Вы находитесь на странице: 1из 32

CAS EC 387 Introduction to Health Economics

Lecture 6: Private Health Insurance

Department of Economics

Spring 2020

Boston University () Lecture 6: Health Insurance Spring 2020 1 / 32


Introduction I

Topics
Uncertainty

Expected utility

Risk aversion

Demand for health insurance

Moral hazard

The welfare loss of excess health insurance

Boston University () Lecture 6: Health Insurance Spring 2020 2 / 32


An Example I

Suppose that you face the risk of incurring a large cost one day with
some probability. What can you do?
rely on savings
rely on friends and family
rely on charity

Suppose 100 people each face the risk of falling ill and incurring a
treatment cost of $50, 000 – large expense for a single individual

Consider this scheme: all 100 people each pay $500 into a pool, and
use it to treat the unhealthy individual

Bene…t: the unhealthy person only pays $500 for treatment

Cost: everybody has $500 less for consumption

Boston University () Lecture 6: Health Insurance Spring 2020 3 / 32


An Example II

The practice of risk sharing is very old

Suppose each of the 100 individuals fall ill with probability p = 0.01

In a large population, this means that (exactly) one individual will fall
sick with a very high probability

Even though the outcome for any one individual is uncertain, the
outcome for the group is predictable

Law of large numbers

Institution for risk pooling – the insurance market

Boston University () Lecture 6: Health Insurance Spring 2020 4 / 32


Random Variables and Gambles I

For any one individual, illness (and therefore healthcare expenditures)


is typically random

Recall that a random variable X takes a set of values with di¤erent


probabilities

The expected value of a random variable is the probability-weighted


average value of all the possible outcomes

If X takes n di¤erent values denoted by xi , for i = 1, .., n and the


probability of each value i occurring is pi 2 [0, 1], then the expected
value of X is:
n
E [X ] = ∑ pi xi
i =1

Boston University () Lecture 6: Health Insurance Spring 2020 5 / 32


Random Variables and Gambles II

Example
What is the expected number of dots in a roll of a die?

Gamble
A gamble/lottery consists of a (monetary) prize and an associated
probability distribution:
G = (w ; p )
Example

Suppose you receive $2 or $0 with equal probability

This gamble is:


G = (2, 0; 0.5, 0.5)

Boston University () Lecture 6: Health Insurance Spring 2020 6 / 32


Random Variables and Gambles III

The expected payo¤ from this gamble is:

E [payo¤ ] = 0.5($2) + 0.5($0) = $1

Fair gamble
Suppose you charge people $1 to take this gamble

Then the gamble is said to be it is actuarially fair

Fair insurance contract


An insurance policy is actuarially fair if the expected gain to the issuer
is equal to its expected cost of providing the policy:

Premium = E [payout ]

Boston University () Lecture 6: Health Insurance Spring 2020 7 / 32


Random Variables and Gambles IV

In the example above, actuarially fair insurance premium is $500 per


person

In reality, insurance contracts are not actuarially fair - there are


loading costs

Loading costs include underwriting

Boston University () Lecture 6: Health Insurance Spring 2020 8 / 32


Risk Aversion I

Who wants insurance? Depends on attitude to risk – risk aversion

Suppose you are asked to choose between:


receiving $1 with certainty, or
participating in the gamble ($2, 0.5; 0, 0.5) above

If you prefer the $1 for sure to the gamble, you are risk averse

Degree of risk aversion may change with stake in gamble

Attitudes to risk are re‡ected in the utility function

Suppose your preferences are u (w ), where w is your wealth

Consider the gamble (x, p; 0, 1 p ) and receiving px for sure

Boston University () Lecture 6: Health Insurance Spring 2020 9 / 32


Risk Aversion II
Letting w0 denote your original wealth, your expected wealth is:

E [w ] = w0 + px

Note that this is so irrespective if you take the gamble or not

The expected utility form the gamble is:

E [u (w )] = pu (w0 + x ) + (1 p )u (w0 )

