This report is released to inform interested parties of ongoing research and to encourage discussion of work in progress. Any views expressed on statistical, methodological, technical, or operational issues are those of the authors and not necessarily those of the U.S. Census Bureau, Cornell University, or any of the project sponsors. This work was partially supported by the National Science Foundation Grant SES-9978093 and SES-0427889 to Cornell University (Cornell Institute for Social and Economic Research), the National Institute on Aging, and the Alfred P. Sloan Foundation. Abowd also acknowledges direct support from the Census Bureau and other NSF grants. All of the data used in this paper are con…dential data. The U.S. Census Bureau supports external researchers ’use of some of these data items through the Research Data Center network (www.census.gov/ces). For public-use data please visit www.sipp.census.gov/sipp/ and click “Access SIPP Synthetic Data.”

We quantify sources of variation in annual job earnings data collected by the Survey of Income and Program Participation (SIPP) to determine how much of the variation is the result of measurement error. Jobs reported in the SIPP are linked to jobs reported in an administrative database, the Detailed Earnings Records (DER) drawn from the Social Security Administration’s Master Earnings File, a universe …file of all earnings reported on W-2 tax forms. As a result of the match, each job potentially has two earnings observations per year: survey and administrative. Unlike previous validation studies, both of these earnings measures are viewed as noisy measures of some underlying true amount of annual earnings.
While the existence of survey error resulting from respondent mistakes or misinterpretation is widely accepted, the idea that administrative data are also error-prone is new. Possible sources of employer reporting error, employee under-reporting of compensation such as tips, and general differences between how earnings may be reported on tax forms and in surveys, necessitates the discarding of the assumption that administrative data are a “true” measure of the quantity that the survey was designed to collect.

In addition, errors in matching SIPP and DER jobs, a necessary task in any use of administrative data, also contribute to measurement error in both earnings variables. We begin by comparing SIPP and DER earnings for di¤erent demographic and education groups of SIPP respondents. We also calculate different measures of changes in earnings for individuals switching jobs. We estimate a standard earnings equation model using SIPP and DER earnings and compare the resulting coefficients. Finally exploiting the presence of individuals with multiple jobs and shared employers over time, we estimate an econometric
model that includes random person and fi…rm effects, a common error component shared by SIPP and DER earnings, and two independent error components that represent the variation unique to each earnings measure. We compare the variance components from this model and consider how the DER and SIPP differ across unobservable components.