Case Study 10-2: The Insurance Paradox — When "Irrational" Is the Rational Choice

Chapter: 10 — Expected Value: How Rational People Think About Risk Theme: Why buying insurance is mathematically negative EV but psychologically and practically rational — and what this reveals about the limits of expected value as a complete decision framework

Priya's Dilemma

It was Priya's third month of job searching, and she was staring at a benefits enrollment form.

She'd finally started doing contract work — freelance content strategy gigs that paid reasonably but came with no benefits. The question in front of her was whether to buy her own health insurance through the ACA marketplace at $380/month, or to go uninsured for a few months while she kept her expenses low.

The math seemed to favor skipping it. She was 22, healthy, rarely sick. The probability of a major health event in the next three months felt vanishingly small. $380/month × 3 months = $1,140. Was she really likely to use $1,140 worth of healthcare services?

She texted Dr. Yuki: "Is buying health insurance when you're young and healthy actually irrational? The EV seems negative."

Dr. Yuki replied with a single line: "The EV in dollars is negative. The EV in utility is positive. Come to office hours tomorrow and we'll talk about why those are different things."

The Basic Math: Insurance Is Always Negative EV in Dollars

Let's establish the baseline. Health insurance (and most consumer insurance) is, by design, negative expected value for the buyer.

Here is why: insurance companies must cover their operating costs, employee salaries, profit margins, and administrative overhead. They do this by charging premiums that exceed the expected payout to any given customer. This isn't predatory — it's structural. If insurance companies paid out more than they collected, they would cease to exist.

A simplified example:

Priya's situation: - Monthly premium: $380 - Probability of needing a major medical service in one month: approximately 3% (rough estimate for healthy 22-year-old) - Average cost of a major medical event (ER visit, imaging, hospitalization): $8,000

Expected value of one month without insurance: EV = (0.97 × $0 medical cost) + (0.03 × −$8,000) = −$240 per month

Expected value of one month with insurance: EV = −$380 per month (premium, regardless of medical events) + (expected value of coverage if something happens)

If we assume the insurance covers the full $8,000 when a major event occurs: Net EV with insurance = −$380 + (0.03 × $8,000) = −$380 + $240 = −$140 per month

So: - Without insurance: Expected cost = $240/month - With insurance: Expected cost = $380/month

Insurance costs Priya an extra $140/month in expected value. The insurance company is capturing that $140 as operating margin.

By pure expected value in dollars, Priya should skip the insurance.

And yet — she shouldn't.

Why Dollar EV Misses the Point: The Utility Asymmetry

The problem with dollar EV in insurance analysis is that losing $8,000 uninsured doesn't *feel* like losing $8,000. It functions like losing far more.

For Priya, who has $6,000 in savings and $28,000 in student loan debt: - A $380 monthly premium is painful but manageable. It represents 6.3% of her monthly gross income. - An $8,000 uninsured medical bill is catastrophic. It would wipe out her entire savings, push her into credit card debt, and potentially disrupt her job search at its most critical stage.

Utility functions, as Bernoulli observed in the eighteenth century, are concave — meaning each additional dollar lost hurts more than the previous one when you're already stretched. The move from "comfortable" to "financially stressed" is a far larger utility loss than the move from "comfortable" to "slightly less comfortable."

Let's try to quantify this:

Dollar losses vs. utility losses for Priya:

When you run the EV calculation in utility rather than dollars:

Expected utility cost without insurance: = (0.97 × 0 utility loss) + (0.03 × 65 utility points) = 1.95 utility points/month

Expected utility cost with insurance: = 3 utility points/month (known, manageable premium cost)

Wait — now insurance does look slightly worse in utility terms (3 vs. 1.95 per month), but only for a three-month window. The critical factor is what happens at the tail:

Without insurance, the worst-case scenario is 65+ utility points of loss. With insurance, the worst case is 3 points of loss per month — a known, bounded cost. That certainty itself has value that pure EV (even in utility terms) doesn't fully capture.

This is the concept of risk premium — the amount a rational, risk-averse person will pay above the expected loss to eliminate variance. For someone with a concave utility function, reducing variance is genuinely worth paying for, even in utility-EV terms.

The Kelly Criterion Lens: Variance Threatens Survival

Return to the Kelly Criterion. Kelly says optimal bet sizing considers not just EV but the probability of ruin.

