Asymmetric Information

Yesterday, I spoke about the Swiss health care system. One of the main attributes of this system is that patients are allowed to choose from any health care plan and the health insurers can not refuse to cover them. Further, since the insurance benefit is mandated by law, there is very little quality difference between plans. Only 8% of health plans restrict provider choice in any way.

An NBER working paper by Frank and Lamiraud analyzes how the number of firms offering health insurance in Switzerland has affected price dispersion and expected price.

Economists generally believe that when more firms/products enter the market, this will increase the probability that a superior health insurer will be available to the public. However, when more products enter the market, search costs increase. Thus, it is theoretically possible for more health insurance offerings to decrease utility.

A paper by Janssen and Moraga-Gonzales (2004) show that an the welfare impact of an increase in the number of sellers depends on the consumers’ search intensity. When consumers search with low intensity, having more firms will reduce search, will not affect expected price and will to greater price dispersion. One can think that inexpensive and/or infrequently purchased items would fall in to this category. When consumers search with high intensity, an increase in the number of firms will increase searching and decrease prices. In the 401(k) market, when employees are offered more than 10 choices, there are reductions in consumer responses (i.e.: less investment switching occurs).

Frank and Lamiraud aim to analyze how the number of firms affects insurance choice in Switzerland. They use a panel survey from the Federal Office for Social Insurance (OFAS). It contains a sample of 2152 individuals and asks about their insurance coverage between 1997 and 2000. Yearly premiums are available at the Federal Office for Public Health (OFSP) website. While there was consolidation in the health insurance market among firms, the average number of health plans offered per canton increased from 39 in 1998 to 52 in 2003. The number of people switching plans was 4.8% in 1997, 5.4% in 1998, 2.7% in 1999 and 2.1% in 2000.

How do people choose their plan?

…40% of people choose a health plan following their parents’ and friends’ choices, and what they see as tradition. Furthermore, as many as 25% individuals declare that they do not strive to pick the health insurance plan with the lowest premium. A substantial number of people explicitly report staying with their health plan based on habit (13.5%) or because they are satisfied with their arrangement (79%)

What do the authors conclude from the data?

First, we show that consumers that switch health plan pay 15% to 16% less in health insurance premiums per month holding ceteris paribus. Second, we show that among consumers expressing dissatisfaction with their health plans those in markets with fewer choices are more likely to express intent to switch. Finally, consumers that used an agent to help them purchase insurance consistently paid significantly lower premiums. This set of results suggests that “mistakes may have been made”.

Despite what orthodox economic theory states, no market is perfect. Understanding these market imperfection imparts important knowledge for economists and health policy makers.

• Richard Frank, Karine Lamiraud, 2008. “Choice, Price Competition and Complexity in Markets for Health Insurance,” NBER Working Paper #13817.

Adverse selection is often seen as a major impediment to the efficient functioning of insurance markets. Rothschild and Stiglitz (1976) create a model where high risk people buy full insurance while low risk individuals buy partial insurance. Yet empirically, one finds that in some insurance markets, low risk individuals purchase more insurance than high risk individuals.

An NBER working paper by Cutler, Finkelstein and McGarry (2008) claims that preference heterogeneity may explain this phenomenon. If low-risk individuals also have a stronger risk aversion preferences, than they may buy more insurance than a high-risk individual who has risk loving preferences. This work is an extension of the Finkelstein and McGarry (AER 2006) article discussed in one of my earlier blog posts.

Using data from the Health and Retirement Study (HRS), the authors measure risk tolerance using the following variables: smoking, having 3+ alcoholic drinks per day, job-based mortality risk, receipt of preventive health services, and seat belt usage. The authors find a negative relationship between individuals who engage in risky behavior (i.e., smoking, drinking, and those working in a high-risk occupation) and the percent who purchase various types of insurance, and a positive relationship between those engage in risk reducing behavior (i.e., preventive medical care, seat belt usage) and the purchase of insurance.

In the above chart, ‘+’ represents a positive statistically significant correlation, ‘-’ represents a negative statistically significant correlation, and o indicates that the relationship is not statistically significant.

The authors can also measure risk preferences based on respondents answers to income gamble questions. There is a weak relationship between risk preferences and risk behavior however.

