What to watch for in the 2019 Census data on earnings, incomes, and poverty

On Tuesday, the Census Bureau will release its annual data on earnings, incomes, poverty, and health insurance coverage for 2019. There are a number of important things to know about this data release, including:

  • These data report household incomes from 2019, not 2020. In short, this data will be silent on what has happened since the pandemic and ensuing recession hit the United States and resulted in widespread job loss and health devastation. It will, however, provide some insights into the labor market as we entered the current recession, which magnified deeply entrenched economic inequalities and racial and ethnic disparities.
  • The data on incomes in 2019 was collected in February, March, and April of 2020. Its collection was likely extremely hampered by the COVID-19 pandemic. This could make it less reliable than previous year’s releases, and biases could be particularly large for low-income households.
  • The data will give us insight into how evenly (or unevenly) economic growth has been distributed across U.S. households in 2019. Other data sources that are released more than once a year too often provide only averages or aggregates into what is happening in a particular month—but next week’s Census release gives a comprehensive picture of how the U.S. economy was working for typical households over the full year, including individual, family, and household level data on annual earnings and incomes.
  • The data is likely to show strong, relatively widespread income growth in 2019. Policymakers should take heed of this, as the roots of this growth was a labor market approaching full employment, with the unemployment rate below 4% for the entire year.

This year’s income data will almost surely be lower quality than usual, despite heroic efforts by the Bureau of Labor Statistics (BLS) and the Census Bureau to maintain quality in the face of a global pandemic. Although the data release includes information about 2019 only, the data was collected between February and April of this year, right as the pandemic began to spread rapidly and most of the country was locked down. This almost surely caused serious issues with data collection, which will likely impact estimates for 2019. In March and April, there was a precipitous drop off in household response rates. This was due, in part, to the Census suspending in-person interviews for health reasons. But even among those households who wouldn’t usually receive in-person visits, for instance, in subsequent months after their initial entry into the survey, there were significant reductions in response rates.

If the drop-off in response rates was random, estimates would be unbiased. However, there’s good reason to believe that the drop-off in response rates was not random. It seems plausible that households facing job losses or health events would be less likely to respond to the survey. My own preliminary analysis of the drop-off in survey responses suggests that’s the case, with a sharper drop for Black and Hispanic workers—who have disproportionately faced economic and health insecurity from COVID-19—and workers with lower levels of education. If that’s the case, then the estimates generated from the remaining respondents may overstate the strength of the labor market in 2019 or, at the very least, rely on fewer numbers of those workers to generate estimates, thereby increasing their uncertainty and widening their confidence intervals.

At the same time, those households that did respond and saw dramatic changes in their employment and/or earnings status between 2019 and the survey date may have let their current situation influence their answers. Although they will have been instructed to provide information about the prior year, some respondents may exhibit recency bias, where they favor their current situation over their past situation. Since the change in employment and/or earnings is more likely to be negative given the downturn of the labor market, this may give a downward bias to their responses for 2019.

Additionally, the extended individual tax filing deadline in 2020 (also caused by the COVID-19 shock) likely will adversely affect data quality. Traditionally, the Census data is collected around March each year precisely because of tax season, when people tend to be more familiar with their previous year’s income as they prepare their taxes. Many people, however, were not preparing their taxes during the data collection window this year. The allowance of non-penalized deferred tax filings may mean that households relied on recall rather than recently compiled income information when answering questions. Taken together, there is some uncertainty about the presence or direction of any bias in the reported data, stemming from lowered response rates coupled with dramatic changes in the labor market and an extended tax filing deadline in the spring.

While out of date for measuring today’s economy, the latest data from 2019 will give us a useful glimpse of what a stronger economy means for working people. In 2019, the unemployment rate averaged 3.7% and wages continued to grow between 2018 and 2019, albeit slowly and unequally. Furthermore, the tightening labor market combined with state-level minimum wage increases translated into some stronger growth for low-wage workers. In next week’s release, we will see how well increases in labor force participation and wages translated into faster annual earnings growth, higher incomes, and lower poverty. Approaching genuine full employment should mean that the benefits of the stronger economy accrue not only to high-wage workers, but also and particularly to middle- and low-wage workers. The benefits should mean faster wage growth as well as an increased ability for workers to secure enough work hours, which, taken together, should translate into increases in household incomes. As a result, I hope we will finally see non-elderly households incomes return to their peak in 2000.

