Professional Documents
Culture Documents
ISBN 978604821338-1
ABSTRACT
This paper investigates the existence and the eciency of a ood risk sharing network in rural Vietnam. We develop a
model where the probability of loss is a subjective risk perception function based on risk information and self-protective
costs. We study the interaction of informal insurance with limited commitment through risk communication inuencing
on households decisions and behaviors where risk information be treated as the public good in a risk sharing network.
apply protective measures, the nal probability of loss actions may lower likelihood of loss, p(C , I ) is non-
will be increasing to both arguments.
p(C ) = p( A1 ... AJ ) = ( -1) J [p( A1 ... AJ )
We introduce two forms of P. In reality, discrete
J
[1] investments in protective measures often obtained from
-p( A j ) + p( Ai A j ) - p( A A i j Ak ) + ...]
j=1 1 j < i < k J 1 j < i J empirical data but not be usually used in theoretical
Where jK
c j C . models in the literature due to the difficult of
discontinuities.
Remark: In the case of independent probability of actions,
Assumption 1. (Discrete values) The probability of loss is
p(C ) = jK p j (c j ) This can be applied when we study
a step function.
only one type of risk which aects to dierent losses2.
Given c 1 ,..., c J 4 and I
If C is known, household then choose which of
{ A1 ,..., AJ } with respective investment costs { c1 ,..., c J } to be p 0 if 0 C c 1
self-protective actions. Household want to minimize 1
p(C ) dened in [1] to obtain the lowest probability of p if c 1 C c 2
p(C , I ) =
loss. She then has to solve the following problem M
J
inf p(C ) such that x c j j C , x j { 0,1} . p if c J C
xj j
jJ
where p [0, pn ] The value c 1 can be considered as the
This problem is well-known and there exits algorism to xed cost: insufficient level of investment cost leads to
solve it. If x j = 1 then A j is applied and not if x j = 0. We inefficient protective measure.
will provide examples at the empirical sections
Assumption 1'. (Continuous values) Assume that p(C , I )
ii) Risk information
is continuously twice differentiable and
Denoted by I the available information of ood risks
p1 < 0, p11 > 0, p2 < 0, p22 > 0 and lim C p = 0 . That is, as
perceived by households who care on protection. Risk
information is related to households perception of risk risk sharing information and the investment in protection
(which eects to the loss probability) and ecient increases, the loss probability decreases at a decreasing
capabilities to take self-protection to reduce the loss rate. And by investing suciently large, the probability
(which eects to utility). People who know deeper risk of loss will close to zero.
information can be able to create higher quality of self-
2.1 Optimal investment
protection. In this model, we assume that household only
get a reward if she has risk information and apply to Denoted M the initial total wealth of the household:
protective measures to reduce the loss when oods Suppose that the utility function u is increasing,
happen. The reward also can be explained, for example, continuous and strictly concave. As a results, if
if a household get a good information of ood risks, with household take no care on protection; her expected utility
the same cost for self-protection measure as others, she will be
can lower ood damage. Moreover, she can learn from
Eun = p n u( M - L ) + (1 - pn )u( M ).
neighbors for more ecient protective measures.
Otherwise, the expected utility can read
In the context of formal ooding insurance is hardly
accessed in Vietnam (Reynaud et al, 2013), informal Eu = p(C , I )u( M - L - C + a ( I )) + (1 - p(C , I ))u( M - C ).
insurance and self-protection can be complement. The In this case, she has to nd a set of optimal investments
insurer may offer different contracts for different type of C * to be the solution of
risks where the premiums depend on the levels of self-
protection taken by of households . So the reward for max{pu( M - L - C + a ) + (1 - p )u( M - C )}. [2]
C
households who have good risk information can help.
Before issue the premium, the insurer may test the risk
perception of households. If the household has good
3
knowledge and information on ood, a premium rebate Moral hazard can happen when insurers do not know
households who buy low premium but do not take or
1
The model is built to be consistent with our database as take fewer protective measures
4
a choice set of several self-protective measures Imagine that a village builds a dike to prevent oods.
2
For example, in a recent our study (Reynaud et al, The feet tall of the dike depends on the money spend on
2013), we focus on three dierent cost invested in investment. And the frequency of ooding depend on
reducing the loss from agricultural productions, home height of its ood barrier. As a result, it depends on the
contents and death. money we invest in.