The expected utility from not taking the gamble:

u (E [w ]) = u (w0 + px )

You are risk averse if:

u (E [w ]) E [u (w )]

Boston University () Lecture 6: Health Insurance Spring 2020 10 / 32


Risk Aversion III

Your utility function is concave

We will assume that loss of health is equivalent to loss of


income/wealth

Hence, utility is not state-dependent

Boston University () Lecture 6: Health Insurance Spring 2020 11 / 32


Numerical Example I

John has preferences represented by the utility function:


p
u (w ) = w

Suppose that his initial wealth is w0 = $25

Suppose that he risks falling ill with probability p = 0.5

If he gets sick, it would cost L = $16 to fully restore his health

His expected wealth is:

E [w ] = pw0 + (1 p ) (w0 L)
= 0.5(25) + 0.5(25 16) = 17

Boston University () Lecture 6: Health Insurance Spring 2020 12 / 32


Numerical Example II

His expected utility is:

E [u (w )] = pu (w0 ) + (1 p )u (w0 L)
p p
= 0.5 25 + 0.5 25 16
= 0.5(5) + 0.5(3) = 4

What if he’d get his expected wealth E [w ] = $17 for certain? Then
his utility is: p p
u (17) = 17 > 16 = 4
His utility of expected wealth is greater than his expected utility of
wealth:
u (E [w ]) > E [u (w )]
So John is risk-averse

Boston University () Lecture 6: Health Insurance Spring 2020 13 / 32


Numerical Example III
Suppose John could purchase an insurance policy that completely
covers his …nancial loss in case of illness
If the insurance contract is fair the premium should equal the
expected payout:
Premium = P = E [payout ] = pL = 0.5(16) = 8
If John gets this policy, his expected utility is:
E [u (w )] = pu (w0 P ) + (1 p ) u (w0 L P + L)
p p
= 0.5 25 8 + 0.5 25 16 8 + 16
p p p
= 0.5 17 + 0.5 17 = 17
No uncertainty - John will have wealth w = 17 regardless of health
status
What is the highest premium that John would accept for this full
insurance policy?
Boston University () Lecture 6: Health Insurance Spring 2020 14 / 32
Demand for Fair Insurance I

Suppose now, more generally, that John’s utility function is u (w ),


where w is wealth

With probability p, he incurs a loss 0 < L < w0 due to illness

Denote by wG his wealth in the good state (no loss), and by wB his
wealth in the bad state (loss)

Without insurance, is wealth is wG = w0 and wB = w0 L

Suppose insurance contracts are actuarially fair. If John selects x


units of insurance, the premium is px

His wealth in the good state is:

wG = w0 px

Boston University () Lecture 6: Health Insurance Spring 2020 15 / 32


Demand for Fair Insurance II
His wealth in the bad state is:

wB = w0 L px + x = w0 L + (1 p )x

How much insurance should John buy?

He wants to maximize is expected utility by selecting a policy (x, px ):

E [U (x )] = (1 p )u (w0 px ) + pu (w0 L + (1 p )x )

At the optimal quantity x , the marginal bene…t of insurance should


equal the marginal cost:

p (1 p )MU (w0 px )∆x = p (1 p )MU (w0 L + (1 p )x )∆x

From this we see that:

MU (w0 px ) = MU (w0 L + (1 p) x )
Boston University () Lecture 6: Health Insurance Spring 2020 16 / 32
Demand for Fair Insurance III
This, in turn, implies that:

w0 px = w0 L + (1 p) x

or:
x =L
With x units of insurance, his wealth in the good state is:

wG = w0 px = w0 pL

And his wealth in the bad state is:

wB = w0 L + (1 p ) x = w0 pL

Hence, John will insure completely against the loss: wG = wB

If insurance is actuarially fair, a risk averse individual will insure


completely
Boston University () Lecture 6: Health Insurance Spring 2020 17 / 32
Moral Hazard I

Moral hazard
In the example above, the loss due to illness was known and …xed at L

This is not realistic - the size of loss is typically uncertain and


endogenous

Actions are not veri…able by insurer (hidden action)

