On the existence of weakly maximal programs for a multisector economy with consumption

On the existence of weakly maximal programs for a multisector economy with consumption

On the Existence of Weakly Maximal Programs for a Multisector Economy with Consumption Nikolaos S. Papageorgiou*‘** National Technical University Depa...

437KB Sizes 0 Downloads 38 Views

On the Existence of Weakly Maximal Programs for a Multisector Economy with Consumption Nikolaos S. Papageorgiou*‘** National Technical University Department of Mathematics, Zografou Campus Athens 157 73, Greece

ABSTRACT In this paper consumption hypotheses

we consider

over on the

an infinite data,

a growth planning

we prove

model horizon

that

there

for a multisector of continuous exists

economy

time.

a weakly

Under

maximal

with mild

feasible

program.

1.

INTRODUCTION

In this paper we consider an infinite horizon, continuous time economic growth problem and we establish the existence of a weakly maximal program (path). Our model also involves consumption. The question of existence of weakly maximal solutions (also known as weakly overtaking solutions), was considered in the past by Peleg [l], Brock and Haurie [2] and Carlson and Haurie [3]. From these works, Peleg considered a discrete time reduced (i.e., no consumption), infinite horizon growth model and proved the existence of weakly maximal stationary programs. On the other hand, Brock and Haurie and Carlson and Haurie concentrated their attention on control systems and Brock and Haurie proved the existence of a weakly maximal trajectory (weakly overtaking, in their terminology), by examining the associated Lagrange problem. Carl-

*Research supported by NSF Grant DMS-8802688. **Presently on leave at the Florida Institute of Technology, Department of Applied Mathematics, 150 West University Blvd., Melbourne, Florida 32901-6988.

APPLZED MATHEMATICS

AND COMPUTATION

48:177-186

0 Elsevier Science Publishing Co., Inc., 1992 655 Avenue of the Americas, New York, NY 10010

177

(1992)

0096-3003/92/$05.00

178

N. S. PAPAGEORGIOU

son and Haurie, on the other hand, are primarily concerned with the derivation of necessary and sufficient conditions for overtaking optimality and corresponding turnpike theorems. Our work in this paper is closer to that of Peleg,

2.

THE

hut our model

is more general

than his.

MODEL

We consider

a model

of economic

growth

with m-commodities,

so our

state space is Rm. As we already mentioned in the introduction, our planning horizon is infinite, i.e., is R,. Our economic process is described by a technology set G z W” x W m x W”. If (k, u, c)EG, then this means that if at a time instant capital k is available, then a level u of investment rate and a level c of consumption rate can be realized. We are also given a function u: W”XRmXW”-+” LJ describing the social utility (welfare) for every combination (k, u, c) of capital-investment rate-consumption rate. An admissible program (path) is a pair of absolutely continuous functions

k: R++WT and c: R++W” such that (k(t), k(t), b(t))~G a.e., k(O)= k,, c(0) = co with k,, ca~W+m and 0 0. Recall that by a well-known

k(s), capital

theorem

of Lebesgue,

the absolutely

continuous

functions

c( *) are almost

everywhere differentiable. Here k(t) denotes the at time t, k(t) is the investment rate and c(t) the rate of the initial endowment of The vector k o E WT represents

stock

consumption.

bundle allocated to capital, the vector CLEWS is the initial commodity consumption and r]&$T is for every time instant t > 0 an upper bound for the consumption vector c(t), which is assumed to be nonnegative in W”. In what follows we will denote the set of admissible programs by F. We can also think

of c(t)

as the

resource

stock

and

-c(t)

as the

resource

depletion rate. Since our time horizon is infinite to avoid issues about the convergence of integrals, which require restrictive hypotheses, we adopt as our criterion in evaluating admissible programs the weak maximality criterion. So a program (k, C)E F is said to be “weakly maximal,” if for any other admissible

program

(k’, c’) (i.e.,

(k’, c’)EF),

-

:li

~b[u(k(t),i(t),e(t))-u(K’(t),ic’(t),6.jf))]

we have

dt>O

This means that for every E > 0 and every b, > 0 we can find b > b, such that 1: u(k( t), k(t), k(t)) dt > 10”u(k’( t), k(t), C(t))] dt - E. In the terminology of [2], the program (k, C)E F is weakly maximal, if it is not overtaken by any other admissible program.

