My research includes a focus on inferring the effects of Medicare payment policy on the behavior of firms offering private plans within the program. One of the chief difficulties in this type of work is controlling for the effects of cost. One expects premiums to go up if payment is reduced, but the cost of care also affects premiums. If one does not control for the cost of care one will obtain a biased estimate of the effect of payment on premiums.
Unfortunately the cost of care and other costs relevant to plan decisions (about where to enter and what to offer) are either not in the public domain or are hard to measure. Relevant correlates and proxies exist and are frequently used, for lack of anything better. However, in 2001 something unusual happened that offered an enhancement in the ability to control for costs in estimating elements of plan behavior.
In December of 2000 the Benefits Improvement and Protection Act (BIPA) was signed into law and created a natural experiment. Medicare plans had already established their premium and benefit structures for 2001 in response to expected costs and the payment rates in force prior to BIPA. BIPA changed those payment rates and plans were permitted to adjust their 2001 premium and benefit levels in response. For each plan, both sets of premium and benefit levels (pre- and post-BIPA) were publicly available to researchers swift-footed (or fingered) enough to download them.
The last-minute change in payment rates and rapid adjustment of premiums and benefits offered the potential to reveal how plans responded to changes in payments over a compressed time frame during which underlying costs could not have changed much. The data that captured these phenomena included variation in payments, premiums, and benefits, but no significant temporal variation in underlying cost. Other dimensions of variation in cost (geographic or plan-level) are relatively easy to control for statistically.
My colleagues Steve Pizer and Roger Feldman and I took advantage of this golden opportunity and published two papers [1, 2]. One set of findings revealed the important role of competitive effects in the Medicare private plan market. These results indicated, for example, that the addition of one more plan to a market evenly divided among three existing plans would approximately offset the effects of a 10-percent reduction in payment rates [1].
Another set of findings lent support to the notion that some benefits offered by plans are less valuable to beneficiaries than they cost taxpayers (a finding also revealed in a later study). Plans are required to spend dollars received above their costs on benefits or premium reductions. In 2001 plans could not lower their premium below zero (i.e. could not offer rebates) as they can today. Consequently, plans at the zero-premium bound were forced to put the extra funds into enhanced benefits. Our statistical analysis indicated that this loss of flexibility was binding. Plans at the zero-premium bound behaved differently–offered different benefits–than plans with a positive premium. This result strongly suggests that the benefits enhancements of zero-premium plans were inefficient. That is, they were not valued by beneficiaries at or above their cost [2].
While these results could probably be obtained with data from a period over which costs changed (using more statistical controls), that they were found in a circumstance where costs did not vary over time increases our confidence in them. Controlling for costs is hard, and the late-year passage of BIPA offered an opportunity to do some important work absent their confounding effects.
[1] Pizer SD and Frakt AB, Payment Policy and Competition in the Medicare+Choice Program,” Health Care Financing Review, 24(1) (Fall 2002): 83-94.
[2] Pizer SD, Frakt AB, and Feldman R Payment Policy and Inefficient Benefits in the Medicare+Choice Program,” International Journal of Healthcare Finance and Economics, 3(2) (June 2003): 79-93.
