Académique Documents
Professionnel Documents
Culture Documents
Keywords: moral hazard model, asymmetric information, Principal-Agent model, optimal contracts, effort
level
1.GENERALREASONSCONCERNINGTHE
INFORMATIONASYMMETRY
2.THEMORALHAZARDMODELCONCERN
INGTHEDISCRETECASE
.
The model studies the behavior of the two part
ners towards risk in conditions of uncertainty, thus ana
lyzing their utility functions. As concerns the effort sub
mitted,theAgentwillberewardedbythePrincipalwith
awageW,anditsutilityfunctionwilldependincreasingly
uponthiswageanddecreasinglyupontheeffortsubmit
ted, meaning ( ) ( ) ( ) , U W e U W V e = , where the
form of the utility function shows that it is additively se
parablerelatedtothetwoargumentsWande.Therefore,
the attitude towards risk will be independent as regards
the level of effort. The wage, on its turn, will represent a
function depending upon the result achieved, mean
ing ( ) W W X = .TheAgenthasriskaversion,andhisutil
ity function will be strictly concave and increasing as re
lated to the wage, which means that 0, 0 U U ' '' > < . A
higher effort of the Agent signifies a reduction of utility,
therefore 0 V' > .Wewillconsiderthemarginaldisutility
oftheeffortasbeingincreasing,meaning 0 V'' > .
The Principal signifies the beneficiary of the re
sult specific to the effort submitted by theAgent, so that
his utility function denoted by ( ) B X W will depend
upon the level of result and, will depend decreasingly
upon the wage paid to the Agent. The Principal is risk
neutral, and we will therefore achieve 0 B'' = , meaning
B' isaconstant.
Asregardstheinformationsymmetry,themodel
will lead us to an optimal wage, depending upon the re
sult,sothattheAgentwillbetheonewhowillchosethe
levelofeffortsubmittedinthefirststageoftheissue,thus
choosingtheminimaleffort.Inthisway,theutilitywillbe
ofmaximalvalue.TheAgentschoicecanbeformallyde
scribedasfollowing:
( ) ( ) ( ) ( )
1
arg max
n
i i
e
i
e p e U W x V e
=
( e
`
)
(1)
Theabovementionedconditionisknownunderthename
of incitation restriction or stimulation restriction, being
the one that shows the moral hazard situation presence.
Regarding the second stage of the issue, by determining
the optimal level of effort that will exist within the con
tract,theAgentdecidesifhewillsignornotthatcontract.
The acceptance will take place if the expected utility will
be at least at the level of reserve utility, meaning: U :
( ) ( ) ( ) ( )
1
n
i i
i
p e U W x V e U
=
( >
. The previous
condition is called restriction of participation or the re
strictionofindividualrationality.
In this way, for the first stage of the game, the
Principal, anticipating the action of the Agent, will pro
poseacontractthatrepresentsthesolutionoftheoptimi
zation issue, maximizing in this way the utility, by the
tworestrictions,meaning:theparticipativeandtheincita
tionone:
( ) { }
( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( )
, 1,
1
1
1
max
arg max
0; 0, 1,
i
n
i i i
e W x i n
i
n
i i
e
i
n
i i
i
i
p e B x W x
e p e U W x V e
p e U W x V e U
e W x i n
=
=
=
=
( e
`
)
( >
> > =
(2)
Themoralhazardmodelontwolevelsofeffort
Wewillstartsolvingthis moralhazard modelin
a discrete variant, by analyzing the situation when the
Agentcanchosebetweentwolevelsofeffort:ahighlevel
ofeffort,denotedbyHandalowleveleffort,denotedbyL.
inthissituation,wewillhave
{ }
,
H L
e e e e andobvious
ly,
( ) ( )
H L
V e V e > .Also,wewillorderincreasinglythe
set X of the potential results that can occur, mean
ing { }
1 2
...
n
X x x x = < < < . We denote by
( ) ( )
( ) ( )
/ 0
, 1,
/ 0
H H H
i i i
L L L
i i i
p p e P X x e
i n
p p e P X x e
= = = >
= = = >
the
probabilities of achieving the result , 1,
i
x i n = in each of
thetwovariantsofeffort.Weassumethattheseprobabili
ties are strictly positive. We also know
that
1 1
1
n n
H L
i i
i i
p p
= =
= =
.
The moral hazard will occur when the Principal
willsolicittotheAgenttosubmitthehighefforte
H
.Con
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 15
trary, by soliciting the low effort e
L
, he will pay to the
Agent an established fix amount from the optimal con
tract, in situation of symmetric information, mean
ing
( ) ( )
1 L L
W U U V e
= + .
ConsideringthatthePrincipalisriskneutral,we
will have constant B' = and therefore,
( ) , 0, B x a x b a b = + > e . The two constants that
occur can be eliminated from the objective function, so
thattheissue(2)canberewritten:
( ) { }
( ) ( )
( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
( )
1,
1
1
1
max
0, 1,
i
i
i
n
H
i i
W x i n
i
n
H H
i
i
n
H L H L
i i i
i
i
p x W x
p U W x V e U
p p U W x V e V e
W x i n
=
=
=
=
( >
(
>
> =
| |
|
|
' = =
|
| |
+ |
|
|
\ . \ .
. We
may notice that if the rate of probability
H
i
L
i
p
p
is decreas
ing in i,
1 2
1 2
...
H H H
n
L L L
n
p p p
p p p
| |
> > >
|
\ .
, then its inverse
L
i
H
i
p
p
will be increasingly in i, and thus, the amplitude of
the function ( )
1
U
' above mentioned will be increasing
in i. But, 0 U'' < since the amplitude is riskphobic, and
therefore U' will be strictly decreasingly, thus resulting
that ( )
1
U
' isastrictlyincreasingfunction.Knowingthat
its amplitude is increasing in i, we will deduct that the
wage W(xi) is increasing in i. As conclusion, if the rate of
probability is decreasing, at higher results, higher wages
willcorrespond.
Themoralhazardmodelonmanyeffortlevels
We will forwards emphasize a generalization on
a discrete case for the moral hazard model, of two levels
of effort likethose mentioned above; these are applicable
in the situation when theAgent can chose between more
levelsofeffort(oractions).
Wethusdenoteby { }
1 2
, ,..., , 3
m
a a a m > theset
of potential actions of the Agent, and by
( )
, 1, , 1,
ij j i
p P X x a i m j n = = = = the probability of
achievingtheresultxj,whentheAgentsubmitstheeffort
ai.As in the previous cases, we assume that these proba
bilities are strictly positive:
0, 1, , 1,
ij
p i m j n > = = (in other situations, some
actionsmightbeexcluded).
ThePrincipal,notbeingabletoobservetheeffort
of the Agent, will offer to him a wage related to the
achievedresult,thissignifyingtheonlyelementfavorable
by the Principal. In order to simplify the annotations, we
will denote by
( )
, 1,
j j
W W x j n = = the wage received
bytheAgent,whentheresultalreadyobservedwasxj.
We will consider an Agent of risk aversion hav
ingautilityfunctionstrictlyincreasingly( 0 U' > ),strictly
concave ( 0 U'' < ) and additively separable, under the
formofU(W)a,andaPrincipalneutraltorisk,havingthe
utility function under the form of B(XW)=XW (we can
assumethis,sincebeingriskneutral,wewillseethat
.
The Principal will decide the level of the wage
offered to theAgent, and he will follow the achievement
of the maximum for the expected utility:
( )
, 1,
1
max
j
n
ij j j
W j n
j
p x W
=
=
(
.
