Вы находитесь на странице: 1из 11

Adverse Selection in Insurance Market

BEPP 305/805 Fall 2012

In our discussion of insurance markets thus far in the course, we have always assumed perfect
information in the market. In other words, the insurance companies know exactly the true
riskiness of all potential customers in the market, and there is no incentive under this setup for
the policyholders to pretend to be someone else. In reality, however, information in the market is
never this transparent. Individuals tend to know much more about their likelihood of incurring
a loss and the severity of such a loss than the insurance companies. This type of asymmetric
information is not unique to the insurance market, as it is seen in financial markets, as well as
many other consumer related markets.
The adverse selection problem we discuss extensively below refers to the situation where one
party to a transaction possesses information that affects the value of the transaction at the time
of contracting. For example, the seller of a used car probably knows more about the true quality
of the car than the potential buyer when the transaction is done. A potential customer for auto
insurance should also know much more about his driving habit than the insurance company thats
going to sell him the policy. Under this scenario, the party with less information may receive an
adversely selected pool of counter parties, and get lower valuation from this transaction than they
originally anticipated.
This set of notes discusses extensively how such adverse selection problem affects the insurance
market, and how an insurance company can design different types of policies to protect itself
against the disadvantage in information.

Benchmark Case: One Risk Type

Lets start our discussion by examining an insurance market with only one risk type. The potential
policyholders are all homogeneous and share the same characteristics below:
an initial wealth W
1

a potential loss L, occurs with probability p


a risk averse utility function u.
The other party in this market is an insurance company that offers an insurance policy that
covers I level of indemnity in case of a loss event. It charges a premium C in exchange for
the coverage. To facilitate the discussion, we further assume that the insurance company is risk
neutral, i.e., it only cares about the expected profits; and there is no administrative cost associated
with underwriting the insurance policies. With these assumptions, the insurance companys goal
is to maximize its expected profit:
C pI
From the individuals perspective, his expected utility without insurance is:
pu(W L) + (1 p)u(W )
If he decides to purchase I indemnity worth of insurance policy offered by the company, his final
wealth in the state of loss and no loss equals to:
wL = W C L + I
wN L = W C
Hence his expected utility with the insurance policy is:
pu(WL ) + (1 p)u(WN L )
The insurance policy is attractive to him if and only if the expected utility from having insurance
is no lower than the expected utility of staying uninsured:
pu(WL ) + (1 p)u(WN L ) pu(W L) + (1 p)u(W )
Lets first study the indifference curve of this individual. By definition, two points along the same
indifference curve give the individual the same level of utility. If we allow very small perturbation
in wealth in the state of loss and no loss at any given wealth combination, it shouldnt change
this individuals utility. Hence:
pu0 (wL )wL + (1 p)u0 (wN L )wN L = 0
Rearranging the terms, the slope of the indifference curve is:
p u0 (wL )
wN L
=
wL
1 p u0 (wN L )
The indifference curve has the following properties:
2

(1)

They are downward sloping - if the individual wants more wealth in the state of no loss, he
will have to give up some resources in the state of loss, and vice versa. In other words, if
he wants to pay lower premium, he should expect to get a lower indemnity in case of a loss
event, given the same level of utility.
The further away it is from the origin, the higher the level of utility - this means that the
individual has more resources to consume when theres loss or no loss, hence the higher
utility.
They are convex, as one is willing to give up more wealth in the state of no loss in exchange
for a unit increment in wealth in state of loss when it is low, and less so when the wealth in
the state of loss is high.

Figure 1: Indifference Curve of a Potential Policyholder

Lets now study the characteristics of the insurance company. The iso-profit curve for the
insurance company plots all combination of insurance policies such that the profit to the insurance
company stays constant, say . Given the expression of wealth in the state of loss and no loss,
the expected profit of the insurance company can be written as:
C p I = W pwL (1 p)wN L pL =
Rearranging the terms, we get:
wN L =

p
W pL

wL
1p
1p

These iso-profit curves have the following properties:


3

(2)

p
They are downward sloping straight lines with slope equals to 1p
. For the insurance
company to collect higher premium in the state of no loss, it needs to provide more indemnity
coverage to the policyholder if loss occurs, to maintain the same level of profitability.

The closer the iso-profit curve is to the origin, the higher the profit level. There are two
ways to show why this is the case.
Mathematically, the intercept of the iso-profit curve on the y-axis is W pL
. There1p
fore, the higher the level of profit, , the lower the intercept, and the closer the curve
is to the origin.
Intuitively, given any point on the iso-profit curve, its x-axis/y-axis combination shows
how much resources is left for the policyholder net of the insurance coverage. The
closer it is to the origin, the lower the amount of resources is left for the policyholders,
and hence the higher the profit for the insurance company.

