Pasar al contenido principal

What to Look for in Next Week’s Census Figures on Poverty, Income, and Health Insurance

On September 10, the Census Bureau will release its official statistics on poverty, income, and health insurance coverage in 2018. Three points are worth noting in advance of the release.

1. Progress on Health Coverage May Have Started to Erode

The Census Bureau will issue national and state estimates of health insurance coverage for 2018, based on both its Current Population Survey (CPS) and American Community Survey (ACS). The 2018 figures in both surveys likely will show an increase in the number and share of Americans without health coverage for the first time since enactment of the Affordable Care Act (ACA), signaling an erosion of some of the historic coverage gains made under the ACA.[1] Nonetheless, the uninsured rate will almost certainly remain somewhere near its historic low because the ACA’s major coverage provisions remain in place, despite the Administration’s continuing efforts to reverse and weaken them.

Importantly, Census is revising the methodology behind the CPS’ uninsured rates this year. This means that the 2018 uninsured rate should be compared to Census’ revised 2017 uninsured rate computed under the new methodology — 7.9 percent — rather than to the somewhat higher uninsured rate originally published last year.[2]

Why is it likely that the data will show an increase in the rate of uninsured? Data from the smaller, previously released National Health Interview Survey (NHIS), which typically align with the Census data, show that the uninsured rate increased by 0.3 percentage points in 2018, amounting to about 1.1 million more uninsured people.[3] Likewise, the Congressional Budget Office (CBO) projects an uninsured-rate increase of 0.3 percentage points in 2018, based on NHIS and other data.[4] While an increase of that magnitude is not statistically significant in the NHIS due to its smaller survey sample, it would be statistically significant in both the CPS and the ACS.

A reversal of progress on health coverage in 2018 would likely reflect, at least in part, the growing impact of ongoing Trump Administration efforts to weaken health coverage under the ACA. Among these efforts, the President and Congress repealed the ACA’s individual mandate (the requirement that most people have coverage or pay a penalty). While repeal did not take effect until 2019, CBO and other experts concluded that it reduced coverage starting in 2018 due to consumer confusion.[5]

In addition, the Administration’s harsh stance against immigrants — including new policies that change how Medicaid eligibility is factored into decisions concerning who can legally enter the United States and who can adjust to lawful permanent resident status — have sowed substantial fear and confusion that has likely increased uninsured rates among individuals in families that include immigrant members.[6] Cuts to outreach and enrollment assistance and support also have reduced enrollment in the ACA marketplaces.[7] And the Administration’s support for new state policies that make it harder for people to enroll or stay enrolled in Medicaid may have contributed to the significant drops in Medicaid enrollment among children and adults in 2018.[8]

These factors will increase uninsured rates, but they likely will reverse only a small fraction of the progress made since the ACA’s enactment. Because of changes in the CPS methodology in 2014 and 2018, the ACS and NHIS are the better surveys for comparing uninsured rates over time. As Figure 1 shows, ACS data show that the uninsured rate plummeted from 15.5 percent in 2010 to 8.6 percent in 2016. Other data show that the uninsured rate in 2016 was the lowest in U.S. history — and that the drop in the uninsured rate as the ACA’s major provisions took effect was the sharpest since the creation of Medicare and Medicaid.[9] Accordingly, if the uninsured rate rises as expected in 2018, it will likely remain near its historic low.

The health coverage gains under the ACA have been greatest in the 32 states (including the District of Columbia) that extended Medicaid to more low-income adults by January 2017, under the ACA’s Medicaid expansion. Had the uninsured rate fallen in non-expansion states as much since 2013 as it did in expansion states, 4.5 million fewer Americans would have been uninsured in 2017.[10] Uninsured rates have fallen for virtually every racial, ethnic, gender, education, and income group that the Census data cover.[11]

Looking ahead, erosion of health coverage is expected to worsen as coverage losses from repeal of the individual mandate and administrative actions weakening the ACA marketplaces and Medicaid continue to grow, especially if various state policies restricting Medicaid coverage — a number of which have so far been blocked by federal courts — take effect. These coverage losses will be partially offset by coverage gains in additional states that have adopted the Medicaid expansion despite Administration efforts to discourage them from doing so.

The greater threat is the Administration’s continued effort to repeal the ACA, whether through litigation or legislation. The Administration recently urged an appeals court to strike down the entire law — eliminating the Medicaid expansion, the health insurance marketplaces and premium tax credits, protections for people with pre-existing health conditions, and all other ACA coverage provisions. If that position were to prevail, the Urban Institute estimates that 20 million people would lose coverage, returning the uninsured rate to slightly higher than before the ACA was enacted. [12]

2. Progress on Income and Poverty Likely Continued

Rising employment and wages in 2018 make it likely that the improvements seen in 2015, 2016, and 2017 — years in which median household income (adjusted for inflation) rose at a healthy pace and the official poverty rate fell significantly — continued last year. This could make 2018 a notable year in some respects:

  • The 2014-2018 period could show the strongest four-year improvement in the official poverty rate since 1969.
  • Median income may have reached a record high in 2018, with data back to 1967.
  • The official poverty rate may have reached its lowest level since at least 2001.

