In the past decade, researchers have made substantial improvements to survey questions that allow them to obtain more accurate information from survey respondents about income and wealth. However, changing survey questions–even for the better–can create problems. For example, if we ask a respondent about his wealth holdings in 1992 and ask him again in 1994 but use a different and improved set of questions, we cannot be sure that changes in his wealth are real because part of the observed change can be due to the fact that we simply got better information the second time we asked. Thus, the cost of improved questions can be inconsistency in the data over time. We refer to this problem as “time-series inconsistency.” In this paper, we describe work that addresses this problem in the Health and Retirement Study (HRS) using data on income from financial assets. We describe a method of computation that allows us to resolve times series inconsistencies.