Zarek Brot

Assistant professor at @UChicago @HarrisPolicy, fellow at @NBERpubs, @HMRLab @Harvard, @ChicagoFed. Health care policy "expert"

[Zarek Brot @zarekcb]

A big new paper with @zackcooperYale @stuartcraig @LevKlarnet, Ithai Lurie, and Corbin Miller: "Who Pays for Rising Health Care Prices? Evidence from Hospital Mergers"

https://t.co/iSZ8A7BRcD

With excellent recent WSJ coverage by @_melaevans: https://t.co/lolw3zmWfZ

[Zarek Brot @zarekcb]

@MattBruenig @ranavain @JoeGrant1900 For workers presented with multiple options, the more generous plan is usually the worse option for everyone (due to slow-moving adverse selection death spirals). https://t.co/5kPArchrO4

[Jessica Dobrinsky @jldobrinsky]

Hospital tax exemption was originally justified on the grounds of charity care: hospitals provided free or reduced-cost treatment to patients who could not pay. This was formalized in IRS Revenue Ruling 56-185, issued in 1956.

But in 1969, the IRS shifted course, replacing the charity care standard with a broader "community benefit" test. Hospitals could now fulfill their tax-exempt obligations by counting research, staff training, professional education, and general operations. The explicit requirement to provide care for patients unable to pay quietly disappeared.

Today, hospital associations defend non-profit status by arguing that tax exemption is what keeps unprofitable service lines—like labor and delivery, behavioral health, and trauma care—open. They claim for-profit hospitals would shutter these services.

On paper, this argument seems reasonable. But in West Virginia, the reality is different. Hospitals continue to claim tax exemption even as they close labor and delivery units and rural service lines. Charity care provided by non-profit hospitals in the state remains below 1% of net patient revenue.

Hospitals keep the exemption, but the services it was supposed to support are disappearing.

[Zarek Brot @zarekcb]

@ScottBarkowski That they receive a high baseline of care. Garthwaite et al show that they are implicitly insured in some important ways. But not necessarily that this is getting them quality care absent insurance.

[Zarek Brot @zarekcb]

Easy to pooh-pooh the earlier belief on this. But it was really not that crazy in a world where there was a strong belief that marginal care was of little clinical value, which is not the craziest view to have.

[Adam Gaffney @awgaffney]

There's been a slew of studies in last decade using quasi-experimental methods — and one experimental study from the IRS! — that have confirmed that gaining coverage affects mortality. But really worth pointing out how utterly bizarre it would be if that *weren't* true —

1/2

[Zarek Brot @zarekcb]

@ScottBarkowski I don’t think I would say that…the recipients also don’t believe it themselves

[Zarek Brot @zarekcb]

@awgaffney (In contrast to some of your interlocutors on this post, I think the new studies are very good and convincing. Though they raise an interesting set of questions.)

[Zarek Brot @zarekcb]

@awgaffney I agree, I am mostly objecting to your seeming view that a belief that insurance would have no effect on health was unjustified ex ante. I think if you looked at corroborating evidence in say 2012 this would have been pretty clearly the way it pointed IMO

[Zarek Brot @zarekcb]

@johnlin08 @ben_ippolito Useful to remember that long-lived repeated interactions are uncommon (see table 2 of https://t.co/H7VvPiCODw)

[Zarek Brot @zarekcb]

@johnlin08 @ben_ippolito This is because r is just our guess of how many failed claims we don't see relative to successful so scales in 1/1-r which is not equivalent to scaling by r.

Our calibration is "for every successful claim how many interactions were there" we assume 1 for each success and failure

[Zarek Brot @zarekcb]

@awgaffney I don't think it's productive to be self-righteous about how medicine obviously works when actual doctor-patient interactions are often unproductive. (I suspect mortality results come from some combo of drug adherence plus new willingness to go to the hospital for emergencies)

[Zarek Brot @zarekcb]

@awgaffney I don't know. If you listen to some of the vignettes of Medicaid recipients from Oregon they have mixed feelings about their own doctor visit experiences being unproductive. A standard checkup is not decreasing mortality for anyone, likely, right? (at least not same year)

[Zarek Brot @zarekcb]

Or doctors can be persistently wrong about clinical value, e.g. this story about cancer treatment

https://t.co/f0h1sdbnm9

[Sarah Miller @smilleralert]

In the 90s, a study showed that combining high dose chemo w/ bone marrow transplant improved survival for patients w stage 4 breast cancer. But insurance companies denied coverage for this new treatment--not enough evidence, they claimed.

[Zarek Brot @zarekcb]

But I think Adam is very silly to say this. Ineffective *on the margin* not the same as *on average*. e.g. on the margin we know lots of care is driven by financial motives, not clinical motives (e.g. Gruber-Owings on C sections and many others)

https://t.co/Fay2gTohTk

[Adam Gaffney @awgaffney]

There's zero debate that the uninsured get very little healthcare. How could it be that zero care had no impact on health? It would suggest that modern medical care is basically ineffective & we might as well go back to leeches, close the hospitals, cancel all doctors, etc.

2/2

[Zarek Brot @zarekcb]

@Jabaluck (Bad enough that I used to put a stylized version of this graph on the first problem set in intro micro and had the students tell me how many different interpretations of the underlying variation you can uncover)

[Zarek Brot @zarekcb]

@Jabaluck The bigger issue is that there is no meaningful question for which the answer is a cross-sectional country-level regression.

[Zarek Brot @zarekcb]

@johnlin08 @ben_ippolito I also don’t really understand the objection about admin costs. Our estimates are just scaled up versions of other time diary estimates. These are the same studies people use to argue admin costs are very high! You cannot throw our results out without throwing out the others.

[Zarek Brot @zarekcb]

@johnlin08 @ben_ippolito I don’t agree. See Sarig 2024 who uses a different instrument in the non LIS pop (https://t.co/TLFLQe0MFP). His reduced-form estimate most comparable to ours (table 1 first column) is surprisingly close.