Information Asymmetry, Adverse Selection & Moral Hazard

Complete vs. Incomplete Information

Complete information: every party knows all relevant facts about every alternative.
- Required for perfectly rational, utility‐maximizing decisions.
- Rare in reality; often information is merely “good enough.”
Information asymmetry (IA): one party possesses superior information relative to another.
- Creates scope for exploitation, market failure, or inefficient allocation of resources.
- Only problematic when the two sides’ incentives diverge.
- If both want the same outcome (e.g., friend fixes your car cheaply), IA is irrelevant.

Wants & constraints framework:
- Less-informed party’s constraint = ignorance.
- Better-informed party’s constraint = moral standards, reputation risk, legal risk.
Unequal knowledge lets the informed side obtain a larger share of surplus—or deter surplus from arising at all.
Leads to two canonical problems:
1. Adverse selection (pre-contract).
2. Moral hazard (post-contract).

Definition: IA about characteristics of goods/people causes some mutually beneficial trades to vanish.
Used-car (“lemons”) model:
- Two qualities: “plums” (high) vs. lemons (low).
- Buyers can’t distinguish quality ⇒ offer an average price P_{avg}.
- Lemons’ owners over-paid → eager sellers; plums’ owners under-paid → exit market.
- Vicious cycle → market increasingly full of lemons ➔ potential complete collapse.
Insurance example:
- Careless drivers know they’ll claim more ⇒ eagerly buy coverage.
- Insurer sets premiums on average risk \bar{r}.
- Higher \bar{r} ⇒ higher premium ⇒ careful drivers exit ⇒ \bar{r} rises again.
- Government fix: mandated participation (e.g., auto liability, Affordable Care Act).
Efficiency loss illustration:
- Seller’s reservation price = \$8{,}000 (plum); buyer would pay \$10{,}000 if certain.
- Efficient trade at \$9{,}000 yields \$1{,}000 surplus each.
- IA makes buyer cap bid at \$7{,}000 ⇒ no deal ⇒ \$2{,}000 total surplus destroyed.

Definition: after a deal, the protected party changes behavior because they don’t bear full cost.
Emerges in Principal–Agent settings.
- Principal = party assigning task (employer, insurer).
- Agent = party performing task (employee, insured).
Workplace: bosses can’t monitor effort; employees may slack off (coffee breaks, online games).
Insurance:
- Fully insured car owner may park carelessly ⇒ higher theft probability.
- Crop-yield insurance made farmers skimp on fertilizer; switch to rainfall-indexed insurance removes behavioral lever (weather can’t cheat).
Mitigation: monitoring, incentive contracts, co-payments/deductibles, reputation systems.

Stage	Hidden Factor	Classic Market
Adverse selection	Hidden type (risk, quality)	Used cars, who buys insurance
Moral hazard	Hidden actions after contract	Driving care, work effort

Less-informed side designs mechanism to extract private info.
- Job interviews, skill tests, reference checks.
- Insurance menus: low vs. high deductibles reveal expected accident frequency.

More-informed party voluntarily conveys information via costly, hard-to-fake signal.
- Certified pre-owned warranty.
- College diploma → signals intelligence, grit, conformity (Bryan Caplan’s view).
- Cost element (tuition, dealership inspection) ensures credibility.

Repeated interactions create an inter‐temporal opportunity cost of cheating.
- eBay feedback, Yelp stars, Better Business Bureau.
- Bad reviews deter future customers ⇒ incentive to forego one-time rip-offs.

Decision maker infers missing info from group averages.
- Choosing Mexican restaurant in “great-Mexican” neighborhood.
- Auto insurers charge young males more because group avg. accident risk ↑.
Ethical/legal concerns: can entrench inequities (racial profiling, employment bias). Often regulated.

Disclosure & Education
- FDA nutrition labels, mortgage APR sheets.
- Risk: information overload; fine print unread ⇒ policy failure (sub-prime mortgages).
Mandated participation
- Auto insurance, health insurance individual mandate.
- Stops adverse-selection spirals; moral-hazard remains.
Information restrictions
- Genetic Information Nondiscrimination Act (GINA) bans insurers from using DNA data.
- Balances fairness vs. insurers’ desire for accurate risk pricing.

$50 “fabulous” tablet scam – perfect metaphor for IA.
Mechanic brake replacement scenario – classic hidden-information abuse.
Hanumanthu’s drought & rainfall insurance – index insurance beats yield insurance.
College wage premium: grads earn 73\% more than HS; degree as signal.
Gun-purchase age limits: statistical discrimination debate (18–21 rifle ban vs. freedom/rights).

Society trades off efficiency vs. fairness: allowing group-based pricing can be efficient yet discriminatory.
Regulation must balance too little and too much information.
Mandates restrict choice but can keep vital markets alive.
Developing human capital (education) yields both knowledge and marketable signals.

Average price under IA: P{avg}=\Pr(\text{plum})\cdot P{plum}+\Pr(\text{lemon})\cdot P_{lemon}
Surplus lost via IA: \Delta S = (CS+PS){complete} - (CS+PS){asymmetric}
Expected claim for insurer: E[Claim] = \sumi pi \times Li where pi = probability driver $i$ has accident, L_i = loss.

IA triggers adverse selection (hidden type) & moral hazard (hidden action).
Market solutions: screening, signaling, reputation, deductibles.
Group-based inference = statistical discrimination (efficient but contentious).
Government fixes: disclosure, education, mandatory buying, info restrictions.
Always weigh marginal benefit of info vs. opportunity cost of acquiring it.