Thursday, September 18, 2008

Correcting Some Misperceptions About McCain's Health Plan

Update, Saturday night September 20: Don Pedro points me to a Brad DeLong post quoting Robert Carroll of the Tax Foundation. According to Carroll, the McCain plan would keep "the payroll tax exclusion for employer-based health insurance". This is news to me, though Don Pedro tells me it's in one of the Tax Policy Center's analyses of McCain's proposals. This means that the McCain plan would be less likely to crash EPHI via adverse selection (how much less is an empirical question the answer to which I do not know). It also means that the McCain plan would be less of a hit to people who continue to receive EPHI than is described below. Over time, the fact that the McCain plan's credit would be indexed to the CPI rather than to health-specific inflation still means that the plan's would tend to be less generous than current policy, though it would take longer for that effect to occur than my discussion below suggests. There are other specific ways in which the analysis below would need to be modified in light of this difference (e.g., which taxpayers would benefit/be harmed least/most), though the qualitative discussion below remains accurate. One last note in light of this new (to me) information is that, at least in the short run, it seems that the McCain plan's fiscal impact would have to be considerably more expensive than current policy. You can't give people more than they get now and save money, unless there are sizable reductions in health expenditures. And any such reductions would likely take some real time to occur--by which point the real value of the McCain credit would have eroded. In any case, my original post follows.

In this post I want to discuss some misperceptions that a number of Democrats and Obama supporters have exhibited concerning McCain's health plan. First, I want to emphasize that it is true that McCain's plan would tax health benefits; there's nothing wrong with pointing that out. Second, though, a key comparison I've seen a number of people make is the wrong one. Much of the criticism I've seen of McCain's plan is based on the observation that the average family HI plan costs something like $12,000, which is more than the $5,000 credit McCain proposes. Brad DeLong makes this argument intemperately here, for example; Matt Yglesias picks it up with some bonus philosophical musings. The discussion below shows that both of these otherwise very smart guys are making a very dumb argument (and doing it in nasty tones, I should add).

This post is very long, so I will summarize my conclusions here. The moral of the story is that if it were implemented, McCain's plan could lead to major changes in the way health insurance is provided in the US. For example, it could lead to huge reductions in the share of people who get coverage through employers, greatly changing the degree of risk pooling and adverse selection, and possibly leading to substantial increases in premium costs for people working for small and mid-sized employers. Without knowing the answers to these questions, it's very hard to know how to evaluate the plan's net costs to any given set of people.

In addition, it seems clear that the plan would have very heterogeneous effects on the out-of-pocket costs faced by different people. Those with middle incomes living in states with high income taxes would be hit hardest, while those living in no-income tax states and having very low or relatively high incomes would benefit in the short run. However, unless health cost inflation is somehow greatly reduced, McCain's plan would be a net loser for virtually everyone in just a few years.

Personally, I see no reason to place the sort of general-equilibrium gamble that McCain proposes. I think the plan's indexing to overall inflation is another example of the McCain campaign's attempt to trick people into thinking they're getting a sports car when they're really getting a jalopy. But that's no reason not to get our facts right.

Anyway, let me turn to the economics of the issue. There are two key economic reasons that health insurance is provided by employers in today's US economy:

  1. In the absence of some other pooling mechanism, risk pooling gives employers an obvious advantage. Thus, employees benefit when health insurance is purchased by employers instead of individually.
  2. The tax deduction provides a very strong incentive to buy health insurance (HI) through employers, with pre-tax dollars.
The features of the McCain plan that I'll discuss here are that it will
  1. Eliminate the tax deduction for employer-provided health insurance (EPHI)
  2. Add a tax credit in the amount of $2,500 for singles and $5,000 for married couples. As I understand it, this credit would be refundable, in the sense that if a person paid less than the credit amount for a policy, the balance could be deposited in a health savings account (HSA).
For the moment, I'm going to ignore reason 1, risk pooling advantages, for employer-provided HI. That's not because risk pooling isn't extremely important, but rather because the confusion about McCain's plan stems from misunderstanding the current tax treatment of EPHI.

