Tuesday, December 25, 2012

Medicare Advantage auctions: Asking too much?

When I teach about auctions, I like to ask students:  What does an auction accomplish, or put differently, what social roles does an auction play?

I point to two major roles:  An auction determines an allocation -- who gets the good being sold, or who is chosen to produce -- and it also determines a price.  Two very important things:  allocation and price.

In Medicare -- a confused and confusing policy area if there ever was one!-- auctions are used in both Medicare Part D (prescription drug coverage) and Medicare Advantage (private Medicare plans).

I am concerned that some policy proposals for Medicare Advantage (MA) are asking too much from an auction, for they add a third role:  determining the subsidy level for subscibers.  This is a complicated issue, requiring auction theory that is at the frontier.  But I think the intuition is pretty clear.  Also, while I will focus on MA here, similar issues arise with Part D plans, albeit somewhat less so because of the way those rules are set.

In a nutshell, here is the way MA plans work now.  Private insurers submit bids to provide health coverage for those over 65, with bids submitted on a county basis.  Folks who qualify for Medicare can either take the standard government-issue Medicare or opt into one of the private MA plans. The private plans are paid by the government a subsidy amount equal to the average per person cost of that county's standard Medicare plan.  If the plan bids more than that, the enrollees in that plan pay the difference between the subsidy and the bid.  If a plan bids less than the subsidy, then enrollees don't pay anything but the plan has to rebate the difference to enrollees as either cash or extra benefits (I do need to verify the specifics of this, but for now I don't think it is crucial).  Importantly, enrollees select which MA plan they want, so choice is a key part of the process.

So this is fine.  The auctions do two things, as above.  They determine which of the private plans provide service (allocation) and they determine a price (the price paid by enrollees).

Note that the subsidy is determined exogenously from the auction -- the average per person cost of standard Medicare.  Granted, there might be some endogeneity here, as the cost of the local Medicare plan depends on who opts into MA plans...but that seems of second order importance.

However, some policy proposals (see Alice Rivlin, for example) will add a third role to MA auctions, that of determining the subsidy.  The typical idea is to set the subsidy at the second-lowest bid of the private insurers.

The first order logic of this is great.  Set the subsidy at that level, and you can be sure that at least two plans will be willing to offer coverage at that subsidy amount.  Even more important, instead of having the MA subsidy set through a political process, it is set in a market mechanism.  What could sound better than that?

Here is my concern, arising from the effect that setting the subsidy in the auction will have on strategic bidding behavior. (Let's be clear that strategic bidding behavior should be expected, that is, insurers will not just put bids in that equal their expected cost, even if that is what the government asks for.  Insurers will put in bids that maximize their expected profit.)

The issue is that by putting in a higher bid, an insurer has a reasonable expectation that it will increase the subsidy (if the bidder happens to be the second lowest bid).  This will increase the subsidy to enrollees and make it less likely that the insurer's bid will result in a net payment by the enrollees.  Also, as the subsidy increases, more people will opt into the MA plan arena.  Seems pretty clear to me that this will result in higher bids.

Amplification of this problem arises because is in MA plans, there is not a standard package of benefits.  By adding benefits, and putting in a higher bid reflecting the higher cost of that expanded package, an insurer minimizes any competitive effect of being a high bidder in the auction while still having a reasonable expectation that the subsidy will be increased.

As all bidders do this, the entire distribution of bids shifts higher.  Studies that have been done on the cost savings from basing the subsidy on the second lowest bid are obviously wrong, as that second lowest bid is going to be higher.

The idea is not that different from shifting from a second-price sealed bid auction to a first-price selaed bid auction (standard auction where something is being SOLD to bidders).  It would seem that taking the highest bid as the price in an auction would clearly be better than taking the second highest.  But as the rules change from second-highest to highest, we have to expect that bidders will lower their bids.  I always ask students:  What do you think is greater -- the second highest out of a distribution, or the first highest out of a lower distribution?

No comments: