Predictive Distributions based on Longitudinal Earnings Data

Consider an individual trying to forecast his future earnings, in order to guide savings and other decisions. We envision an individual seeking advice from a financial planner. The individual provides data on his earnings history and on various personal characteristics such as age and education. The planner has access to longitudinal data sets that provide data on earnings histories and personal characteristics for samples of individuals. We devise optimal ways to combine the individual's information with the survey data in order to provide the individual with a conditional distribution for his future earnings. We work with data from the Panel Study of Income Dynamics, and our main modification of previous models is to allow for heterogeneity in volatility. This has important consequences, in that the spread of the predictive distribution becomes sensitive to the variability in the earnings history.