CPS Earnings Data for the Self-Employed: Words of Caution for Use and Interpretation

Article type: Research Article

Authors: Devine, Theresa J.^a

Affiliations: [a] The Population Research Center of the University of Chicago and NORC 1155 East 60th Street Chicago, IL 60637, USA. Phone: 312-667-9584/753-7394, Internet: terri@cicero.spc.uchicago.edu

Abstract: This paper examines several problems that one faces when using self-employment earnings data from the Annual Demographic Files for the 1976–91 March CPS, including: (i) variation in earnings definitions across different classes of workers (wage-and-salary, unincorporated self-employed, and incorporated self-employed workers), (ii) changes in earnings definitions between years, (iii) relatively high nonresponse rates for the self-employed (about 30 percent, versus 17–19 percent for wage-and-salary workers), and (iv) inaccurate coding of earnings imputation flags in the public use files. In short, the evidence presented here suggests that how one handles each of these problems can have a significant effect on estimates of the returns to time spent self-employed. The author thanks John Coder, Enrique Lamas, Chuck Nelson, and Ed Welnaik for their willingness to answer her long list of questions about Census procedures, and Joyce Mlakar, John Morgan, and Amy Pavelko for their help in sorting out the tax code. Support from the Census Research Fellow Program and the National Science Foundation (SES 90–23–776 and SES 90–103–07) is gratefully acknowledged.

DOI: 10.3233/JEM-1995-21303

Journal: Journal of Economic and Social Measurement, vol. 21, no. 3, pp. 213-251, 1995