Abstract

This paper exploits a psuedo-natural experiment provided by the pattern of Social Security tax withholding to test whether household consumption responds to expected changes in take-home pay. In the U.S., Social Security taxes cause predictable swings in after-tax income. The tax rate has been increased 6 times between 1980 and 1993. Further, individual earners pay Social Security taxes on only their first 50,000 dollars or so in annual earnings. After earning this cutoff amount in a given year, an individual's take-home pay increases by around seven percent until January. Analysis reveals large and significant violations of consumption smoothing: consumption increases a half a percent for every one percent of additional after-tax income. I find little evidence for or against the hypothesis that households with fewer liquid assets respond more to these income changes. However, consistent with some models of bounded rationality, the violations of consumption smoothing are most pronounced for semi-durable goods.

Published version from JSTOR. 
Revised version for AER. 
AER version.

To replicate the results of the paper: 
Programs for replication from the raw CEX data. The CEX data are available from the ICPSR or the BLS