Summary and Conclusions Although almost half of all United States taxpayers employ a paid tax return pre- parer, little IS known about the motivations’ for preparer use. Given the influence of preparers on a wide range of taxpayer de- cisions, including compliance, a better un- derstanding of the determinants of pre- parer usage has relevance to both tax policy and aldministration. Our empirical results are based on a bal- anced panel of tax returns for 1982-4, which spans the tax rate changes phased in by ERTA, in contrast to prior research that is primarily based on cross-sectional tax returns or surveys. We employ an esti- mation technique that controls for the unobserved individual-specific characteris- tics that are likely to affect preparation mode choice but that are constant over ttme. We find evidence of considerable heterogeneity across the taxpayers in the panel, which suggests that the panel data estimates are preferable to the cross- sectional estimates. For the individual variables, our results pro- vide sever.31 interesting findings. First, the time-cost of tax preparation is an impor- tant determinant of preparer usage, al- though its marginal effect is small. Time- cost is statistically significant even after controlllng for complexity (modeled as the type of ta.lc return filed and the number of additional forms and schedules attached to the return) and source of income. Second, we find that being self-employed signifi- cantly increases the probability of using a preparer, a result consistent with the con- sensus in 1 he literature. Unlike previous studies, however, we find that preparer usage iS negatively associated with the number of dependents and inferences about other demographic variables depend on itemization status. Finally, unlike pre- vious cross-sectional studies, we do not de- tect an ass,ociation between income or tax rate and preparer use after controlling for time-cost, complexity, and sources of in- come. Although panel data offer promise in providing evidence on the separate ef- fect of income and tax rates, we are un- able to examine this issue satisfactorily, be- cause the correlation between these variables ir: our panel