The optimal contract of moral hazard according
toanactionaiestablished,of 1, i m e ,willberepresented
by the optimal solution of the optimization issue, that
followsthemaximizationoftheutilityfunctionspecificto
the Principal, by meeting the incitation restrictions and
theparticipationrestriction:
( )
( )
( )
( )
( )
( )
( )
, 1,
1
1 1
1
max
, 1, ,
j
n
ij j j
W j n
j
n n
ij j i kj j k
j j
n
ij j i
j
p x W
p UW a p UW a k mk i
p UW a U
=
=
= =
=
( (
> = =
(
>
(3)
From the first order optimal condition, as related to the
wageWj, 1, j n = ,wewillachieve:
( ) 1
1
1 , 1,
m
kj
k
k
ij j
k i
p
j n
p U W
=
=
( | |
= + = ( |
|
'
(
\ .
(4)
The incitation restrictions saturated from issue (3) will
have multipliers strictly positive, and those actions will
bringthesameutilityastheactionai.
Theoptimalcontractisgivenbyrelationship(4),
wherethereport
( )
1
j
U W '
isincreasinglyasrelatedtoWj,
since U' is strictly decreasing. The optimal wage can be
deducted from the previous relationship under the form
( )
1
1
1
, 1,
1
j
m
kj
k
k ij
k i
W U j n
p
p
=
=
| |
|
|
|
' = =
| ( | |
+ | ( |
|
| (
\ .
\ .
.
Onemightnoticeatthiscontractthatthewageisnotcon
stant, as in the situation when the Principal notices the
actions of theAgent, so that the two partners share their
risk. The variation of wage Wj will depend upon the
probabilityreport,meaning
kj
ij
p
p
.
Hypothesisofthemodel
We consider that the progress of the gross in
come of the Principal is described by a continuous ran
dom variable x , whose sharing is determined by a varia
ble of effort e, controlled by the Agent, but which the
Principal cannot see. We will denote by F(x,e) the reparti
tion of the variable x and by f(x,e) its density, assuming
thatthesupportofthedistributiondoesnotdependupon
e.
We assume that the Agent is neutral or has risk
aversion, having the utility function strictly increasing,
concave and additively separable, under the form of
( ) ( ) ( ) , U W e U W V e = (therefore, 0, 0 U U ' '' > s ,
0, 0 V V ' '' > > ).
The Principal will be also neutral or of risk aversion, with
theutilityfunctionstrictlyincreasingandconcave,depend
inguponthelevelofresultandbythewagealreadypaid
to the Agent, B(XW). We will obviously
achieve 0, 0 B B ' '' > s .
The Agent will accept the contract proposed by
the Principal, if the utility expected will reach at least the
levelofreserveU .
Weassumethattheresultxdependsupontheef
fort submitted e and upon the random nature states,
meaninge :x=x(e, e ),with
| |
1 2
, x x x e .
Asconcernstheacceptance,theAgentwillchose
the effort e submitted, the nature being involved by its
random state e and will be remunerated by a wage
of
| | ( )
1 2
: , , W x x W W x
+
= .
Thecaseofsymmetricalinformation
IfthePrincipalcannoticeandcontroltheefforte,
accomplished by the Agent, we reach the situation of
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 17
symmetricalinformationmodel.Thecontractthatwillbe
proposedtotheAgentwillberepresentedbytheoptimal
solutionofthemaximizationissue,asregardstheestima
tion of utility, in conditions of participation restriction
conditions:
( )
( ) ( ) ( )
( ) ( ) ( ) ( )
,
max
e W x
E B x W x
E U W x V e U
>
. From the
conditions of I
st
order optimum, as regards the wage W
andtheefforte,wewillachieve:
( ) ( )
( ) ( )
| )
( ) ( ) ( ) ( ) ( ) ( ) ( )
2 2
1 1
1 2
, ,
, ,
x x
e e
x x
B x Wx
x x x
UWx
Bx Wx f xe dx UWx f xe dx V e
'
= e
'
( ( ' '
' + =
} }
(5)
The third condition of first order (as related to )
0
L
c
=
c
will ensure the saturation of the participation
restriction.
The first condition of relationship (5), for two
random values
| |
1 2
, , x x x x ' e might be writ
ten:
( ) ( )
( ) ( )
( ) ( )
( ) ( )
| |
1 2
, , ,
B x W x U W x
x x x x
B x W x U W x
' '
' = e
' ' ' ' '
.
