Working Paper: CEPR ID: DP10271
Authors: Lucia Corno; Aureo De Paula
Abstract: Self-reported data on sexual behaviors have been criticized to be unreliable. In recent studies, risky sexual behaviors have therefore been measured using biomarkers for curable sexually transmitted infections (STIs). Nevertheless, no previous research have tested how reliable such data are. In this paper, we first build an epidemiological model to assess the relative performance of biomarkers versus self-reported data. We then suggest an econometric strategy that combines both types of measures, biomarkers and self-reported data, to improve the estimation of correlates of risky sexual behaviors. Using the Demographic and Health Survey from Zambia, we calibrate the model and provide conditions under which self-reported data are a better proxy for risky sexual behaviors than biomarkers. In countries with low STIs prevalence, the biomarker has a higher probability of misclassification of risky behaviors than self-reported answers. Finally, we apply our estimation strategy to these data.
Keywords: biomarker; misclassification; risky behaviour; self-reported
JEL Codes: C25; I12; I15
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| self-reported data on risky sexual behaviors (C83) | more reliable than biomarkers for STIs in populations with low STI prevalence (C83) |
| low STI prevalence (J80) | self-reported data provides better proxy for risky sexual behaviors (C83) |
| high STI prevalence (I14) | biomarkers yield more accurate classifications of risky behaviors (D91) |
| exclusive reliance on either self-reported data or biomarkers (I10) | biased estimates of risky sexual behavior (C83) |
| self-reported data and biomarkers (I10) | improve estimation accuracy of risky sexual behaviors (C83) |
| conditions of STI prevalence (I12) | accuracy of measures used for risky sexual behavior (C83) |