The Value of Medicaid: A Conversation with Nathan Hendren

Nathan Hendren, one of the authors of the two “The Value of Medicaid” papers that I explored last week, kindly reached out on Friday. We had an illuminating conversation, which he has nobly granted me permission to share in its entirety.

Much more informative than the Chat exegesis, IMHO. Assuming I’m now correctly understanding Hendren, the key issue is that “The Value of Medicaid” finds that $1000 of Medicaid spending crowds out $600 of other spending. So $1000 of Medicaid generates only $400 of marginal health spending, which recipients value at $200-$400. Hence $1000 of Medicaid spending only delivers $200-$400 of value, leading to the controversially low estimate in the abstract of the original paper.

P.S. True to form, Nathan has nobly agreed to respond to your comments.

Nathan Hendren, MIT

---

Hendren

Hi Bryan (cc Tyler),

I hope you don’t mind me emailing, but I saw Bryan’s post about my Medicaid paper with Amy and Erzo and felt I needed to help clarify this one. The 0.2-0.4 is the value of medicaid to the recipients per dollar of budgetary cost of Medicaid. As we noted in the original version of the paper and alluded to in the original abstract, this 0.2-0.4 is not a ratio of benefits per dollar of health care. Medicaid also provides about 60 cents of a transfer to hospitals and other providers of uncompensated care. The final version reported the benefits to recipients per dollar of health care (the 0.5-1.2). So, the 0.5-1.2 statistic is the correct statistic if you’d like to ask whether the recipients valued the cost of their health care (as opposed to the cost to the government of their health insurance).

Both numbers are interesting and I still think the 0.2-0.4 statistic is important for people to keep in mind when debating Medicaid…as you note, it suggests Medicaid beneficiaries would rather have the cash cost of Medicaid to the government (and providers of uncompensated care would likely be unhappy about that shift). But, the difference in numbers reported in the abstract was not due to any methodological refinements or interpretive shift (items 2 and 3 in your post). The shift in numbers in the abstract was done to prevent the type of misleading/incorrect translation that was made in your sentence after referencing the 0.2-0.4. The 0.2-0.4 does not mean individuals do not value the resource cost of the health care they obtain through Medicaid. The correct statistic for that particular statement is the 0.5-1.2. The 0.2-0.4 is per dollar of government spending, and that spending also generates that 0.60 to the providers of uncompensated care. All of these results were available in both the original and final versions of the paper.

I hope that helps clarify. I should say that I am very sympathetic to the spirit of your post and the importance of addressing motivated reasoning and social desirability bias. One thing this experience showed me is that it is important to standardize how we construct benefit-cost ratios and these types of comparisons in the literature to limit researcher influence of normative comparisons. To that aim, you might be interested in the efforts we have been making at Policy Impacts to promote a standardization in the way in which we do this type of policy analysis for policies like the Oregon experiment and others that have been rigorously evaluated in the literature.

Happy to chat more if helpful.

Best,

Nathan

---

Caplan

It’s a great pleasure to hear from you, Nathan. Questions:

1. So would you say that the Chat summary of how the results changed is false, misleading, or what?

2. If there’s 60 cents of transfer, why didn’t the range go from .2-.4 to .8-1.0?

3. The story is that Medicaid gives providers 60 cents per dollar over their opportunity cost?

---

Hendren

Thanks for your kind and generous response.

1. I would say it’s misleading because you’re suggesting that people value $1 of health care resource spending at 0.2-0.4. We never said that, but we recognized after writing the paper that that was how it was being interpreted. In a sense, we changed the abstract because we found people were getting confused and making statements suggesting $1 of medical care was valued at 0.2-0.4 by beneficiaries.

2. This is because we chose to focus on the fraction of the resource cost they value — i.e. what is people’s value of the cost of the health care they receive. The $1 of spending is only 0.3-0.4 dollars of increases in resource cost (i.e. increases in health care resources used by people on Medicaid). Taking the ratio of 0.2/0.4 gets you the 0.5. Taking the ratio of 0.4 to 0.3 gets you closer to the 1.2. The exact numbers are accounting for the uncertainty a little bit more accurately, but you can see how this generates a range from 0.5-1.2.

3. The story is that $1 of spending on Medicaid is actually giving 60 cents to providers of uncompensated care holding fixed their behavior (i.e. they would have been providing this care in an uncompensated form had they not gotten paid by Medicaid). This bucket of uncompensated care is a complicated mix of hospitals, non-profits, bankruptcy charge-offs, and state Medicaid backup funds. But, the point remains that when we say we’re going to spend $1 more on Medicaid, over half of the benefits go to these providers of uncompensated care (who now get paid for this care), not the beneficiaries themselves.

---

Caplan

Thanks, Nathan. Just so I’m clear, suppose Medicaid recipients were offered $1000 of treatment. How much would they pay out of their own pockets?

---

Hendren

We estimate that Medicaid recipients value $1000 of health care between $500 and $1,200. We estimate that Medicaid recipients value a $1000 expansion of Medicaid funding at $200 to $400 (while providers of uncompensated care would value it around $600). Does that make sense?

Sure, the entire exchange would be fine to post from my perspective ~

Best,

Nathan

---

Caplan

Sorry, I’m even more confused now. Is the distinction that matters “$1000 of health care versus $1000 of Medicaid funding”? Or “$1000 of health care versus a $1000 expansion of health care”?

---

Hendren

Both statements are about an expansion of Medicaid (akin to those who got coverage in Oregon through the experiment). The distinction that matters is health care versus Medicaid funding. Using the loose numbers, $1000 of Medicaid funding generates about $400 in increased health care usage (and $600 in transfers to those who would have provided that care anyway). Of that $400, we estimate that recipients value that increased health care at about $200-400.

Let me know if that makes sense?

---

Caplan

Last question: If the health care providers would have delivered the health care anyway, in what sense is quantity higher?

---

Hendren

Of the $1000, $600 is spending on health care they get anyway; $400 is the increase in quantity.

---

Caplan

And the idea is that the recipients would have somehow gotten the $600 even if Medicaid didn’t exist?

---

Hendren

Yes, exactly — that’s the idea/result that Medicaid crowds out other forms of uncompensated care.