Shocking news — work requirements don’t work

Benjamin Sommers** and others published an important study with a completely expected result on Wednesday in the New England Journal of Medicine.  They surveyed Arkansas to evaluate what was happening with employment and insurance coverage as a result of the state’s decision to implement work requirements for Medicaid:

We conducted a telephone survey to compare changes in outcomes before and after implementation of the work requirements in Arkansas among persons 30 to 49 years of age, as compared with Arkansans 19 to 29 years of age and those 50 to 64 years of age (who were not subject to the requirement in 2018) and with adults in three comparison states — Kentucky, Louisiana, and Texas. ….

Our study had three primary outcomes: the percentage of respondents with Medicaid, the percentage of respondents who were uninsured, and the percentage of respondents reporting any employment. Secondary outcomes were the number of hours worked per week, the percentage of respondents satisfying any category of community engagement requirement (described below), the percentage of respondents with employer-sponsored insurance, and two measures of access to care — the percentages of respondents having a personal physician and reporting any cost-related delays in care….

estimate of changes in the percentage of respondents who were not insured was an increase of 7.1 percentage points (95% CI, 0.5 to 13.6; P=0.04).

Uninsurance rates increased more for the work requirement cohort than other cohorts.  This is not an unexpected result.  Almost every pre-waiver approval analysis projected significant enrollment drops due to increased paperwork friction.  The amount of friction would be a function of how user friendly the roll-out and implementation was; it was not a particularly user friendly process as the reporting system was online only with limited professional office hour availability that made reporting extremely difficult and unlikely for people who did not have reliable internet or worked jobs that did not neatly map to a 9-5 assumption.

Overall, more than 92% of the respondents in all four groups — and nearly 97% of the respondents 30 to 49 years of age in Arkansas — were already meeting the community engagement requirement or should have been exempt before the policy took effect.

Work requirements are targeted at an incredibly small cohort of people who might be able to work but don’t.  This is very wide spread pain to sort out the “deserving” vs “undeserving” working poor.

Employment declined from 42.4% to 38.9% among Arkansans 30 to 49 years of age, a change of −3.5 percentage points. The three comparison groups had similar decreases, ranging from −2.9 to −5.7 percentage points.

And work requirements did absolutely nothing for employment.

None of this is particularly surprising.  It is good that we have very firm evidence of the obvious as this type of evidence raises the bar in future litigation against arbitrary and capricious waiver approvals.  The current federal district court judge who is overseeing lawsuits against work requirements has held that work is not a fundamental purpose of Medicaid.  If the study had shown absolutely minimal to no net coverage loss as people shifted to exchange or employer sponsored insurance and significant income gains, then the administration’s argument that this was an evidence based experiment with plausible real gains could hold some water.  Instead, this study shows that work requirements are fundamentally paperwork requirements that culls enrollment without producing employment effects.




** DOI: 10.1056/NEJMsr1901772


Oklahoma Medicaid expansion is on the ballot

Oklahoma activists are going the same route as Utah, Idaho and Nebraska activists successfully used in the 2018 election cycle: They are trying to get enough signatures to get Medicaid expansion on the ballot.



If you live in Oklahoma, this question needs slightly more than 177,000 valid signatures to appear on the 2020 ballot.

Odds are that even if it passes, there will be follow-on shenanigans as we have seen in Utah, Nebraska and Idaho to either delay or water down the expansions. In my opinion, a bad expansion is better than a perfect non-expansion. I assess the counterfactual as no expansion instead of a full expansion so people with different reasonable counterfactuals will vehemently disagree with me.

The ballot box is not the only way that Medicaid expansion of some sort may come to Oklahoma. There is a bananpants county level expansion proposal floating out there.

Here the scheme would be two or more bordering counties could expand Medicaid. The state share of the expansion (10% of costs) would be funded by local taxes. This would be wonderful for health and public finance economists and a complete cluster for everyone else.

Revisiting Graham-Cassidy

President Trump has stated that he wants a new healthcare proposal. His budget called from major Medicaid and exchange cuts as well as the Graham-Cassidy framework of state blockgrants to cover far fewer people in the exchange and Medicaid expansion populations.

In the short run this is irrelevant. There probably are fifty one votes in the Senate for something like Graham-Cassidy to pass. The Majority Leader would be willing to schedule that vote.

There are not 218 votes in the House to pass Graham-Cassidy. Nor is there a Speaker willing to schedule a vote on Graham-Cassidy if it was likely to pass.

However, it is worthwhile to look at the logic of the plan. It is a major cut to federal spending and a major redistribution of federal spending. Right now, more federal money goes to states that aggressively implement the Affordable Care Act or have very high cost markets. That means states like California which aggressively outreach for every single possible enrollment and expanded Medicaid will get more federal ACA money than states like Mississippi or South Dakota which have not expanded Medicaid and have not aggressively pushed enrollment on the Exchanges.

