As a consequence of the government's firm policy aimed at lowering fertility in China, there is considerable interest in measuring trends and differentials in desired family size. Accordingly a large number of surveys have included questions on family size preferences, but there have been few attempts to gauge the validity of the data so generated, despite the obvious possibility that some respondents may not report freely their true preferences. This article, exploiting the existence of two surveys conducted in Shanghai that used different approaches for assuring confidentiality, estimates the degree of underreporting, identifies the groups most likely to underreport, and traces the consequences of underreporting for modeling desired family size.