This paper demonstrates that administrative data may be inferior to survey data under particular circumstances. We examine the effect of state laws governing the minimum age of marriage in the United States. The estimated effects of these laws are much smaller when based on retrospective reports from census versus administrative records from Vital Statistics data. This discrepancy appears due to systematic avoidance behavior of two kinds. Some young people marry in states with less restrictive laws; others appear to have misrepresented their age on their marriage certificate. Our results have important implications regarding legal avoidance and the use of administrative data.