Under the current system, employees receive some compensation through cash and some through fringe benefits. For simplicity, assume the only fringe benefit is HI.

Let's call the (pre-tax) value of EPHI, B (for benefits); for an employer paying 75% of the average HI premium of $12,000, the value of B would be B=$9,000, with the rest of the premium being paid for out of cash compensation by the employee. Let's call the pre-tax amount of cash compensation C; as a concrete example, I'll use the value C=$50,000. For the moment, I will ignore the fact that we have progressive taxation and assume that there is a single tax rate, T. I'll also assume that employees place a value of exactly one dollar on each dollar spent on EPHI. All of the above assumptions are empirically wrong, but making them will simplify the discussion, and I'll return to them below.

An employer's cost of paying an employee C in cash and B in benefits is C+B, or $59,000 in our example. This means that an employer is equally happy paying $59k in cash or $50k cash plus $9k in HI. If there were no EPHI subsidy and no risk-pooling gains from purchasing HI through employers, then our employer in this example could just as well pay $59k cash and continue to attract the same workers.

Now, the pre-tax value to workers of the 50cash+9fringe package is higher than $59k. To see why, note that under the current system, the value of *after*-tax compensation is in our concrete example is

(1)        A = (1-T)*$47,000 + $12,000

The taxable compensation here is $47,000 because workers must pay tax on only the part of cash compensation that isn't spent on EPHI, i.e., $50,000-$3,000=$47,000. The $12,000 part in equation (1) arises because that's the value of the EPHI plan ($9,000 comes directly from the employer, and $3,000 is paid by the employee out of pre-tax compensation).

Now consider a company that paid only cash, i.e., made no employer contributions to EPHI, so that B=0 for this employer. Imagine that employees have access to the very same plan that they did above, but they have to pay for it out of pocket, with no tax subsidy. If this company paid the worker $50,000, then the value of the worker's after-tax compensation would be only (1-T)*$50,000, which is less than A in equation (1) as long as the tax rate is greater than 0 (which it is).

The proper inference to draw from this example is that a company that didn't want to pay for HI would have to pay a higher salary to its workers. If it didn't, it would not be able to attract workers of the same quality. The next question is, What is the amount of fully taxable cash compensation, C', that would leave a worker exactly as well off as if she received A in equation (1)? To answer this, we simply set

(1-T)*C' = (1-T)*$47,000 + $12,000

and solve for C'. The answer is that

(2) C' = $47,000 + $12,000/(1-T).

Suppose that the tax rate is 0.4 (more on this below). Then 1-T=0.6, and $12,000/(1-T) = $20,000. And this implies that C'=$67,000. In other words, the compensation package that costs the employer $59k has a pre-tax value of $67k to workers, for a difference of $8k in pre-tax value. Since the tax rate is 0.4 in this case, the after-tax value of the additional $8k is (1-0.4)*$8k, or $4,800. What this means is that if the worker is paid $59k in cash, all taxable, and is then given $4,800 in a (nontaxable) tax credit by the government, the worker will be just as well off as with the cash-and-fringe compensation package.

Another, much simpler, way to see this is to note that the value of the current tax deduction is simply the product of the tax rate, T, with the amount of compensation that is not taxed. Since this amount is $12k in our example, at a tax rate of 0.4, the value of the subsidy is 0.4*$12k=$4,800. So, in this case, the employee's net cost is $7,200, not $12,000. So if the price of the HI policy were the same under either policy (a huge if to which I return below), the net cost to the employee would be $7,000 with McCain's credit of $5,000 and $7,200 under the current system. In other words, the proper comparison in this example is $4,800 with $5,000, not $12,000 with $5,000. More generally, the comparison should be T*$12,000 with $5,000, where T is the worker's marginal tax rate.

This discussion shows that the 12-versus-5 critique of McCain's plan compares the wrong numbers. It assumes that the EPHI subsidy has the value of the full $12k in policy costs, but the value of the tax subsidy is much less: the difference between the out-of-pocket cost of the policy with and without the subsidy.