Sources of moral hazard in healthcare


Quantity demanded of medical care greater than amount consumer
would purchase if he/she paid full cost

Individual less likely to engage in preventive behavior and/or more


likely to engage in unhealthy behavior

Boston University () Lecture 6: Health Insurance Spring 2020 18 / 32


Moral Hazard II

Individual demands higher quality/more costly types of care than


he/she would in the absence of insurance

Individual’s incentives to monitor health care providers lower than in


absence of insurance

Individual’s incentive to search for lower prices lower in than in the


absence of insurance

Moral hazard is fully consistent with rationality

If demand for health care is not inelastic, moral hazard will occur – if
the insurance policy is complete, the marginal cost of consumption is
zero

Example insulin vs. skin lotion

Boston University () Lecture 6: Health Insurance Spring 2020 19 / 32


Moral Hazard III
Moral hazard leads to ine¢ cient use of medical care and higher cost
of insurance

Policy: prioritize/more complete coverage for medical care with more


inelastic demand

Example
Suppose Sara’s demand for doctor’s visits when sick is:
p
y (p ) = 10
10
The probability of illness is q = 0.5

How many visits will she consume when sick if she has no insurance
and the price of a visit is $50? How much does she spend on medical
care when sick?

Boston University () Lecture 6: Health Insurance Spring 2020 20 / 32


Moral Hazard IV

What is the expected spending on physician visits when she has no


insurance?

Explain whether an insurance company will be able to make a pro…t


by o¤ering full insurance coverage for all physician visits and charging
a premium of $125

How many visits will Sara consume when she has full insurance when
sick, so that the demand price is zero? If the supply price of a visit
remains at $50, what is the total cost to the insurer when Sara is
sick? What is the actuarially fair premium?

Boston University () Lecture 6: Health Insurance Spring 2020 21 / 32


Coinsurance I

Ways to mitigate moral hazard


Coinsurance is a cost sharing arrangement – the insured pays a
fraction (a percentage) of the medical cost

The share that the insured individual pays is called co-payment

A deductible is a …xed sum that the insured pays out-of-pocket

Purpose is to counteract moral hazard by making consumes aware of


the cost

Example
Suppose an insurance contract stipulates a $500 deductible and a
coinsurance rate of 20%

Boston University () Lecture 6: Health Insurance Spring 2020 22 / 32


Coinsurance II

If medical cost is $500 or less, patient pays $500 (100% of total


medical cost)

If medical cost is $1, 000, patient pays:


$500 + 0.2($1, 000 $500) = $600 (60% of total cost)

if medical cost is $10, 000, patient pays:


$500 + 0.2($10, 000 $500) = $2, 400 (24% of total cost)

Boston University () Lecture 6: Health Insurance Spring 2020 23 / 32


Coinsurance III

Coinsurance makes demand more inelastic – demand curve rotates


clockwise

Demand w/o insurance is Q0 at price P0 . Coinsurance rate c < 1


makes demand higher at every price level: consumer pays P1 = cP0
and demands Q1

Extra cost is
P0 ( Q 1 Q0 )
Welfare gains and losses – individual does not internalize full cost

Insurance subsidizes care covered by insurance – lifestyle choices?

Boston University () Lecture 6: Health Insurance Spring 2020 24 / 32


Feldstein’s Study of Moral Hazard in the US I
Feldstein: The Welfare Loss of Excess Health Insurance (1973)
Structural model of demand for health care – estimation of demand

Also examines the dynamic interaction between insurance and the


demand and supply for health care

Estimation of the welfare gains from increasing coinsurance rates

Main point: Americans are purchasing too much health insurance –


reducing coverage would be welfare enhancing (> $4 billion!)