The authors confirm that risk behavior lead to an increased probability of adverse events which would be covered by insurance. For instance, smoking increases mortality which would lead to an earlier life insurance payout. Increased preventive health activities decrease the probability of a nursing home stay.

Thus the authors conclude the following:

Our analysis yields two main findings. First, in all five markets, we find that individuals who engage in what are commonly thought of as risky behaviors (smoking, drinking, or prior employment in jobs with higher mortality rates) or who do not take measures to thought to reduce risk (preventive health activities or wearing of a seat belt) are systematically less likely to hold each of these insurance products. Second, we find that these same individuals tend to have higher expected claims for life insurance and long term care insurance, but lower expected claims for annuities; for Medigap and acute health insurance, there is no systematic relationship between the behavior measures and expected claims.

These results can help to explain the puzzle of insurance we started with: why is adverse selection not more common? In annuity markets, there is clear evidence of adverse selection: people who live longer are more likely to buy insurance. The standard adverse selection model is one explanation for this, but so is variation in risk tolerance; people who have less risky behaviors live longer and are more likely to buy annuities. In life insurance, our results suggest that differential risk tolerance can help explain why people with lower mortality rates have more insurance. Similarly, in the case of long-term care insurance, people who use more preventive care or are more likely to wear seat belts buy insurance more readily but also stay out of nursing homes.

• David Cutler, Amy Finkelstein, Kathleen McGarry. 2008. “Preference Heterogeneity and Insurance Markets: Explaining a Puzzle of Insurance,” NBER WP #13746.
• Finkelstein, Amy and Kathleen McGarry. 2006. “Private information and its effect on market equilibrium: new evidence from long-term care insurance.” American Economic Review 96(4): 938 – 958

Many people believe that better information technology (IT) can help improve the quality of medical care in the U.S. and around the world. For instance, if a doctor prescribes a drug which interacts harmfully with a drug the patient is already taking, a computer program could notify the doctor of this problem. If a patient has a certain disease, a computer program could help to narrow down the physician’s treatment options.

The problem with this top-down, one-size-fits-all doctoring is that every patient is unique. Of course, there are medical standards which should generally be followed, but often these standardized treatments have to be altered slightly or completely depending on patient wishes, the prevalence of multiple diseases, or economic constraints. Further, the physician typically has superior information regarding the patient’s condition than a standardized computer algorithm.

An NBER working paper by Javitt, Rebitzer and Reisman investigates a randomized trial of the implementation of a new IT system at a commercial HMO plan. The program was ‘turned on’ for some patients for each physician in the HMO and ‘turned off’ for other patients. The program would give three care consideration (CC) alerts: stop a drug, do a test, or add a drug. Each of these warnings had 3 severity levels.

The authors compare data on the CC alerts which occurred for the ’switched on’ patients with the hypothetical CC alerts which would have occurred if the software had been running. The authors found that the IT program lead to 6% lower charges in the turned on group compared to the control. The authors interpret this to be due to lower hospitalization rates and medication complications in the treatment group.

One key aspect of the program is that the CC alerts could be ignored by doctors.

“Physicians often have better information about their patients than does the error detecting software. Actions that look like a misstep to the computer may in fact be the result of informed physician choice, informed patient choice and/or patient non-compliance. For this reason, the HMO and the software company viewed CCs as recommendations that physicians were free to ignore if they disagreed.”

Physicians seemed to heed these CC alerts more than other studies had previously thought they would. Why?

Our results, however, suggest something more: the messages in the CCs seem to have had a bigger effect on physicians than the conventional medical channels used to promote the HOPE trial findings. It seems plausible that the CCs were influential because they linked a general recommendation (“take ACE inhibitors�?) to a specific patient [my italics] and a specific cite to the medical literature.”

To me, this seems to be the most appropriate way to implement medical IT.

• Javitt JC, Rebitzer JB, Reisman L, (2008) “Information Technology and Medical Missteps: Evidence from a Randomized Trial” Journal of Health Economics, 27(3): 585-602.
• Javitt JC, Rebitzer JB, Reisman L, (2007) “Information Technology and Medical Missteps: Evidence from a Randomized Trial” NBER WP #13493.