Tax Data show evidence of strong income gains for higher-income families and only muted decreases for lower- and moderate-income families in 2020



Tax data through June 7, 2021 give yet one more indication that this has been a recession like no other. In a typical recession, declines in income across the income distribution lead to increases in income tax refunds and decreases in annual tax payments during the subsequent tax filing season. Those annual tax payments, along with quarterly payments made during the year that generally reflect non-wage income, are called nonwithheld payments to distinguish from the tax payments that most people have withheld from their paychecks throughout the year. Compared to recent recessions, the recent tax data showed relatively strong nonwithheld payments and a more modest increase in refunds. That is perhaps not surprising because we already know, thanks to data releases from the Bureau of Economic Analysis, that aggregate income held up well.

However, analyzing data from the Daily Treasury Statements during tax season can help us gain a better understanding of the impact of the recession on households and business owners across the income distribution. Overall, the tax data appear to reflect the strong stock market gains and much higher trading volume, and only muted decreases in income for 2020 as a whole among lower- and moderate-income tax filers.

In particular, the preliminary data point to several conclusions (subject to uncertainty—particularly in understanding the refund data):

  • Taxable income among higher-income households and nonwithheld payments were boosted significantly by increases in the stock market and trading activity.
  • Most of the increase in refunds was likely driven by tax law changes.
  • The remaining increase in refunds likely reflects, for the most part, the effects of a drop in taxable income among lower- and moderate-income households—driven by declines in labor market income. Nonetheless, the refund data suggest that taxable income among those filers held up well compared to outcomes in the two preceding recessions. Note that taxable income excludes most but by no means all of the recent unemployment insurance benefits.

Nonwithheld tax payments and refunds in 2002 and in 2009

To understand how tax payments changed in recent recessions, we examine nonwithheld tax payments and refunds in the 2002 and 2009 tax filing seasons (each of those occurring one year after a recession began) and we compare them to the prior year. In those years, as is usual in the year following a recession, payments fell (largely reflecting changes in payments to higher-income filers) and refunds—excluding any temporary tax rebates enacted by Congress—increased (reflecting higher refunds across the distribution).[1]

As shown in figure 1a, in 2002 and 2009, nonwithheld payments from February to May (largely reflecting tax liability owed for the prior tax year) fell by more than one-third.[2] he declines in 2002 and 2003 in nonwithheld payments largely reflected steep drops in net income from capital gains of almost 50 percent in 2001 and almost 30 percent in 2002, which significantly affected higher-income taxpayers.[3]

As shown in figure 1b, tax refunds rose by roughly 25 percent in 2002 and 20 percent in 2009. In 2001 and 2008, the typical workers who became unemployed had too much in income taxes withheld from their paychecks during the year, because the amounts were calculated assuming that they would be employed for the full year and would therefore be in higher tax brackets. Those taxpayers likely received the overwithheld amount back as refunds when they filed their tax returns.

In addition, the same drop in capital gains income that caused declines in tax payments also brought about increases in refunds for those higher-income taxpayers who made unnecessarily large quarterly estimated payments. In fact, a disproportionate amount of the overall refund increases went to higher-income households in both years. For example, in tax year 2008 (2009 filing season) tax refunds for households with more than $100,000 in adjusted gross income (AGI) rose by over 40 percent, while refunds for households with lower amounts of income rose by about 15 percent.

Decreases in tax payments and increases in refunds helped to buttress household income during those recessions. Some of that happened automatically because of the nature of the tax system. To the degree that the drop in incomes caused withholding amounts to fall and people to lower their quarterly estimated payments, it occurred quickly. For many, the decrease in payments and increase in refunds was delayed until taxpayers filed their tax returns for the previous year. In addition, as is typical during recessions, policymaker-enacted tax cuts helped to maintain aggregate demand when household incomes fell.

Complications in comparing recent tax data to prior periods

Assessing tax collections this year is more difficult than usual for several reasons. Most importantly, income tax filing deadlines were postponed in the 2020 and 2021 filing seasons. In 2020, the annual filing deadline moved from April 15 to July 15, and due dates for the first two quarterly estimated payments were moved from April 15 and June 15 to July 15 as well. In 2021, only the annual filing deadline was moved to May 17. As a result, in 2019 we saw, as usual, the major amount of nonwithheld payments with tax returns by the end of April, a couple of weeks or so after the filing deadline. However, the delayed deadline in 2020 means that we did not observe the amounts paid with tax returns or the commingled amounts of the second quarterly estimated payments normally due in mid-June, until late July. As a result, we will not be able to compare 2021 tax collection data to 2020 data until August, and such a comparison would also include two estimated quarterly payments.