2
Under assumption 1, if if c i C < c j then p(C ) = p i It is Eu[M - C - p(L - a ) + X ]=u[ M - C - p(L - a ) - p ( p )].
easy to see that Hence
Eu = pu( M - L - C + a ) + (1 - p )u( M - C ) pu( M - C - L + a ) + (1 - p )u( M - C ) = u( M - C - p(L - a ) - p ( p )).
= p u( M - L - C + a ) + (1 - p )u( M - C )
i i [3]
p i u( M - L - c i + a ) + (1 - pi )u( M - c i ) = Eu(c i ).
Since y be the money that household is ready to pay for
Therefore, the solutions of [2] belong to {0, c 1 , c 2 , ..., c J } . escape the risk, u( M - C - y ) = u( M - C - p(L - a ) - p ( p )),
Notice that when C * = 0 then no investment, household
y = p( L - a ) + p ( p ) [4]
only take protective action if Eu(C *) > Eu(0) = Eu n .
Similarly, if denote by p ( p ) the risk premium when the
n
3
iii) If T - p y L - p ( p y ) + a p y < C then Moral hazard can happen when insurers do not know
households who bought low premium but do not take
u ( M - C - p L - p ( p ) + a p ) < u( M - T ) : the households
y y y
protective measures. In this case a = 0 in ( I , N ) .
will buy insurance.
As we know, if the premium is independent to 2.3 Income effects
investment in self-protection, insurance may cause a
negative incentive to self-protection. Moral hazard also Suppose that now information is costly. Let dene the
occurs when households are fully covered by insurers, budget constraint M = tx X + Y + tc + tI I where tx , tc are
they do take protective measures, even they cause the prices for non-market goods X , self-protection cost and
loss then get indemnity. We now apply our model to information. Y is other composite goods. Consider the
explain a kind of risk sharing arrangement as the maximized problem of household above
informal insurance in a small group of households. A
max {p(C , I )u1 ( X , Y ) + (1 - p(C , I )u 2 ( X n , Y )}
premium rebate is oered to low risk households I ,C , X ,Y
then the insured always takes self-protection. Otherwise, she p1u1 - p1u2 - mtc = 0
does not invest in self-protection. p2 u 1 - p 2 u 2 - m t I = 0
Proof: pu11 + (1 - p )u12 - mt x = 0
We restrict our model where two groups can buy pu21 + (1 - p)u22 - m x = 0
insurance. The households who have good risk
information and take self-protection with low premium Then
and those who do not take self-protection with a high C 1
u1 - u 2
= = 1 [7]
premium. As in the model setting, if households are not p p1 pu2 + (1 - p )u22
interested in insurance, they can choose self-protection or
Note that
not. So far, we have 4 groups, which will be denoted as
(I,Y), (I,N) (NI,Y), (NI,N) where I=Insurance; NI= No u( X , Y )
u( X , M - t x X - tcC - t I I ) dM u( X , Y )
= = [8]
insurance, Y=Yes to protection, N= No to self-protection. Y M dY M
Denote by B the indemnity from insurers. The utilities of Moreover, our maximized utility can b e read
4 cases are
EU = pV 1 ( M , tx , t y , t I , L) + (1 - p )V 2 ( M , t x , t y , t I , L )
u( I , Y ) = pu ( M - C - L - (T - a ) + B + (1 - p )u( M - C - (T - a ))
u( I , N ) = p nu( M - L - (T + a ) + B + (1 - p n )u( M - (T + a )) Where
u(NI , Y ) = pu ( M - C - L ) + B + (1 - p )u( M - C - (T - a ))
V 1 ( M , t x , t y , tI , L ) = u1 ( X * ( M , tx , ty , t I , L),Y * ( M , tx , t y , t I , L )),
u(NI , Y ) = p nu( M - L) + (1 - p n )u( M )
V 2 ( M , tx , t y , t I , L ) = u 2 ( X * ( M , t x , t y , t I , L ), Y * ( M , tx , ty , tI , L ))
From Lemma1, we will use the equation
pu( x0 - x ) + (1 - p )u( x0 ) = u( x0 - px - p ( p)) where p ( p )
M V1 -V2
Therefore, =
is the risk premium when the loss probability is p p p 1
V /M + (1 - p )
V2 /
M
evaluated at x0 Then we have
C M
It follows that =
u( I , Y ) = u ( M - C - (T - a ) - pL + p B - p ( p ))
n p
p
u( I , N ) = u( M - (T + a ) - p nL + p nB - p ( p n )). Thus we see that marginal self-protection expenditure
equal to WTP for risk reduction. Similarly,
If u( I , Y ) > u( I , N ) then C < p ( p ) - p ( p ) + ( p - p )L + a .
