This paper uses an innovative household level database to ask two basic questions related to the well-known issue of income underreporting in household surveys in developing countries: (a) The extent of this underreporting in practice, and (b) whether and how it varies systematically with respondent, household, income, and survey design features. Drawing on rural household data from 17 developing and transition countries, our results indicate that the observed differences between income and consumption are extremely large, being on average 31 percent. We also find evidence of the underreporting being systematically associated with key individual, household and survey characteristics. Agricultural income is the component suffering more than any other from underreporting. Implications for policy analysis and for future research are drawn.