Projecting Premiums Under the New Health Care Program (Warning! Long, and Pretty Wonky)

The Atlantic takes Columbus Day off, so I am late to the game on the dueling analyses of future health care premiums.  A PWC report commissioned by the health insurance industry says that premiums will skyrocket.  Meanwhile, Jonathan Gruber of MIT says no, they'll plummetEzra Klein says it's hard to say who's right.  I don't find it hard to say that both analyses have serious problems.

The liberal blogosphere has made much of the fact that the PriceWaterhouseCoopers report assumes that the excise tax is applied to plans, rather than estimating the change in behavior.  This is a completely defensible choice, because there is no good way to figure out how much behavior is going to change.  The CBO does the same thing when it refuses to score costs--or savings--it regards as sufficiently difficult to calculate.  Unless you have a good data set on the corporate income elasticity of healthcare benefits, any number that you assign to this change in behavior are simply a wild assed guess.

But that's going to bias PWC's numbers upward, which they sort of forget to mention.  How much?  By much less than people may think just reading the complaints.  This quote from PWC has been extensively highlighted:

We have estimated the potential impact of the tax on premiums.  Although we expect employers to respond to the tax by restructuring their benefits to avoid it, we demonstrate the impact assuming it is applied.

Reading this, you might think that everyone is going to structure their benefits to get around this tax.  But the CBO expects us to collect quite a bit of money from this tax.  Interestingly, we have some insight into the thoughts of the CBO and Joint Committee on Taxation on the income elasticity, because the Baucus bill changed two things about the excise tax between proposal and final amendment:  it increased the excise tax from 35% to 40%, and it eased the threshold for retirees and workers in high risk professions.  The net change is less than $14 billion over the ten years.  By 2019, the tax is still supposed to be collecting over $45 billion a year.

Even if you assume that all of the change between the original and the amended bill came from the increase in the excise tax--and it most definitely doesn't--this implies a relatively small change in behavior on the part of companies even for a sizeable increase in the excise tax.  If the change in behavior is big enough to actually invalidate the PWC analysis, then we're screwed, because we just opened up an annual hole in the program of $50 billion or so.  I'm more worried about the fact that they estimated the prevalence of "Cadillac plans" from COBRA data, since I can think of a bunch of ways in which the COBRA sample might not match the general population of people with employer insurance--but then, I don't know how the CBO or the JCT produce their estimates, so it's hard to complain too much.

(As an aside, wouldn't it be great if we made the CBO and the JCT show their work more?  An easily accessible trove of subtables would now be easy to post online, and it would be wonk heaven . . . )

PWC estimates that the Baucus bill will increase large employer health insurance costs by about 11%.   5% of that is supposed to come from the excise tax.  (Other employers also take a 5% hit).  How much should we shave from this to account for the change in employer behavior in response to the excise tax?  That is hard to say.  But here's one possible way you could calculate it.  The National Coalition on Healthcare cites the Robert Woods Johnson foundation as saying that under current law, total employer health care costs could reach $850 billion in 2019.  The $46.3 billion that we are expecting to raise that year from the excise tax is . . . 5.4% of that.

Of course, employers share costs with their employees; on average, they pay about 80% of the total premiums.  So bump that number up to $1.1 trillion, in order to account for the employee share.  It's still 4.3%.  This suggests that calculating dynamic effects for the excise tax would lower PWC's final estimates only slightly.

There are other complaints:  PWC anticipates that Medicare cuts will result in cost-shifting to the private sector, which is questionable, and impossible to estimate accurately.  But that's not the lion's share of the cost increase.  Most of the premium increase they estimate result from things that we pretty much all agree are going to happen:  taxes are going to be levied on insurance companies and various providers, and passed through to the customers; and there is going to be adverse selection in the insurance market without a strong enough mandate.  I think it's pretty hard to argue at this point that the mandate is strong enough--the Massachusetts penalty now is higher than the Baucus penalty will be in 2016.  But I'm sure there are plenty of people willing to tell me why I'm wrong.

So how does Gruber get such a different result?  For one thing, he looks at the cheapest possible plan under the new structure.  This is not unreasonable, because my understanding is that the uptake of the premium (silver and gold) packages in Massachusetts has been pretty limited; most people take the cheapest premiums.  These plans have high deductibles and co-pays and limited doctor networks, so of course they cost less.

This analysis is slightly problematic, because he uses the Baucus bill's target of a Bronze plan worth 65% of the actuarial value of the most expensive plan.  The problem is, there are hard limits on how much cost sharing these plans can engage in; they cannot have out-of-pocket expenditures higher than the HSA limits (plus a health care inflation index), and small employers, who are expected to purchase a number of these plans, cannot offer any plan with a deductible higher than $2,000 for individuals, $4,000 for families.  So these calculations do not necessarily represent the experience that most people will actually have in the health care market.  But I don't know how you model that.  Right now, it's my understanding that the average private policy has a deductible about $3,000 for individuals, so maybe it doesn't matter.  I just don't know.