Broad-based growth should also mean the narrowing of racial earnings and income gaps. As of 2018, racial and ethnic income gaps persisted as household incomes experienced uneven growth. The figure below shows real median household income by race and ethnicity from 2000 to 2018. In 2018, the median Black household earned just 59 cents for every dollar of income the median white household earned, while the median Hispanic household earned 73 cents for every median white household income dollar.

Figure A

Real median household income by race and ethnicity, 2000–2018

Year White  Black  Hispanic  Asian  White-imputed   Black-imputed  Hispanic-imputed  Asian-imputed  White  Black  Hispanic  Asian  White  Black  Hispanic  Asian 
2000 $66,712 $43,380 $48,500 $69,069  $44,614  $46,989 
2001 $65,835 $41,899 $47,721 $68,161 $43,091 $46,234
2002 $65,646 $40,839 $46,334 $73,660 $67,965 $42,001 $44,890 $79,501
2003 $65,388 $40,633 $45,160 $76,231 $67,698 $41,789 $43,753 $82,276
2004 $65,178 $40,292 $45,670 $76,631 $67,481 $41,438 $44,247 $82,708
2005 $65,458 $39,898  $46,360  $76,873 $67,771  $41,033 $43,846 $84,991 
2006 $65,449 $40,116 $47,169 $78,291 $67,762 $41,257 $45,699 $86,560
2007 $66,676 $41,388 $46,958 $78,343 $69,032 $42,565 $45,495 $86,616
2008 $64,923 $40,154 $44,326 $74,913 $67,217 $41,296 $42,945 $82,824
2009 $63,895 $38,423 $44,628 $74,982 $66,153 $39,516 $43,238 $82,901
2010 $62,857 $37,114 $43,433 $72,402 $65,078 $38,170 $42,080 $80,048
2011 $62,001 $36,215 $43,217 $71,139 $64,192 $37,245 $41,870 $78,653
2012 $62,465 $36,945 $42,738 $73,415 $64,672 $37,996 $41,406 $81,169
2013 $62,915 $37,547 $44,228 $70,687 $65,138 $38,615 $42,850 $78,153 $65,138 $38,615 $42,850 $78,153
2014 $63,976 $37,854 $45,114 $78,883 $63,976 $37,854 $45,114 $78,883
2015 $66,721 $39,440 $47,852 $81,788 $66,721 $39,440 $47,852 $81,788
2016 $68,059 $41,924 $49,887 $85,210 $68,059 $41,924 $49,887 $85,210
2017 $69,806 $41,584 $51,717 $83,314 $69,806 $41,584 $51,717 $83,314 $69,851 $40,963 $51,390 $83,376
2018 $70,642  $41,692  $51,450  $87,194 


ChartData Download data

The data below can be saved or copied directly into Excel.

Note: Because of a redesign in the CPS ASEC income questions in 2013, we imputed the historical series using the ratio of the old and new method in 2013. Solid lines are actual CPS ASEC data; dashed lines denote historical values imputed by applying the new methodology to past income trends. The break in the series in 2017 represents data from both the legacy CPS ASEC processing system and the updated CPS ASEC processing system. White refers to non-Hispanic whites, black refers to blacks alone or in combination, Asian refers to Asians alone, and Hispanic refers to Hispanics of any race. Comparable data are not available prior to 2002 for Asians. Shaded areas denote recessions.

Source: EPI analysis of Current Population Survey Annual Social and Economic Supplement Historical Poverty Tables (Table H-5 and H-9)

Copy the code below to embed this chart on your website.

Also of note in the figure above are the dotted lines, which represent an imputation of historical values due to changes in the Census’ data processing system. When corrected for the break in the series, it is clear that Black households have failed to regain the incomes lost during the slow early 2000s recovery and expansion as well as during the Great Recession’s prolonged recovery. Since white households have seen faster growth—though not fast by most benchmarks—the Blackwhite gap in household incomes has actually widened over time. We will look to the latest Census data to see if any progress has been made in closing those gaps as well as achieving decent earnings and income growth across all demographic groups. The growing economy along with a significant drop in inflation in 2019 (1.8% in 2019 as compared to 2.4% in 2018) makes real widespread growth in living standards more likely.

A widespread increase in living standards in 2019 would be promising and should highlight how valuable truly tight labor markets are to most families. There was nothing about the 2019 labor market that was unsustainable economically—core inflation was tame. Federal Reserve Chair Jerome Powell recently affirmed this view, noting that the full-employment economy of the late 1990s did not lead to spiraling inflation as he stressed the importance of allowing the recovery to fully develop to achieve better outcomes for those too often left behind. Next week’s data release will hopefully inform policymakers how to get us back on the road to full employment sooner rather than later.