(As the Technical Note below explains, recent methodological changes to the survey data will make it harder to judge whether these milestones have truly been reached.)

These would be welcome milestones. None, however, would signify that 2018 itself was unusually strong for a year in which the overall economy grew. Instead, they would reflect the cumulative effect of a record-long economic recovery that started in 2009 and included some years of exceptionally large drops in poverty and strong income gains. For example, in 2015 the poverty rate fell more than in any other year since 1969: from 14.8 percent to 13.5 percent, making it the only year in the last 50 in which the decline exceeded 1 percentage point. Also in 2015, median household income rose by the largest amount on record — more than 5 percent, after adjusting for inflation — with data back to 1967. (See Figure 2.)

Moreover, reaching a record high median income level is unsurprising in a growing economy. Median inflation-adjusted family income reached a high in 32 of the 60 years preceding the Great Recession. It should also be remembered that the strong gains of recent years continue a long recovery from the depths of the severe downturn that began in December 2007. Even in 2017 — the eighth full year of economic recovery from the Great Recession — median income and poverty had not quite returned to their pre-recession (2007) levels. Modest improvements in 2018 would be a welcome development but would still leave these measures barely improved over 2007. (See Table 1 in Technical Note.)

3. Supplemental Census Data Will Show Impact of Anti-Poverty Programs

In addition to the official poverty figures, Census will release tables on September 10, using the Supplemental Poverty Measure (SPM), that show how many people various government programs lift above the poverty line.

Unlike the official poverty measure, which only counts cash income, the SPM counts various other forms of assistance, including refundable tax credits for working families (the Earned Income Tax Credit and Child Tax Credit), SNAP (formerly known as food stamps), rental assistance, and the Low-Income Home Energy Assistance Program (LIHEAP). (The SPM also accounts for families’ out-of-pocket medical and work expenses, uses a slightly different poverty line, and makes other modifications to the poverty measure.) The SPM tables will show how many people, including children and seniors, each of these programs lifted above the poverty line in 2018.

Those tables will also show how many people are lifted above the poverty line by programs that the official poverty measure does count, because they provide benefits in cash. These include Supplemental Security Income (SSI) for the elderly and disabled poor, Temporary Assistance for Needy Families (TANF) for very low-income families with children, Social Security, and unemployment insurance.

The President’s 2020 budget[13] proposes deep cuts in a number of the low-income assistance programs, including SNAP, SSI, and TANF, as well as drastic cuts in health insurance.[14] While extremely unlikely to be enacted this year or next, the proposals reflect where the President and some congressional leaders would take the country in the years ahead if they have the opportunity to do so. The President’s budget calls for: 1) cutting SNAP by about 30 percent over ten years; 2) cutting the Department of Housing and Urban Development by $9.7 billion, or 18 percent, including eliminating rental vouchers for 140,000 low-income households and raising rents on 4 million other low-income households; 3) eliminating LIHEAP; and 4) repealing the ACA.

These and other proposals in the budget would increase poverty, widen income and racial disparities, and sharply increase the number of uninsured.

Technical Note: Survey Changes May Hamper Multi-Year Comparisons

As noted, Census survey improvements in recent years may complicate some multi-year income, poverty, and health insurance comparisons.

Income and Poverty

Starting with 2013, the Census Bureau altered survey language in the CPS to capture more household income (e.g., additional forms of retirement income).[15] As a result of these changes, caution is necessary when comparing income levels before and after 2013. Census also made additional changes to CPS data-processing procedures for 2017 and subsequent years, though these have smaller effects.[16] And the changes in both years have only a small impact on the overall poverty rate.[17]

On September 10, Census will show CPS income and poverty data for 2013 and 2017 in two versions: one comparable to earlier years and one meant for comparison with later years.

A separate Census Bureau survey, the American Community Survey (ACS), has not substantially changed its method of estimating income and poverty since 2006, so it can be used for direct comparisons across longer periods, such as 2007-2017. Census is not due to release ACS income and poverty data for 2018 until September 26.

Table 1 below presents apples-to-apples data on income and poverty changes over various time periods using ACS and CPS data. Some CPS-based comparisons are not available because of the changes in survey methodology. These data can help provide longer-run context for the data that Census will release on September 10 and can help in comparing the improvements in 2018 with previous years.

Apples-to-Apples Changes in Income and Poverty
  Percent Change in
Median Household Income (Inflation-Adjusted)
Percentage-Point Change in Official Poverty Rate
2016 to 2017 +1.8 +2.5 -0.4 -0.6
2015 to 2016 +3.1 +2.0 -0.8 -0.7
2014 to 2015 +5.1 +3.8 -1.3 -0.8
2007 to 2017 N/A +0.3 N/A +0.4

Source: U.S. Census Bureau. Additional calculations by CBPP. All changes shown are statistically significant.

