Healthcare Demand, available at: https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf
Toxvaerd, F. (2020): Equilibrium Social Distancing, available at: https://www.inet.econ.cam.ac.uk/research-papers/wp-abstracts?wp=2008
About the author
Dr Flavio Toxvaerd, Visiting Fellow
Dr Flavio Toxvaerd is a University Associate Professor at the Faculty of Economics, University of Cambridge and a fellow of Clare College. He holds degrees from the University of Copenhagen (BSc Economics, MSc Economics), the London School of Economics (MSc Econometrics and Mathematical Economics) and ... Learn more