As noted above, though, McCain's plan would index the value of his credit to the overall inflation rate, not to health inflation. According to this organization, "Since 2000, employment-based health insurance premiums have increased 100 percent, compared to cumulative inflation of 24 percent". If the same thing happened over the next 7 years, the cost of a $12,000 HI plan would become $24,000, while the value of the McCain tax credit would rise to only $6,400. With a value of T=0.4, this means that the current subsidy value is $9,600 (0.4 times $24k), which is $3,200 more than McCain's subsidy in new prices (and roughly $2,600 in today's). That's a huge hit.

There are several further issues to discuss:
  1. By all accounts, the value of the McCain credit would be indexed to the consumer price index (CPI), not a health-costs price index. This is very important because health costs have typically risen much more quickly than the CPI. For example, one estimate I've seen is that health inflation for 2007 was 6.9%, or roughly twice the CPI. No doubt, McCain's advisers would claim that by indexing only to the CPI, they are going to induce people to be more price-sensitive. There's obvious logic to this argument, but it's hard to explain the fact that health inflation regularly has exceeded overall inflation with this sort of logic. The current system has been in place for many years, and if anything, in terms of services (not expenditures on them), it's become less generous, not more, in recent years. So it seems very likely that health costs will continue to rise more quickly than the CPI, in which case the real value of the McCain credit would shrink. See point 5 below for more on this issue.
  2. Under the McCain plan, what would happen to wages? The current EPHI subsidy is a subsidy for a particular good, health insurance expenditures, not a wage subsidy. The only way to get the subsidy is to purchase HI through employers, which helps explain the prevalence of EPHI: even ignoring risk pooling, why buy something in an unsubsidized way when you can buy it in a subsidized way? Since the current system is not a wage subsidy, eliminating it should not affect employer costs (except through whatever general equilibrium effects arise from changes in the use of health care, costs of policies, etc). If employers did stop offering HI plans, in the new labor market equilibrium they would simply pay workers the total employer cost of all compensation previously paid, i.e., $59k in the above example. If employers continued to offer HI plans, say due to risk pooling advantages, one would expect them to pay a total of $59k in any combination of cash and fringe, all of which would then be taxed.
  3. Under the McCain plan, would employers continue to offer HI policies? This is a difficult question to answer in the absence of detailed evidence. If there were no adverse selection problems, then one would expect employers to continue offering HI plans if employers have a cost advantage in purchasing them relative to individual employees. However, the main argument for an equilibrium cost advantage is the *existence* of adverse selection problems. Therefore, it seems reasonable to think that some employers, especially small but possibly mid-sized and larger ones, would stop offering HI policies. The adverse selection argument here is standard. If employers can get a better deal, it's because there's reason to think that low-risk people don't opt out of EPHI. The current reason for this is easy: the tax subsidy. Even if you're receiving less than the full actuarial value of the cost of an EPHI plan, the tax subsidy encourages you to retain coverage. Once the subsidy is gone, one can expect lower-risk, healthier people to exit employers' group plans and look for cheaper plans on the individual market (if these exist; more below). Thus while I'm not certain that McCain's plan would destroy the EPHI system, it strikes as completely plausible that it could, at least for some employer-size classes.
  4. Under the McCain plan, would overall health costs fall? By making its tax credit refundable, the McCain plan is essentially the same as saying "here's money, spend it how you like", provided that people buy some sort of HI plan, however minimal its coverage. That means that people will pay $1 for $1 worth of HI coverage, rather than the current (1-T)*$1. The higher a person's current marginal tax rate, the greater the incentive to reduce her HI plan costs. This means that the McCain plan will encourage those with relatively high incomes to buy so-called catastrophic plans, those with high deductibles and copays. To the extent that people then choose to decline treatment or search out less expensive treatments than those prescribed by their doctors, total health care utilization would fall, lowering costs. One would also expect more risk-loving people with lower incomes to do the same thing. People who don't fully appreciate the health risks they face could find themselves under-insured, leading to the Samaritan's dilemma: people show up at doctors' offices and emergency rooms without the ability to pay and are treated anyway. This kind of situation is the classic privatized gains-socialized losses picture and would tend to increase costs if people delay care until it is more costly.
  5. Does McCain's plan pay for itself? The answer to this question depends on two key things. First, if McCain's plan does not affect health costs but reduces the total out-of-pocket costs to consumers of buying HI, then it will necessarily add to the deficit: you can't give people more than they're getting now without reducing costs and save money in the process. Second, if the plan reduces health costs, then in principle it could both reduce out-of-pocket costs and reduce the deficit. The opposite is also true of course: if the plan screws up the HI market via increased adverse selection, we could wind up with increased total costs of a given quality of care, both increasing out-of-pocket costs and increasing the deficit.
  6. Would the McCain plan increase or decrease people's net costs of health insurance? Essentially, this question asks whether Obama partisans' criticism that the McCain plan taxes health insurance (which it certainly does) is appropriately balanced by a defense that McCain's plan reduces net costs. Leaving aside the issues of adverse selection and health costs discussed in points 2 and 3 above, there are two key issues here:
A. What is a person's marginal tax rate?

