Professional Documents
Culture Documents
Implicit Incentives
Lecture 13
I Explicit incentives:
I they are provided by explicit contractual commitment by
principal
I these have been the focus so far
I Implicit incentives:
I arise when a principal or principals have some ex post
discretion how to respond to agent’s performance
I agent has implicit incentives to change performance to
in‡uence principals’discretionary decisions
I these are necessarily intertemporal incentives
Implicit incentives
Overview
Simple model
I Players: one manager (A) and many identical risk-neutral
…rms
I Two periods: t = 1, 2
I A’s performance is publicly observable in each period
yt = a + et
I et : e¤ort =) privately chosen by A (hidden action) at a cost
c ( et )
=) cost is strictly increasing and convex
I a: A’s ability =) initially unknown to everyone (no hidden
information)
=) prior: a N (a, σ2a )
I Key assumption 1: a is unknown to anyone
I F (a ) –prior distribution of a, common knowledge
explicit incentives
I Each period wage is determined in a competitive market for A
Marta Troya Martinez, NES Implicit Incentives 6 / 33
Career concerns Simple model Noisy performance Political cycles Simple model
Simple model
I Agent’s preferences:
I Per period utility: wt c (et ) - what’s the agent’s risk
attitude?
I Total expected utility: E [w1 c (e1 ) + δ(w2 c (e2 ))] where
δ < 1 is discount factor
I Firms preferences:
I Competition draws wages to the expected output =) in
equilibrium …rms break-even
I The wage is determined at the beginning of each period
through competition to recruit the worker
1 = c 0 (etFB )8t
I What is e1 ?
I Firms form a belief e e1 about e1 =) e e1 has to be correct in
the equilibrium: ee1 = e1
I w2 = E [y2 j y1 , e
e 1 ] = E [ a j y1 , e
e 1 ] + e 2 = y1 e e1 =) a is
in fact deterministic, given y1 and e e1
I Agent chooses e1 to maximize
w1 c (e1 ) + δE [a + e1 e
e1 ] = w 1 c ( e1 ) + δ [ a + e1 e
e1 ]
I Equilibrium:
δ = c 0 (e1SB ) and e1SB = e
e1
I First period e¤ort is positive despite the …xed wage!
Intuition: e¤ort increases the market’s belief about agent’s
ability and hence the future wage (career concerns)
I How does it compare to e1FB ? there is underprovision of e¤ort
New model
σ2 σ2
e1 ] = a 2 ε 2 + (y1
w2 = Ea [a j y1 , e e1 ) 2 a 2
e
σa + σ ε σa + σ ε
I Agent chooses e1 to maximize: w 1 c (e1 ) + δEy1 [w2 ]
σ2 σ2
w1 c ( e1 ) + δ a 2 ε 2 + ( a + e1 e1 ) 2 a 2
e
σa + σ ε σa + σ ε
I Equilibrium:
σ2
δ 2 a 2 = c 0 (e1SB ) and e1SB = e
e1
σa + σ ε
Marta Troya Martinez, NES Implicit Incentives 12 / 33
Career concerns Simple model Noisy performance Political cycles Noisy performance
I
suppose workers have only correlated shocks ρ = corr (εti , εtj )
=) Firms can use ytj to reduce the variance of "noise" εti
=)then place more weight on yti in estimating ai =)the
worker i makes more e¤ort =) CPI a good idea!
I suppose the workers only have correlated abilities
η = corr (ai , aj ) =) Firms can use ytj to reduce "prior"
variance of ai =)then place less weight on yti in estimating ai
=) the worker i makes less e¤ort =) CPI not a good idea!
=) CPI may be bene…cial but also have unexpected negative
consequences=)Meyer & Vickers (1997, J. of Pol. Economy )
Marta Troya Martinez, NES Implicit Incentives 15 / 33
Career concerns Simple model Noisy performance Political cycles Noisy performance
I Employee (A) works for employer (P) for 2 periods, but only
1-period contracts can be written
I A performs the same task in both periods and A’s performance
in period 1 provides information about A’s ability, or about
di¢ culty of task, or about environment, and this information
is useful to P for evaluating A in period 2: Cov (y1 , y2 ) > 0
I P can contract on yt
I At start of period 2, P will want to use this information:
I Good period-1 performance y1 will induce P to judge A’s
period-2 performance y2 against higher standard (“ratcheting
up” of standards)
I Higher y1 implies that for any given y2 , P will attribute less to
A’s e¤ort and more to ability/ease of task/environment, so
that for any given y2 , w2 will be less
Political cycles
I 2 periods: t = 1, 2
I Voters’preferences:
wt = y (1 τ ) + gt
y –income, τ –tax rate (exogenous), gt –public good provision
I Government budget constraint:
gt = η (τy rt )
I rt –rents extracted by incumbent politician =) rt r < τy
I η –incumbent’ s competence which is unknown to everyone
h i
=) η 1 1
U 1 2σ , 1 + 2σ and E (η ) = 1
I Incumbent’s choice is whether to use the tax revenues to
provide gt
I Voters observe gt but not rt and are obviously better o¤ with
a competent politician
I If a new politician is appointed, η is drawn randomly
Marta Troya Martinez, NES Implicit Incentives 19 / 33
Career concerns Simple model Noisy performance Political cycles Political cycles
Political cycles
v = r1 + pδ(R + r2 )
Political cycles
I What is r1 ?
I Voters form a conjecture about the rents in period 1 e
r1
I Then they estimate the talent
g1
e
η=
τy e
r1
I They reappoint incumbent if he/she is more talented than the
expected opposition e
η E (η ) = 1
I Since g1 = η (τy r1 ), e
η 1 is equivalent to
τy e
r1
η τy r1 =)reappointment probability:
τy e
r1 1 τy e
r1
p = prob η = +σ 1
τy r1 2 τy r1
Equilibrium
I Incumbent solves:
max r1 + pδ(R + r )
r1
1 τy e
r1
s.t. p = +σ 1
2 τy r1
I Solution:
τy e
r1
1=σ δ (R + r )
(τy r1 )2
and e
r1 = r1
I Assuming an interior solution:
1
r1 = τy σδ(R + r ) and p =
2
I Equilibrium rents are smaller:
I If the value of winning the elections is large δ(R + r )
I If the uncertainty about the talent is small (σ large)
Marta Troya Martinez, NES Implicit Incentives 22 / 33
Career concerns Simple model Noisy performance Political cycles Political cycles
Testable implications
Hypotheses
Data
Conclusions