In this project, the researcher will analyze techniques for combining different datasets. The general purpose of this research is to build on the distinct strengths of the different datasets. Some datasets such as the Current Population Survey (CPS) are particularly representative but may contain only limited variables. Other datasets, such as the National Longitudinal Survey (NLS) or the Panel Study of Income Dynamics (PSID) may be less representative than the CPS but have the advantage of more detailed information. In addition, datasets such as the NLS and PSID have information on the same units over longer periods of time (panel data) rather than only once (cross-sections). Methods will be developed, extending those previously developed in econometrics and statistics, that allow combining such datasets. They will be connected to standard methods such as weighting and interpreted in the context of missing data problems. This research will lead to a better understanding of the links between techniques used in econometrics, such as generalized method of moments estimation, and techniques in statistics such as raking and weighting. These methods will then be applied to artificial as well as real datasets to evaluate their performance in a realistic setting.
