The recent Census Bureau report on poverty for 2021 raises questions about the survey’s accuracy during COVID-19. U.S. poverty levels were found to have dropped substantially — due, in no small part, to the stimulus payments and temporary expansions of cash transfer programs enacted across various relief packages. However, the assessed impact of pandemic unemployment benefits seems mysteriously underwhelming.

While the stimulus checks and expanded Child Tax Credit lifted 8.9 million and 5.3 million people out of poverty, respectively, unemployment insurance (UI) brought only 2.3 million people above the federal poverty line (For each government benefit, the antipoverty impact is evaluated by counting the number of persons in households whose total income is below the poverty line with and without the dollars delivered by that specific program). Although all three policies made significant contributions towards reducing poverty, it is still surprising that unemployment benefits did not have an even larger effect given the scope and scale of the federal expansion.

UI’s anti-poverty effect is likely understated 

For context, over 10 million workers received UI benefits every week in 2021 until the federal enhancements expired for good in September. Those typically ineligible for UI – gig workers, self-employed individuals, etc. – were able to qualify, and regular weekly benefit sizes were topped up by $300 for all recipients. Federal UI expansions were a clear pandemic success story, helping millions of families put food on the table during a period of mass job dislocation.

Yet evidence suggests the Census Bureau numbers may be underestimating UI’s success. In a new paper by Jeff Larrimore of the Federal Reserve Board, data reporting issues associated with the Bureau’s Current Population Survey (CPS) are shown to have a major impact on the Census UI estimates for 2021. IRS tax data, derived from actual administrative records, indicates that over $300 billion in unemployment benefits were distributed among 26 million people in 2021. But the CPS data the Census used for its annual poverty report appears to shortchange the size of the UI program by nearly $200 billion.

Data collection methods may be a potential source of this difference. Underreporting social insurance program benefits, such as UI, is a longstanding issue of household surveys like CPS. When respondents are asked about their household income, they may forget to include UI payments or lowball amounts received due to the stigma associated with the benefits. Shortcomings with survey reporting could have been exacerbated further by the scale of the COVID-19 UI programs and difficulty reaching prospective participants during the pandemic. While the IRS records are not perfect, the results are based on administrative tax data that line up much closer with other federal agency numbers.

The data divergence isn’t limited to last year

These reporting disparities stood out in the first year of the pandemic as well. According to IRS data analyzed by Larrimore and his colleagues, Jacob Mortenson and David Splinter, there were 45.4 million total UI recipients receiving a combined $565 billion in benefits in 2020. In contrast, CPS data indicates 23.6 million beneficiaries received $218 billion. The bottom two income deciles alone were recorded as having received $168 billion less in UI benefits relative to the IRS figures for 2020. 

In short, the aggregate Census data are nowhere near the IRS tax collection numbers, especially for low-income workers. A similar discrepancy can be found when looking back at data from the Great Recession. During that period, UI receipts and benefits distributed went up, but the CPS data failed to appropriately measure the increase among those with less income. A large portion of the benefits distributed to those near the poverty line are simply missed.

IRS data can improve pandemic poverty estimates

The assessed antipoverty effects of UI are understated because they are based on data that underreports the amount of cash support delivered to beneficiaries. When adjusting for the benefit levels seen in the IRS data, the supplemental poverty measure (SPM), which factors in tax credits not included in the official poverty rate, drops from 9.1 percent to 7.3 percent in 2020 (a 20 percent reduction) and from 7.8 percent to 6.6 percent in 2021 (a 15 percent reduction), as shown in the chart below.

These adjustments are presented with population totals below. Officially, unemployment insurance is given credit for lifting 5.5 million people above the SPM poverty line in 2020 and another 2.3 million people in 2021. When IRS data on unemployment insurance recipiency and benefit levels are applied using Larrimore’s estimates, an additional 5.9 million and 3.9 million individuals are kept out of poverty in 2020 and 2021, respectively. In other words, the poverty reduction impact of pandemic unemployment insurance was more than twice as large as the Census data suggest when augmented with national tax data.

An important caveat is necessary to contextualize these results. Just as the CPS data understates the benefits distributed, the IRS data could overstate the numbers. The UI program has struggled with identity theft and false payments during the COVID period. Notably, the Department of Labor’s Employment and Training Administration assessed there was an improper payment rate of 18.7 percent

Factoring that false payment rate into the IRS-adjusted calculations lowers the poverty reduction totals to 9.3 million people in 2020 (initially 11.4) and five million in 2021 (initially 6.2). Yet, under those conditions, the poverty reduction levels from unemployment benefits would still be 83 percent greater than the Census estimates. Even if you assume false payments went to one of every three reported beneficiaries, the antipoverty effects of UI would still be 50 percent larger than the official numbers indicate.

The takeaway

Misleading data estimates can have large effects on people’s perceptions and, eventually, policy decisions. To date, policy organizations and media outlets have taken the Census numbers at face value. These misconceptions risk ingraining an inaccurate portrait of UI’s anti-poverty impact as the consensus. Discounting over half the real impact made by unemployment programs may then cause policymakers to underrate the importance of UI reform in the list of legislative priorities. This does a disservice to workers. For the sake of defending future expansions, it is essential that we get these measurements right going forward.

Photo Credit: iStock