Underreporting of income by self-employed? : a meta-analysis studies building on Pissarides & Weber 1989
Abstract
This meta-analysis has investigated the relative income underreporting (k) in a sample of 30
international empirical studies, adding up to 342 estimates, which are based on Pissarides & Weber’s
expenditure-based approach. The effect size, k, is compiled by the exponential function of gamma over
beta, and thus do not automatically fit the meta-analysis methodology. Meta-regression analysis shows
that publication selection bias is likely present in the literature, implying that researchers and editors
systematically select larger estimates to report in their studies. Investigation of the heterogeneity in the
literature uncovered diverse characteristics among the studies affecting the estimates. We found that
using instrument variables and a proxy for permanent income seem to give systematically lower
underreporting of income, though weaker evidence than expected. Published studies tend to report
higher estimates than unpublished studies. An economy’s tax level is, on the other hand, not important
for the estimated underreporting of income in said economy.
When correcting for publication selection bias and applying our assessment of “best practice” in the
estimation process, we find that 11.3% of income are left unreported by self-employed from an
underreporting factor of k = 1.128.