For Priya's situation, "ruin" doesn't mean literal bankruptcy. It means a financial shock severe enough to derail her job search, force her to take any available job out of desperation rather than making a strategic choice, and add months of debt-servicing pressure to an already difficult transition period.

The probability of financial ruin without insurance over a three-month period: - ~3% chance per month of a major event × 3 months ≈ ~9% cumulative probability - At 9% probability of catastrophic disruption, going uninsured is accepting a near-one-in-ten chance of financial derailment

Kelly thinking says: if this bet can bankrupt you relative to your position in a game you care about winning (the job search), the EV calculus changes dramatically. You should pay to reduce variance even at a cost to expected value, because you cannot afford to go bankrupt at this stage.

Generalizing the Lesson: When Insurance Logic Applies

The insurance paradox extends far beyond health coverage. The underlying principle — pay a known, manageable cost to avoid an unknown, catastrophic one — applies in multiple domains:

Career Insurance

Priya was also offered an expanded job-search strategy by a career coach at $200 for a package of resume reviews and interview prep. She could skip it and save the money.

But the job search is itself an insurance problem: - Expected cost of the coaching: $200 (certain) - Expected benefit: uncertain, but potentially months off the job search timeline - Cost of an extended job search: real income loss, psychological stress, growing desperation

The coaching fee was career "insurance" against a longer, more damaging search. By the logic above, even if the coaching had a negative dollar EV (which it may not — good coaching genuinely helps), the reduction in variance and the worst-case duration of the job search could make it rational.

Emergency Funds as Self-Insurance

Financial advisors consistently recommend keeping 3–6 months of expenses in liquid savings, even though cash earns minimal returns and the "expected" outcome is that you'd have done better investing that money in the market.

The emergency fund is self-insurance: a known cost (foregone investment returns, roughly 7–10% per year) in exchange for reduction of catastrophic downside (being forced to sell investments at a low, take out a high-interest loan, or go into credit card debt during an emergency).

Under the same utility analysis: the foregone returns have low utility cost because they're foregone gains (never felt as real money). The catastrophic borrowing has high utility cost because it's a real crisis. The asymmetry makes self-insurance rational.

Startup Insurance: Not All Eggs in One Basket

Marcus faced a version of this when considering whether to put all his development resources into a single new feature. The insurance logic said: diversify. Build a small version of the feature first. If it fails, the cost is a few weeks, not the entire product roadmap.

This is variance reduction through staged betting — a structural form of self-insurance that maintains optionality while limiting catastrophic downside.

Research Spotlight: Prospect Theory and the Pain of Loss

Kahneman and Tversky's 1979 prospect theory provides the empirical backbone for why the insurance paradox is real, not just theoretical.

Their finding: losses feel roughly twice as painful as equivalent gains feel pleasurable. This is called loss aversion. The utility loss from losing $8,000 is approximately twice the utility gain from receiving $8,000 as a gift.

When this asymmetry is properly incorporated into the utility function, insurance becomes rational for anyone with: 1. Significant exposure to catastrophic loss 2. Limited capacity to absorb that loss (savings below the potential loss amount) 3. High sensitivity to financial disruption (e.g., a job search, a startup runway, school enrollment)

Notably, loss aversion also means that people over-insure in some domains and under-insure in others. Extended warranties on $200 electronics are typically negative EV in utility terms as well as dollars — the "catastrophe" of a broken microwave is genuinely not catastrophic. But health, housing, and income protection involve potential losses that are genuinely catastrophic in utility terms, making insurance rational even at negative dollar EV.

The Twist: When Insurance Becomes Irrational

Having established that insurance is often rational despite negative dollar EV, we should also acknowledge when it's not:

Small-Ticket Extended Warranties

An extended warranty on a $15 pair of earbuds is almost certainly negative EV in both dollars and utility. The worst-case scenario (losing $15) doesn't threaten survival, doesn't require bankruptcy-level debt, and doesn't disrupt any life domain. There is no catastrophic tail to insure against. You are paying a significant premium (relative to the product value) to avoid a manageable, small loss.

The utility asymmetry that justifies health insurance does not apply here. The rational choice is to self-insure — to mentally set aside the possibility of replacing the earbuds and accept the variance.

Insurance Against Tiny Probabilities of Minor Losses

Travel insurance for a $300 domestic flight (when you have sufficient savings to absorb a missed flight) is similarly questionable. The loss is real but not catastrophic. The premium is proportionally high. The variance you're eliminating is not survival-threatening.