Graham-Cassidy wants to give block grants to states that over time converge to a narrow band on a per-capita basis. It reduces the overall pool of money available and then shifts the remaining funds to states that have done opposed the ACA’s implementation. There were variants where money would be freed up to throw at Senators from states that had implemented the ACA and Medicaid Expansion aggressively but whose votes might be needed to pass the bill.

During the summer of 2017, I tracked the outside evaluation of federal fund flows to states in 2026 under the counter-factual of Graham-Cassidy being implemented and current law of the ACA with CSR funding as the baseline. The coastal states got hammered while the Great Plains, Mountain West and the Deep Confederacy did well.

Circumstances have changed. The three major changes are more states have expanded Medicaid since September 2017, the termination of CSR payments increased effective net subsidies for more people and the elimination of the individual mandate probably depressed enrollment. The 2017 scores will need to be updated, but I think a 2019 score of Graham-Cassidy would be similar.

Idaho, Partial Medicaid Expansion and the 400% FPLers

Medicaid is primarily health insurance for poor people or very sick people.

Idaho’s legislature is monkeying around with the voter approved straight-up Medicaid expansion to 138% of the Federal Poverty Level (FPL).


This will harm middle class Idaho families who need community rated, guaranteed issue insurance from the individual market.

How does that work if Medicaid is health insurance for poor people?

Cost Sharing Reduction (CSR) work-arounds of Silverloading and differential morbidity matter.

Adrianna MacIntyre and I argued in a Health Affairs blog that full expansion has two paths to decreasing premiums for people earning over 400% FPL that are not available if a state elects and receives a waiver for a partial expansion to only 100% FPL.

 evidence found that Medicaid expansion improved the risk pool of state individual markets, suggesting that the population between 100 and 138 percent FPL is sicker and more expensive, on average, than other exchange enrollees. Insuring this cohort through Medicaid is associated with a seven to eleven percentage point decrease in individual market premiums. …

household incomes between 100 percent and 150 percent FPL, those that would be eligible for 94 percent AV silver plans.  This income bracket overlaps the Medicaid expansion income group significantly.  States that fully expand Medicaid end up with far fewer people in the most generous CSR bucket, as they have moved the 100-138 percent population to Medicaid

CSR 94 Enrollment by all APTC receiving enrollees 2018

Keeping a cohort that is more expensive than the rest of the ACA individual market risk pool in the risk pool raises premiums. Pulling the 100-138% population out of the ACA risk pool lowers market premiums as long as this group is more expensive than average. Furthermore while Idaho has engaged in the Silver Switcheroo, Silverloading increases premiums for folks who want a Silver plan and buy it on Exchange either because they don’t know if they will be just over or just under the subsidy cut-off point of 400% FPL or they can’t access an off-Exchange plan that meets their requirements.

Full Medicaid expansion reduces the premium pain of the middle class. Partial expansion continues the pricing pain for the middle class.

The Forever Bullshit War on Abortion

The article John posted yesterday from Jezebel about the New York Reproductive Health Act (RHA) is worth a read. It addresses the cloud of bullshit around abortion after 24 weeks of pregnancy which, as anyone who cares about women’s health instead of policing uteri knows, is a rare procedure only used in cases of severe, unsurvivable fetal abnormality or risk to the mother’s life:

The RHA now ensures that people in New York will have their constitutional right to an abortion; that includes the right to abortion after the 24th week in pregnancy if the pregnant person’s life or health is threatened by the pregnancy, or if the fetus has a condition incompatible with life. Prior to this legislation, I have had patients find out about a fetal anomaly who then had to travel long distances to other states to get the care they needed. Abortion later in pregnancy is not what patients anticipate for themselves; it’s not how they see their pregnancies unfolding. I had one patient who couldn’t afford to travel outside of the state and so she continued the pregnancy and the baby died shortly after birth due to a brain malformation. Years later, she is unable to tell her story without tears.

Before Ralph Northam’s blackface revelations, he was part of a faux controversy over a similar law in Virginia.  Northam is a pediatric neurosurgeon, and here’s the answer he gave to a radio station on the issue of abortion after 24 weeks:

“[Third trimester abortions are] done in cases where there may be severe deformities. There may be a fetus that’s nonviable. So in this particular example, if a mother is in labor, I can tell you exactly what would happen,” […] “The infant would be delivered. The infant would be kept comfortable. The infant would be resuscitated if that’s what the mother and the family desired. And then a discussion would ensue between the physicians and the mother.”

I’m sure it could have been phrased better, but he basically spoke the truth.  Babies who are severely deformed may live for a few minutes, hours or days, but they’re not going to live long. It’s a sad fact of life, but just speaking it summons a screeching hoard of anti-abortion dementors who pick apart the wording of whoever spoke the awful truth.

These zealots are winning – soon the Supremes will effectively do away with affordable abortions, if not all abortions, in red states.  They’ve have been enabled by people like Trump, who know that their children or mistresses will be able to get a safe, legal abortion because they can pay for it.  It’s only the poor who suffer, as always.