Are there more self-employed people, or not? IRS data shows a significant increase in the portion of the workforce reporting positive net income from self-employment on their tax returns. It rose by about 20 percent after 2000, peaking in 2014 at just under 12 percent. However, annual labor force surveys suggest that the self-employment rate has been flat since 2000. How can these two results be reconciled? This question motivates a terrific new paper by Andrew Garin, Emilie Jackson, and Dmitri Koustas entitled, New Gig Work or Changes in Reporting? Understanding Self-Employment Trends in Tax Data.
The authors mine IRS and linked Social Security Administration data to explore two hypotheses. The first hypothesis is that gig work is driving the increase in self-employment. The second is that income-based incentives in the tax code are causing individuals to report more self-employment income. The authors quickly reject the first hypothesis. They use a methodology that allows them to identify whether tax-reported earnings are from an online platform economy (OPE) firm or not. They conclude that OPE work does not explain the increase in taxpayer-reported self-employment. (P. 2.)
Then, the authors explore the tax code’s incentives to report additional income. For individuals with dependents and earned income in the phase-in range for tax credits like the Earned Income Tax Credit (EITC) and Child Tax Credit, reporting additional income results in a larger refundable credit. This could motivate individuals to report self-employment income who otherwise would have reported none. To investigate this, the authors look at the growth in the propensity to report earnings from self-employment across different groups of taxpayers over time. They find that for EITC recipients with children, the propensity to report self-employment income grew from 14 percent in 2000 to over 20 percent in 2010. For childless EITC recipients who lack similar incentives to report additional income (because they face positive marginal tax rates throughout the credit’s phase-in range), there were only modest changes in propensity to report. Further, most of the growth was found among women at the time of first childbirth and was concentrated among households combined wages below the first EITC kink point, where incentives to report are highest. (P. 11.)
The correlation between reporting more self-employment income and eligibility for larger refundable tax credits could mean either a real labor supply response (individuals actually working more in a self-employed capacity) or a pure reporting response, whether related to fake or real self-employment income. To disentangle real labor supply from pure reporting effects, the authors surmise that parents expecting a first child around the end of the year will face uncertainty about the year in which the birth will occur—and the corresponding EITC-eligibility status for that year—until the end of the year, at which point labor supply decisions for that year have already been made. They then use a regression discontinuity approach in which they compare reported self-employment earnings for parents with first births that occur right before the end of a given tax year to those with births that occur right after the end of the tax year.
The new parents with first births in December provide strong evidence that there is a pure reporting response. The authors’ main result is visually striking: the propensity to report self-employment income (but not 1099-third-party-reported income) is higher before January 1 and is concentrated among individuals with W-2 wages in the EITC phase-in range. (Figure 4). By contrast, there is no effect for individuals with wages in excess of the EITC phase-in range. For individuals with an incentive to report, the share reporting self employment earnings increases by an average of 4.6 percentage points from a baseline rate of 10.8 percentage points (43%). (P. 16.) The authors also find that this reporting effect has increased approximately fivefold from 2000 to 2015. This is staggering: among taxpayers with first births in December and January, the entire increase in the propensity to report self-employment income after first births can be accounted for by a pure reporting effect that has grown dramatically over time. (P. 23.)
The next step is to investigate the source of this pure reporting effect. Are individuals fabricating self-employment income on their tax returns to maximize refunds? Or are individuals with actual self-employment income that would have otherwise gone unreported properly reporting that income? (P. 18.) Using data from National Research Program audits of randomly-selected 1040s, they find “that while almost no one incorrectly reported self-employment income in 2001, the share doing so has risen substantially over subsequent years, but only among individuals with a tax incentive [to report fake income].” (P. 19.) However, reporting fake income can explain only a minority (20 percent) of the rise in reporting implied by the authors’ regression discontinuity estimates. (P. 19.) The rest can be attributed to a decrease in under-reporting; that is, an increase in reporting of real self-employment income that otherwise would have gone unreported.
Finally, the authors explore two explanations for the increase in the pure reporting effect: EITC policy expansions that increased incentives to report, and possible changes in individuals’ knowledge of the incentives at stake. They reject the first explanation by looking at a significant EITC expansion in 2009; they find no discontinuity in reporting around that policy change. To explore the second, the authors build on earlier studies showing that the propensity to report exactly the amount of self-employment income that qualifies a taxpayer for the maximum benefit (known as “sharp-bunching”) has spread geographically and increased over time. They use the share of sharp-bunching behavior within 3-digit zip codes as a proxy for knowledge of EITC incentives and find that the propensity to report self-employment income in response to incentives grows significantly with greater knowledge.
In terms of implications, the authors position their results as a cautionary tale about over-reliance on administrative data as compared to survey data. Beyond methodological implications, however, I worry that this paper might be seized upon by those who seek to justify high audit rates of EITC recipients. I take the opposite view: the authors’ estimate that only 20 percent of the pure reporting effect was from outright fabrication (rather than reduced underreporting of real income) seems low, not high. It is evidence that individuals have learned over time how the EITC system works, and and that they overwhelmingly comply. This seems like a very good thing for the tax system. The credits are being delivered in the vast majority of cases as Congress intended because self-employment income that should have been reported in the first place is now being reported. This might help taxpayers form a habit of full reporting of self-employment income that will hold even when tax credits are not at stake. Moreover, the IRS should want to understand the process by which EITC knowledge has disseminated over time so it can try to use those same channels for encouraging uptake of other government programs.
One can only hope that interested policymakers might read this paper carefully before jumping to conclusions about tax credit policy. It would be well worth their time.