Figure 2: Iso-Profit Curve for the Insurance Company

Now lets combine the information of the insurance company together with that of the potential
policyholders to see whats the optimal policy to offer from the insurance companys point of view.
A potential policyholders default position is no-insurance, where his wealth is wN L = W , and
wL = W L. The shaded region in Figure 3 below are all the acceptable policies to this individual,
as they represent equal or higher level of utility compared to the default position of no-insurance.
The insurance company can offer policy B shown in Figure 4 below to the policyholder. It
gives the insurance company more profit compared to policy A, where the individual essentially
chooses to stay uninsured. The insurance company can further improve its profitability by moving
downwards along the individuals indifference curve, as this brings its iso-profit curve closer to
4

Figure 3: Acceptable Policies

the origin. Its profit is maximized where the iso-profit curve is tangent to the indifference curve
of the individual, shown as policy C in Figure 4 below. Beyond this point, the insurance policy
is no longer acceptable to the individual when comparing to his no-insurance default position.
Hence, the tangency condition for optimal insurance policy from the insurance companys view is
the slope of the indifference curve equals to the slope of the iso-profit curve:

p
p u0 (wL )
=
1 p u0 (wN L )
1p

(3)

u0 (wL ) = u0 (wN L )
wL = wN L

Notice an important assumption in the setup of the model: there is only one insurance company
in this market, and therefore it acts as a monopoly, and extracts as much profit as possible from the
policyholders. To see why this is the case, lets compare different options from the policyholders
point of view. All the insurance policies offered by the in surance company move along the
indifference curve of the individual where the no-insurance option locates. The insurance company
has no incentive to offer policies that will allow the policyholder to achieve a higher level of utility
than point A, as it faces no competition. The optimal policy, policy C, offers the policyholder
the same level of utility compared to that of staying uninsured.

Figure 4: Polices that can be offered by the insurance company

Two Risk Types

Now suppose that the potential policyholders have two different risk-types: high-risk types (H)
and low-risk types (L):
High risk types have initial wealth W and can incur a loss L with probability pH
Low risk types have initial wealth W and can incur a loss L with probability pL
pH > pL
Similar to the one risk type benchmark case, the slopes of the indifference curve for the high risk
types and the low risk types are defined as the followings:
slopeH =

pH
u0 (wL )
1 pH u0 (wN L )

slopeL =

pL u0 (wL )
1 pL u0 (wN L )

Given the fact that pH > pL , we can conclude that |slopeH | > |slopeL |. This means that the high
risk types value more the wealth in the state of loss compared to the low risk types, and they
are willing to give up more wealth in the state of no loss (in the form of paying higher premium)
in exchange for additional increment in their wealth in the state of loss (in the form of receiving
higher indemnity).
From the insurance companys point of view, the slopes of the iso-profit curves for these two
types of potential customers are:
pH
slopeH =
1 pH
6

slopeL =

pL
1 pL

And again given the fact that pH > pL , we can conclude that |slopeH | > |slopeL |. The steeper isoprofit curve for the high risk types is equivalent to saying that it is more costly for the insurance
company to provide additional coverage to the high risk types compared to the low risk types.
Therefore, in order to stay the same level of profitability, it needs to charge the high risk types a
higher premium.

2.1

Symmetric Information

If the true riskiness of all potential policyholders is 100% observable to the insurance company,
then we can solve the case for two risk types exactly the same fashion we did for the one risk
type. The optimal policies are characterized by two sets of tangent points: one for the iso-profit
curve and indifference curve of the low risk type, and one for that of the high risk types:
H:

pH
u0 (wL )
pH
=

1 pH u0 (wN L )
1 pH

L:

pL u0 (wL )
pL
=

1 pL u0 (wN L )
1 pL

H = w H and w L = w L . In
Solving the algebra here, we see the tangency condition leads to wL
NL
L
NL
other words, the insurance company maximizes its profit by offering two sets of full insurance
policies to high risk types and low risk types respectively, assuming that the true riskiness is
observable.

Figure 5 below illustrates the optimal polices graphically. Since the optimal policies are offered
H = w H and w L = w L ), these policies should lie on the 45
in the form of full insurance (given wL
NL
L
NL
degree line. Low risk individuals are offered full insurance policy B, and high risk individuals are
offered full insurance policy C. It is not surprising to see that policy B offers the policyholders
more resources in both states (loss or no loss) compared to policy C. This is due to the fact that
low risk types have lower probability of incurring a loss, and the insurance company charges a
lower premium when the true riskiness is transparent.1
1

Notice here that we are not making any comparison across the two iso-profit curves for different risk types. The
profitability of policy B and C is essentially determined by the intercepts of these iso-profit curves on the y-axis.
Recall that the intercepts are expressed as:
H:

W pH L H
1 pH

L:

W pH L L
1 pL

One cant make an easy comparison between H and L by just eye balling these iso-profit curves.

