Ever since the early 1990s, a number of influential economists have argued that many epidemiological models need to be modified to account for endogenous changes in transmission. Traditional epidemiology assumes that individual choices can be ignored: models do not allow for individual agency, and the people being modeled just act like marbles bouncing of each other at random. More recently, the rules for these agency-free models have gotten more sophisticated in separating out different types of individuals, but the current paradigm in epidemiology (which I was exposed to during Jim Koopman’s excellent course on the topic in Fall 2011) still doesn’t have the people being modeled actually making any decisions.
Economic epidemiology has continued to develop as its own cross-disciplinary field, and has consistently focused on HIV as its most important case study. Starting with Philipson and Posner, economists have argued that models that account for risk compensation do a better job of forecasting the spread of HIV than those that ignore it. This pattern is stronger among gay men in the US than in sub-Saharan Africa, where responses tend to be small or even statistically indistinguishable from zero. However, recent work by my advisor and her coauthors (Godlonton, Munthali and Thornton 2012) and Dupas (2011) has shown that people in Africa do respond by changing their sexual behavior when they are taught facts about the relative risks of HIV transmission across population groups. At the same time, there’s increasing evidence (from Godlonton et al. and other work by Anglewicz and Kohler) that people in Malawi badly misunderstand HIV risks across all dimensions: they overestimate the prevalence of the virus, it’s transmission rate, and how quickly it kills you. I’m not aware of evidence that’s quite as systematic for developed countries, but preliminary research by Thornton, Foley and myself looking at US college students finds similar overestimates.
This is perplexing: it’s easy to see how the respondents in the Godlonton et al. and Dupas studies could respond to risks, because they were actively taught what those risks were. But how can we explain the famous* example of gay men in San Francisco reducing their sexual risk-taking in line with rising HIV prevalence if, as I suspect, they weren’t really aware of what the prevalence was? I suspect that many studies that compare individual-level behavior with factual information about disease risks are in fact picking up an intermediate variable, which is policy responses to the epidemic. The one group who is likely to understand the actual risks is the health authorities, who can enact provisions in response such as ad campaigns that aim to change social norms.
If this is what’s really going on, it would go a long way toward reconciling the finding of small risk responses in Africa (e.g. Oster 2012) with the larger ones seen in the US. Maybe individual responses are always small on average, and public health authorities are just much more active and responsive in the developed world (which would be consistent with their relative levels of funding).
This would also change the whole discussion of what economic epidemiologists have been measuring. If we are picking up responses by health officials, rather than individuals, then we can’t argue that our results are policy-invariant and hence a guide to optimal disease prevention.
NB I’ve been writing this on my iPhone as I weather a lengthy transit delay, thus the lack of supporting links. I hope to go back and throw some in later on.
*A textbook example of Kerwin’s Razor, which states that anything that needs to be titled as “famous” cannot in fact be famous. You wouldn’t say “the famous singer, Justin Timberlake”, because you don’t need to point out his fame. Everybody knows who JT is: he’s famous.