We propose an alternative two-step method to correct the bias in corruption indicators associated with the presence of reticent respondents in firm surveys. Our method uses indirect and direct questions on bribery payments. An indirect question is used to identify reticent respondents, which makes it possible to adjust responses to a direct question assessing bribery activities. A simple theoretical model is presented to examine respondent behavior to sensitive questions using direct and indirect formulations. Applying the two-step method to a survey of 382 newly created firms in Madagascar, we find that the frequency of bribery measured by the standard way of measuring corruption activities ignoring reticence is underestimated by 47 percent.