As noted above, the value of the current subsidy is the product of a person's marginal tax rate and the total (employer-paid plus employee-paid) cost of the HI plan. I'll continue to focus on a family plan, assuming that the total cost is $12,000. The relevant marginal tax rate is the sum of the employee's federal income-tax marginal rate, her state rate if applicable, and the payroll tax.

The possible federal income-tax marginal rates are 0, 0.10, 0.15, 0.25, 0.28, 0.33, and 0.35. State marginal income tax rates range from 0, since many states have no income tax, to roughly 0.10 (e.g., places like California and DC). Let's split the difference and consider the case of 0.05. The payroll tax is a total of 0.153. Only half of this is paid out of employee funds, with the other half paid by employers. It's a truism of introductory microeconomics that this distinction is irrelevant; what matters is the total impact of the payroll tax, which is 0.153. (There's a much deeper question that has to do with whether one should regard payroll taxes as being taxes or simply forced investment in/purchase of Social Security and Medicare coverage; it's true that some people's SS benefits would rise due to taxation of their now-cash compensation for HI, but I will ignore that for this discussion.) Anyone who earns more than the cap has a payroll tax of 0.029 (the Medicare part only); everyone with a federal marginal rate above 0.25 will be over the cap, and virtually everyone with a 0.10 rate will be below the cap.

Putting all this together means that our lowest total marginal rate is 0.153 (federal rate of 0.10, no state income tax, and payroll tax of 0.153), and our highest is 0.503 (federal=0.25, state=0.10, payroll=0.153). I note that because the people with the highest federal rate don't pay the full payroll tax, their marginal rate for HI subsidy purposes will be between 0.379 and 0.479 (the lower being with no state income tax, the higher being with a state mtr of 0.10).

Someone with a marginal rate of 0.153 gets a subsidy value of roughly $1,800 from a $12,000 plan, while someone with a marginal rate of 0.503 gets a subsidy value of roughly $6,000 from this plan. Thus even if we assume all plans are priced the same after McCain's plan takes effect, there is a LOT of variation in the net effect of the McCain plan, both in dollar terms and in whether his tax credit makes up for the loss of the current tax subsidy. One thing worth noting, though, is that (a) few people in the 0-rate federal bracket likely receive EPHI benefits, which makes McCain's plan even better for them, but (b) these folks' ability to pay more than the amount of the McCain credit for a plan is likely very limited due to their low incomes, so they are likely to get a very low-quality plan (still better than nothing, of course).

The typical person paying more under McCain's plan would be middle-income (federal marginal rate of 0.25), with earnings below the payroll tax cap (i.e., below roughly $104k), and living in a state with a sizable income tax. The typical person paying less would be living in a state with a low income tax rate, or with either very low or high income (in the first case, the federal marginal rate is low, and in the second case the person is above the SS cap).

Make sure to read point C below!

I note also that the $12k figure is an average, which means there's lots of room for people to do better or worse than the above analysis discusses.