Our solution is to compare the tax data that we have on hand for 2021 (through June 7, 2021) with tax data for 2019 through June 7, 2019. That comparison can tell us whether the tax payments and refunds this year look large or small relative to a recent pre-pandemic year. If tax policy were the same during both periods, the differences in payments and refunds should be informative about changes between 2018 and 2020 in taxable income across the income distribution. However, tax policy did change, so we must also consider how those policy changes affected the tax data.

In most cases, we believe the effects of policy changes reduced tax payments or increased refunds, but the size of the effects is not clear.

The following three main policy changes reduced revenues or increased refunds:

  • Rebate checks. Since March 2020, the three rounds of “rebate checks” are estimated to cost $868 billion in total once all payments are issued. Two of those rounds applied to the 2020 tax year. If the Internal Revenue Service issued a person a check, it was recorded in the federal budget as an outlay instead of a reduction in revenue. However, in the 2021 tax filing season, filers were able to claim these 2020 rebate amounts as a part of their tax calculation if they had not been issued to the qualifying individual or if the amount was less than the individual was due. In response to the last two recessions, rebate checks were also issued to people in 2001, 2003, and 2008, but they were recorded either totally or largely as reductions in revenue. In order to facilitate comparability and focus on tax liability as it relates to taxable income, we strip all of these identifiable “rebate checks” out of the tax data in all years. Among the rebates we are not able to strip out of refunds, we estimate that they increased 2021 refunds by several billion dollars.
  • Business loss provisions. For business owners, the loss limitations for non-corporate entities were loosened under the CARES Act to allow immediate use of certain losses incurred from 2018 to 2020. The Joint Committee on Taxation (JCT) estimated that the provisions would reduce revenues by $64 billion in fiscal year 2021. We expect a portion of this to show up as reduced nonwithheld payments in 2021. In addition, we expect that this provision boosted refunds since February by roughly $5 billion, although that amount is very uncertain.
  • Lookback Provision for Refundable Tax Credits. The Consolidated Appropriations Act, 2021, gave individuals the option to substitute their 2019 earned income for the 2020 amount if doing so increased their earned income tax credit and child tax credit amounts when they filed taxes for 2020. We estimate that this change boosted refunds by a couple of billion dollars in 2021.

The following policy changes or associated anticipatory effects have mixed and less definitive effects:

  • As a result of the substantial expansion to unemployment benefits—allowing previously ineligible workers to claim benefits, increases in the benefit amounts, and an increase in the length of time individuals are allowed to use their unemployment benefits—some indeterminate tax implications may transpire. The American Rescue Plan Act of 2021 included a retroactive $10,200 exclusion for taxpayers with below $150,000 modified AGI on 2020 unemployment compensation. For filers who paid their taxes early in 2021, before that change in the law, that initial taxation and then retroactive exclusion probably boosted nonwithheld payments and refunds by several billion dollars; some of the refunds are still being processed by the Internal Revenue Service.
  • Timing factors such as the election results and the anticipation of tax rate increases in 2021 may have prompted taxpayers to shift income into 2020, boosting revenues in the 2021 tax filing season.
  • Similarly, tax law changes enacted in late 2017 may have resulted in income shifting into 2018, potentially making payments with tax returns in 2019 look stronger when compared to 2021.

In addition to those policy changes, several additional factors make it difficult to make comparisons between years. It is not uncommon to have filing extensions for taxpayers in locations experiencing extreme weather, for example taxpayers in Texas, Louisiana, and Oklahoma had an extended deadline this year to mid-June (from mid-May) as a result of winter storms.  Finally, the IRS has experienced significant processing delays due to the pandemic. We estimate that a small portion of the refunds the IRS has sent out in the past few months, perhaps upwards of a couple of billion dollars, relate to returns filed in 2020.

Comparing 2021 to 2019 and prior years

As shown in figure 1a, nonwithheld payments from February to May 2021 were 21 percent higher than in the same period in 2019. They were much higher compared to the same period in 2020, but, as explained above, that comparison is not informative. The post-recessionary increase is all the more extraordinary given the changes in tax law that work in the other direction. Overall, the unusually large increase in nonwithheld payments likely stems, at least in large part, from capital gains income being unusually strong in 2020 when the stock market increased overall and trading activity rose sharply. In addition, proprietors’ income in 2020, as measured in the National Income and Product Accounts, was only about 2 percent below trend—a considerably more modest shortfall than after the 2008 recession.

Relative to 2019, refunds were up nearly 5 percent between February and May 2021—somewhat smaller than is typical after a recession begins (as shown in figure 1b). Including refunds paid through early June, the increase is almost 8 percent, or about $20 billion. Importantly, the strength in nonwithheld payments suggests a very different composition than what we usually see driving the post-recession increase in refunds. Recall that in 2002 and 2009, refunds were boosted by excessive estimated payments and overwithholding among higher-income taxpayers who likely experienced a drop in capital gains income in those periods. That factor is probably not boosting refunds now.

