This paper has two goals. First, it illustrates the importance of panel data with examples taken from research in progress using the U.S. Census Bureau's Longitudinal Research Database ( LRD ). Although the LRD is not the result of a "true" longitudinal survey, it provides both balanced and unbalanced panel data sets for establishments, firms, and lines of business. The second goal is to integrate the results of recent research with the LRD and to draw conclusions about the importance of longitudinal microdata for econometric research and time series analysis. The advantages of panel data arise from both the micro and time series aspects of the observations. This also leads us to consider why panel data are necessary to understand and interpret the time series behavior of aggregate statistics produced in cross-section establishment surveys and censuses. We find that typical homogeneity assumptions are likely to be inappropriate in a wide variety of applications. In particular, the industry in which an establishment is located, the ownership of the establishment, and the existence of the establishment (births and deaths) are endogenous variables that cannot simply be taken as time invariant fixed effects in econometric modeling.