Recall optimal insurance coverage

Health expenditures not exogenous - moral hazard

Health care is endogenous: evidence that physicians and hospitals


increase prices (and services) with increasing insurance coverage

Boston University () Lecture 6: Health Insurance Spring 2020 25 / 32


Feldstein’s Study of Moral Hazard in the US II
The welfare trade-o¤: demand distortion and risk spreading – some
coinsurance is optimal

New de…nition of health care (and higher prices) lead to increasing


demand for insurance (reinforcement)

Spillovers to the uninsured (more expensive)

Demand for health insurance

Insurance not a …nal good – intermediary good

Coinsurance is a single quantity measure

Price of insurance and demand for insurance – premium re‡ects


insurance price and health care prices

Increasing health care prices ! increasing expenditures

Boston University () Lecture 6: Health Insurance Spring 2020 26 / 32


Feldstein’s Study of Moral Hazard in the US III
Income e¤ects (risk attitudes, tax breaks)

Other e¤ects (demographics, supply), habit formation in demand

Empirical strategy
Data for hospital insurance – cross section of time series for US states
1959 - 1965

Variables:
ENR enrollment (estimated)
COINS average coinsurance rate (estimated)
QINS = 1/[(1 ENR ) + ENR COINS ] quantity insurance
PINSA price of insurance
PCARE price of care (average cost in short-term general hosp)

Boston University () Lecture 6: Health Insurance Spring 2020 27 / 32


Feldstein’s Study of Moral Hazard in the US IV

INC average household income, CPI de‡ated


GROUP group insurance
DEM demographic variables
PDD patient days demanded
USEX composite price sensitivity variable (estimated)

PDD is estimated as:

PDD = F (PCARE , INS, INC , DEM, GROUP )

Parameterized habit formation in enrollment (proportional adjustment


model):
λ
ENRit ENRit
=
ENRi ,t 1 ENRi ,t 1

Boston University () Lecture 6: Health Insurance Spring 2020 28 / 32


Feldstein’s Study of Moral Hazard in the US V

Estimation method: regress enrollment and demand for insurance:

qinsit , enrit
= λ [ β0 + β1 pinsit + β2 pcareit + β3 incit ] +
λ[ β4 Usexit + β5 pddit + β6 groupit ] + (1 λ) enri ,t 1

pins, pcare, and usex are endogenous – need for IV (lagged variables)

Logs of variables give elasticities

Boston University () Lecture 6: Health Insurance Spring 2020 29 / 32


Feldstein’s Study of Moral Hazard in the US VI
Results
Demand for insurance
pcare is positive – increased expenditure risk outweighs tendency to
purchase less care
pins insigni…cant (unclear)
group is positive – premium and tax advantages clearly increase
demand for insurance
inc ambiguous e¤ect– risk tolerance and normal good
slow adjustments (habit persistence)

Hospital choice of quantity and quality given cost structure

Increase in insurance changes the hospital’s product =) demand up


=) hosp increases both quantity and quality (if both are normal)

Boston University () Lecture 6: Health Insurance Spring 2020 30 / 32


Feldstein’s Study of Moral Hazard in the US VII
Relation between price and insurance
Stability issues
Welfare e¤ects of reduced insurance
Recall trade-o¤: risk bearing and price distortion (deadweight loss)
Note that quality changes may reduce the deadweight loss
Gain from reduced price distortion
Average coinsurance rate for hospital stay 30% (1970)
Feldstein calculates that increasing coinsurance to 50% implies
welfare gain between 1 and 3 billion; to 67% perhaps almost 5 billion

More recent estimates of welfare loss


$33 $109 billion (8.9% 29.1% of total exp) (1984)
Manning/Marquis (1996) – optimal coinsurance rate 45%
Boston University () Lecture 6: Health Insurance Spring 2020 31 / 32
Health Insurance Elasticities I

Rand Health Insurance Experiment: natural experiment with varying


degrees of co-payment

Demand for all medical services respond to coinsurance changes

Dental services and prescription drugs – fully insured used 76% more
than those with 95% co-payment

Deductible: very small deductible (< $50) has large e¤ect, medium
deductible(< $500) has small e¤ect, large deductible has large e¤ect

Rural China study

Boston University () Lecture 6: Health Insurance Spring 2020 32 / 32

Вам также может понравиться