2015) use government changes in Medicare Part D prescription drug coverage as an instrument to study consumer responses to cost sharing, while Scoggins and Weinberg (2016) use plan characteristics as an instrument for the actual observed cost share in claims data. Furthermore, our data do not include general plan parameters, such as copayments and deductibles, that would allow us to predict what the cost share would be in the absence of expenses. We implement this by using the actual monthly cost share for each subsequent month until another month of positive spending occurs, at which point the backward myopic cost share is revised again.
Here we assume that consumers expect the cost share for their first visit to be the average of the cost share of health care plans for all other consumers in the first month of seeking care. If the year ends with a period of one or more months without visits, we will continue the most recent month's cost share for the remainder of the year. In Panel A, the retarded myopic individual made no visits in January, and therefore January's average cost share is assigned.
In March, this individual receives care and pays only 75% of the cost, so this new lower cost share applies from April to September. Finally, in October, the consumer takes into account the 20% share of costs from the previous month and adjusts accordingly for the rest of the year. From April onwards, the individual expects that the share of costs will be the same as it turned out to be in September, namely 20%.
5 We also conducted overidentification tests for our models, using the cost shares of inpatient, outpatient, and pharmacy services as instruments, as well as the cost shares specific to each type of service.
Results
The lowest line in December in each figure is that of the HMOs, which is just below that of the HDHPs. At the other extreme, costs are highest for the non-managed care comprehensive health plan, which also has the oldest enrollees and lowest average cost share. HDHPs and HMOs are at opposite ends of the spectrum of the monthly cost share variation.
Since the average cost sharing in an HMO is almost constant from the beginning to the end of the year, the variation is almost zero, making the imputations very close to the actual values. Myopic backward and forward prices are above actual values for most of the year. This is particularly striking in the case of hospital spending, where cost share imputations are almost constant over time and show very little year-to-year variation.
The lack of month-to-month variation within a plan in hospital cost sharing is a precursor to why we have difficulty accurately estimating the impact of cost sharing across inpatient categories, such as room and board, as well as other categories. services, even in our extremely large sample. Separate estimates are presented for the myopic backward and forward cost share models.8 The aggregate demand elasticity is estimated to be -0.44 using the myopic backward cost share expectations and -0.41 for the myopic forward cost share expectations. , which are reliable and consistent with most of the previous literature.9. Our model estimates almost identical elasticities of the -0.29 (backward myopic) and -0.28 (forward myopic) models. 2016) estimate an average drug elasticity of −0.24 for Medicare Part D, a population older than our sample.10 Hospital expenditures are estimated to have a statistically significant elasticity of −0.30 using lagged pricing. myopic versus an incredible -2.35 for myopic price futures.
This is a particular problem for myopic primes, so we mostly focus on myopic lagged results in the remainder of this section. Emergency room (ER) spending has a remarkably low elasticity of -0.04 using myopic lagged prices, as does maternity (-0.09), consistent with our expectations. 8 We performed F tests of our first-stage instruments, and the minimum F statistic was 14.65 in the posterior myopia specification and 9.03 in the anterior myopia specification, with most values greater than 50.
9 Kowalski (2016) uses a different IV approach to estimate aggregate demand elasticity of -1.49 at the average percentage. Surgical procedures and Dialysis, two types of very expensive services, are of the wrong sign and not statistically different from zero. Consider drugs, goods, and services such as statins (for high blood pressure), insulin, oxygen, home health visits for the chronically ill, and long-term behavioral, physical, and occupational therapies.
Extensions and sensitivity analysis
This may reflect the current trend that most people receiving professional mental health care receive only drugs or MH/SA treatment in primary care (McGuire, 2016), which is consistent with the low level of specialist mental health care spending , which is observed in our data. Most people with expenses for these types of services are likely to be associated with high annual expenses and thus exceed any reasonable deductible or stop loss and face very little price variation. This is also a good argument why cost-sharing on such services will be an ineffective tool for cost containment, as it carries financial risk with little effect on the expenditure on those services.
Spending on these services is very predictable for many patients, and if they are expensive, consumers can easily expect to exceed their deductibles. Our approach may nevertheless have some strength, as even users of these services may be surprised by other types of spending that affect the cost share. Adults in single and family contracts are similar in demand responsiveness, while children are less price responsive.
HMO enrollees are less price responsive than HDHP enrollees for outpatient spending, but not for pharmacy spending. This result is also consistent with the results of Einav et al. 2015) documenting the heterogeneity of Medicare pharmacy demand. Enrollees in low-cost-sharing firms have more inelastic demand than those in high-cost-sharing firms.
We present this summary of the effects of cost-sharing on various population and employer subgroups without attempting to interpret them, primarily as examples of the power of using a readily available instrument in large samples to explore differences along such dimensions. A concern with our IV is that cost shares are correlated with the degree of consumer foresight. Here, the highest value RRS corresponds to seven times the average consumption, while the lowest RRS range is .1, which is only 10 percent of the population average.
The analysis shows that, as expected, backward and forward shortsighted prices have a sharp downward slope: sicker people are more likely to exceed any deductibles or stop losses and pay lower cost shares. Ideally, there would be no relationship between the “leave one out” average cost shares and the RRS. The resulting elasticity estimates are less stable (especially for rarely used services), which we believe is because the linear probability model performs extremely poorly. Table A-8) predicting log(expenditures) are also problematic, detecting both positive and negative price effects, as might be expected given the large selection effects in the first stage.
Conclusions
An innovation of the paper is that we estimate two different kinds of spot prices: backward myopic prices, where consumers never expect future consumption, and forward-looking spot prices, where they fully anticipate actual consumption in the current month when making consumption decisions. Our approach holds great promise in potentially providing a new instrument—the employer's average monthly cost share of a service of interest—that can be used in other studies looking for an instrument for consumption or utilization of a service. If consumers are diagnosed with condition 123 at different times of the year and face changing cost shares over time, our approach can be used to generate a reasonable instrument to assess the effectiveness of each of these procedures, tests, or drugs in treating condition 123 .
Moral hazard in health insurance: are dynamic incentives important? The review of economics and statistics. Alternative transformations to deal with extreme values of the dependent variable.” Journal of the American Statistical Association. Identification of four unique spending patterns among older adults in the last year of life challenges standard assumptions.” Health matters.
The Response of Drug Expenditures to Nonlinear Contract Design: Evidence from Medicare Part D.” The Quarterly Journal of Economics. Rational Behavior in the Presence of Coverage Ceilings and Deductibles.” The Rand Journal of Economics. The Demand for Treatment Episodes in a Health Insurance Experiment.” Journal of Health Economics.
Note: Shown are the expected cost shares for backward myopia and forward myopia for a single, hypothetical consumer who visits only in March and September, and who experiences an average cost share of 0.75 in March, and 0.2 inches. The average cost share for someone who visits for the first time in January for this employer*annual plan is assumed to be .90. The forward myopic consumer anticipates the cost shares in March and September and uses them until new information becomes available.
Note: Each panel presents the average share of actual cost for the services indicated by month, averaged over the seven-year period for each plan type for the sample of 73 employers. Note: Each panel presents monthly expenditures for services indicated by month, averaged over the seven-year period for each plan type for the sample of 73 employers. Note: Each panel shows the actual, myopic backward, and forward cost portions for three types of plans: HMO, PPO, and HDHP.
Each model includes individual* year fixed effects, 12 monthly time dummies and uses the employer's average cost share for it. Elasticities are estimated at the average cost share for that type of service and S.E is obtained using the delta method.
