Insurance

HPR 501: Economics of Health Care Delivery

Mark Pauly
 Insurance: rational and behavioral
• Reference: Kunreuther and Pauly. Insurance and Behavioral
Economics: Improving Decisionmaking in the Most
Misunderstood Industry
• Rational: Buyers and sellers of insurance against any risk
make choices to maximize expected or average utility
• Behavioral: Anything else.
• What we observe: “Many people” buy insurance when they
should not, do not buy insurance when they should, choose
too much insurance coverage, choose too little insurance
coverage, choose the wrong insurance coverage, overclaim
on their insurance, underclaim on their insurance
• Collectively chosen (government) insurance plans “many
times” do the same thing.
• But some of the people do the rational thing all of the time
and most of the people do the rational thing some of the time.
 Rational or behavioral?

If your car is more than 4 years old…. When it needs expensive repairs—and
it will—you will be grateful you bought Car Shield insurance!
 The Rational Value of Insurance
Assume that insurance does not itself affect the probability or amount of
“loss.” (We will modify this later.) Treat bills for medical care as a
random loss that threatens wealth, and wealth is all the matters.

Insurance moves money from one prospective “state of the world” to the
other. Compared to no insurance, buying a policy lowers your wealth
in the no loss state and raises your wealth in the loss state. “Fair”
insurance (premium 50 cents) on a 50-50 risk of losing $1 changes
wealth in both states by 50 cents. You end with the same wealth
regardless—so risk disappears.

This will increase your utility if the marginal value (utility) of the dollars
you give up in no loss state is less than marginal value of dollars you
gain in the loss state: The big winner (not always the rich guy) buys
the drinks. You are “risk averse.”

You will be willing to pay something more than fair for this service if
you experience diminishing marginal utility of income/ wealth. You
will pay more the greater the divergence between the different MUs

Why don’t widows buy nursing home insurance? Low MU of money in

the sick state relative to well state? Why don’t people (always) buy
coverage for preventive care? No risk. Risk neutral for small losses?
 Insurance and Health Reform
• Health reform should ideally correct three things:
no insurance, underinsurance, and
overinsurance.
• Actual ACA bill tried to do so with subsidies and
mandates for the uninsured and the “Cadillac tax”
on high cost health plans.
• Probably (insufficient) steps in the right direction,
depending on imperfect implementation
 Two Views of “Health” Insurance
• The financial view: health insurance
protects the household‘s wealth against
unusually high medical spending.
• The access view: insurance increases the rate at
which people will choose to consume medical
services that will improve their health
• Tradeoff: the financial view motivates voluntary
insurance buying. However, the access view
motivates (most) policy interest. Better access
may threaten finances through higher premiums.
W(c)ould the average person buy the insurance
that leads to cost effective outcomes?
 In econ theory, why non-poor people do and
do not demand conventional health
insurance.
They buy insurance because…they prefer a small certain
payment to the risk of a large loss
and….they cannot afford a costly but
rare expense. Other reasons?
They do not buy because…
1. At the premium they face, insurance is a bad deal
because of…
 administrative expenses (loading)
 imperfect risk rating (adverse selection)
2. Insurance causes higher expenses (moral hazard)
3. Not because they cannot afford it.
 “Affordability” and Insurance
• We do not have a technical definition of
"affordability.“ What does it mean to
you?
• People say they cannot afford insurance.
• But then they can even less well afford to be
uninsured
• Unless they expect someone to help
them out with big uninsured costs or
needs
 Why People Buy Health Insurance
Assume probability of loss (illness) is 0.5, and treatment cost is given at L.
U(YN)

U(YN - .5L)
A
EU B

U(YL)

YL YN - .5L YN

AB is the risk premium or the maximum “loading” you would pay.

