Abstract:
|
In this talk, I demonstrate that data fusion (or statistical matching) methods may be used to predict long-run impacts when only impacts on shorter-term measures are available. Data fusion concatenates individual-level data on the short-term effects with auxiliary data relating the intermediate and final outcomes of interest, and thereby accounts for effect heterogeneity and allows for rigorous inference. We use simulations to examine the performance of data fusion, and illustrate its use to estimate the impact of health insurance on long-term housing equity by fusing data from the Oregon Health Insurance Experiment (OHIE) lottery on long-term housing equity with observations from the Panel Study of Income Dynamics (PSID).
|