This dissertation examines voluntary HIV testing in three specific contexts. The first chapter evaluates the effects of written informed consent regulations on HIV testing rates using a difference-in-differences approach to analyze a natural experiment that occurred when New York State weakened its HIV testing written informed consent regulations. The study finds that streamlining the consent process in New York increased HIV testing rates about 1.7 percentage points, which is a large relative increase of approximately 31%.

The second chapter uses methods from the econometric literature on partially identified models to estimate HIV prevalence from biometric survey data in which some fraction of respondents opt out of the HIV testing module and are not tested for HIV. The method is applied to biometric survey data collected in four African countries: Kenya, Ethiopia, Lesotho, and Zimbabwe. A key finding is that the estimates of HIV prevalence that are standard in the literature represent an optimistic interpretation of the data. The biometric survey data are consistent with a range of levels of HIV prevalence, and standard estimates typically are much closer to the lower end of this range than they are to the higher end.

The third chapter reports reanalyzes data from a field experiment conducted in Malawi (Thornton, 2008). The experiment was intended to evaluate the effect of participation in voluntary counseling and testing (VCT) on condom purchasing. The experiment employed a non-standard design. Subjects were randomly assigned one of 27 differently valued VCT participation incentives that were meant to induce self-selection into treatment and control groups. The design makes it difficult to understand what treatment effects are uncovered by the data. The reanalysis uses statistical tools from the literature on treatment effect heterogeneity to clarify what the incentives experiment reveals about the causal effects of VCT. It also evaluates the strengths and weaknesses of the incentive based design. One interesting result is that VCT participation caused an increase in condom purchasing among the initial people induced to participate in VCT. But expansions of the program to cover more than about 65% of the population actually reduced condom purchasing among the new program entrants.