Many Working People Could Lose Health Coverage Due to Medicaid Work Requirements
April 11, 2018
Recent Trump Administration guidance for the first time lets states take away Medicaid coverage from people who are not working or participating in work-related activities for a specified number of hours each month. The guidance allows states to impose work requirements on adult Medicaid enrollees other than those who are 65 or older, pregnant, or qualify for Medicaid because they receive federal disability assistance. The first work requirement policy that the Centers for Medicare & Medicaid Services (CMS) approved was Kentucky’s, under which enrollees can lose coverage if they fail to document 80 hours of work or work activities each month. CMS has also approved work requirement policies in Indiana and Arkansas, and additional states have proposals pending at CMS or under discussion.
Medicaid work requirements will almost certainly cause many low-income adults to lose health coverage. Coverage is most obviously threatened for enrollees who are not already working, who make up roughly 40 percent of those potentially subject to work requirements. More than 80 percent of these enrollees report that they are unable to work due to an illness or disability, caregiving responsibilities, or because they are in school. While state policies include exemptions for people with disabilities and certain caregivers and students, some people with serious barriers to work will not meet the criteria for exemptions or will struggle to overcome the bureaucratic hurdles to document that they qualify.
Less understood, many working people also will likely lose coverage due to work requirements.Less understood, many working people also will likely lose coverage due to work requirements. Most non-elderly adult Medicaid enrollees work, but in low-wage jobs that generally do not offer health insurance and are often unstable, with frequent job losses and work hours that can fluctuate sharply from month to month. As a result, our analysis finds that 46 percent of low-income workers who could be affected by Medicaid work requirements would be at risk of losing coverage for one or more months under a work requirement policy like Kentucky’s. Even among people working 1,000 hours over the course of the year — about 80 hours per month, meeting Kentucky’s standard on average — 1 in 4 (25 percent) would be at risk of losing coverage for one or more months because they would not meet the 80-hour requirement in every month. (See Figure 1.)
These estimates suggest that, if all states adopted work requirements along the lines of Kentucky’s, millions of working people could lose coverage or face interruptions in coverage due to unstable employment or hours. In addition, some enrollees who are actually meeting Medicaid work requirements will likely lose coverage because they get tripped up by the paperwork required to prove it.
State Work Requirement Policies
Approved and proposed state work requirements vary significantly in their details, but all would put coverage at risk for workers with unstable employment and hours.
Kentucky’s work requirement was the first approved by CMS. Its policy requires enrollees to document 80 hours of work or qualifying work activities, such as volunteer work or job training, each month. Enrollees who fail to meet the requirement in a given month have one month to make up the missed hours by working those hours on top of the standard requirement (e.g., working 120 hours if they worked 40 hours the previous month). Otherwise, their coverage is suspended the following month and reinstated only in the month following a month in which they again meet the work requirement. (Enrollees can also prevent a suspension of coverage or regain eligibility by completing a health or financial literacy course, but can only exercise this option once per year.)
Other states’ work requirements are more or less stringent than Kentucky’s in various respects. Under Arkansas’ approved policy, for example, enrollees will lose coverage if they fail to meet the 80-hour monthly work requirement for any three months during a calendar year. However, once they lose coverage, they are locked out of Medicaid until the subsequent year (up to nine months), with no option to regain their health insurance. Most states require or are proposing to require at least 80 hours of work or work activities per month (i.e., 20 hours per week), but New Hampshire, for example, is proposing to ultimately require 30 hours per week, Michigan is considering requiring 29 hours per week, and Alabama is proposing to require 35 hours per week for parents of children age 6 and older (20 hours per week for parents of younger children).
As noted, work requirement policies generally allow enrollees to meet the requirement by completing activities other than paid employment, although the specific set of activities varies by state. In theory, an enrollee who is out of work or unable to find sufficient hours of work in a given month could meet the requirement with other activities. But in practice, this will likely prove challenging or impossible for most workers. Those with irregular hours may not know they will come up short until the end of the month, and their employers may require them to be available for work much of each week even without guaranteeing them a particular number of hours. Likewise, a worker who loses his or her job will generally find it difficult to immediately line up sufficient volunteer or job training hours. Notably, approved and pending state work requirement proposals do not guarantee volunteer or job training slots to enrollees who want them, or even guarantee any assistance finding a slot.