Maximize: EU = pU(YI) + (1 - p)U(YN), where p = probability of loss.
With higher loading, one is less likely to buy coverage.
Numerical Example of Pooling, Or “Law of Large Numbers

loss probability = 0.5, loss amount = $1,000

Risk faced by Crusoe or Friday
No Pooling
Event Probability Loss/Person
Nc or Nf 0.5 -0-
Lc or Lf 0.5 1,000
2-person Pool
Nc, Nf 0.25 -0-
N c, Lf 0.25 500
Lc , Nf 0.25 500
L c, Lf 0.25 1,000

Probability of losing $1,000 or $0 is cut in half, probability of losing $500 is 0.5.

Loss/person will converge to $500 with minimal variance as number of pool members:
grows. Probability that average loss differs from $500 can be made miniscule.
Insurance is costless, requires no difference in tastes, but does require uncertainty (but not
really homogeneity).
Implication:
1. Can you gain competitive advantage by how you pool risks?
• Size • Underwriting • Pricing
2. How “rich” can health insurance companies be? Not very, no special talent needed.
What if the loss probabilities differed?
• There would still be gains from pooling
• But the premium would have to be higher for the higher
risk person for both to be willing to buy coverage.
“Pooling” reduces risk by providing coverage but it need
not average risk by varying premiums. Steven Brill
(America’s Bitter Pill) is wrong again.
• Takeaway: efficient insurance pools the risk of what is yet
to happen, not the “risk” that has already occurred.
Risk pooling is voluntary and efficiency improving. Risk
averaging may be equitable but must be compelled so is
not necessarily efficiency improving.
Willingness to pay for insurance is greater, for a
given fair premium, the higher is the loss and the
lower is the probability of loss.
FB GB

FA
GA

YL YL
B
YN  P YN
A

If probability is pB  1/ 2, WTP is FB B (Loss is YN - YL )

B

G
If probability is pA  1/3, WTP is FAG A (Loss is YN - L )
A
 What is the price of insurance?
• Answer
– The premium is not the price. The price is
the “loading” per dollar of coverage. Reform’s
upper bound on MLR incorporates this.
– Loading = premium - expected benefits
• Illustrative example
– Assume: Average “insurable” expense for 20-25
year olds = $1000 per year; in-patient hospital
expense is 30% of total.
 If you knew you were an average
risk, which policy would you prefer?
• A
– Covers in-patient expense only
– Premium = $500
• B
– Covers all types of expenses
– Premium = $1100

Insurance is more like banking than

widget making.
 Definitions and Propositions in the
Theory of Insurance
• Expected value of a loss L with probability p=pL.
• (Actuarially) fair premium: premium equals
expected loss.
• A risk averse person will buy exactly full coverage
of all risky losses if premiums are fair and
expected loss is fixed. No “Aflac!”
 An Example of Moral Hazard: Impacts on Level of
Insurance Coverage or/and Managed Care Standards
Assumptions Two illness states possible:
“moderate” and “severe”
PM = PS = ½

The key problem: Insurer cannot tell which type of illness occurred
Two basic polar types of insurance are possible:
“fee-for-service” with cost sharing but no quantity limits
“HMO” with no cost sharing but quantity limits

Gross Price = 1
Points on Demand Curves
QM = (Full price) = 50
QM = (P=0) = 150
QS = (Full price) = 100
QS = (P=0) = 300
With no insurance, expected expense is:
0.5(50) + 0.5(100) = 75

Fair premium for full coverage insurance is:

0.5(150) + 0.5(300) = 225
Propositions: Risk averse person would prefer to pay
a 75 premium to facing a prospect with expected
loss of 75. But may prefer no insurance to paying
225 for full coverage. Why? You get 225 worth of
care. Ans: It’s not worth 225; it’s worth 75 +
0.5(150), or 150. There is a “welfare cost” of 75.
 D Moral Hazard
M DS

DM B A
P P = MC

Marginal
welfare
cost

= MB
P 50 100 150 200 250 300 Units
ˆ
$
Marginal benefit from risk
reduction
MWC
MWC   Pˆ  P
MC  Q

Marginal welfare cost

 PUser
ˆ Pr ice

Pˆ
0 P* 100% cost sharing
P P

MWC on “Prime” is smaller than “no prime”

Why the optimal level of use is Q* = (50,100) approximately:
• stays off the “shallow of the curve”
• what cost-benefit analysis would say are the best levels to offer.

