Moral Hazard -...
Transcript of Moral Hazard -...
Moral Hazard
Julio Davila
2009
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫πf (π, e)dπ <
∫πf (π, e)dπ
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that
πF (π, e)
]ππ
+
∫F (π, e)dπ <
∫F (π, e)dπ + πF (π, e)
]ππ
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that
πF (π, e)
]ππ
+
∫F (π, e)dπ <
∫F (π, e)dπ+πF (π, e)
]ππ
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫
I the principal wants the agent to choose e
I the agent wants to choose e
what would the principal offer to compensate the agent forchoosing e instead?
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫I the principal wants the agent to choose e
I the agent wants to choose e
what would the principal offer to compensate the agent forchoosing e instead?
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫I the principal wants the agent to choose e
I the agent wants to choose e
what would the principal offer to compensate the agent forchoosing e instead?
Julio Davila Moral Hazard
Principal-agent problem
1 two individuals: a ”principal” and an ”agent”
2 the principal gets a surplus π if the agent makes some effort e
3 π is unknown and its density f (π, e) depends on the effort e
4 e ∈ {e, e} with e < e, and F (π, e) < F (π, e) for all π
so that ∫E (π | e) < E (π | e)
∫I the principal wants the agent to choose e
I the agent wants to choose e
what would the principal offer to compensate the agent forchoosing e instead?
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed
max maxe,w
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ = e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ
u(w(e))
∫f (π, e)dπ = e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(e)) = e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
w(e) = u−1(e)
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed
maxe
maxw
∫[π − u−1(e)]f (π, e)dπ∫
w(e) = u−1(v(e))
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed
maxe
maxw
{∫πf (π, e)dπ − u−1(e)
}∫
w(e) = u−1(v(e))
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: observable effort
if both π, e can be observed
e∗ = arg maxe
{∫πf (π, e)dπ − u−1(e)
}∫
w∗ = u−1(e∗)
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed
max maxe,w
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed
max maxe,w
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC−f (π, e) + λeu
′(w(π, e))f (π, e) = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))!!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))!!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))!!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w(π, e))f (π, e)dπ ≥ e
FOC1
λe= u′(w(π, e))!!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ
u(w)
∫f (π, e)dπ = e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
u(w) = e
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxe
maxw
∫[π − w(π, e)]f (π, e)dπ∫
w = u−1(e)
FOC1
λe= u′(w(π, e))
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
e ∈ arg maxe′
{∫u(w(π))f (π, e ′)dπ − e ′
}
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
e ∈ arg maxe′
{∫u(w(π))f (π, e ′)dπ − e ′
}
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC
f (π, e) = λeu′(w(π))f (π, e) + µeu
′(w(π))[f (π, e)− f (π, e ′)
]
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC
f (π, e) = λeu′(w(π))f (π, e) + µeu
′(w(π))[f (π, e)− f (π, e ′)
]
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe + µe
[1− f (π, e ′)
f (π, e)
]
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe + µe
[1− f (π, e ′)
f (π, e)
]with λe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe+ µe
[1− f (π, e ′)
f (π, e)
]with λe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
assume, for all π,f (π, e) ≥ f (π, e ′)
and for some π ∈ [π′, π′],
f (π, e) > f (π, e ′)
then ∫[f (π, e)− f (π, e ′)]dπ = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
assume, for all π,f (π, e) ≥ f (π, e ′)
and for some π ∈ [π′, π′],
f (π, e) > f (π, e ′)
then ∫[f (π, e)− f (π, e ′)]dπ = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
assume, for all π,f (π, e) ≥ f (π, e ′)
and for some π ∈ [π′, π′],
f (π, e) > f (π, e ′)
then ∫[f (π, e)− f (π, e ′)]dπ = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
assume, for all π,f (π, e) ≥ f (π, e ′)
and for some π ∈ [π′, π′],
f (π, e) > f (π, e ′)
then∫ π′
...[f (π, e)− f (π, e ′)]dπ +
∫ π′
π′[f (π, e)− f (π, e ′)]dπ+∫ ...
π′[f (π, e)− f (π, e ′)]dπ = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
assume, for all π,f (π, e) ≥ f (π, e ′)
and for some π ∈ [π′, π′],
f (π, e) > f (π, e ′)
then∫ π′
...[f (π, e)− f (π, e ′)]dπ +
∫ π′
π′[f (π, e)− f (π, e ′)]dπ+∫ ...
π′[f (π, e)− f (π, e ′)]dπ>0!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
then, for almost all π,f (π, e) = f (π, e ′)
and henceE (π | e) = E (π | e ′)!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
then, for almost all π,f (π, e) = f (π, e ′)
and henceE (π | e) = E (π | e ′)!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
let ∫f (π, e)dπ = 1 =
∫f (π, e ′)dπ
then, for all π in some [π′, π′],
f (π, e) < f (π, e ′)
and henceE (π | e) = E (π | e ′)!!
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe+ µe
[1− f (π, e ′)
f (π, e)
]with λe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))<0, for all π in some [π′, π′]!!
with λe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe + µe
[1− f (π, e ′)
f (π, e)
]with λe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe + µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
u(w(π))
∫f (π, e)dπ − e ≥ u(w(π))
∫f (π, e ′)dπ − e ′
FOC1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))− e ≥ u(w(π))− e ′
∫FOC
1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫− e ≥ −e ′
∫FOC
1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫e ≤ e ′
∫FOC
1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0?
