Should we worry about the observer effect? evidence from Pelotas

Most social sciences and medical studies assume that being observed does not affect subjects' behavior. However, interviews may cause changes in individuals' behavior or may inhibit changes which would occur if they were not being observed. If being observed changes the behavior of the studied population, the sample ceases to be representative of the population. In this paper, I investigate whether individuals periodically interviewed in a longitudinal epidemiological research conducted in Pelotas, Brazil, are affected along relevant dimensions, in particular, education and health. I find only a significant effect on ENEM score.