State work requirement policies vary in the exemptions they allow. The CMS guidance requires states to exempt people who cannot work due to illness or disability, but states have significant scope to set the criteria and procedures for those exemptions. States are also required to exempt individuals who are meeting or exempt from work requirements in the Supplemental Nutrition Assistance Program (SNAP) or Temporary Assistance for Needy Families (TANF). Kentucky provides an exemption for one parent or caregiver of a minor child or person with disabilities, while Indiana exempts only parents of young children, and several states are proposing to exempt only parents of infants. States also vary in their treatment of students and people with substance use disorders and in whether they apply work requirements to older adults.
Many Low-Wage Workers Experience Spells of Unemployment or Low Hours
To analyze the impact of Medicaid work requirements on working people, we use the Census Bureau’s Survey of Income and Program Participation (SIPP). The SIPP is unique among federal surveys in the detailed longitudinal data it collects on households’ wages and work hours over several years. The most recent suitable 12 months of data spans June 2012 to May 2013, before the Affordable Care Act’s (ACA) expansion of Medicaid to cover more low-income adults. Nonetheless, we can use the SIPP to analyze work patterns among adults who would likely have been eligible for Medicaid if expansion had taken effect: namely, those with monthly family income below 138 percent of the federal poverty line.
Specifically, we examine what share of adults would fail to meet Kentucky’s 80-hour-per-month work requirement at some point during a year. We focus on Kentucky’s requirement because it was the first approved and (as discussed above) is fairly typical of state policies, some of which are harsher and others more lenient.
Our analysis examines the population of likely Medicaid-eligible adults who could be included in state work requirement policies under the CMS guidance: adults age 19 to 64 whose Medicaid eligibility is not based on receipt of federal disability assistance. For several reasons, we do not attempt to exclude other individuals who might be exempt under state policies. First, as noted, these policies vary significantly across states. In addition, criteria for some exemptions (e.g., inability to work due to illness or disability) are not well defined, even in approved state waivers, and in some cases are not observable in our data. Moreover, some individuals eligible for exemptions will likely have difficulty gathering and submitting the evidence to prove they qualify and thus will still be at risk of losing coverage if they cannot meet the work requirement standard. Because we cannot model all features of state policies, however, our analysis identifies the share of working enrollees at significant risk of losing or experiencing interruptions in coverage, not those who would necessarily do so. (For additional details of our analysis, see the Appendix.)
Consistent with other research, we find that most (about two-thirds of) non-elderly low-income adults who had incomes below the current Medicaid eligibility limit in some month worked at some point during the 2012-2013 year. Of those, most worked substantial hours over the course of the year, suggesting strong attachment to the labor force. Nearly 9 in 10 worked at least 500 hours, and over 70 percent worked at least 1,000 hours, or an average of about 80 hours per month.
But while most adults in low-income families worked substantially, many did not work continuously and for a consistent number of hours each month. As a result, 46 percent of likely Medicaid-eligible low-income working adults would have been at risk of losing their Medicaid coverage for at least one month during the year under Kentucky’s work requirement. Among those working at least 500 hours, 38 percent would have been at risk of losing coverage for at least one month. And even among those working 1,000 hours (i.e., meeting the Kentucky work requirement standard on average over the course of the year), 25 percent would have been at risk of losing or seeing an interruption in coverage. (See Table 1.)
|Work Participation Among Low-Income Adults Potentially Subject to Medicaid Work Requirements|
|Worked fewer than 80 hours in at least one month|
|Total number of workers (millions)||Number (millions)||Share|
|At least 500 hours||18.1||6.9||38%|
|At least 1,000 hours||15.0||3.8||25%|
More Recent Data on Medicaid Enrollees
When we use the Census Bureau’s Annual Social and Economic Supplement to the Current Population Survey (CPS ASEC) to look at adults actually enrolled in Medicaid in 2016, we similarly find that large shares of working enrollees would be at risk of losing coverage under state work requirement policies. The CPS ASEC does not track monthly hours and earnings, so we cannot directly evaluate whether participants would have met Kentucky’s Medicaid work requirement standard in any given month. However, it does ask participants about their total earnings over the course of the year, number of weeks of work, and usual weekly hours, as well as whether they were enrolled in Medicaid. Also, because the survey is conducted annually, more recent data are available, from a period after the ACA’s Medicaid expansion took effect. We therefore use the CPS ASEC to look at patterns of work participation among adults who report being enrolled in Medicaid at some point during the year.