Why don’t people do as they ought, and avoid wasteful use?

The restaurant check problem writ large.

Almost all economics-based managerial issues in health insurance relate to

control of moral hazard, to getting Q’s closer to Q*
Solutions:
First best (often infeasible)
Determine severity exactly and then either,
a) pay 50 if m occurs, 100 if s occurs, period, or
b) pay full cost but limit quantity (via practice parameters) to 50 and
100.
I.E., “Pure Indemnity” or “Perfect Managed Care”
(if you can do one you can do the other). So what’s the problem?
Other solutions:
Cost sharing (demand side limits)
Imperfect Managed care (supply side limits)
Optimal Insurance in the Presence of
Moral Hazard
• Start at zero coverage and consider 1% increments.
• The person thinking of paying more for more
coverage trades off the marginal value of risk
protection (MVRP) (good) per increment against
marginal welfare cost (bad).
• Beginning at zero insurance, MVRP is high and MWC is
low (see diagram).
• But the first falls and the second rises as coverage is
increased, so optimal coverage is where they are equal.
• Optimal coverage is less generous the higher the
responsiveness of quantity demanded to user price. But
not if there are externalities.
 Other Real World Influences

• Community rating or “non-risk” rating

• Tax subsidy
• Costliness of administration
• Community Rating
– Charge everyone the same premium regardless of
risk
• If the premium covers insurer costs, it exceeds the
value of expected expenses for the young and
healthy (YH), and falls short for the older and
sicker (OS)
• Result: a really stupid way to try to do good:
– YH’s don’t want to buy
– Insurers don’t want to sell to OS’s
– Without mandate it probably increases number of
low risk uninsured and total head count of uninsured
but may increase number of high risk insureds
 Alternatives to Community Rating
with No Exclusions
• If you pay the same for instant insurance no matter
what, you rationally should wait until you get sick
to buy.
• CR taxes insurance bought by low risks to make a
transfer to high risks.
• Improvement 1: high risk pool funded by general
taxes. Does not discourage low risks
• Improvement 2: guaranteed renewability at class
average rates. Encourages most to buy
regardless of future risk; avoids
“reclassification” risk.
 The Tax Exclusion
Assume that your boss chooses the insurance you want and takes the
cost out of your pay Assume you value the insurance.
Employer premium payments are excluded from:
• employee income tax 150M people
• employee payroll tax
• employer payroll tax

Under a Cafeteria (section 125) Plan: employee premium payments can also
be excluded from income tax About 100 M people

Flex spending account can shield out-of-pocket payments up to $3000/year

About 30M people
• used only by about 1/3 of firms with employee premiums

Tax deductible payments on the income tax 5M people

 Value of Tax Exclusion (Federal)
in 2006
• Total: $200B+
$170B: Federal.
• Exceeds Federal Medicaid in cost, a loophole
without much justification.
• In 2002, those with incomes below $15K get an
average subsidy of $96; those with incomes above
$100K get a $2500 subsidy.
 Effects of Tax Subsidy
1) The tax subsidy causes excessive loading
• Example: expense of $80 with probability 1,
loading of 10% of benefits. P = $88. Marginal
tax rate = 30%, employer labor cost $20,000.
Have Employer Buy
Pay Out-of-Pocket Coverage

