Adverse Selection in Insurance Markets

Why Risk Information Changes an Insurance Pool

Insurance pools work by collecting premiums from many people and using them to pay claims. The insurer’s expected cost depends not only on the claims each person might make, but also on which people choose to enroll. When individuals know more about their own risk than the insurer does, enrollment can change the pool’s risk mix: people expecting higher claims may be more likely to buy coverage at a given premium.

This pattern is called adverse selection. It can raise the average expected claim cost of the insured group, leading an insurer that must cover its costs to raise its premium. If the higher premium then causes lower-risk people to leave, the pool can become still more expensive. The calculations below focus on this enrollment-and-premium mechanism; they assume that coverage does not change claim behavior.

Definition: Adverse selection in an insurance market occurs when people’s private information about their expected risk affects whether they enroll, so the people who buy coverage differ systematically in risk from the broader population or the group the insurer initially expected to insure. A premium based on the wrong expected pool can therefore be too low to cover claims or high enough to discourage some potential enrollees.

Expected Claims and a Pooled Premium

Suppose there are risk types $j=1,\ldots,k$, and type $j$ has $N_j$ people. Let $X_j$ denote one person’s claim over the coverage period, and let $E[X_j]$ be that type’s expected claim. If everyone is enrolled and claims are paid in full, the average expected claim per enrollee is the weighted average of the type-specific expected claims. The weights are the types’ shares of the enrolled population.

\bar{C}=\frac{\sum_{j=1}^{k}N_jE[X_j]}{\sum_{j=1}^{k}N_j}

Assume the insurer charges a proportional loading $\lambda$ on expected claims to cover administrative expense and other costs included in this simplified model. The break-even premium is then $P=(1+\lambda)\bar{C}$. This is a model of expected cost, not a guarantee that realized claims in a particular period will equal their expectation. It also assumes that no other revenues or costs affect the calculation.

P=(1+\lambda)\frac{\sum_{j=1}^{k}N_jE[X_j]}{\sum_{j=1}^{k}N_j}

The insurer may not know each person’s risk type. In the examples, individuals know their own type and decide whether to enroll by comparing the quoted premium with a stated maximum amount they are willing to pay. That amount summarizes the person’s valuation of the coverage; it is given in each problem, not inferred from expected claims alone. A risk-averse person may value insurance above the expected claim, but willingness to pay is not identical to expected claim for every person.

Why the Pool Average Uses Enrollment Shares

The pooled cost formula follows directly from adding expected claims. If type $j$ has $N_j$ enrollees, the type’s total expected claims are $N_jE[X_j]$. Adding those totals gives expected claims for the whole pool. Dividing by the total number of enrollees converts that total into an expected claim per enrollee. Thus the weights are based on the people who actually enroll—not necessarily on the type shares in the original population.

\frac{\sum_jN_jE[X_j]}{\sum_jN_j} =\sum_j\left(\frac{N_j}{\sum_rN_r}\right)E[X_j]

The expression in parentheses is type $j$’s share of the insured pool. If a type leaves, its $N_j$ becomes zero for the recalculation. That changes the weights even though the expected claim of each remaining type has not changed.

Worked Example: A Pooled Quote and a Risk-Group Exit

A hypothetical insurer considers covering 1,000 people: 200 high-risk people and 800 low-risk people. Each high-risk person has a claim of $8,000 with probability 0.06 and no claim otherwise. Each low-risk person has a claim of $8,000 with probability 0.015 and no claim otherwise. Claims are independent across people. The insurer applies a 25% loading to expected claims. At the quoted premium, a low-risk person is willing to pay at most $220 and a high-risk person at most $700. Calculate the all-person premium, determine who enrolls at that quote, and recalculate if only high-risk people enroll.

Method and claim assumptions. For each type, the claim is a two-outcome random variable: $X=8{,}000$ with the stated probability and $X=0$ otherwise. There is no investment income or discounting in this one-period calculation. Expected claim equals claim amount multiplied by claim probability.

For a high-risk person, $E[X_H]=0.06(8{,}000)+0.94(0)=\$480$. For a low-risk person, $E[X_L]=0.015(8{,}000)+0.985(0)=\$120$.

If all 1,000 people enroll, the group’s total expected claims are $200(\$480)=\$96{,}000$ for high-risk people and $800(\$120)=\$96{,}000$ for low-risk people. The total is $\$192{,}000$, so the average expected claim is $\$192{,}000/1{,}000=\$192$ per person.

The loaded pooled premium is $(1+0.25)(\$192)=1.25(\$192)=\$240$. At this quote, low-risk people decline because $\$240>\$220$, while high-risk people accept because $\$240\leq\$700$.

If only the 200 high-risk people enroll, the average expected claim is $\$480$. The revised loaded premium is $1.25(\$480)=\$600$. High-risk people still accept, since $\$600\leq\$700$, so a high-risk-only pool is consistent with the stated enrollment limits.

Check and common mistake avoided. The initial average must weight the type-specific expected claims by the population shares: $0.2(\$480)+0.8(\$120)=\$96+\$96=\$192$. Adding $\$480+\$120$ without weights would give $\$600$, which is not the average claim per person. The recalculated high-risk-only premium also correctly uses only the people still enrolled.

