This project will make both substantive and methodological contributions to the study of idiosyncratic income risk. Substantively, the project will use a panel dataset on labor earnings from administrative records that has three key advantages: (i) a very large sample size with a long average time span, (ii) minimal measurement error, and (iii) no top-coding. These features of the dataset will allow us to relax a number of restrictive assumptions that previous researchers were forced to make. Methodologically, the project aims to provide an alternative approach for estimating income risk, by focusing on matching moments of the data whose economic significance is more immediate than the ‘covariance matrix of income residuals’ used in the extant literature. The gap between these two approaches can be substantial. Overall, this project intends to provide a reliable ‘user’s guide’ for applied economists, who can find precisely estimated parameters of income specifications most suitable for their models.
What Do Data on Millions of U.S. Workers Say About Life Cycle Income Risk?
Jae Song, Fatih Guvenen,2013