HEALTH CARE AND CONSUMER SHOPPING
Much of the rhetoric about the health care market today is about the soaring cost and the increasing share of GDP of medical care. Regarding the second point, we should not care about the share of GDP going to any particular product market. If consumers and producers are engaging in voluntary exchange and there are relatively few impediments to the types of contracting in which consumers can engage, then the amount of consumption in that product market reflects the fulfillment of consumers’ desires, and there is no necessary share that one product market should have of GDP. It is the soaring costs that may reflect that all may not be well in the medical care market. Here I provide an overview of empirical findings related to consumer shopping behavior in both health care and health insurance markets.
In analyzing consumer shopping behavior in the health care market, we must remember that consumers’ incentives to shop may be attenuated, depending on the type of insurance coverage a consumer has. Many studies on consumer shopping utilize differences in the behavior of those with a high- deductible health plan (HDHP)8 and those with traditional health plans. The idea is that high deductibles give consumers “skin in the game.” In HDHPs, insurance does not begin to cover the costs of health as soon as traditional insurance plans, which increases the marginal cost a consumer faces each time she seeks medical attention. This should incentivize the consumer, at the margin, to consider shopping before deciding on a particular health care provider. Sinaiko, Mehrotra, and Sood (2016) report findings from a survey of approximately two thousand health care consumers. The consumer demographics were broadly similar, with HDHP consumers having slightly higher incomes, and beliefs and attitudes toward health care were also similar. When consumers were asked whether they would use price information if it were available, there was no statistical difference in the percent of “very likely” and “likely” responses between the consumers, with HDHPs (56 percent) and consumers with traditional health plans (50 percent). When asked about their most recent use of medical care, only 10 percent of traditional health plan consumers and 11 percent of HDHP consumers responded that they had considered other health care professionals, and only 4 percent of HDHP consumers and 3 percent of traditional health plan consumers responded that they considered other health professionals and compared costs of those health professionals.
From these findings, it does not appear that consumers who face higher marginal costs engage in different shopping behavior than those who face lower marginal costs. Several factors could explain these findings. Search cost may be too great. Shopping for price requires calling different providers, asking whether they take a particular insurance, then waiting for the care provider to respond with what the price of the care will be, which depends on the negotiated contract between the insurer and the provider. Also, given that the consumers with HDHPs have higher incomes, the relative search costs for the HDHP group are somewhat higher. However, the fact that only half of each group would use price information if it were available suggests that there is a low level of interest in shopping around.
Brot-Goldberg et al. (2015) compare the behavior of health care consumers from a large employer that switched from a plan with no cost sharing to an HDHP. As would be expected, spending decreased substantially, approximately 12 percent over a two-year period. The authors decompose this spending reduction to measure the effect that comes from price shopping, quantity reduction, and quantity substitution. The price shopping effect is found by comparing whether consumers shifted toward a lower-cost provider, conditional on a given procedure. The total quantity effect is measured as the total change in health care expenditure minus the change in expenditure caused by provider price inflation and consumer shopping. Quantity reduction is just the measured change in the number of procedures, and quantity substitution is the part of the total quantity effect not explained by quantity reduction. They find that 90 percent of the spending reduction in the first year after the switch comes from quantity reduction, with the remainder coming mostly from quantity substitutions. The mix of procedures chosen by consumers suggests that consumers, overall, chose higher-priced providers after switching to the HDHP. The authors also estimate the potential reduction in medical expenditure if those consumers who choose providers with higher than median prices choose providers with median prices.9 This would reduce expenditure by approximately 20 percent.
These findings do not necessarily suggest that consumers are acting irrationally. The analysis does not account for the quality of the provider, and depending on the correlation between price and quality, reducing expenditures solely through quantity reductions may be optimal. That quantity reduction accounts for almost the entire reduction in expenditure may be a sign that moral hazard was driving previous health care consumption decisions. This analysis also only considers the effect of the decisions of a subset of consumers on provider prices. If the quantity reduction were market wide, we would expect providers to reduce prices. That the mix of providers whom consumers chose became more expensive after the switch does, however, suggest that consumers may not be shopping effectively.
These studies suggest that consumer shopping for medical providers may not be effective in enhancing competition in health care markets. There are factors that are not taken into account: search costs incurred every time a consumer shops for a provider, the quality of the doctor, and the switching cost of choosing a new provider. These are difficulties that economists encounter when analyzing the shopping behavior of consumers in the health care market. We may be able to avoid these difficulties by analyzing how consumers choose prescription drug plans (PDPs).
