Even when good census data are available, as in California, other factors such as the proportion of people commuting to and from the affected area may greatly affect the size of the population present at the actual time of the earthquake. Thus, even estimating the population at risk may be difficult, let alone selecting appropriate control subjects (133). As a result, almost all of the published epidemiologic studies on earthquake-related injuries are descriptive rather than the analytic studies necessary to test hypotheses that particular types of exposures or hazards (e.g., collapsing buildings) are associated with injuries.