Again, we find that most non-elderly adult Medicaid enrollees not receiving federal disability assistance worked over the course of the year (over 60 percent). But well over a third of workers would have been at risk of losing or seeing interruptions in coverage under a work requirement policy like Kentucky’s. Among all enrollees with earnings, 37 percent were either out of work at some point during the year or typically worked fewer than 20 hours per week. Even among those with earnings over $7,250 — the equivalent of half-time work at the federal minimum wage — 26 percent had either a spell of unemployment or usual hours under 20 per week, putting them at risk of losing coverage for some period.
The 2016 CPS results also confirm that our key findings hold in an economy closer to full employment. (The unemployment rate averaged 4.9 percent during 2016, compared to 7.8 percent during the 2012-2013 period covered by our SIPP data.)
Findings Reflect Instability of Low-Wage Labor Market
Approved and pending state work requirement policies are based on the assumption that people who want to work can find steady employment at regular hours. This assumption is out of step with the realities of the low-wage labor market.
Many Medicaid enrollees work in industries in which both employment and hours are volatile. The two industries that employ the most Medicaid enrollees potentially subject to work requirements are restaurant/food services and construction. Grocery stores, department and discount stores, and the home health industry also employ large numbers of Medicaid enrollees. Jobs in these industries are characterized by:
- Unstable employment. The construction, retail, and accommodation/food services industries all have above-average job separation rates; the latter two industries also have substantially below-average median job tenure. While many factors contribute to this instability, one is the lack of flexibility in these and other low-wage jobs. Workers may lose their job if an illness or family emergency or a disruption in child care arrangements forces them to take even a short amount of time off from work. Among working adults in our sample, common reasons for spells of unemployment include an inability to find work or a layoff from a previous employer (43 percent), caregiving responsibilities (18 percent), and chronic illness or temporary illness or injury (9 percent).
- Volatile hours. Majorities of workers in the food services or production (71 percent), retail/wholesale trade (63 percent), and construction (54 percent) industries report that their hours vary significantly from week to week.  The retail and leisure/hospitality industries also have high rates of involuntary part-time work (i.e., when workers want full-time work but can only obtain part-time work), and the rate is above-average in construction. Moreover, the retail industry and other sales occupations also have above-average rates of irregular scheduling, in which employees may be expected to be “on call” and available for work much of the week but may receive well under 20 hours of work.  Under these circumstances, workers may find it nearly impossible to consistently meet 80-hour-per-month work requirements.
In addition to job conditions, other challenges faced by low-income workers, such as unreliable transportation arrangements or unstable housing, may also contribute to employment volatility. Recent research finds that low-income renters who experience a forced move (such as following an eviction) are likelier to be laid off from their jobs than otherwise-similar renters who did not experience a forced move.
In short, even people with strong labor force attachment, who are highly motivated to find and maintain employment, will often experience periods during the year when they are not working or are not working enough hours to meet state work requirements.
Burdensome Reporting Requirements Will Also Cause Coverage Losses
Low-wage workers may also have trouble documenting their compliance with work requirements even in the months when they meet them. While some states may be able to verify compliance for some enrollees electronically, other enrollees may have to submit paystubs, timesheets, or other documentation each month, sometimes from multiple employers. This will be challenging and time consuming for enrollees and creates many chances for people to lose coverage due to inadvertent paperwork errors — theirs, their employer’s, or the state’s. State backlogs in processing paperwork could also cause people to lose coverage or experience coverage interruptions.
Moreover, some states plan to impose even more onerous reporting requirements. Arkansas, for example, will require enrollees to report their hours for the previous month by the fifth day of each month; workers who miss that deadline will be considered out of compliance for the month. Arkansas will also require enrollees to report work participation through an online portal, a major challenge for enrollees without Internet access. (In Kentucky, 19 percent of non-elderly adult Medicaid enrollees lack any Internet access and 42 percent lack broadband access, recent research shows; Internet access in Arkansas is likely similar.)
Studies consistently show that red tape and paperwork requirements reduce enrollment in Medicaid. The same will likely be true for working people newly subject to work requirements.
Interruptions in Coverage Due to Work Requirements May Impede Work
The Administration’s argument for Medicaid work requirements rests on the premise that, by strengthening incentives to work, these requirements will increase labor force participation. Evidence from other programs casts doubt on that premise, with studies finding that work requirements generally have only modest and temporary effects on employment, failing to increase long-term employment or reduce poverty. For enrollees who already work substantial hours, thereby demonstrating strong labor force attachment, Medicaid work requirements are particularly unlikely to have a positive impact on work. Instead, they will add to the challenges of low-wage work by taking away workers’ health coverage during periods when they cannot find work or sufficient hours of work.