Enrollment and Repricing

A quoted premium can change the group that buys insurance. If the insurer responds by repricing for the actual enrollment, the premium may change again. To analyze this process, calculate the premium for the assumed pool, compare it with each type’s stated willingness to pay, remove any type that declines, and recalculate using only the remaining enrollees. Then check that the remaining types still accept the revised price.

Specify expected claims by type.
Use the claim probability and claim amount for each type to calculate its expected claim.

Calculate a premium for the assumed pool.
Weight expected claims by the types’ shares among those assumed to enroll, then apply the stated loading.

Apply enrollment limits.
For each type, compare the premium with its maximum willingness to pay. A type whose limit is below the premium does not enroll.

Recalculate and verify.
Recompute the premium for the remaining pool and check whether each remaining type still chooses to enroll at that price.

This procedure is a simple way to test whether a proposed pool and premium are consistent with the stated choices. It does not imply that every insurance market adjusts prices immediately or that a unique outcome always exists. Rules on pricing, participation, and coverage can alter the result.

Worked Example: Repricing After Low-Risk People Decline

A hypothetical insurer considers a one-period policy for 600 low-risk and 200 high-risk people. Each low-risk person has a claim of $10,000 with probability 0.03; each high-risk person has a claim of $10,000 with probability 0.15. Otherwise, each person has no claim. Claims are independent. The loading is 10%. Low-risk people will pay at most $700, and high-risk people will pay at most $1,800. If the insurer first quotes a premium based on all 800 people, determine enrollment and the break-even premium for the resulting pool.

Method and claim assumptions. Claims have two outcomes: $10,000 with the stated type-specific probability, or zero otherwise. Expected claims are the claim amount times the probability. The premium includes the stated 10% loading, with no other cost or investment adjustment.

The low-risk expected claim is $0.03(\$10{,}000)+0.97(0)=\$300$. The high-risk expected claim is $0.15(\$10{,}000)+0.85(0)=\$1{,}500$.

For the initial all-person pool, total expected claims are $600(\$300)=\$180{,}000$ for low-risk people and $200(\$1{,}500)=\$300{,}000$ for high-risk people. The total is $\$480{,}000$ across 800 people, giving an average expected claim of $\$480{,}000/800=\$600$. The initial loaded quote is $1.10(\$600)=\$660$.

At $660, low-risk people enroll because $\$660\leq\$700$, and high-risk people enroll because $\$660\leq\$1{,}800$. Under the stated enrollment limits, all 800 people accept this quote. Therefore, if the insurer can maintain that quote for the coverage period, the all-person pool is consistent with the stated choices.

Now consider the repricing outcome if the insurer instead charges the break-even premium for the actual pool after low-risk people have declined—for example, because the initial quote was higher than their limit. A high-risk-only pool has an average expected claim of $1,500 and a loaded premium of $1.10(\$1{,}500)=\$1{,}650$. High-risk people still enroll, since $\$1{,}650\leq\$1{,}800$; low-risk people do not, since $\$1{,}650>\$700$. This high-risk-only pool is also consistent with the stated limits.

Check and common mistake avoided. The all-person premium is based on the weighted average $0.75(\$300)+0.25(\$1{,}500)=\$225+\$375=\$600$, not the unweighted sum of the two type averages. Also, the $660 quote calculated for the initial pool and the $1,650 quote for a high-risk-only pool describe different enrollment assumptions; they are not interchangeable.

Required Participation and Cross-Subsidies

Voluntary participation is not the only possible arrangement. If a rule requires everyone in a defined group to enroll and the insurer charges one common premium, the low-risk people cannot leave solely because their expected claims are below the pool average. The premium can then reflect the whole required pool. Relative to type-specific loaded expected costs, this arrangement transfers part of the cost from lower-risk to higher-risk people. Whether such a rule is desirable depends on broader policy objectives and constraints; the arithmetic alone does not settle that question.

Worked Example: A Common Premium With Required Participation

A hypothetical plan covers 900 low-risk people and 100 high-risk people, and participation is required for everyone in the group. Each low-risk person has a claim of $5,000 with probability 0.06; each high-risk person has a claim of $5,000 with probability 0.30. Otherwise, the claim is zero. Claims are independent, and the plan applies a 5% loading. Calculate the common premium and compare it with the loaded expected claim for each type.

Method and claim assumptions. For either type, an individual claim is the stated $5,000 amount with the stated probability, or zero otherwise. Expected claims equal probability times claim amount. Required participation means all 1,000 people remain in the pool for the premium calculation.

The low-risk expected claim is $0.06(\$5{,}000)=\$300$. The high-risk expected claim is $0.30(\$5{,}000)=\$1{,}500$.

Expected total claims are $900(\$300)=\$270{,}000$ for low-risk people and $100(\$1{,}500)=\$150{,}000$ for high-risk people, for a total of $\$420{,}000$. The average expected claim is $\$420{,}000/1{,}000=\$420$. Applying the 5% loading gives the common premium $1.05(\$420)=\$441$ per person.

