I just published a paper on CON laws and spending in Contemporary Economic Policy. As frequent readers of this blog will know, CON laws in 34 states require healthcare providers in 34 US states to get permission from a state board before opening or expanding, and one goal of the laws is to reduce health care spending. The contribution we aim for in this paper is to lay out a theoretical framework for how these laws affect spending.
There have been many empirical papers on this, typically finding that CON laws increase spending, but the only theory explaining why has been simple supply and demand. Health care markets are hard to model for a few reasons, but one big one is that most spending is done through insurers, so the price consumers pay is typically quite a bit lower than the price producers receive. This leads to “moral hazard”- i.e. overuse and overspending by consumers. Normally economists hate monopolies because they lead to underproduction, so in a market with overuse its fair to ask (as Hotelling did about nonrenewable resources)- could two market failures (moral hazard overuse and monopoly underuse) cancel each other out?
This is part of the idea behind CON laws restricting competition, but in this case it appears not to work. Moral hazard exists for a reason- people value the risk protection they get from insurance, and higher medical prices from imperfect competition raise this risk more than they reduce overuse, harming consumer wellbeing. That’s the result of Gaynor, Haas-Wilson and Vogt’s 2000 JPE paper “Are Invisible Hands Good Hands? Moral Hazard, Competition, and the Second-Best in Health Care Markets“, which thoroughly models how health insurance fits into the market for healthcare.
Its a great paper as far as it goes, but its focus is on how competition affects welfare in the market for healthcare; we take their models and solve them instead for total spending. We get the same answer as the simple supply and demand models give: it depends on elasticities, but the demand for health care tends to be inelastic, so CON should typically increase spending. We then extend the model to show how this effect can vary based on the health of the consumer and the costs of the producer. We test the predictions of our models using data, confirming that patient health matters in the way we expect, but finding that producer costs don’t (at least not the proxy we have data for):
Our own empirical work finds that CON is associated with 3% higher overall per capita health spending, but larger changes in the distribution of spending. We find that CON has no significant effect on the spending of those with excellent health while increasing spending by the less healthy as much as 12%.
I’ve been saying “we” did all this, so I should note my co-author on the paper was Tom Hamami, who we just hired at Providence College. Economists always teach the importance of specialization, but usually end up working with people very similar to ourselves; most of my publications have been with other empirically-focused microeconomists. This paper by contrast was all about specialization- I had spent a month trying to extend the Gaynor, Haas-Wilson and Vogt model myself and was getting nowhere, so I asked Tom for help. He’s a real theorist and knocked the whole thing out in a week like it was nothing. Meanwhile I did all the empirical work; I was using the restricted MEPS-HC so Tom wouldn’t have been allowed to see the data even if he wanted to, which he didn’t. So there were huge gains from trade here and I’m glad to get the publication. But our theory isn’t the final word here, and we conclude with some ideas for how future work could improve on it.
On the empirical side, the data could be better:
We find that neither state health wages nor its interaction with CON are statistically significant at traditional levels. Given that our proxy for marginal cost is far from perfect, we do not see the lack of statistical significance as a major threat to the relevance of our theoretical model, but we do think it indicates the necessity of future empirical work that incorporates supply‐side data to better proxy for marginal cost
On the theoretical side, modeling insurers as imperfectly competitive would match reality better:
Our theoretical model assumes that the insurance market is competitive, but in many areas the private insurance market is quite concentrated, and public insurance programs might behave in entirely different ways. The challenges of modeling non‐competitive or non‐profit‐maximizing insurers are substantial, but so too would be the payoff to a successful model.