Note: I don't think the point about geographical variation has been much noted. A related question is whether states would follow the federal lead in eliminating the EPHI tax subsidy, as I've assumed. If they did, they'd have increased tax revenue. If not, more employees would benefit from the McCain plan.

B. What happens to the out-of-pocket costs of a health insurance plan?

Quite apart from the questions discussed in points 3-4 above, one would expect the price of HI plans to change under the McCain plan. This is because the current system subsidizes each dollar spent on HI, which shifts the demand for HI (actually, it rotates the HI demand curve out and to the right from its horizontal-axis intercept). In a competitive market, this could be expected to increase the equilibrium price of HI plans. No one seriously thinks this market is competitive, but no doubt we see at least some of this effect relative to a no-subsidy world. Because the McCain credit is similar to cash, it would create something a lot like a no-subsidy world, and therefore we would expect the supplier's price to fall relative to the current supplier's price (with today's subsidy in place, the supplier's price exceeds the consumer's price by the amount of the effective subsidy, i.e., T times the cost of the plan). How much this price would fall depends critically on whom the marginal consumer is, and I have no special insight on this front.

Equally important is the adverse selection issue. If McCain's plan induces many employers to drop coverage and/or causes many employees to seek coverage in the nongroup market, then that would put upward pressure on the consumer price paid for HI plans.

C. What happens to health and overall inflation?
Let's follow the post-2000 numbers cited above and assume that health cost inflation rises 100% over the next 7 years, and overall inflation by only 24%. This means that the value of the McCain credit would be $6,400, while the cost of today's $12,000 HI plan would be $24,000 (other things equal).
Because $6,400 is less than 0.27 of $24,000, if health inflation continues apace, then other things equal, everyone in the US making enough money to have a marginal rate of 0.27 or higher--which is to say, just about anyone paying any federal income taxes at all--will be at best no better off under McCain's plan, and most people would be worse off. 


Jennifer Imazeki said...

Thank you for an excellent, excellent post! I will be talking about both candidate's health care plans in my micro principles class and you just made my job so much easier!

lerxst said...


I thought about blogging on this too. I did a total double take when I read Brad's post --couldn't believe he made that mistake.

Great post.

Don Pedro said...

Great post, which clears up a few things for me. A few things:

1) One point you don't mention that Goolsbee and Furman have harped on is that the switch from tax-exempt health benefits would cause some people to jump brackets. I describe this issue in this post:

I calculated that the bracket jump would mean an extra tax increase of up to $900 for some people, although I realize now I forgot to take into account the fact that the employee-paid portion of EPHI is pre-tax. So in fact, the tax increase would be up to $1200 (correct me if I have this wrong.)

According to G&F there's a CBO study about this, but I haven't found it.

2) I think the key point in terms of the costs in the scenarios you lay out is the tremendous uncertainy. This was emphasized by Lerxst and by Henry Aaron in the post he links to here:

3) I think you're too generous in evaluating whether the McCain plan will kill EPHI. It's true there's tremendous uncertainty in whether employers will drop EPHI and how soon, but the point of the whole proposal is to kill off EPHI, and that is the hope and the expectation of the proposal's designers. Give the adverse selection incentives, it seems very likely that in the long-term at least (don't ask me how long that is) the large majority of employers will drop EPHI. One complication: it's anyone's guess what would happen with public employees. Would McCain kill EPHI for federal employess, for example?

4) Ultimately, the big problem with the McCain proposal is that it would leave many on their own in the private market and unable to buy insurance. Dealing with these people was an afterthought in the proposal--I don't think there was even any mention at all in their original version--and now the budget proposed to subsidize insurance for these people is ludicrously inadequate. I'll write a post about this issue in the next couple days.

Don Pedro said...

By the way, you should make sure Brad and Mark get this.

Jonah B. Gelbach said...

thx all for the comments. don pedro, sorry i didn't provide those links -- was short on time, but i recommend people read the posts there.

jennifer: really glad you think this will be useful in class! wait til you see my principles textbook....:-)