The loaded expected claim for a low-risk person is $1.05(\$300)=\$315$, so that person pays $\$441-\$315=\$126$ above that type-specific amount. The high-risk loaded expected claim is $1.05(\$1{,}500)=\$1{,}575$, so a high-risk person pays $\$1{,}575-\$441=\$1{,}134$ less than that type-specific amount.

The group-level transfer amounts balance: $900(\$126)=\$113{,}400$ in total amounts paid above the low-risk loaded expected claims, and $100(\$1{,}134)=\$113{,}400$ in total amounts below the high-risk loaded expected claims. The comparison is a breakdown of the common premium, not a claim that each person’s realized claims equal their type’s expectation.

Check and common mistake avoided. The common premium uses all 1,000 people: $0.9(\$300)+0.1(\$1{,}500)=\$420$, then $1.05(\$420)=\$441$. Using only the high-risk expected claim would incorrectly treat the required pool as if low-risk people had left.

Common Mistakes and Why the Topic Matters

The premium and enrollment calculations show why an insurer cannot always take the population’s average risk as the insured pool’s average risk. The relevant composition is the composition after people respond to the price. If the insurer cannot observe types and voluntary enrollment is sensitive to the quote, a premium calculated from the full population may not describe the people who ultimately buy coverage.

Using unweighted type averages. Add the expected claims after multiplying by each type’s number of enrollees, then divide by total enrollment. Do not simply add the type averages.
Using population shares after enrollment changes. Recalculate shares from the people who remain in the pool. An exiting type has no weight in the new pool.
Assuming higher risk guarantees purchase. A high-risk type may still decline if the quoted premium exceeds its stated willingness to pay. Check every type’s enrollment condition.
Confusing risk classification with adverse selection. A premium based on each person’s known risk is different from a single premium based on a mixed pool. The adverse-selection mechanism arises when private information influences who enrolls at the available price.
Ignoring the model’s limits. These examples hold coverage and claim probabilities fixed. They do not model changes in behavior after insurance, uncertain expenses beyond expected claims, or rules that prevent repricing or require enrollment.

Key takeaway: To analyze adverse selection in insurance, calculate expected claims by type, weight those costs by the types actually enrolled, apply the pricing rule, and test whether each type still accepts the premium. A change in enrollment can change the pool’s average risk and its break-even price.

Check Your Understanding

For each problem, use the stated one-period claim model and enrollment limits. Assume the insurer applies the stated loading to expected claims and has no other costs or revenues.

A pool has 700 low-risk people and 300 high-risk people. Each low-risk person has a $4,000 claim with probability 0.05; each high-risk person has a $4,000 claim with probability 0.25. Claims are otherwise zero. The loading is 10%. Find the premium if all enroll. If low-risk people will pay at most $450 and high-risk people at most $1,200, who accepts that quote? (Answer: The all-person expected claim is $0.7(\$200)+0.3(\$1{,}000)=\$440$, so the premium is $\$484$; low-risk people decline and high-risk people accept.)
A group has 300 low-risk people whose expected claim is $500 each and 100 high-risk people whose expected claim is $2,000 each. The insurer loads expected claims by 8%. Low-risk people will pay at most $700, while high-risk people will pay at most $2,200. What is the all-person premium, and what is the high-risk-only premium if low-risk people decline? Check whether high-risk people accept the revised quote. (Answer: The all-person average is $(300(\$500)+100(\$2{,}000))/400=\$875$, giving a premium of $\$945$; low-risk people decline. The high-risk-only premium is $1.08(\$2{,}000)=\$2{,}160$, which high-risk people accept.)
A plan requires 900 low-risk and 100 high-risk people to enroll. Their expected claims are $300 and $1,500, respectively, and the loading is 5%. Calculate the common premium and each type’s loaded expected claim. How much above or below that type-specific amount does each type pay per person? (Answer: The common premium is $1.05(0.9(\$300)+0.1(\$1{,}500))=\$441$. The loaded type-specific amounts are $315 for low-risk people and $1,575 for high-risk people; the common premium is $126 above the low-risk amount and $1,134 below the high-risk amount.)
Low-risk people have an expected claim of $250 and a maximum willingness to pay of $440; high-risk people have an expected claim of $1,250. The insurer applies a 10% loading and quotes one premium based on a pool in which the high-risk share is $q$. What is the largest high-risk share at which low-risk people still accept the quote? (Answer: The premium is $1.10(\$250+\$1{,}000q)$. Setting this equal to $440 gives $\$250+\$1{,}000q=\$400$, so $q=0.15$, or 15%.)
A voluntary pool initially includes 120 low-risk people with expected claims of $400 each and 80 high-risk people with expected claims of $1,600 each. The loading is 5%. Low-risk people will pay at most $800 and high-risk people at most $1,800. Find the initial all-person premium, determine who accepts, then calculate and check the premium if only high-risk people remain. (Answer: The initial average expected claim is $(120(\$400)+80(\$1{,}600))/200=\$880$, so the premium is $\$924$; low-risk people decline and high-risk people accept. The high-risk-only premium is $1.05(\$1{,}600)=\$1{,}680$, which remains below their $1,800 limit.)

What You'll Learn