Taxable Income $20,000 $19,912

Taxes $6,000 $5,973.60
Medical Cost $80 -0-
“Disposable” $13,920 $13,938.40
Income
2) The tax subsidy causes excess moral hazard
• Because it subsidizes people to buy insurance
with lower cost sharing than they would
otherwise choose so the additional use is not
fully offset by risk reduction benefits.
• So removal would improve efficiency, reduce the
deficit, and reduce health care spending.
• Otherwise it is a terrible idea.
 Value-Based Cost Sharing
• If patient’s MB curve, after seeing doc, is the true curve,
ideal coinsurance (c*) varies only with elasticity.
• If patient MB is below “truth,” c* also varies with effect of
ignorance but still depends on elasticity.
• Maybe better to leave patients in blissful ignorance than to
inform them.
• Why initially offer insurance that incentivizes non-
compliance through cost sharing?
• But will I want to buy an insurance that overrules
my mistaken but firmly held choices?
• Don’t canonize but give it a try.
 Removal of Tax Subsidy Would...
1. Improve efficiency
• lower total spending by up to 10% 12-25% reduction in coverage
causes a 10-20% fall in
spending on covered services
• Possibly lower future growth rate
• Allow us to be comfortable with whatever happens to spending growth

[Link] Equity
• Benefit now goes to higher income people
• Benefit is uneven with income categories
• But this conclusion depends on what we do with the money - since
the benefit increases with income but is “regressive”
• Would it hurt to reduce the employment-based system? Could
you cash out the tax loophole?
 Some Comparison Loading %’s
• Old Medicare (excluding supp and extras): 2-3%
• Medicare (including supp and extras): 4.5-6%
• Medicaid: 5-6%
• Large private group: 5-6%; Private Medicare 14%
• Individual: as high as 50% but is about12-18%
with subsidy.
• N.B. Medicare and Medicaid usually have a tax
distortion cost of at least 25% per dollar.
 The Incidence of Employer-Paid Health
Insurance (or Any Benefit)
Assume:
All employers required to provide a new benefit. Employers are profit
maximizers over the relevant range; the number of persons seeking work is
not affected by the levels of wages or fringes.

Local labor markets are competitive before the mandate

Given the wage WB, a representative firm hires the profit-maximizing number of
workers LB. The last worker just pays his/her way (VMP or MRP=WB)

Impose the mandate, which costs $P per worker. Employers will want fewer
workers, since the cost is now WB + P. Will cut back employment.
(Unemployment (workers seeking jobs without jobs) occurs.)
Money wages start to fall. They will only stop falling
when there are no workers seeking jobs without jobs.
But the compensation cost per worker at which there
are just enough jobs is $[Link] money wages must
fall to WA=WB - P before they can reach equilibrium.

THE FULL COST FALLS ON WORKERS!

• Even if supply is responsive to real wages, some of

the cost still falls on workers
• If workers value benefits at exactly their cost, full
incidence is on wages even if supply is responsive
 Some Complicated Questions in
Benefits Management
• Why do employers think it is their money?
• Who bears the cost shift from the uninsured or low-
paying-government insurers? (Doctors, hospitals
and insured workers)
• Empirical Evidence on incidence?
* “As if” demand
* Workers comp, older workers, and maternity
benefits
* Larger worker premium shares
Incidence of Required Health Insurance Costs
S
W
S'

W0
P
W1

D' D0

L0 L*
D0 is the demand curve for labor (in terms of the money wage W) if no
coverage is required. Initial wage is W0. Now let a premium of $P be
imposed. Demand curve shifts down by P to D', and labor demanded falls to
L0. Unemployment pushes money wage down to W1=W0-P.

If supply curve is like S', some of cost falls on employers or consumers.

 (Altruistic) demand for insurance for
other people
• Assume that the choice by a state of Medicaid depends on the preferences
of the median person.
• Who is not poor, pays taxes, and is not minority.
• Altruistic insurance is a normal good with a positive income elasticity—
rich states do more
• Altruistic insurance demand slopes downward and to the right.
Generosity increases when the “price” of coverage bought by your extra
tax dollar falls with matching
• Altruistic demand is affected by taxpayer “tastes”: South, racial
homogeneity, distribution of income among taxpayers
• Implications: block grants of equal amounts will reduce spending;full
Federalization will lead to a change, but what? Saying more requires a
theory of politics. Doing more requires much better evidence on addl.
benefit!