Meanwhile, coverage losses or interruptions in coverage can have serious adverse impacts on health. While losing health coverage worsens health for all groups, it is especially harmful for people with chronic health conditions and other serious health needs, who require regular access to medications and other treatments to maintain their health. Low-income people have above average rates of chronic conditions, and even working Medicaid enrollees — who are generally healthier than other enrollees — have high rates of serious health needs. For example, a study of working adults enrolled in Michigan’s Medicaid expansion found that more than half had a serious physical health condition such as heart disease, asthma, or diabetes and 25 percent had a mental health condition, often depression.
For such individuals, disrupting access to care may not only worsen their health but also impede their employment. Majorities of working people who gained coverage through the Medicaid expansion in Ohio and Michigan reported that it made them better at their jobs or made it easier for them to keep working. Likewise, a long-term randomized trial found that providing older adults with regular care for heart disease increased their earnings, likely by reducing their time out of work due to illness.
In contrast, Medicaid work requirements are likely to set off a vicious cycle for some working enrollees. Health problems are a common cause of job loss among low-wage workers, in part because (as discussed above) low-wage jobs often offer little flexibility to take time off due to illness. In states with work requirements, some workers who lose their jobs due to health setbacks may then lose their health coverage and access to treatment as well, which would make it far harder for them to regain their health and their employment. Similarly, loss of coverage due to failure to document sufficient hours of work may lead to deteriorating health, causing job loss.
Primary SIPP Analysis
This analysis uses the 2008 panel of the Census Bureau’s Survey of Income and Program Participation (SIPP) to assess the work rates of individuals in low-income families. The SIPP is a large-scale, national survey that collects information about household and individual income, program participation, labor force activity, and demographics, including monthly information on employment, hours, and earnings.
The 2008 SIPP panel followed households from 2008 through most of 2013. For this analysis, we focus on the final 12 months in which all households were included in the survey, June 2012 through May 2013. We include only individuals who responded to the survey in all 12 months.
Our analysis focuses on non-elderly adults in families with monthly incomes below 138 percent of the federal poverty line. We define a “non-elderly adult” as anyone who is 19-64 years old throughout the period. To capture the population potentially affected by work requirements, we exclude those who received disability income from Supplemental Security Income (SSI) or Social Security Disability Insurance (SSDI) at any point during their 12 months in our sample.
For our primary analysis, we focused on adults in families with monthly incomes below 138 percent of the poverty line. We weight all results by the number of months in which a person’s income fell below the Medicaid limit, meaning that estimates are applicable to the population likely eligible for Medicaid in a typical month. (As discussed below, results are similar if we proxy for Medicaid eligibility based on annual income.)
We calculate monthly work hours by multiplying an individual’s usual weekly hours worked in a given month by the number of weeks he or she reported working in that month. We then sum this over a person’s 12 months to get annual hours worked. We classify individuals as working if they worked at any point over the course of the year. We classify individuals as at risk of losing or experiencing an interruption in coverage under Kentucky’s work requirement if they have fewer than 80 hours of work in one or more months (from June 2012 to May 2013) when they would also have been eligible for Medicaid.
One well-known limitation of the SIPP is what is known as “seam bias.” Since households are surveyed once every four months, they sometimes report the same answers for each of those four months. For example, rather than accurately reporting their hours for each of the four months, they may accurately report their hours for the most recent month and then report the same hours for the previous months. Or, they may report an approximate average of their hours over the period as their hours for each of the four months.
Because we focus on whether households fail to meet the Kentucky 80-hour-per-month work requirement in at least one month during the year, seam bias will generally cause our analysis to understate the share of households at risk of failing the requirement. That’s because seam bias usually manifests itself in one of two ways. In the first case, in which an individual misreports one month of less-than-half-time work as four months of less-than-half-time work, we will correctly characterize the individual as failing the work requirement in at least one month. In the second case, in which an individual misreports one month of more-than-half-time work as four months of more-than-half-time work, we may mistakenly categorize the individual as meeting the work requirement in all four months.
In addition to our primary analysis, we consider several alternatives, which produce similar findings.
