Prediction bias in times of Covid-19: use of a linear infection-prediction model when the true model is exponential

The set of measures designed to halt the unrelenting transmission of Covid-19, prescribed by the World Health Organization and widely disseminated by local governments, include frequent washing of hands, use of hand sanitizers and face masks, social distancing and self quarantine, in case of potential exposure to Covid-19. Most countries in advanced stage of the disease and have now locked down parts of (or even entire) cities. Despite the widespread lockdown, there is a sense that people in general and policy makers in particular realized the scale of the issue a little too late, which resulted in delayed action. One possible reason for such delay is the mental model we turn to when estimating the likelihood of infection in the future. In the case of Covid-19, the true underlying data generating process governing the growth of coronavirus cases is exponential in nature but policy makers may end up delaying their action, if they apply a linear model to predict the scale of the problem in a week or two’s time. We conjecture that people in general use linear models to predict the future when the true underlying DGP is in fact exponential in nature, resulting in systematic underprediction or behavioral bias. Our second related conjecture is that those with a larger bias are more likely to underestimate the danger and less likely to observe the prescribed best practices by the WHO.