Separating Hyperplanes

Here's a graph of capital gains tax revenue, along side the federal deficit for reference

What stands out to me is that the deficit is a lot more volatile (and almost always larger) than cap gains revenue. The relative stability of capital gains revenue is not exactly what I'd expect considering the erosion over time of the tax base due to the growing number of tax-privileged vehicles like 401k, IRA, HSA, 529 and others. In the most recent year of data, 2014, cap gains revenue amounted 25.4 percent of the deficit. Essentially, selling bonds is four times as big a source of funds for the federal government as taxing capital gains. Here's what that ratio looks like over time:

Data came from FRED and treasury.

1/26/2018 01:14:00 PM

There's been a debate on twitter lately about the proper role of insurance in healthcare. Alex Tabarrok summarizes the proposition being debated:

The basic idea Tabarrok is getting at is that insurance is a security that reduces the variance in your expenses, while a scheme that reduces the expected value of your expenses is more properly understood as redistribution (or, in non-econ speak, welfare). Those are fine definitions, I suppose, even though they do not conform to everyday usage of the terms. I think we can understand community-rated insurance like employer-sponsored plans or ACA exchange plans as containing a mix of both health-based redistribution and pure economic insurance. Lots of people in these plans would not be able to afford them without the "redistribution" if that's what you want to call it.

Still, philosophically, this is a weird distinction to make. Analytically, insurance and redistribution are the exact same thing. You can take the exact same security and it will be either insurance or redistribution depending on the state of mind of the policy holder and insurer. In a risk-rated insurance market, if I have leukemia and don't know it yet, then I can afford to buy insurance that covers leukemia treatment, so this is a true Tabarrokian insurance contract. But if I have leukemia and do know it, then insurers will bake the cost of leukemia treatment into my premium and I won't be able to afford it. The exact same contract is insurance if I'm sick and I don't know it, or welfare if I'm sick and I know it. What a strange distinction!

This is where John Rawls was coming from. The moral case for insurance coverage of leukemia treatment shouldn't depend on who knows you have leukemia. This is why we often refer to welfare as "social-insurance." All welfare is an insurance contract, but re-imagined under an alternative information set.

Information sets do matter, economically, of course. Adverse selection happens when insurance customers have more information about their future health costs than insurers do. Community rating forces insurers to price their products as if they know nothing about individual-specific health risks, but because customers do have information about their own health risks, we're at risk of adverse selection. Whether we are talking about insurance or redistribution, information sets constrain the set of possible outcomes. If you want both a healthy person who knows he is healthy and a sick person who knows he is sick to be able to have insurance, you must ensure that the design of this scheme includes enough incentive to persuade the healthy person to buy it. This is called an "incentive compatibility constraint." The ACA does this by taxing people without insurance, which is implicitly a subsidy for healthy people who buy insurance. Risk-rating schemes do this by charging more to sick people, which is implicitly a subsidy for healthy people who buy insurance. The process of designing a scheme to achieve the best outcome possible without violating the incentive compatibility constraints given the real-world information set is called "mechanism design." The scheme, whether it is a private insurance contract or welfare, is called a "mechanism."

I don't particularly care whether we call this mechanism "health insurance" or "redistribution." Tabarrok draws a distinction based on information sets, which do matter for the feasibility of the design of the scheme, but the moral case for each is identical.

9/22/2017 08:57:00 AM
9/07/2017 02:05:00 PM

Auto-enrollment is a wonky health reform idea that never seems to quite die, despite being too complicated for our current politics to implement. The idea most recently graced the pages of the Wall Street Journal, which proposes that as part of the AHCA, uninsured individuals should be auto-enrolled in mini-med health plans whose premiums don't exceed the size of the AHCA tax credit. There are a number of flaws with this plan, such as the fact the AHCA won't be enough to cover the entire premium for most people, but one of the problems is, as David Anderson says, the insurance market has more churn than a butter factory. This makes it very hard for the government to know who to auto-enroll into insurance plans. Remember that there are special enrollment periods that could strike any time of year, so that decision needs to be made for every single individual in the US every single month of the year.

The WSJ plan is bad. But I think we actually already have all the tools in place to do a workable form of auto-enrollment. We already have refundable tax credits for insurance through the ACA. We already charge taxes on a monthly basis. We already have an annual filing process for retroactively correcting errors in monthly tax payments. We already have a requirement to report month-by-month health insurance status on your taxes. What more do we need? Here's my proposal:

The government would let private insurers bid on pools of uninsured people. The lowest bid wins the contract to cover uninsured individuals in the pool. Premiums (net of premium tax credits) are automatically added to your monthly tax bill unless one of the following apply:

  • you have ACA-compliant insurance in that month and did not use the auto-enroll plan
  • you opted out for that policy year during an open enrollment or special enrollment period

If neither of those apply, then you'd owe the premiums (minus the premium tax credit) as part of your monthly tax bill, and in exchange you can show up to hospitals and receive care with bronze-level coverage.

The nice thing about this is it sidesteps Anderson's concern about the difficulty of tracking who becomes uninsured each month, by making it all retrospective. Insurers don't have to actively enroll anyone—they enroll in the auto-enroll plan upon showing up at a healthcare provider without any other form of insurance. When they file taxes on April 15th, it will become clear which months the individual was on the auto-enroll plan and which months he wasn't. As with all taxes, it's the individual's responsibility to pay the right amount each month, or pay the late penalty if they under-payed in any months. So for example, if an individual had no insurance, didn't opt out of the auto-enrollment plan, but also didn't pay the auto-enroll premiums each month, then they'd owe those premiums (minus the amount of the premium tax credit) on April 15th, plus a percentage penalty.

There's obviously a risk premium arising from the fact that insurers who bid on the uninsured pools won't know exactly how many they might have to cover. That likely means most uninsured people could find a better deal by actively shopping on the ACA market. If we were really clever, we'd put the opt-out form on, at the bottom of the page showing all the insurance plans and prices available to them in their market. Make them scroll through all their other options before opting out entirely.

4/20/2017 09:54:00 AM