Using annual rather than monthly income as a proxy for Medicaid eligibility. Medicaid eligibility is based on monthly, rather than annual, income, and so individuals in families with incomes over 138 percent of the poverty line for the year as a whole may still be eligible for Medicaid for part of the year. However, people who are eligible for Medicaid for only a few months of the year are probably less likely to enroll. As an alternative to our main analysis, we examine individuals whose annual family income falls below 138 percent of the poverty line. As Appendix Table 1 shows, among those in this group who worked at some point during the year, 49 percent worked fewer than 80 hours in at least one month, as did 27 percent of those who worked at least 1,000 hours during the year.
Excluding caretakers of young children. As explained in the main text, we generally did not attempt to evaluate who in our sample might be exempt from work requirements under state exemption criteria. However, because most states exempt or propose to exempt at least primary caregivers of children age 6 and under, we examined how excluding such individuals would affect our results. As shown in the table, results are similar to the results for the full sample.
Analyzing Data from the 2014 SIPP
As this report explains, the more recent 2014 SIPP would only extend our analysis by a few months and is still being revised. For these reasons, we have chosen to conduct our analysis using the last year of the earlier panel. Nonetheless, as a robustness check, we replicated our analysis using these data, which cover the 2013 calendar year. As the table shows, results are similar to those from the earlier SIPP panel.
|APPENDX TABLE 1|
|Share of Workers With Fewer Than 80 Hours Of Work in at Least One Month Across Various Specifications|
|Primary results||Annual instead of monthly income||Excluding caretakers of young children||Using the 2014 SIPP panel*|
|At least 500 hours||38%||41%||36%||44%|
|At least 1,000 hours||25%||27%||24%||28%|
Because the SIPP data predate the ACA’s expansion of Medicaid to cover low-income adults, we can only use the SIPP to examine work participation among low-income adults who would likely be eligible for Medicaid under the expansion. To confirm that our key findings would hold among adults actually enrolled in Medicaid, we use the Census Bureau’s 2017 Annual Social and Economic Characteristics supplement to the Current Population Survey (CPS) to assess the work rates of Medicaid-enrolled individuals in low-income families. The 2017 CPS includes information for 2016.
The CPS — which, like the SIPP, is a large-scale, national survey — is the preferred source for information on health insurance coverage and also includes robust information about household and individual income, labor force activity, and demographics. However, the CPS does not include the detailed monthly data included in the SIPP; in particular, it includes usual weekly hours for the year as a whole, but not for each month. In addition, respondents are asked to remember their work history for an entire year, instead of being surveyed multiple times during the year. For these reasons, we cannot implement our full SIPP analysis in the CPS; notably, we cannot identify people who worked each month and had usual weekly hours of more than 20 hours per week, but still fell short of 80 hours in some month. The CPS analysis may also be less complete if participants are worse at recalling employment history over longer time periods. Nonetheless, we can still use the CPS to identify workers at risk of losing or seeing interruption in coverage under work requirement policies.
Specifically, we utilize information on Medicaid enrollment, family income, personal earnings, weeks worked in a year, and hours worked in a typical work week. Like our SIPP-based analysis, our CPS-based analysis focuses on non-elderly adults and excludes those who receive SSI or SSDI disability assistance. In this case, we restrict the population to those who report being enrolled in Medicaid at some point during the year. We examine work experience for (1) those with any earnings, and (2) those with earnings equivalent to working 20 hours per week (or half-time) for the full year at the federal minimum wage ($7,250 in annual earnings).
We calculate annual work hours by multiplying an individual’s usual weekly hours by the number of weeks worked in a year. We classify an individual as likely failing Kentucky’s work requirement at some point during the year if they worked fewer than 50 weeks or had usual hours under 20 per week. It is standard to use 50 weeks of work annually as an indicator of full-time work, but our results on the impact of Kentucky’s work requirement would not materially differ if we instead used a lower threshold, such as 45 weeks.
As described in the main text, we find that 37 percent of working Medicaid enrollees with any earnings would have been at risk of losing coverage under a work requirement policy like Kentucky’s in 2006, as would 26 percent of those with at least $7,250 in annual earnings (the equivalent of half-time work at the federal minimum wage).
 The authors thank Stacy Dean, Rachel Garfield, Arloc Sherman, and Judith Solomon for helpful comments.