Such relationship shows the optimal distribution of the
riskbetweenthetwopartners:themarginalratesofsubs
titutionoftheincomeareequal,nomatterthestateofthe
nature.
The second relationship of (5) can also be written under
the fol
form
( ) ( ) ( ) ( ) ( )
( )
( )
2
1
,
x
e
x
B x W x U W x f x e dx V e
'
( ' + =
}
,
and can be seen by the equality between: the marginal
disutilityoftheeffort(theleftpart)andthetotalmarginal
utility,weightedby ,inconditionsofsaturatingthepar
ticipationrestriction.
Wewillforwardsanalyzethewaytheincomesof
theAgent and of the Principal related to the variation of
result, in the situation when both partners have risk
aversion.
Analyzing the first relationship of (5), we can
write ( ) ( ) ( ) ( )
B x W x U W x ' ' = and, deriving as
related to the result x, we deduct that
( ) ( ) ( ) ( )
1
dW dW
B x W x U W x
dx dx
| |
'' '' =
|
\ .
.
Replacing with the expression achieved in (5),
wewillachieve:
( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )
B x Wx U Wx
dW
dx B x Wx U Wx B x Wx U Wx
'' '
=
'' ' ' '' +
.(6)
Since both partners have risk aversion, their utility func
tions will be strictly increasing and strictly concave, and
we will therefore achieve:
0, 0, 0, 0 B B U U ' '' ' '' > < > < . We therefore reach
that 0
dW
dx
> , sothat an increase of the result willdeter
mine (as it should be achieved) an increase of the Agent
income.
Analyzing the Principal, his income is xW(x),
and therefore we will achieve
( ) ( )
1 0
d dW
x W x
dx dx
= > (we used 1
dW
dx
< , since
therightpartreportofrelationship(6)hasthedenomina
torhigherthanthenominator).
Thecaseofasymmetricalinformation(moralhazard)
WeforwardsassumethatactionsoftheAgentare
no longer observable by the Principal. Therefore, the ef
fort e is chosen by the Agent, in order to maximize the
estimated utility,
( ) ( ) ( ) ( )
( )
2
1
arg max ,
x
x e
e U W x V e f x e dx
'
( ' ' e
}
,
and the problem of optimization at the Principals level
willbethefollowing:
( )
( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( )
2
1
,
max
arg max ,
e W
x
x e
E B x W x
E U W x V e U
e U W x V e f x e dx
'
>
( ' ' e
.
At this issue, one might notice the incitation re
strictionthatisonitsturnaproblemofoptimization,fact
that can be solved in some hypothesis, as in the discrete
situation,by usingthemethodoffirstorderapproxima
tion. We will replace the incitation restriction from the
previous problem, with a condition of first order, thus
reachingtothefollowingissue:
( )
( ) ( ) ( )
( ) ( ) ( ) ( )
( )
( ) ( ) ( ) ( )
2
1
2
1
2
1
,
max ,
,
, 0
x
x e W
x
x
x
e
x
B x W x f x e dx
U W x V e f x e dx U
U W x f x e dx V e
( >
( '
' =
}
}
}
(7)
Writing down the first condition of first order, as related
to the wage W, and the effort e, for the problem (7), we
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 3, MARCH 2012, ISSN 2151-9617
https://sites.google.com/site/journalofcomputing
WWW.JOURNALOFCOMPUTING.ORG 18
willdeduct
( ) ( )
( ) ( )
( )
( )
,
,
e
B x W x f x e
f x e U W x
' '
= +
'
.
The second condition of first order can be written by us
ingthesecondrestrictionof(7):
( ) ( ) ( ) ( ) ( ) ( ) ( )
( )
2 2
2
1 1
, , 0
x x
e
e
x x
Bx Wx f xe dx UWx f xe dx V e
( ( ' ''
'' + =
} }
.