Over-Insurance Through Risk Aversion Bias

Kahneman and Tversky's research also showed that people often over-insure because of loss aversion, not because insurance is objectively rational. The mere framing of a loss (vivid catastrophe narrative) triggers disproportionate willingness to pay to avoid it, even when the utility calculation would not support that premium.

A self-aware decision-maker asks not just "does this catastrophe feel awful?" but "does the dollar premium genuinely justify the utility benefit of avoiding this specific risk?"

Priya's Final Decision

In Dr. Yuki's office hours, Priya worked through the analysis.

"The dollar EV says skip the insurance," Priya said. "But the utility EV probably says buy it."

"What's the tipping point?" Dr. Yuki asked.

"Whether I can actually absorb an $8,000 hit without derailing the job search."

"And?"

"I can't. My savings wouldn't survive it. And the job search is the whole game right now. I'd be taking on debt at the worst possible time, probably taking any job I could get instead of waiting for the right one. The insurance is cheap relative to that risk."

"What did you just do there?" Dr. Yuki asked.

Priya thought. "I... ran the Kelly Criterion? I assessed whether the downside threatens my survival in this context."

"Exactly. You used EV correctly — not as a formula to maximize dollar return, but as a framework for thinking about what outcomes actually mean to you at this specific stage of your life." Dr. Yuki leaned back. "The insurance paradox isn't really a paradox. It's a reminder that expected value is a tool for thinking, not a substitute for it. The math tells you the average. Your job is to translate the average into your actual life."

What the Insurance Paradox Tells Us About Luck

The insurance paradox has three implications for the science of luck:

1. Variance management is a legitimate strategy. People who consistently make "lucky" decisions often do so partly by eliminating the catastrophic tail risks that can derail them. Insurance, emergency funds, diversified portfolios, and staged experimentation are all forms of variance management that preserve the ability to keep playing.

2. Rational behavior is context-dependent. The "rational" choice for Priya is not the same as the rational choice for someone with $500,000 in savings. Utility-based rationality is personal and situational. Anyone who tells you "the math says X is always the right choice" without asking about your utility function is telling you less than half the story.

3. The goal is not to maximize expected value — it's to maximize expected utility. This distinction sounds subtle but it changes everything. Maximizing dollar EV leads to taking every positive-EV bet regardless of variance. Maximizing utility EV leads to taking positive-EV bets where the variance is appropriate for your situation — and buying protection against the ones that aren't.

Luck science isn't about taking every positive EV bet. It's about building a life architecture where variance works for you rather than against you — where the risks you absorb are ones you can actually survive, and the rewards you seek are ones your current position can actually capture.

Priya bought the insurance.

Discussion Questions

1. The Self-Insurance Question. At what savings level do you think Priya would be rational to skip health insurance? (In other words: at what wealth level does the utility argument for insurance weaken?) Defend your answer with reference to the utility curve concept.

2. Extended Warranty Rationality. A store clerk offers you a 2-year extended warranty on a $400 laptop for $80. The laptop has a 12% chance of failure in years 2–3, with an average repair cost of $250. Is this warranty positive or negative EV in dollars? In utility? Should you buy it? (Your answer may depend on your personal financial situation — state your assumptions.)

3. Insurance and Risk Tolerance. Some financial advisors say young people should skip term life insurance because they have no dependents and healthy risk tolerance. Others say they should buy it because it's cheapest when you're young and healthy. Apply the utility framework to adjudicate this debate.

4. Startup Insurance. How might Marcus apply the insurance principle to his startup's operational decisions? Identify at least three types of "insurance" (not necessarily purchased financial products) he could buy for his company — and evaluate whether each is justified given the utility analysis.

5. The Systemic Question. If individual people rationally buy insurance (paying more than expected payout), and insurance companies rationally sell it (collecting more than they pay out), who "wins"? What does this tell you about how risk is redistributed in society — and whether that redistribution is luck-increasing or luck-decreasing at a systemic level?

Key Takeaway

Insurance has negative expected value in dollars but positive expected value in utility — and this distinction is not a flaw in the framework, it's the framework working correctly. The lesson is that rational decision-making requires utility-based EV, not dollar-based EV. For any decision involving potential catastrophic loss, variance matters independently of the average outcome. Smart luck management is not just about maximizing expected return — it's about structuring your exposure to risk so that you remain in the game long enough for the law of large numbers to work in your favor.