Wednesday, July 15, 2020

How many newly uninsured? Families USA and the feds paint different pictures

Click here to subscribe to xpostfactoid

Families USA released a report this week estimating that 5.4 million adults newly unemployed by the pandemic as of May are also becoming uninsured.

The estimate is based on data and analyses concerning  1) the number of people who have lost jobs (BLS), 2) the likely percentage of newly unemployed who have lost access to health insurance, and 3) the percentage of newly unemployed in the ACA era (2014-2018) who have found other insurance, including Medicaid and marketplace (Urban Institute).

The analysis boils down to a fairly simple equation: about a quarter of those who lose jobs do not find their way to job-based insurance through a family member, Medicaid, marketplace coverage or COBRA. Much higher percentages of the newly unemployed are estimated to become uninsured in states that have refused the ACA Medicaid expansion (42.5%) than in expansion states (22.6%).

5.4 million uninsured adults equates to about a 2.6 percentage point increase in the uninsured rates for Americans aged 18-64.  Based on the most recent National Health Interview Survey (Jan-June 2019), that would suggest an uninsured rate of about 16.3% for ages 18-64. As I've noted recently, more immediate survey data seems to indicate less severe increases in the uninsured population. A Commonwealth Fund survey of about 2300 adults conducted May 13 through June 2 seemed to indicate that a bit less than 2% of the population had recently become uninsured.

More strikingly, the experimental new Household Pulse Survey updated weekly by the CDC -- a joint project of the Census Bureau and National Center for Health Statistics -- shows comparatively little movement since its launch in the week of April 23, when it recorded a national uninsured rate of 12.6%  for ages 18-64 -- more than a percentage point lower than the 2019 NHIS estimate. The weekly updates have been quite volatile but on the whole have ticked upward, reaching 13.5% in the week of June 25 (there'll be another update tomorrow).  Estimates for individual states, shown below, average three percentage points higher in Pulse than in the FUSA study, and vary widely.
Click here to subscribe to xpostfactoid

Caveats abound. The Pulse survey is conducted online only, which may skew responses toward the more affluent, and response rates are low. The confidence interval for the national rate extends to 14.1% in the week of June 25, and for individual states the much wider intervals mostly encompass the Families USA estimate.

Still, self-reported insurance status seems to show less dire losses than estimates based on past behavior -- not only Families USA's, but the Urban Institute's, the Kaiser Family Foundation's, and others. That doesn't mean that the current survey results paint a more accurate picture, however.

More generally, current survey data may not provide a more accurate picture than analysis based on historical patterns. I asked Stan Dorn, author of the Families USA report about this. While acknowledging that the historical patterns that the FUSA analysis relied on may not apply in the current crisis (a social science version of "past performance does not guarantee future results"), he noted:
Coverage losses typically take time to materialize during a downturn. During the Great Recession, for example, many workers were originally furloughed, receiving benefits without pay, with both firms and workers optimistic about an early return to work. As time went by, temporary suspensions turned into full employment terminations...The same pattern appears to be emerging under the current pandemic-driven recession. Put differently, many who are becoming uninsured due to job loss are in their final stages of receiving employer-based coverage.
The uninsured rate does seem to be a lagging indicator behind unemployment. As I noted recently, a United Hospital Fund study of Medicaid enrollment in New York in the 2008-9 recession found that peak Medicaid enrollment occurred seven months after peak unemployment filing.  In addition to furloughed workers eventually losing their jobs and job-coverage, a fair number of people may pay COBRA premiums for a spell, particularly as buoyed by Covid relief checks and enhanced employment insurance payments, which are set to expire soon (The Commonwealth Fund survey showed more of the newly unemployed taking up COBRA than Medicaid or marketplace). A laid off employee has 60 days to elect COBRA and another 45 days to pay the first premium. As coverage is retroactive once the premium is paid, it can be used as a free placeholder if never effectuated.

These factors suggest that the Families USA report may be more of a forecast than a May snapshot. Tragically, had the U.S. used the shutdown from March through May to crush the curve on Covid-19 infection, as most wealthy countries did,  the apparent relative modesty of immediate health insurance losses may have had more long-term salience. Furloughs may not have turned into long-term unemployment at rates now likely; COBRA might have plugged the gap for more people; recent gains in employment may have been sustained.  In insurance as in employment and education, the current surges in Covid infections threaten to lock in and extend the damage.

All that said, the gap between Families USA estimates and Pulse Survey results is worth noting. Below, I've set the Families USA state-by-state estimates beside a three-week average (June 11--June 30) of Pulse Survey results (which can fluctuate pretty wildly week-by-week on a state level).

Uninsured Population Aged 18-64: Pulse Household and Families USA Estimates
Pulse Survey weeks 7-9 (June 11-30, 2020); Families USA May 2002 estimate

The average national uninsurance rate for adults in the Pulse Survey for the weeks cited, June 11-13, was 13.2%. From May 7 - June 2, the nearest match to the Families USA study period, the average Pulse rate was 12.8%, so the contrast would have been a bit sharper.

For the three states missing data in one week (Hawaii, Iowa, Maine, Rhode Island), I used a two-week average.

Update, 11:30: I thought the Pulse survey updated on Thursdays, but I just went to check something, and Week 10 (July 2-7) is up. The national uninsurance rate has spiked to 14.0% -- the first time it's risen above the Jan-June 2019 NHIS estimate. Score one for the lagging indicator hypothesis?

Update 2: Stan Dorn points out to me that the Pulse data includes 18 year-olds, who are Medicaid-eligible and so, like children generally, have a lower uninsured rate. For 2020, the FUSA study relies on Urban Institute data for 19-64 year-olds.

Click here to subscribe to xpostfactoid

No comments:

Post a Comment