 Rachel Garfield, Robin Rudowitz, and Anthony Damico, “Understanding the Intersection of Medicaid and Work,” Kaiser Family Foundation, January 5, 2018, https://www.kff.org/medicaid/issue-brief/understanding-the-intersection-of-medicaid-and-work/.
 For example, consider an enrollee who works 40 hours in January. Unless she works 120 hours or more in February, she will lose coverage in March. If she works 80 hours or more in March, she will regain coverage in April; otherwise, her coverage will remain suspended. For details of Kentucky’s policy, see https://www.medicaid.gov/Medicaid-CHIP-Program-Information/By-Topics/Waivers/1115/downloads/ky/ky-health-ca.pdf.
 For a comparison of state proposals, see MaryBeth Musumeci, Rachel Garfield, and Robin Rudowitz, “Medicaid and Work Requirements; New Guidance, State Waiver Details, and Key Issues,” Kaiser Family Foundation, January 16, 2018, http://files.kff.org/attachment/Issue-Brief-Medicaid-and-Work-Requirements-New-Guidance-State-Waiver-Details-and-Key-Issues.
 While the Census Bureau recently released data from the first wave of the 2014 SIPP panel, covering 2013, the new panel now only asks questions every 12 months (a response to budget cuts) and for some purposes may be worse at tracking monthly variation in outcomes. Since it is still undergoing revision and would currently only extend our analysis by a few months, we conducted our analysis using the earlier panel. We focus on the final 12 months of the full sample to capture a period well after the end of the Great Recession. Our results would be similar using preliminary data from the 2014 SIPP panel, covering calendar year 2013 (see Appendix).
 Medicaid eligibility is based on monthly, rather than annual, income, and so individuals in families with incomes over 138 percent of the poverty line for the year as a whole may still be eligible for Medicaid for part of the year. (Adults with family income above that level may also be eligible for a transitional period when their incomes increase.) Our main analysis includes non-elderly adults who do not receive disability income and whose family income falls below 138 percent of the poverty line for any months during the year. We consider whether these individuals would have failed the Kentucky work requirement standard in some month of the year during which they were also eligible for Medicaid. All results are weighted by the number of months in which individuals were eligible for Medicaid. For additional details, see the Appendix. The Appendix also presents alternative results limited to people in families with annual incomes below 138 percent of the poverty line; these results are broadly consistent with, but show slightly higher failure rates than, our main analysis.
Because we assume that adults with family incomes up to 138 percent of the poverty line would be eligible for Medicaid, our results are most applicable to work requirements in states that have adopted the ACA Medicaid expansion. We do not attempt to limit our sample based on other requirements for Medicaid eligibility, such as immigration status.
 The CMS guidance also precludes imposing work requirements on pregnant women, whom we are not able to exclude from our analysis. It is unlikely that excluding this group would meaningfully change the results.
 Center on Budget and Policy Priorities, “How Medicaid Work Requirements Will Harm Seniors and People with Disabilities,” January 26, 2018, https://www.cbpp.org/research/health/how-medicaid-work-requirements-will-harm-people-with-disabilities-and-serious.
 The Appendix also examines how our results would change under alternative approaches to identifying adults likely eligible for Medicaid and if we exclude parents of young children; the results are qualitatively similar.
 These and all other results are weighted by months of Medicaid eligibility. They thus reflect the share of adults eligible for Medicaid in a typical month who worked at some point during the year.
Similarly, Kaiser Family Foundation researchers find that 60 percent of adult Medicaid enrollees who could be subject to work requirements under the CMS guidance work. See Garfield, Rudowitz, and Damico, op cit.
 For a more in-depth discussion of how labor market volatility affects employment patterns among low-wage workers, see Brynne Keith-Jennings and Raheem Chaudry, “Most Working-Age SNAP Participants Work, But Often in Unstable Jobs,” Center on Budget and Policy Priorities, March 15, 2018, https://www.cbpp.org/research/food-assistance/most-working-age-snap-participants-work-but-often-in-unstable-jobs.
 Garfield, Rudowitz, and Damico, op cit.
 Bureau of Labor Statistics, “Annual Total Separation Rates By Industry and Region, Not Seasonally Adjusted,” https://www.bls.gov/news.release/jolts.t16.htm and Bureau of Labor Statistics, “Employee Tenure in 2016,” https://www.bls.gov/news.release/pdf/tenure.pdf. For a given industry, the job separation rate is the number of people leaving their job during the year (voluntarily or involuntarily) as a share of annual average employment. Median job tenure refers to the median number of years that wage and salary workers have been with their current employer.
