Information asymmetry means one side of a transaction knows more than the other. Sellers know their car's history. Employees know their own effort level. Insurees know their own risk. This imbalance causes markets to malfunction: good products get driven out, risky people get insured, and effort drops when unobserved.
The market for lemons
Akerlof's 1970 insight: if buyers can't distinguish good used cars from bad ones ("lemons"), they offer an average price. Sellers of good cars find this too low and withdraw. The average quality drops, so the price drops further. In the worst case, only lemons remain. Bad quality drives out good: adverse selection.
The lemons problem generalizes: whenever one side has private information before a transaction, the wrong types tend to participate. Health insurance attracts the sick. Loan markets attract risky borrowers. The uninformed side anticipates this and adjusts terms, which drives away more good types.
Scheme
; Adverse selection in insurance; High-risk people are more likely to buy
(define (expected-claim risk) (* risk 10000))
(define population
'((0.05"low-risk")
(0.15"medium-risk")
(0.30"high-risk")))
; Average risk if everyone buys
(define avg-risk
(/ (apply + (map car population)) (length population)))
(define premium (expected-claim avg-risk))
(display "Fair premium (all buy): $")
(display (exact->inexact premium)) (newline)
; But low-risk won't buy at this price
(define buyers
(filter (lambda (p) (>= (expected-claim (car p)) (* 0.8 premium)))
population))
(display "Who buys at $") (display (exact->inexact premium))
(display "? ") (newline)
(for-each (lambda (b) (display " ") (display (cadr b)) (newline)) buyers)
; New average risk is higher
(define new-avg
(/ (apply + (map car buyers)) (length buyers)))
(display "New average risk: ") (display (exact->inexact new-avg)) (newline)
(display "Adverse selection: pool gets riskier")
Moral hazard
After the transaction, the informed party may change behavior. Insured people take more risks. Employees with guaranteed pay shirk. The problem is hidden action, not hidden information. Solutions include deductibles, co-pays, monitoring, and performance-based pay.
Scheme
; Moral hazard: behavior changes after insurance
(define base-accident-prob 0.10)
(define loss-if-accident 10000)
; Without insurance: careful
(define careful-prob base-accident-prob)
(define expected-loss-careful (* careful-prob loss-if-accident))
; With full insurance: reckless (no personal cost)
(define reckless-prob 0.25)
(define expected-loss-reckless (* reckless-prob loss-if-accident))
(display "Expected loss (careful): $")
(display (exact->inexact expected-loss-careful)) (newline)
(display "Expected loss (reckless): $")
(display (exact->inexact expected-loss-reckless)) (newline)
; Solution: deductible
(define deductible 2000)
(display "With $2000 deductible:") (newline)
(display " Personal risk if reckless: $")
(display (exact->inexact (* reckless-prob deductible))) (newline)
(display " Personal risk if careful: $")
(display (exact->inexact (* careful-prob deductible))) (newline)
(display " Incentive to be careful: $")
(display (exact->inexact (* (- reckless-prob careful-prob) deductible)))
Signaling and screening
Spence's signaling model: the informed party takes a costly action to reveal their type. A college degree may not teach much, but it signals ability because it's cheaper for high-ability people to obtain. Screening is the reverse: the uninformed party designs a menu that makes types self-select (insurance deductibles, warranty options).
Wikipedia: Information asymmetry