Read on Twitter
New Paper Alert "The Liquidity Sensitivity of Healthcare Consumption: Evidence from Social Security Payments"
joint w/ @talgross and Daniel Prinz (Who is on the job market! Hire him!)
@nberpubs WP released today: https://www.nber.org/papers/w27977
A thread:
We study the effects of health insurance copayments on low-income households.
Typically, we think about copayments as counteracting moral hazard, or "over-consumption" of healthcare goods and services due to the subsidized prices under insurance.
Typically, we think about copayments as counteracting moral hazard, or "over-consumption" of healthcare goods and services due to the subsidized prices under insurance.
But what if consumers don't have enough cash to pay those copayments?
The household finance lit has shown that ppl delay their consumption until their income arrives. But does this apply to healthcare where people are often insured and the stakes of delaying can be high?
The household finance lit has shown that ppl delay their consumption until their income arrives. But does this apply to healthcare where people are often insured and the stakes of delaying can be high?
Until now, there has been no evidence that we're aware of.
In this paper, we leverage quasi-random assignment of social security check "paydays" to provide evidence that Medicare Part D beneficiaries do indeed delay filling prescriptions until they get paid.
In this paper, we leverage quasi-random assignment of social security check "paydays" to provide evidence that Medicare Part D beneficiaries do indeed delay filling prescriptions until they get paid.
Here's the money graph.
In it we plot prescription fills around payday for people about to receive their checks relative to fills for people who won't receive their checks for 2 weeks.
Red = ppl paying $2-5 copays per script
Blue = ppl whose copays are fully subsidized
In it we plot prescription fills around payday for people about to receive their checks relative to fills for people who won't receive their checks for 2 weeks.
Red = ppl paying $2-5 copays per script
Blue = ppl whose copays are fully subsidized
We find a spike in fills of around 11% on payday for ppl paying copays. Nothing for ppl not paying copays.
Spike is bigger after 5 week pay periods than after 4 week pay periods, bigger for more expensive drugs, bigger in lower-income zip codes (see this figure).
Spike is bigger after 5 week pay periods than after 4 week pay periods, bigger for more expensive drugs, bigger in lower-income zip codes (see this figure).
Spikes are not just driven by refills but also by new fills, where stockpiling is not possible.
We see spikes for drugs like blood thinners, insulins, anti-biotics, where docs believe short-term non-adherence can have serious health consequences.
We see spikes for drugs like blood thinners, insulins, anti-biotics, where docs believe short-term non-adherence can have serious health consequences.
We then show that the spikes go away when people move into a program that fully subsidizes their copayments, indicating that it really is the copayments that drive this behavior, not some kind of "bus fare" effect.
Finally, we use the transition to the program that fully subsidizes copayments to estimate price elasticities separately for "payday fillers" and "non-payday fillers".
The elasticities for the payday fillers are 2x the size of the average elasticity, suggesting that the payday spikes are not just driven by temporal substitution but also by these liquidity sensitive consumers forgoing filling scripts altogether.
So, what are the implications?
When it comes to low-income households, we need to think differently about health insurance.
As @ProfKMEricson and @JustinSydnor have pointed out, for these households insurance isn't just about risk protection vs. moral hazard.
When it comes to low-income households, we need to think differently about health insurance.
As @ProfKMEricson and @JustinSydnor have pointed out, for these households insurance isn't just about risk protection vs. moral hazard.
For these households, insurance also allows people to consume drugs when they want/need them rather than when they have cash.
Further, our results suggest that the conventional way of thinking about demand-response of healthcare consumption to insurance as inefficient "moral hazard" or "over-consumption" may be incomplete here.
In the paper, we develop a simple graphical framework and use our empirical estimates to illustrate this. Check it out!
The paper is out as an NBER Working Paper today. Comments welcome!
And thanks to Julia Yates for awesome research assistance on the paper. She's applying to grad school this year. Accept her. You won't regret it!
And thanks to Julia Yates for awesome research assistance on the paper. She's applying to grad school this year. Accept her. You won't regret it!
