Behavioral Economics Confounds Health Insurance Design and Pricing

By Roy Goldman

Health Watch, March 2025

An earlier version of this article appeared in the January and April 2023 issues of The Wharton Healthcare Quarterly under the title “Behavioral Economics in Consumer Purchases of Health Insurance.”

When designing and pricing a health care product, not only is it important to know how to model frequency and severity, but it is essential to understand human behavior. Of course, the issue of anti-selection in medical insurance is well known, as individuals know more about their health and prior expenditures than an insurer does. This type of behavior is certainly rational and to be expected.

However, one must also anticipate behavior that would seem to be financially suboptimal from an actuarial perspective. There are three main drivers of this behavior. First is the preference for loss avoidance. Many individuals will pay extra to avoid a possible, but unlikely, large loss in favor of a more likely small gain.[1] Second is the difficulty most people have in understanding statistical thinking, especially situations involving conditional probabilities. Third is the reluctance to change.

From my career as a health insurance actuary, I’ve chosen five examples that show the importance of anticipating consumer behavior when designing and pricing health insurance products.

Reluctance to Switch Carriers in a Group Insurance Environment

Only a small minority of individuals change carriers when purchasing individual coverage during Medicare or Affordable Care Act (ACA) open enrollment periods, which is understandable because the coverages are complicated and hard to compare. As a new Fellow at the Prudential Insurance Company of America (Pru), I learned there’s another reason why individuals are reluctant to change.

In October 1980, Pru earned a major endorsement by the American Association of Retired Persons (AARP) to market the association’s medical plans to its members, as well as to members of the National Retired Teachers Association (NRTA).

In the late 1950s, Colonial Penn Insurance Company was founded to provide medical insurance to NRTA because no existing health insurer thought “seniors” were insurable. Subsequently, Colonial Penn created AARP; solicited new members to both organizations; and sold them indemnity health insurance, life, and property and casualty (P&C) products.

Typically, in group insurance the winning carrier acquires all members from the previous carrier. But in this instance, Pru only won the right to compete with the prior carrier. Between May and July 1981, policyholders had three opportunities to choose a carrier for their current plans. The plans were identical and easy to understand. What do you think happened?

We were pleased that two-thirds of the members chose Pru, but both companies were surprised to see that Colonial Penn’s one-third share generated two-thirds of the claims. Subsequent analysis revealed that individuals who had claims in the past were more reluctant to change carriers. Interestingly, in the absence of a coordination of benefits clause, some members enrolled with both carriers and received double payments.[2]

Product Selection in Medicare Supplements

I was transferred to Pru’s AARP office in January 1981. On my first day, I was asked to design new Medicare Supplement plans that met the minimum standard as defined by the Baucus Amendment (adopted by Congress in June 1980). The SOA syllabus had taught me about anti-selection. When given obvious choices, consumers in better health choose the less expensive and comprehensive option. Claims experience in these plans tends to be better than average, while the experience in the more expensive plan tends to be worse. As the plans are re-rated in future based on experience, the premium differentials increase, causing the more expensive plan to lose all but the least healthy members.

We tried to avoid that mistake by designing a hierarchy of four plans that filled an increasing number of Medicare gaps, but we also added desired non-Medicare benefits—such as vision or private duty nursing—to the lower-priced plans. Plan choice became less obvious, and those plans remained pretty much intact until new legislation in 1992 created 10 standard plans. In my opinion these standard plans err by building on one another in obvious ways, which leads to anti-selection.

For example, the only difference between standard plans A and B is that B covers the inpatient hospital deductible (IHD). Yet, in my zip code, even though the national average annual hospital admission rate is about 30% for people 65 and older,[3] the difference in the annual premium between A and B for some insurers is nearly 100% of the IHD. Clearly, plan B members must be incurring greater claims in all coverages than plan A members.[4]

Plans F and G illustrate the same phenomenon. Their high-deductible counterparts, F* and G*, pay the same benefits, but only after the insured reaches an annual out-of-pocket deductible for covered Medicare expenses ($2,800 in 2024). In my zip code, the annual difference in premium (by the same insurer) is $3,720 between F and F* and $3,108 between G and G*. Thus, the difference is greater than the annual deductible the insured is trying to avoid.

In the absence of medical underwriting, who is choosing plan B over A or plan G over G*? I suggest three scenarios. Each one contributes to adverse selection:

1. Less healthy enrollees whose costs will exceed the deductible.
2. Risk-averse enrollees willing to pay more to avoid a deductible, some of whom will feel entitled to use more services.
3. Enrollees who lack sufficient savings to cover an unplanned expense. For them, avoiding the deductible is a rational decision if they have the cash flow to afford the additional monthly premium. Insufficient savings, though, may have forced some to avoid getting medical advice or annual tests in the past; therefore, when problems arise now, they may be more severe

Product Selection in Open Enrollment with Multiple Carriers

Selection can drive huge differences in premium even when the benefit differences are relatively small. A great illustration is in the design of plans offered to state and public employees in Missouri circa 2003. Insurers were asked to submit premium rates for two plan options, Low and High. Members could choose either plan from any approved carrier. There were only two small differences between the plans: (1) the Low plan had a $15 physician copay vs. $10 for the High; and (2) copays for prescription drugs were $5 to $10 higher with the Low plan.