2 1 n
4
Consider a group of n households indexed by Proposition 5 With risk sharing information, each
i = 1,2,..., n. In this setting, when formal insurance household's optimal cost of self-protection C i* is less than or
mechanisms are hardly accessed, we develop a mo del equal to the optimal cost of self-protection C i without risk
where household invest in self-protection but also in risk sharing information, C i* C i . If p12 (C , I ) > 0 we have
sharing information. It is costly to form such an C i* < C i .
relationship as the expenditures of monitoring and
Proof: Consider problem with sharing information
maintaining the network and the cost for risk sharing
information as the private provision of public good.
max{p(C i , I )ui ( M - Li - Ci - ( I i - I -* i ) + a ( I ))
Ci , I
Suppose that household i shares risk information to other
+ (1 - p(C i , I )ui ( M - C i - ( I i - I -* i ) + a ( I ))}
member in the network and the expenditure of this
information is I i . The total shared information I = i = 1 I i
n
The rst order condition yields
can be considered as the total supply of provision of
0 = p1ui 1 - puci 1 - p1ui 2 - (1 - p )uci 2
public good. Assume that the loss probability of
household i is p(C i , I ) satises the following assumption + p2 ui 1 - (1 - a 'I )uIi 2 - (1 - p)(1 - a 'I )uIi 2
where u = u ( M - Li - C i - ( I i - I - i ) + a ( I )),
i1 i *
The utility maximization of household i
ui 2 = ui ( M - C i - ( I i - I -* i ) + a ( I )), ui 2 > ui 1 .
max{p(C i , I )ui ( M - Li - C i - I i + a ( I ))
Ci , Ii
From F.O.C we know
+ (1 - p(C i , I )ui ( M - C i - I i + a ( I )))}
Denition 1 The Nash-equilibrium in this model is a vector of (1 - a 'I )uIi 1 + (1 - p )(1 - a 'I )uIi 2
p(C i* , I * ) = - p2 -
risk information investment (I 1* , I i* ,..., I n* ) such that for each ui 2 - u i 1
household i solves the problem i1
puc + (1 - p)uc i2
-
max{p(C i , I + I -* i )ui ( M - Li - C i - I i + a ( I )) ui 2 - ui 1
Ci , Ii
Consider problem without sharing information
+ (1 - p(C i , I + I -* i )ui ( M - C i - I i + a ( I )))}
n max{p(C i ,0)ui ( M - Li - C i ) + (1 - p(C i ,0)ui ( M - C i )}
where I - i = j = 1, j i I - j .
* *
Ci , I
s.t C i + I - a ( I ) = M - Li + I *
-i C i* C i .
*
I I -i
Empirically based on a survey conducted in Vietnam
Let denote the demand function for G of household i ; (Reynaud and Nguyen, 2013), we nd the evidence on
then G = max{ f i ( M - Li + I - i ), I - i } + a ( I ) the existence of the network as farmers form
relationships to mitigate risks. The willingness to pay for
By using a xed point theorem, we have ood risk reduction in the network and the behavior
Proposition 3 If a '( I ) < 1 then there exists a Nash changes of farmers investing in self-protection and risk
equilibrium information seeking as the effective mechanisms are
evidences of networks efficiency.
Prop osition 4 There exist a constant b [ p ,1] such that
REFERENCES
-b p(1 - a '( I ))
= p1 (C i* , I * ) + p2 (C i* , I * ) - . Mai, C. V., M. J. Stive,and P. H. VanGelder (2009):
Li Li Coastal Protection Strategies for the Red River Delta,
If household is risk neutral, we have Journal of Coastal Research, pp. 105116
1 - a '( I )
p1 (Ci* , I * ) + p2 (C i* , I * ) = - Pilarczyk, K. W.,and N. S. Nuoi(2005): Exp erience and
Li
Practices on Flood Control in Vietnam, Water
Assume that p12 (C , I ) 0, it means the spill-over from International, 30(1), 114122.
increased total sharing information decreases or leaves
unchanged the marginal benet of self-protection Reynaud,A et al (2013): Living with o o ds: o o d
expenditure for reducing loss probability. When I * = 0, protective behaviours and ood risk perception of
when dene C as the optimal cost of self-protection Vietnam households, The Geneva papers of Risk and
without risk sharing information. Insurance, vol. 38, July 2013, p. 547-579.
5
Reynaud,A and Nguyen M.H (2013): Flo o d Risk
Reduction: Results From a Choice Exp eriment in
Vietnam, VCREME working paper, 01-2103.