So then, what accounts for the roughly $20 billion increase in refunds, and how does that compare to previous recessions? We estimate that about three-quarters of the increase is explained by the changes in tax law and other factors described above. That stands in stark contrast to 2002 and 2009, where we estimate that law changes did not substantially boost refunds. The remainder, roughly $5 billion, likely reflects the effects of a drop in taxable income among lower- and moderate-income households—driven by declines in labor market income. Note that the increase in refunds we are attributing to a decline in taxable income among lower- and moderate-income filers is much smaller than what we saw in the 2002 and 2009 filing seasons.[4] That comparison suggests that taxable income among lower- and moderate-income households held up well compared to outcomes in the preceding two recessions.

One way to think about that finding is to look at changes in the sum of wages and salaries and government social benefits, which includes benefits such as Social Security payments and unemployment insurance benefits. That aggregate rose 4 percent in 2001 and in 2008 (and fell in 2009). In 2020, after excluding 70 percent of unemployment insurance benefits—likely the upper end of what was excluded from taxable income—that aggregate rose 6 percent. That comparison provides some indication that taxable income among lower- and moderate-income households may indeed have held up relatively well.

Comparing 2021 to 2020 and 2019

Examining figures 2a and 2b gives another perspective on how unusual the tax data have been. Here, we compare cumulative nonwithheld payments and refunds from February to June 7, 2021 to those we see from February to August in 2019 and 2020. Doing so allows us to see the full effect of the recession on the tax data in 2020, given that filing deadlines were so significantly delayed. For both 2019 and 2020, the figures reflect the annual tax return data and two quarterly payments. For 2021, which extends through June 7, 2021, the figures reflect the annual tax return data and only one quarterly payment.

Comparing 2020 to 2019, cumulative refunds were roughly unchanged by the end of August and nonwithheld payments were lower through that period. Those refunds—which exclude the rebate amounts in 2020—largely reflect economic activity from 2019 and 2018, respectively, when economic growth was relatively stable.

Nonwithheld payments in 2020 reflected not only economic activity in 2019, from payments with tax return filings, but also 2020 activity from quarterly estimated payments. The first two estimated payments for 2020, normally due in April and June but delayed until July, were probably lower than in 2019 as a result of the severe downturn. Those quarterly payments were paid together with amounts from tax return filings and cannot be identified separately. We can observe that estimated payments rebounded in September 2020 and then especially in January 2021, the last quarterly estimated payment for the 2020 tax year.

Comparing 2021 to the two prior years, cumulative nonwithheld payments from February to early June 2021 were larger than those from February to August 2020 and roughly equal to the seven-month period in 2019. That is true despite the fact that 2019 and 2020 include an extra quarterly payment. These data suggest that taxable income, particularly for higher-income filers, was quite strong in 2020 relative to 2018 and 2019.

Refunds by early June have moved above those through the same period in 2019, likely reflecting a number of factors, as described earlier. Those factors include a drop in income among lower- and moderate-income households in 2020 and the effects of tax law changes. The IRS is still processing refunds from returns timely filed this tax season, and the amount of refunds should continue to climb somewhat through the end of June.


We see two main takeaways from the tax data over the last few months. First, taxable income among higher-income households and nonwithheld payments were boosted significantly by increases in the stock market and trading activity. Second, the data suggest that taxable income among lower- and moderate-income households held up well compared to outcomes in the preceding two recessions.

A more complete picture of household income will be available from Census later this year, including an accounting not just of how households have done in the labor market and the stock market—the focus of this piece—but also how their incomes were affected by the extensive fiscal supports enacted since the onset of the pandemic. Such information may indicate, for example, how useful recent overall refund data are for identifying income movements across the distribution in 2020.


[1] In a typical year, nonwithheld income tax payments with tax returns are disproportionately from higher-income taxpayers, whereas lower and moderate-income taxpayers more typically receive tax refunds. For example, according to the IRS, 89 percent of nonwithheld payments made with tax returns in 2019 (for liabilities incurred in 2018) were made by filers with income above $100,000, who represent only 19 percent of total tax filers. In contrast, 74 percent of the dollar amount of refunds went to filers with AGI below $100,000, and taxpayers with AGI below $100,000 were about 81 percent of total tax filers in the country.

[2] In comparing payments across years in this figure, it is important to note that these data do not account for changes in inflation over time or population growth. Those factors lead to higher payments between, for example, 2001 and 2008.

[3] In recent years, roughly 90 percent of capital gains income have gone to taxpayers with over $100,000 AGI.

[4] Comparisons to the 2002 filing season are clouded by substantial tax law changes that included changes to tax rates.


  • Acknowledgements and disclosures

    We thank Jason Furman and William Gale for their helpful feedback. Moriah Macklin, Winnie Yee, and Elisabeth Raczek provided excellent research assistance. Veronica Clevenstine and Sophia Mariam provided superb research support.