In 2006, the federal government first began expanding Medicare coverage to include prescription drugs using the Medicare Part D program. According to one report, Part D will cost taxpayers $47 billion in 2007.
Yet it is possible that Medicare Part D could actually save taxpayers money. If prescription drugs and other medical care are substitutes, then increasing funding for lower cost pharmaceuticals could actually save taxpayers money on the more expensive hospital stays (covered by Medicare Part A) and physician visits (covered by Medicare Part B). For instance, it is possible that regularly taking beta blockers may reduce the chance that one needs an expensive heart surgery.
On the other hand, if pharmaceuticals and other medical care are compliments, than increasing Part D funding, could increase the total spending in Medicare Parts A and B. For instance, individuals taking prescriptions drugs may need to go to the doctor more often–covered by Part B–in order to have their pharmaceutical usage monitored.
So how does Medicare Part D affect other Medicare spending?
This is the question Baoping Shang and Dana Goldman investigate in their NBER Working paper “Prescription Drug Coverage and Elderly Medicare Spending.”
Data and Methods
Shang and Goldman use data from the 1992-2000 Medicare Current Beneficiary Survey (MCBS) and compares Medicare spending differentials between individuals who have a Medigap policy with drug coverage and individuals who have a Medigap policy without drug coverage.
Since Medicare spending–like most health care spending–is right skewed with a large mass at zero expenses. The authors use a two-part regression structure. In the first regression, the the authors use a probit regression to determine the probability an individual had any health care spending. In the second regression, Shang and Goldman utilize an OLS (an later an IV) structure to find the impact of Medigap drug coverage on total spending, conditional on the fact that the individual had some spending. Mathematically, the two regressions look as follows:
- p* = β0 + β1*d +β2(d*Income) + ε
- ln(Y|Y>0) = γ0 + γ1*d +γ2(d*Income) + ν
p* is the probability of any spending, d is a dummy variable if the individual has drug benefits, and Y is total Medicare spending.
This econometric structure could lead to incorrect inferences if selection bias were present. In fact, “[c]ompared to those with prescription drug benefits, Medicare beneficiaries without drug benefits tend to be older, less educated, less likely to be in an urban area, and poorer. They are sicker in term of both self-reported overall health and histories of chronic diseases.”
In an attempt to eliminate selection bias, Shang and Goldman employ state reforms in the health insurance markets as instrumental variables. These reforms include the following:
- “Guaranteed issue requires health plans to offer coverage to all individuals, regardless of their health status or claims experience.
- Rate rating includes rating bands, very tight rating bands, and community rating. Rating bands restrict health plans’ use of experience, health status, or duration of coverage in setting premium rates for individuals. Very tight rating bands allow very limited adjustment for experience, health status, and duration. Community rating prohibits health plans’ use of experience, health status, or duration of coverage in setting premium rates for individual coverage.”
For their instrument, Shang and Goldman look at states with 1) both guaranteed issue and rate rating, 2) states with only rate rating, and 3) states with neither. Since MCBS is a panel, the authors employ a discrete factor model to control for three different levels of unobserved heterogeneity directly and allows some correlation of these fixed effect terms with the error terms.
A simple two part model finds that the “prescription drug benefits increase drug spending by $157, reduces Medicare Part A spending by $135, and increases Medicare Part B spending by $31”–a net $104 reduction in Medicare spending. The more complicated structural model using structurally estimating unobserved heterogeneity parameters finds that the drug benefit increases drug spending by $170 (or 22%). However, “prescription drug benefits decrease Medicare Part A spending by $350 or 13%; and prescription drug benefits decrease Medicare Part B spending by $74 or 4% although the estimates are statistically insignificant.”
Healthcare Economist comment
Even for those who oppose government provided health insurance, few would argue with the statement that given Medicare’s existence, it is important to be sure it operates in the most efficient way possible. This paper demonstrates that Medicare Part D may be cost saving. Leaving out prescription drug benefits may lead patients to choose expensive surgeries–which are free to them since they are covered by Medicare –over taking prescription drugs–which are costly without Medicare Part D. The authors sum up their findings in a compelling manner: “…it appears that Medicare beneficiaries may have been overinsured with respect to medical services, and underinsured with respect to prescription drugs.”
Shang, Baoping; Goldman, Dana; (2007) “Prescription Drug Coverage and Elderly Medicare Spending” NBER WP #13358.