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫e ≤ e
∫FOC
1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ = e
∫e ≤ e
FOC1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
w = u−1(e)
∫e ≤ e
FOC1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫e ≤ e
∫FOC
1
u′(w(π))= λe +µe
[1− f (π, e ′)
f (π, e)
]with λe > 0, and µe = 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
maxw
∫[π − w(π)]f (π, e)dπ∫
u(w(π))f (π, e)dπ ≥ e
∫u(w(π))f (π, e)dπ − e ≥
∫u(w(π))f (π, e)dπ − e
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e
λe + µe
[1− f (π′, e)
f (π′, e)
]<[1− f (π, e)
f (π, e)
]µe + λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e
λe + µe
[1− f (π′, e)
f (π′, e)
]<[1− f (π, e)
f (π, e)
]µe + λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
λe + µe
[1− f (π′, e)
f (π′, e)
]<[1− f (π, e)
f (π, e)
]µe + λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
λe + µe
[1− f (π′, e)
f (π′, e)
]>[1− f (π, e)
f (π, e)
]µe + λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
λe + µe
[1− f (π′, e)
f (π′, e)
]>[1− f (π, e)
f (π, e)
]µe + λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
λe+µe
[1− f (π′, e)
f (π′, e)
]>[1− f (π, e)
f (π, e)
]µe+λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
λe + µe
[1− f (π′, e)
f (π′, e)
]>[1− f (π, e)
f (π, e)
]µe + λe
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
1
u′(w(π′))>
1
u′(w(π))
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, i.e.
1
u′(w(π′))u′(w(π)) > u′(w(π′))
1
u′(w(π))
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, then
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
FOC1
u′(w(π))= λe + µe
[1− f (π, e)
f (π, e)
]with λe > 0, and µe > 0
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
• if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, then
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
• if only π, e can be observed, for each e
w(π) = u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
• if only π, e can be observed, for each e
let π, π′ be such that π′ is relatively likelier if e, then
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
• if only π, e can be observed, for each e
w(π) = u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
• if only π, e can be observed, for each e
let π, π′ be such that π < π′ , then(?)
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
• if only π, e can be observed, for each e
w(π) = u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
• if only π, e can be observed, for each e
let π, π′ be such that π < π′ , then
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
only if f (π,e)f (π,e) is decreasing in π
• if only π, e can be observed, for each e
w(π) = u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
• if only π, e can be observed, for each e
let π, π′ be such that π < π′ , then
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
only if f (π,e)f (π,e) is decreasing in π
• if only π, e can be observed, for each e
w(π) = u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: non observable effort
• if only π, e can be observed, for each e
let π, π′ be such that π < π′ , then
1
u′(w(π′))u′(w(π)) < u′(w(π′))
1
u′(w(π))
only if f (π,e)f (π,e) is decreasing in π (Monotone LR Property)
• if only π, e can be observed, for each e
w(π) = u−1(e)
Julio Davila Moral Hazard
Principal-agent problem
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫u(w(π))f (π, e)dπ = e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫u(w(π))f (π, e)dπ = e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫u(w(π))f (π, e)dπ = e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫u(w(π))f (π, e)dπ = e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that
u
(∫w(π)f (π, e)dπ
)> e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫w(π)f (π, e)dπ > u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ−
∫w(π)u−1(e)dπ<
∫πf (π, e)dπ − u−1(e)
∧∫πf (π, e)dπ−
∫w(π)f (π, e)dπ<
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ<
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ<
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ<
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ<
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ<
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ<
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ <
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ <
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it can be implemented (even if non observable)at the same cost u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ <
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it can be implemented (even if non observable)at the same cost u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ ?
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ >
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may still be implemented if non observablebut at a higher expected cost
∫w(π)f (π, e)dπ
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ >
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may still be implemented if non observablebut at a higher expected cost
∫w(π)f (π, e)dπ (inefficiency)
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
∫w(π)f (π, e)dπ
Julio Davila Moral Hazard
Principal-agent problem: risk-averse agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
∨∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
∫w(π)f (π, e)dπ (ineff.)
Julio Davila Moral Hazard
Principal-agent problem
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫u(w(π))f (π, e)dπ = e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that
u
(∫w(π)f (π, e)dπ
)= e
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
1 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w = u−1(e)
2 if e is not observable, implementing e costs w = u−1(e)
if e is not observable, implementing e costs w ∈ RΠ such that∫w(π)f (π, e)dπ = u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
∫πf (π, e)dπ−
∫w(π)u−1(e)dπ<
∫πf (π, e)dπ − u−1(e)
q∫πf (π, e)dπ−
∫w(π)f (π, e)dπ<
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ <
∫πf (π, e)dπ − u−1(e)
q∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ <
∫πf (π, e)dπ − u−1(e)
q∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it can be implemented (even if non observable)at the same cost u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ <
∫πf (π, e)dπ − u−1(e)
q∫πf (π, e)dπ −
∫w(π)f (π, e)dπ <
∫πf (π, e)dπ − u−1(e)
if e is optimal, it can be implemented (even if non observable)at the same cost u−1(e)
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
q∫πf (π, e)dπ −
∫w(π)f (π, e)dπ ?
∫πf (π, e)dπ − u−1(e)
if e is optimal, it may not be implemented if non observablebecause of a too high expected cost
Julio Davila Moral Hazard
Principal-agent problem: risk-neutral agent
∫πf (π, e)dπ −
∫w(π)u−1(e)dπ >
∫πf (π, e)dπ − u−1(e)
q∫πf (π, e)dπ −
∫w(π)f (π, e)dπ >
∫πf (π, e)dπ − u−1(e)
if e is optimal, it can be implemented (even if non observable)at the same (expected) cost u−1(e)
Julio Davila Moral Hazard