A thorough understanding of the health implications of unwanted and unintended pregnancies is constrained by our ability to accurately identify them. Commonly used techniques for measuring such pregnancies are subject to two main sources of error: the ex post revision of preferences after a pregnancy and the difficulty of identifying preferences at the time of conception. This study examines the implications of retrospective and prospective measurement approaches, which are vulnerable to different sources of error, on estimates of unwanted and unintended pregnancies. We use eight waves of closely-spaced panel data from young women in southern Malawi to generate estimates of unwanted and unintended pregnancies based on fertility preferences measured at various points in time. We then compare estimates using traditional retrospective and prospective approaches to estimates obtained when fertility preferences are measured prospectively within months of conception. The 1,062 young Malawian women in the sample frequently changed their fertility preferences. The retrospective measures slightly underestimated unwanted and unintended pregnancies compared to the time-varying prospective approach; in contrast the fixed prospective measures overestimated them. Nonetheless, most estimates were similar in aggregate, suggesting that frequent changes in fertility preferences need not lead to dramatically different estimates of unwanted and unintended pregnancy. Greater disagreement among measures emerged when classifying individual pregnancies. Carefully designed retrospective measures are not necessarily more problematic for measuring unintended and unwanted fertility than are more expensive fixed prospective ones.