Figure 5: Optimal Insurance Policies When Information Is Symmetric

2.2

Asymmetric Information

In real life, however, potential policyholders tend to have much better knowledge of their true
riskiness than the insurance company. Therefore, it is generally difficult for the insurance company
to tell who are the high risk types and who are the low risk types in the pool of potential customers.
If the insurance company continues to offer two full insurance policies B and C to this opaque
group of customers, it will end up seeing everyone purchasing the policy B, originally designed
for the low risk types, and no one will choose to purchase policy C. The logic is as the followings:
low risk types will choose to purchase policy B, as they are indifferent with this policy compared
to the case of no insurance. High risk types now have all the incentive in the world to pretend to
be low risk types and purchase policy B instead of policy C, as theres no way for the insurance
company to discover their true riskiness and policy B offers them a higher level of utility. The
insurance company will suffer as a result of this asymmetric information, and therefore it is not
wise to offer two full insurance polices.
To protect its interest, the insurance company can decide to offer only one full insurance
policy, policy C, under the asymmetric information setup. When this is the case, the low risk
types will decide to stay uninsured, as purchasing policy C will result in a lower level of utility
compared to point A. This decision is optimal if the proportion of low risk types, lets call it ,
is low in the entire market.
When is high in the population, it is obviously no longer optimal for the insurance company
to give up the market for low risk types. To get both risk types to participate in the insurance
market, the insurance company can offer plans with different coverage to let policyholders self8

select into the right plan based on their true riskiness. The optimal policies to offer under
asymmetric information when is high consists of a partial insurance and a full insurance. The
goal is to get both risk types participate in the insurance market, and yet no one has the incentive
to pretend to be the other type.
The insurance company can offer a partial insurance E to entice the low risk types. Notice that
policy E gives the low risk types the same level of utility comparing to the no-insurance position,
point A, therefore the low risk types are willing to purchase the policy. When this partial insurance
is available, the insurance company can no longer offer the same old full insurance policy C to the
high risk types, as they will have the incentive to purchase the partial insurance as well, because it
will provide them with a higher level of utility than policy C. To prevent the high risk types from
pretending to be the low risk types, the insurance company has to give up some of the profit it
can extract from the high risk type, and offer full insurance policy D instead. Notice at policy D,
the high risk types enjoy the same level of utility had they chosen the partial insurance policy E,
so they would have no incentive to disguise as the low risk type. From the high risk individuals
point of view, they are better off at policy D than policy C, as they enjoy a higher level of utility
at policy D.
Figure 6: Optimal Insurance Policies When Information Is Asymmetric

The insurance company is essentially making some trade off when a partial insurance and a
full insurance are offered: in order to incentivise the low risk types to purchase some insurance
(at policy E), the insurance company has to give up some of the additional profit it can extract
from the high risk types, therefore it has to offer full insurance policy D instead of policy C,
such that the high risk types have no incentive to disguise as the low risk types. Again this
trade off is worthwhile only if the proportion of low risk types in the population, ,
9

is high. When is low, it is optimal for the insurance company to offer only policy
C to extract the highest profit possible from the high risk types.
One last point before we wrap up the discussion. You may wonder why the insurance company
is only offering a partial insurance E such that the low risk types are barely indifferent between
having this insurance and no insurance? If the insurance company were to offer a partial insurance
that gives higher level of utility than no insurance point A, the low risk types are definitely happy
to take it. To prevent the high risk types from pretending to be low risk types and taking this
new partial insurance, the insurance company would have to offer an even more generous full
insurance policy than policy D, so this new policy would lie somewhere along the 45 degree line
between policy D and policy B. This separation mechanism works fine except for one important
point: the insurance company wants to maximize its profit. By offering a more generous partial
insurance than policy E and a more generous full insurance than policy D, the insurance company
is giving up more of its profit potential, and this is obviously not an optimal decision.

Conclusion

We have discussed the optimal insurance policy under information asymmetry. We start with a
benchmark model where information is symmetric and show that it is optimal for a risk neutral
insurance company to offer full insurance to maximize its profit. This is the case independent
of the number of risk types in the market, as long as they are all observable to the insurance
company.
When the information is asymmetric between the potential policyholders and the insurance
company, the insurance company can use a partial insurance as a screening device to distinguish
the different riskiness.
When the proportion of low risk types is low in the pool of potential policyholders, it is optimal
to just offer one full insurance policy, which targets the high risk types and allows the insurance
company to extract all possible profit from this group of policyholders. Under this scenario, the
low risk types will stay out of the insurance market and choose to remain uninsured. This is one
type of market inefficiency, as the low risk types are pushed entirely out of the insurance market
due to information asymmetry.
When the proportion of low risk types is high in the market, then it is optimal for the insurance
company to offer a partial insurance and a full insurance. Low risk types will choose to purchase
the partial insurance that makes them equally happy as staying uninsured. High risk types will
choose to purchase the full insurance, which benefits them as it offers a higher level of utility than
staying uninsured. Note that this separating equilibrium is another form of market inefficiency,
10

as under the information asymmetry, the low risk types no longer have the option to purchase
full insurance that would be available to them otherwise under the symmetric setup.

11

Вам также может понравиться