Are Consumers Forward-Looking in Responding to Health Care Prices?

08/30/2012
Featured in print Bulletin on Aging & Health

The rising cost of public and private health insurance is a cause of mounting concern for policy makers, employers, and individuals. One frequently mentioned demand-side approach to controlling cost growth is to increase consumer cost-sharing, for example by incorporating larger deductibles or coinsurance into the design of insurance plans.

A number of past studies dating back to the landmark RAND Health Insurance Experiment of the 1970s have established that demand for health care is sensitive to price, suggesting the existence of what is often called "moral hazard" in health utilization. Yet these studies typically look at the relationship between demand and a single consumer price, even though the health insurance price schedule is usually non-linear, with consumers facing different prices at different points in the year depending on whether they have satisfied their deductible or hit their out-of-pocket maximum for the year.

This observation is the motivation for a new study Moral Hazard in Health Insurance: How Important is Forward-Looking Behavior? (NBER Working Paper 17802) by researchers Aviva Aron-Dine, Liran Einav, Amy Finklestein, and Mark Cullen. The authors ask whether consumers who face the same health care price today but are likely to face a different price at the end of the year make different choices in their usage of health care today. If so, this would suggest that consumers are forward-looking in their health care consumption decisions. On the other hand, if the future price is unrelated to current behavior, this would suggest that consumers are myopic in health care decisions (alternatively, consumers could be poorly-informed about the price schedule or liquidity constrained).

The challenge is to find a source of variation in future health care prices that is unrelated to health care consumption today (except through the desired price channel). For example, sicker individuals will face a lower price at the end of the year because they are more likely to satisfy their deductible and reach the out-of-pocket maximum, yet their higher health care consumption today may be more a function of their illness than a response to the lower future price.

The authors' solution is to compare workers who are hired (or join the company's health insurance plan) at different points in the year. A worker who starts work later in the year will face a pro-rated premium, but the same deductible and coinsurance amounts as a worker joining the firm on January 1. As a result, the worker who is hired later in the year is less likely to satisfy the deductible and therefore faces, on average, a higher expected end-of-the-year price. For their empirical analysis, the authors use data on medical claims for 7,000 employees of Alcoa, Inc. hired over the period 2004 through 2007, as well as similar data for nearly 100,000 employees hired by two other large firms in the early 2000s.

The study's results reject the hypothesis of completely myopic behavior. Health care utilization appears to respond to the future price of health care, with a ten percent increase in future cost resulting in a 4 to 6 percent decrease in utilization today. To put these results in context, the authors develop and calibrate a dynamic model of individual behavior. This exercise "suggests that, at least in the populations we study, individuals may be far from fully forward looking, but that, nonetheless, the extent of forward looking behavior we detect has a non-trivial impact for forecasting how medical spending will respond to changes in non-linear health insurance contracts."

The authors conclude, "these findings have important implications for estimating or forecasting the impact of alternative health insurance contracts on medical spending, which is a topic that receives great interest and attention both by academics and in the public policy arena."


The authors gratefully acknowledge financial support from: the National Institute on Aging (R01 AG032449); the National Science Foundation (SES-0643037, Graduate Research Fellowship); the John D. and Catherine T. MacArthur Foundation; Alcoa, Inc.; and the U.S. Social Security Administration through a grant to the NBER as part of the Retirement Research Consortium.