l3 = θlι + (1
ʌ
— θ)l2 is feasible as the choice set of possible loadings is unbounded.
[T (m)](X3 )
max { kl - X32 + δ ʃ
∕m(c(X3, l) + v + u)f1(u)du g(v)dv
-∞
c 00
δ /
⅛c-
У m(c(X3,l)+ r + v + u)f2 (u)du g(v)dv j∙
Xc
kl3 - X32 + δ
J-00
[ m(c(X3, l3) + v + u)f1(u)du g(v)dv
— —o
c 00
δ /
⅛c-
/00
m(c(X3,13 ) + r + v + u)f2(u)du g(v)dv
k [θl^χ + (1 - θ)⅛] - [θX1 + (1 - θ)X2]2
-sr λ , -sr f ∙. ,→ zʌ. , -sr f ∙.
/■Xc
+δ
— — o
ec(X1,l1) (1 +c(X2,l2) y∞ m (θ [c(X1,ι^1 ) + υ + U +(1 - θ) [c(X2 ,l^2) + v + u]) f1(u)du g(v)dv
C 00
+δ /
Xc
θc(X1∕1b (1 — ff)c(X2,t2)
J' m (θ [c(X1,l^1) + r + v + u] +(1 - θ) [c(X2,l^2) + r + v + u]) f2(u)dug(v)dv
The second line uses the fact that l3 is feasible and hence the value under the optimum by definition
ʌ
has to be at least as high. The third line uses the definition of X3 and l3. Using Proposition 1 in the
above equation, namely that c(X1, l1) = c(X2, l2) = c we get (the second line utilizes the fact that
both m and — x2 are concave functions).
[T(m)](X3) > k [θl^1 + (1 - θ)⅛] - [θX1 + (1 - θ)X2]2
^ ,J- . . , i-.
Xc
+δ
— — о
Ω ~/ V 7.∖ /1 Ω∖^δV. 7.∖ . „
У m (θ [c(X1, /1 ) + V + u] +(1 - θ) [c(X2, Z2) + V + u]) f1(u)du g(v)dv
C 00
+δ /
Xc
ʌ ʌ
θc(X1,l 1) — (1 — ff)c(X2,l 2)
J' m (θ [c(X1,l^1) + r + v + u] +(1 - θ) [c(X2,l^2) + r + v + u]) f2(u)du g(v)dv
12 - (1 - θ)X22
ʌ ʌ I
> θklι + (1 - θ)k^2 - θX
+δ [θm ^c(X1 ,l^1)+ v + u) +(1 - θ)m (c(X2,Z^2 ) + v + u)] f1(u)du g(v)dv
+δ J' J' [θm (c(X1,l^1)+ r + v + u) +(1 - θ)m (c(X2 ,l^2)+ r + v + u)] f2 (u)dug(v)dv
θ fkl^1 - X12 + δ [ [ m fc(X1,l"1 ) + v + u) f1 (u)du g(v)dv
—00 ——∞
+δ J^ J^ m (c(X1,l^1)+ r + v + u) f2(u)dug(v)dv]
+(1 - θ) kl2
- X2 + δ [ [ m (c(X2,l^2) + v + u) f1(u)du g(v)dv
— — — —
+δX J^ m (c(X2,l^2) + r + v + u) f2(u)dug(v)dv]
θ[T(m)](X1) + (1 - θ)[T(m)](X2)
The last two lines are simple rearrangements and definition of the value function. We hence know
that the unique attractor, the value function V(X) is concave.
28