ESTIMATING THE DETERMINANTS OF TAXPAYER COMPLIANCE WITH EXPERIMENTAL DATA*** JAMES ALM*, BETTY R. JACKSON*, AND MICHAEL MCKEE** ABSTRACT income accounts.' While creatively de- The fundamental difficulty in empirical rived and used, such surrogate measures work on taxpayer compliance is the ab- are necessarily aggregate and approxi- mate. Still other research has used data sence of detailed and reliable information generated from surveys of taxpayers. 5 on individual compliance choices. This Unfortunately, these surveys are subject paper uses data from laboratory experi- to a number of methodological problems ments to estimate individual responses to that make their data highly suspeet.6 The tax, penalty, and audit rate changes, as well difficulties with the various data sources as to changes in government expenditures. have meant that no one single study has The empirical results confirm some (al- been able to construct and use measures though not all) theoretical predictions, and of tax, penalty, and audit variables .7 It compare qualititatively with other empir- seems unlikely that the quality of field ical work. Taxpayer reporting increases data will soon-if ever-improve. with greater audit and penalty rates; how- There is, however, another source of in- ever, these responses are not large. Com- formation on the compliance choices of in- pliance is also greater when individu@'s dividuals: data from laboratory experi- face a lower tax rate and when they receive ments. There is a growing literature that something for their taxes. uses experimental methods to examine the compliance behavior of individuals.' S INCE Clotfelter (1983), the empirical However, these data have not been fully analysis of taxpayer compliance has utilized in the estimation of individual attracted increasing attention. However responses to government policies. In par- despite these efforts, our understanding' ticular, no work estimates individual re- of compliance remains surprisingly lim- sponses to a broad range of fiscal vari- ited. The fundamental difficulty in em- ables. pirical work is the absence of detailed and This paper provides such an analysis. reliable information on individual com- Data from a series of laboratory experi- pliance choices. By its very nature, people ments are collected, and these data are have an incentive to hide information on then used to estimate the individual re- their evasion behavior, and this conceal- sponses to various policy changes. Unlike ment makes empirical work quite diffi- measures of evasion in previous empirical cult. Most work in the United States has work, those that emerge from the labo- used information from the Taxpayer ratory are accurate and unambiguous Compliance Measurement Program measures of individual noncompliance; (TCMP) of the Internal Revenue Service these measures are also derived in a set- (IRS).' However, TCMP data have some ting that controls explicitly for extra- well-recognized deficiencies,' and until neous influences on individual behavior. recently the IRS has only made these data There are reasons for caution in using available to researchers in aggregate form these data (see below). However, the lab- for the year 1969.3 Partly in response, oratory offers perhaps the only opportu- other work has attempted to use various nity to investigate in a controlled envi- indirect measures of evasion, such as the ronment individual responses to tax, discrepancy between income reported on penalty, and audit rate changes, as well tax returns and income in the national as to other policies for which field data do *University of Colorado, Boulder, CO 80309-0256. not readily exist, such as the role of fiscal **University of New Mexico, Albuquerque, NM exchange in tax compliance. 87131. The estimation results confirm some 107