Several studies show that upon enrollment in Medicare Part D, consumers choose PDPs that are not optimal by not choosing the lowest-cost alternative. Abaluck and Gruber (2011) show that only 12 percent of consumers chose the lowest-cost plan in their state (with a mean savings of 30 percent possible), and if only plans with non-increasing variance in expenditure are analyzed, 70 percent of consumers did not choose the lowest cost plan, conditional on realized expenses. These authors also found that consumers put more weight on plan premiums than on expected out-of-pocket costs.
This may be a somewhat unfair test, since consumers cannot perfectly foresee their prescription usage throughout the year. Heiss et al. (2013) provide benchmarks for various types of consumers: perfect foresight, rational expectations (using past data plus current information to forecast costs), adaptive expectations (assuming next period’s cost will equal last period’s cost) minimum premium, and random. The authors calculate the optimal plans for individuals using these benchmarks and compare them to consumers’ actual choices and find that only compared to randomly choosing a prescription drug plan do more than half of consumers choose a lower cost plan than the benchmark. The percent of consumers choosing a plan with total higher costs than the benchmark varies from 66 percent when compared to the cheapest premium benchmark to 93 percent with the perfect-foresight benchmark.10 So even when compared to rules-of-thumb behavior, consumers overspend.
Zhou and Zhang (2012) and Patel et al. (2015) find similar results of consumer overspending and non-enrollment in lowest-cost plans, and Ho, Hogan, and Morton (2015) find that consumers’ inattention blunts competition to reduce prices among plan providers. Evidence on consumer learning is mixed with Ketcham et al. (2012), finding significant learning, with 81 percent of the sample benefiting from reduced overspending, while Abaluck and Gruber (2016) find little evidence of learning. When provided with information about potential savings, many consumers switch to the lower cost plan (Patel et al. 2009).
There is some evidence that consumers can effectively shop under different payment contracts. Robinson and Brown (2013) examine a change in payment arrangements for knee and hip replacement surgery within the California Public Employee’s Retirement System (CalPERS). CalPERS switched to a reference price system in which CalPERS set a maximum amount it would pay for knee and hip replacements. Patients could choose any number of hospitals to receive the knee or hip replacement. If the negotiated price was higher than the reference price, the patients had to cover the difference. Before CalPERS put this payment arrangement into effect, high-cost (priced above the reference price) hospitals served 52 percent of CalPERS knee and hip replacement patients. After the change, high-priced hospitals operated on only 37 percent of CalPERS patients. In addition, those high-priced hospitals reduced their prices by an average of 26 percent after CalPERS switched to reference pricing. This suggests that for certain procedures, increasing the marginal cost the consumer faces may lead to increased shopping. I should note that, in addition to switching to reference pricing, CalPERS also provided patients a list of hospitals that charged less than the reference and were of sufficiently high quality. Without this information, it is unclear how well patients would have been able to shop for the procedure.
This brings up the idea that consumers will be able to shop better for some health care based on the urgency of treatment. If consumers can shop for some services, such as hip and knee replacements, we may expect that prices are less variable. Frost and Newman (2016) analyze a nationally representative data set of health insurance claims and categorize the services into shoppable and non-shoppable.11 They find that shoppable services have a smaller variation in prices than non-shoppable services. This is crude evidence that shopping does occur in the health care market. The authors also calculate that about 43 percent of health expenditure is on shoppable services and that only 7 percent of out-of-pocket spending is on shoppable services.
This suggests that under current institutional arrangements the range of shopping (whatever the effectiveness) in the health care market is rather limited.
This review of the literature suggests that health care consumers face difficulties in shopping for health care. This could be because there are switching costs involved in changing providers, high search costs due to the Rube-Goldberg nature of the health care system, cognitive limitations that impede proper calculation, or consumers not bearing the full marginal cost of care.12 While there is some evidence to suggest that health care consumers can effectively shop for health care services, under most institutional arrangements we see today, shopping remains marginal and muted. There are, however, historical examples of various institutions, such as mutual aid societies, and contemporary arrangements, for example, medical tourism, that provide insights into how exchange in the health care market can be structured to allow for beneficial competition. These show that increasing the range of exchange relationships, rather than constraining them, can lead to a more well-functioning market.