If priced separately, assuming everyone was enrolled in that plan, the difference in actuarial value was only about $8 a month. However, when pricing the two plans with member selection, we assumed that members choosing the High plan would have greater utilization of all services. We strove to be the lowest bidder on the Low plan and the highest bidder on the High plan. We ended up with a majority of the Low-plan members, who had fewer claims than expected, while the carriers with more attractive High-plan premiums incurred much greater than the expected claims. Within two years, members faced a $100 monthly premium difference between the Low and High plans. Eventually, Missouri moved all members to one plan.

Product Selection in an Employer Group Plan

About five years later, I observed a similar situation at Geisinger Health System. Employees were covered by the system’s own insurer, of which I was the chief financial officer. Historically, employees were offered a choice between a high and low option plan. The benefit differences were significant enough that employees choosing the high option had greater plan contributions.

I asked my actuarial team to build a model for Human Resources to show the likely consequence of continuing to raise the high-option contributions to match the worsening experience. As the required high-option contributions increase, the healthiest individuals in the high-option plan switch to the low option, leaving the high-option pool with even greater expected costs per insured. At the same time, the average health of the low-option plan also deteriorates, but not as much as that of the high option. As the cost differentials increase in the following year, the situation repeats. The healthiest in the high option move to the low option. The expected costs per insured rise in both plans, with the high-option plan increasing more than the low option. Observing how the situation eventually becomes untenable, Human Resources switched the system to one insurance pool with one plan and one employee contribution rate for everyone.

Product Selection in Medicare Prescription Drug Plans

Since Medicare does not cover retail prescription drugs, members without coverage through an employer or a Medicare Advantage plan, can purchase a prescription drug plan (PDP) from a private insurer under Part D of Medicare. This program began in 2006, and from the very beginning, members’ choices have led to unanticipated results.

My experience at Humana showed that Medicare members with the fewest prescriptions choose the lowest priced plans, while members with many prescriptions choose higher cost plans even though they could receive similar benefits in cheaper plans.

Members can easily go online at Medicare.gov, input their prescriptions, and receive a list of plans available in their zip code in descending order of total out-of-pocket costs (i.e., premium plus copayments). But, as observed earlier, insureds are reluctant to switch carriers when they have received good claim service from their current carrier. In addition, the lowest priced plans have preferred pharmacy networks that may not be convenient to all members. Still, an excellent topic for a research paper would be why so many insureds are not purchasing the most economically prudent plan.

In 2006, the first year of the PDP program, Humana offered the lowest priced plan. It garnered a huge market share and had a favorable medical loss ratio. When I joined Humana in 2010, it no longer had the lowest priced plan, and its market share had decreased. When a new lowest priced plan was introduced using Walmart as the preferred network, Humana again experienced significant growth with favorable claim experience. Indeed, our models implied that the lower the price, the more favorable the expected experience but with some sacrifice of market share (assuming some buyers would be wary of an extremely low price).[5] Humana’s strategy has been copied by other carriers, so it now seems that every year there is a different carrier with the lowest priced plan.

In 2008, Humana showed how difficult it is to anticipate member behavior correctly. To attract healthier members, it lowered copayments on tier 1 and tier 2 generic drugs and assumed that current members would not shift from tier 3 and 4 brand drugs with higher copayments. However, that assumption proved faulty, as members switched to the lower tiered drugs, leaving the company with a higher share of the costs than expected.[6]

Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries, the editors, or the respective authors’ employers.

---

Roy Goldman, Ph.D., FSA, MAAA, CERA, is a retired actuarial executive. Roy can be reached at roygo@earthlink.net.

Endnotes

[1] Paradoxically, individuals with a likely small loss are willing to gamble on getting a large gain. See, for example, Daniel Kahneman, Thinking, Fast and Slow (New York: Farrar, Straus and Giroux, 2013), or Michael Lewis, The Undoing Project: A Friendship That Changed Our Minds (New York: W. W. Norton, 2017).

[2] Prudential, however, did not make a windfall profit as its winning bid guaranteed AARP at least a 75% loss ratio and an expense ratio below 25%. Refunds to AARP based on these guarantees reduced premium rates in future years.

[3] U.S. Infectious Disease Statistics, “Hospital Admission Rates by Age and Sex 2021,” Ministry of Health Singapore, Mar. 18, 2022, https://www.moh.gov.sg/others/resources-and-statistics/healthcare-institution-statistics-hospital-admission-rates-by-age-and-sex-hospital-admission-rates-by-age-and-sex-2021.

[4] Members who can satisfy underwriting requirements and want to avoid the deductible can purchase lower-priced B plans, but the likelihood is that the premium difference between plans A and B will grow.

[5] Due to the risk corridor program, insurers did not retain all the earnings when the medical loss ratio (MLR) was much lower than expected. Nor did they suffer extraordinary losses if the MLR was unfavorable, as the federal government shared in the gains and losses.

[6] Russ Brit and Dinah Wisenberg Brin, “Humana’s Net Sinks Amid Portfolio Woes,” The Wall Street Journal, Oct. 28, 2008, https://www.wsj.com/articles/SB122510550123371629?mod=Searchresults_pos1&page=1.