Estimating Earnings and Labor Supply Responses to Income Taxation with RCS Data

National Bureau Of Economic Research Inc, Cambridge MA

Investigators

Abstract

The labor supply response to income taxation is a venerable topic in labor economics and public economics. It has experienced a revival in the last decade with a literature that has refocused attention on responses in earnings rather than hours of work, and with a focus on high-income individuals rather than the full distribution of workers. Generally speaking, somewhat higher elasticities for high-income earners have been .fund than were .apparent from the older hours-ofwork literature for the workforce as a whole. A major methodological issue that has emerged in the newer literature relate, to the intersection between difference-in-difference (DID) methods and panel versus repeated cross section (RCS) data. The recent literature has used DID methods for the most part because there is no plausible cross-sectional identification for assessing the effects of the federal income tax, which applies equally to all individuals in the country at a point in time. DID methods compare changes over time, from before to after a change in marginal tax rates, for high-income taxpayers to changes for taxpayers at lower levels of income. While these methods face some difficulties when applied to panel data, they face greater difficulties when applied to RCS data because income itself is an endogenous, time-varying variable. Forming "treatment" groups and applyingWaldtype estimators to groups defined by income leads in principle to inconsistent estimates. This is a serious problem because RCS data are far more common than panel data in this area of research and are a major continuing database for tax analysis. The research outlined in this proposal will investigate methods for the estimation of federal income tax effects on earnings and labor supply with RCS data. The starting point for the research is based on the P.I.'s own work on RCS methods. The P.I.'s research has sought to clarify and delineate the identification conditions for consistent estimation with RCS methods and to propose alternative estimation methods. The p.rs work has also made clear the central importance of age ami cohort in the use of kCS methods--that is, following cohorts of individuals from one cross-section to another as they age--a focus which is missing from existing tax analysis with RGS data. The research will also be based on the P.L's :_~~'.= in the tax effects area using panel data, work which has sought to clarify the identification conditions for DID methods in general. The project will investigate the estimation of tax effects with several cross-sections of the Survey of Consumer Finar.z:s (SCF), used in common in some parts of the analysis with several crosssections of SOI microdata tax files as well as several cross-sections of SOI microdata tax files. Some comparative analysis will be conducted with a 1983-1989 panel of the SCF as well.