Health Insurance

Census also made alterations to the CPS survey language starting in 2013 in order to reduce the underreporting of health insurance coverage. To increase recall accuracy, the survey now asks both about an individual’s current source of health insurance coverage and about the individual’s sources of health coverage over the entire year.

In addition, Census has made changes to the methods it uses to impute health insurance coverage status to individuals with missing information in the survey. The revised imputation process is more comprehensive and is based on the health insurance unit, enabling it to more accurately reflect how families typically purchase health insurance coverage. For the first time, in the upcoming Census Bureau release, health insurance coverage estimates will reflect the impact of both the redesigned health insurance questions and the improved imputation process. Consistent estimates will be available for 2017 and 2018.

Due to these changes, historical comparisons of health insurance coverage data based on the CPS should not be made with data for years prior to 2017. Those interested in longer-term health insurance coverage trends should use the ACS instead as the preferred data source.

End Notes

[1] The ACS, but not the CPS, showed a statistically significant increase for 2017 as well.

[2] For the Census Bureau’s explanation of recent revisions to the processing system for Current Population Survey data, see The 2017 uninsured rate was 8.8 percent under the former system. The Census Bureau’s existing published materials include this estimate, rather than the 7.9 percent figure Census has computed under the revised system.

[3] Robin Cohen et al., “Health Insurance Coverage: Early Release of Estimates from the National Health Interview Survey, 2018,” Centers for Disease Control and Prevention, May 2019,

[4] “Health Insurance Coverage for People Under 65: Definitions and Estimates for 2015 to 2018,” Congressional Budget Office, April 2019,

[5] Congressional Budget Office, “Federal Subsidies for People Under Age 65: 2019 to 2029,” May 2019,

[6] Hamutal Bernstein et al., “One in Seven Adults in Immigrant Families Reported Avoiding Public Benefit Programs in 2018,” Urban Institute, May 2019,

[7] Centers for Medicare & Medicaid Services, “Trends in Subsidized and Unsubsidized Enrollment,” August 12, 2019,

[8] Matt Broaddus, “Research Note: Medicaid Enrollment Decline Among Adults and Children Too Large to Be Explained by Falling Unemployment,” Center on Budget and Policy Priorities, July 17, 2019,

[9] Council of Economic Advisers, “2017 Economic Report of the President,” December 15, 2016,

[10] Matt Broaddus, “Census: States Not Expanding Medicaid Lagging on Health Coverage,” Center on Budget and Policy Priorities, September 12, 2018,

[11] Center on Budget and Policy Priorities, “Coverage Gains Under the ACA Have Been Widely Shared,” March 23, 2018,

[12] Linda Blumberg et al., “State-by-state Estimates of the Coverage and Funding Consequences of Full Repeal of the ACA,” Urban Institute, March 2019,

[13] Richard Kogan et al., “Cuts to Low-Income Assistance Programs in President Trump’s 2020 Budget Are Wide-Ranging,” Center on Budget and Policy Priorities, May 15, 2019,

[14] The SPM does not treat health insurance as income (or calculate the number of people lifted out of poverty by health coverage programs). Having health insurance can affect a family’s poverty status indirectly under the SPM, however, by reducing its out-of-pocket medical expenses.

[15] Census reported a median income level for 2013 that was 3.2 percent higher (about $1,780 higher in 2018 dollars) for households asked the new survey questions as compared to households asked the old survey questions. Carmen DeNavas-Walt and Bernadette D. Proctor, “Income and Poverty in the United States: 2014,” U.S. Census Bureau, September 2015,

[16] The 2017 changes lowered median income for that year by 0.4 percent, or about $238. U.S. Census Bureau, Income and Earnings Summary Measures by Selected Characteristics: 2018 CPS ASEC Production File and 2018 CPS ASEC Bridge File,, adjusted for inflation to 2018 dollars.

[17] The 2013 changes to the CPS did not significantly alter the poverty rate, which was 14.5 percent based on the Census Bureau’s traditional survey questions. Using a redesigned survey and a relatively small survey sample, Census estimated that the 2013 poverty rate was approximately 14.8 percent — not statistically significantly different from 14.5 percent —and would have equaled about 14.5 percent if the new questions had been applied in a somewhat larger CPS survey sample. See DeNavas-Walt and Proctor, “Income and Poverty in the United States, tables D-2 and E-2.

The 2017 changes did not alter the Census Bureau’s estimate of the overall poverty rate for 2017, which was 12.3 percent under both the old and new processing systems. U.S. Census Bureau, “Income and Earnings Summary Measures by Selected Characteristics: 2018 CPS ASEC Production File and 2018 CPS ASEC Bridge File,”