The other problem is that he makes some strong assumptions.  For example, he assumes that the actuarial value of a "young invincible" plan is .5, versus .65 for the "Bronze" plan.  But as far as I can tell, the "Young Invincible" plan is the bronze plan, minus preventive care.  But catastrophic coverage insurers routinely throw in cheap preventive care visits as a freebie, on the grounds that an annual checkup may save them a pricey hospital visit later.  So it's not clear how much these policies will actually save.

Meanwhile, I simply have no idea where his family numbers come from. The CBO estimate of the cost of a "silver" plan for a family of four (unsubsidized) is $14,400; Gruber's estimate of the premium for a family of four with a head of household aged 45 is $8,340.  The actuarial value of the bronze plan is lower, of course, but changing the actuarial value from .7 to .65 only brings it down to $13,392.  He says he's working in 2009 dollars, which could explain part of it, but to bring the CBO's estimates in line with his, I need an inflation rate of over 7%.  There's something here that I'm not understanding.  I've emailed Jonathan Gruber, and will report back if I hear where my error may lie.

Quibbling about the models aside, I think the real problem with his analysis is that he ignores the employer market.  This is not a small omission.  Employer insurance is the lion's share of the coverage market; the CBO shows that in the nonelderly market of 267 million, 150 million are covered by employer insurance.  The "nongroup" market, aka individual insurance, is 13 million.  Non-employer, non-government health insurance is projected to cover around 30 million by 2019, against about 155 million in employer plans.

As many reform advocates keep telling us, this plan does almost nothing for or to employer plans.  The delivery system reforms are all in the Medicare/Medicaid market, while the subsidies and pooling changes are in the exchanges.  This plan doesn't have any mechanism to keep their premiums down, or control the costs that accrue to their employees; any cost savings there are occur in Medicare or the non-group market.  What it offers them is . . . an excise tax on high cost plans.  Yes, yes . . . healthy workforce!  Preventative care!  All I can say is, these marvelous savings do not seem to be accruing to employers in Massachusetts--or for that matter, people on the exchange, whose premiums are apparently increasing even faster this year than they did before reform.   Since employees pay about 20% of that cost directly, and (most economists agree) 100% indirectly, increasing their premiums by an average rate of 5-10% is a big deal.

For that matter, this health care plan doesn't really have much designed to control costs in the individual market, beyond initial savings in administrative costs.  I'd like to see Jonathan Gruber, and the CBO, explain why their projections for the individual market differ so greatly from what is actually happening in Massachusetts.  Most analyses include big drops in administrative costs to bring costs down--but they don't include, well, whatever is making premiums rise so fast in Massachusetts.

After all, Massachusetts has a plan that is very similar in its core elements, with a stronger mandate.  It doesn't have our 40% excise tax, but as we've seen, the excise tax is more of a revenue mechanism than a cost control mechanism.  I've heard some explanations, like the fact that Massachusetts has a lot of teaching hospitals--but Massachusetts had a lot of teaching hospitals before reform, so why are premiums increasing more rapidly now? Moreover, Masachusetts also had some advantages, like a smaller uninsured population, and a wealthier, more educated citizenry than the national average.

The theory that drawing all the missing low-cost users into the pool through a mandate would help lower the average cost was a fine theory, but I don't see it panning out in practice.  So I don't really understand why Gruber is projecting that it will suddenly start working at the federal level--there may be a reason, but I wish I could have it explained to me.  The countries that have lower health costs do not rely on pooling to get that effect.  They resort to price controls.

At some level, of course, this is all quibbling.  The beauty of projecting into the future is that you can get any result you want with rather small tweaks to the model, any of which are probably defensible--will lab costs rise 0.5% a year, or 1.5% a year?  At the end of ten years, that's the difference between a 5% increase in those costs, and a 16% increase.  It's no surprise that people who support health care reform tend to get absurdly precise numbers claiming costs will go down, and people against it tend to get absurdly precise numbers claiming costs will go up.

Gruber's numbers are well below the CBOin some way that I don't understand, which worries me, and PWC's numbers are well above their estimates of things like the increase in premiums resulting from all the new provider taxes (1% from the CBO; 2.5% from PWC).  I'm not sure there's any good reason to rely very much on either of them.  The fact is, we can outline the various reforms, and which way we think each of them will push insurance prices, but there's no way to get a number that's even moderately accurate.  Would either AHIP or Jonathan Gruber like to wager any sum of money that it would actually hurt them to lose on their estimates coming in within 5% of the actual premiums in 2016?  I wouldn't bet a dime on either projection.

That is not, mind you, because I'm accusing either side of bad faith.  Jonathan Gruber is a very smart economist.  PWC is a pretty good accounting firm.  But the problem is indeterminate.  Premiums for your average individual plan under the exchanges probably won't be $3,000 or $20,000.  Beyond that, who knows?

At least with the CBO, you're (sort of) comparing apples to apples--their estimates may not be especially accurate, but they're done in a fairly consistent way that allows us to benchmark the cost of bills, even though reality may differ substantially because of unanticipated factors.  Whatever the good intentions of the authors, these dueling private estimates just end up being used the way drunks use lampposts--for support, rather than illumination.

Update:  I have more insight into the CBO/JCT calculation of the excise tax that gives the criticisms more bite . . . but also raises some questions about the revenue predictions.