 For workers in our sample who report at least one month not working (in a month in which their family income is below the Medicaid income limit), we look at the reason given for not working in the first month out of work.
 Golden, op cit.
 Matthew Desmond and Carl Gershenson, “Housing and Employment Insecurity Among the Working Poor,” Social Problems, Vol. 63, Issue 1, February 2016, https://academic.oup.com/socpro/article-abstract/63/1/46/1844105.
 Judith Solomon, “Arkansas’ Harsh Medicaid Work Requirement Jeopardizes Recent Progress,” Center on Budget and Policy Priorities, March 5, 2018, https://www.cbpp.org/blog/arkansas-harsh-medicaid-work-requirement-jeopardizes-recent-progress.
 Anuj Ganopadhyaya and Genevieve Kenney, “Who Could Be Affected by Kentucky’s Medicaid Work Requirements, and What Do We Know about Them?,” Urban Institute, February 2018, https://www.urban.org/sites/default/files/publication/96576/2018.02.15_ky_medicaid_numbers_finalized_1.pdf.
 Margot Sanger-Katz, “Hate Paperwork? Medicaid Recipients Will Be Drowning in It,” New York Times, January 18, 2018, https://www.nytimes.com/2018/01/18/upshot/medicaid-enrollment-obstacles-kentucky-work-requirement.html.
 CMS letter to state Medicaid directors (18-002), January 11, 2018, https://www.medicaid.gov/federal-policy-guidance/downloads/smd18002.pdf.
 LaDonna Pavetti, “Work Requirements Don’t Work,” Center on Budget and Policy Priorities, January 10, 2018, https://www.cbpp.org/blog/work-requirements-dont-work.
 Renuka Tipiereni, Susan D. Goold, and John Z. Ayanian, “Employment Status and Health Characteristics of Adults with Expanded Medicaid Coverage in Michigan,” Journal of the American Medical Association, December 11, 2017.
 Ohio Department of Medicaid, “Ohio Medicaid Group VIII Assessment: A Report to the Ohio General Assembly,” January 2017, http://medicaid.ohio.gov/Portals/0/Resources/Reports/Annual/Group-VIII-Assessment.pdf. See also Kara Gavin, “Medicaid Expansion Helped Enrollees Do Better at Work or in Job Searches,” University of Michigan Health Lab, June 27, 2017, http://labblog.uofmhealth.org/industry-dx/medicaid-expansion-helped-enrollees-do-better-at-work-or-job-searches.
 Melvin Stephens, Jr. and Desmond J. Toohey, “The Impact of Health on Labor Market Outcomes: Experimental Evidence from MRFIT,” National Bureau of Economic Research Working Paper No. 24231, January 2018, http://www.nber.org/papers/w24231?utm_campaign=ntw&utm_medium=email&utm_source=ntw.
 The Census Bureau typically calculates poverty thresholds at the family level. Related subfamilies, which are immediate families that include someone related to a person in the primary family, are typically assigned the family income and poverty thresholds of the primary family. Since Medicaid eligibility is based on the immediate family, we instead calculated income and poverty thresholds at the subfamily level. We calculated subfamily poverty thresholds using the U.S. Department of Health & Human Services’ poverty guidelines, see https://aspe.hhs.gov/poverty-guidelines.
 Results are also similar if we do not weight by the number of months in which a person’s income fell below the Medicaid income limit.
 This calculation effectively treats a person’s “usual hours of work” as their average hours of work (during the weeks they were working). If people instead report their usual hours of work, ignoring weeks in which their hours were especially high or low, that could lead us to either over- or underestimate the share of workers failing to meet an 80-hour-per-month requirement.
 Seam bias is also likely to cause us to further understate the share of people working more than 1,000 hours per year who fail the work requirement in at least one month. To the extent we mistakenly classify people as meeting the work requirement in all months (because they overreport their hours in some months), we also likely misclassify them as working more than 1,000 hours.
 Specifically, we focused on the heads or spouses of subfamilies and we excluded single heads of subfamilies living in units with children under 6 or one co-parent of a child under 6 (we excluded the parent with fewer hours of work or, if they worked equal numbers of hours, the head of the unit). We focused on subfamilies to capture those families where a grandparent or other relative is taking care of a young child. We repeated this analysis instead matching children to their parents and found similar results.