Holmstrom and Shavell (1979) have shown that
if ( ) , 0
e
F x e s and 0 U'' < (the Agent has risk aver
sion),then 0 > .Inthissituation,thedistributionofrisk
isnolongerParetooptimal.Iftheparticipativerestriction
isexpostmet,theAgentwillachieveanestimatedutility
higher that the reserve utility U (U will be achieved
onlyinunfavorablemarketconditions).
4.CONCLUSIONS
Inthesituationofdiscretemodelwithtwolevels
ofeffort,theparticipationrestrictionoftheAgentissatu
rated,andthewageachievedwithintheoptimalcontract
is no longer constant, in contradistinction to the symme
trical information. Such situation proves that the moral
hazardbringstothePrincipalastrictlypositivecost,and
implicitly a lower profit than in the situation of the sym
metricalinformationcontract.ThewageoftheAgentwill
depend upon the level of result for such contract, where
thepaymentdependinguponresultwillbedoneinorder
to motivate the Agent. The advantage of establishing a
wage related to result is represented for the Principal in
finding out some information as regards the effort sub
mittedbytheAgent.Theincomedependsuponinforma
tionandwillincreasedependingupontheresult,ifabet
terresultisassociatedtoahighereffort.
Concerning the continuous model, in the situa
tion of the symmetrical information, the distribution of
the risk between the two partners will be optimal: the
marginalratesofsubstitutionoftheincomeareequal,no
matterthestateofthenature.Iftheinformationisasym
metrical,thedistributionofriskwillbenolongerPareto
optimal.Iftheparticipativerestrictionisexpostsatisfied,
theAgentwillachieveanestimatedutilityhigherthanthe
reserve utility U (U is achieved for only unfavorable
marketconditions).
REFERENCES
1. Besanko, D., Braeutigam R. (2005) Microeconomics, 2
nd
edition,JohnWiley&Sons,NewYork,ISBN0471457698
2. Bolton,P.,DewatripontM.(2005)ContractTheory,Massa
chusettsInstituteofTechnology,ISBN0262025760
3. Crawford, T. W., Kuperman A. (2006) Gambling on Hu
manitarianIntervention:Moral Hazard,RebellionandInternal
War,NewYork:Routledge,ISBN0415379466
4. Dembe,A.E.,Boden,L.I.(2000)MoralHazard:AQuestion
of Morality?, New Solutions, vol.10, nr. 3, pag. 257279,
ISSN10482911
5. Gravelle, H., Rees R. (1992) Microeconomics, 2
nd
edition,
LongmanGroup,London,ISBN0582023866
6. Holmstrom, B. (1979) Moral Hazard and Observability,
BelleJournalofEconomics,vol.10,pag.7491
7. Laffont, J. J. (1995) The Economics of Uncertainty and In
formation, Massachusetts Institute of Technology, ISBN 0
262121360
8. Laffont,J.J.,MartimortD.(2002)TheTheoryofIncentives:
ThePrincipalAgentModel,Princeton,ISBN0691091838
9. MachoStadler,I.,PerezCastilloJ.D.(1997)Anintroduc
tiontotheeconomicsofinformation,OxfordUniversityPress,
ISBN0199243271
10. Marinescu, D., Marin D. (2009) Microeconomie avansat.
Economia informaiei, ed. ASE, Bucureti, ISBN
9786065051744
11. Pauly, M. V. (1968) The Economics of Moral Hazard: Com
ment, American Economic Press, vol. 58, nr. 3, pag. 535
536,ISBN0201055244
12. Salanie,B.(2005)TheEconomicsofContracts:APrimer2nd
edition,TheMITPressBooks
13. Spence, A. M., Zeckhauser R. (1971) Insurance, Informa
tionandIndividualAction,AmericanEconomicReview,vol.
61,nr.2,pag.380387,ISSN00028282
14. Zeckhauser, R. (1970) Medical Insurance: A Case Study of
theTradeoffbetweenRiskSpreadingandtheAppropriateIncen
tives, Journal of Economic Theory, nr. 2, pag. 1026, ISSN
00220531