179

Weak1y Maximal Program Our goal is to establish

the existence

To this end, we will need problem.

of a weakly maximal pair (k, C)E F.

the following

hypotheses

on the data of our

G c W” x W” x W” is nonempty,

HYPOTHESIS (Hl).

closed

and con-

vex.

If k&3:,

HYPOTHESIS (H2). 4( Ilk 11)with 4: R++W+

HYPOTHESIS(H3). (k, u, c)eG,

(k, u, C)EG implies

continuous,

There

exists

monotone

Xe(0,

that (lvI[ + 11cII < and 4(O) = 0.

1) such that for every

k~R;l,

if

then v+ AkeRy.

HYPOTHESIS(H4).

There

exists a positive

constant

M > 0 such that if

IlkI1 2 M, k&2?, th en for all (k, u, c)EG, we have (k, u) < 0 (here (m, *) denotes the inner product in Rm). These hypotheses represent the technological constraints in our economy and are

more

or less

standard

in problems

of growth

theory.

So

hypothesis (H2) means that the investment and consumption rates are determined by the availability of capital. Hypothesis (H3) says that net investment cannot be lower than loss due to depreciation and finally hypothesis (H4) implies that if capital exists in sufficiently large quantities, then loss due to depreciation exceeds production.

HYPOTHESIS(H5). (i) (ii)

There

exists (I,

0, O)EG such that few:

u(k, 0, c)
and

k = x.

This hypothesis implies that utility level U = u(& 0, 0) can be maintained forever with no investment or consumption change if capital x is available. So capital g is a substitute for investment and consumption. Then hypothesis (H5) (i) says that utility level U cannot be exceeded without any investment, whereas hypothesis (H5) (ii) guarantees the uniqueness of capital stock Z.

180

N. S. PAPAGEORGIOU

Following there

exists

[2] we say that an admissible

program

(k, C)EF is “good,”

if

r9~W such that

jy[u(k(t), k(t), c(t))+] for all b > 0. This is equivalent

&

dt>e

to saying that

/+[u(k(t),ic(t),k(t))-ii]

dt>-a~.

0

b+m

In the terminology “eligible.”

of Takayama

There

HYPOTHESIS(H6).

[4, page 5891, such a program

exists at least one good program

is called

(k, c)EF.

This hypothesis implies that the underlying technological basis is rich enough to sustain a utility (welfare) level close to U. The next hypothesis is also about the richness

of the technology

set of our economy.

HYPOTHESIS(H7). For every peRrn \ {0}, sup{( p, u): (k, u, c)EG} Finally we will need a hypothesis about the utility function.

> 0.

HYPOTHESIS(H8). u: Rrn x W” x W” -+ R is a positively homogeneous, continuous concave function such that there exist a, b, CE L’(W+, RI”‘) =

L’,(W+ ) for which denotes

the concave

/a” u*( a( t), b(t), c(t)) dt > - 00, where u*( *, -, *) conjugate (i.e., concave Fenchel transform) of u( *, -,

-) (see 151). 3.

MAIN RESULT We will start with an auxiliary

feasible

result

that gives us some bounds

for the

programs.

LEMMA 3.1.

If hypotheses

suchthatllkIIm, llkllo~

( Hl)-(

H4)

llcll,, llCllo< M,

hold, then there exists M, > 0 for all (k, c)EF.

Weakly Maximal Program PROOF.

Let

181

(k, C)E F. Then by definition (k(t),

k(t),

we have a.e.

i(t))~G

Then by hypothesis (H3), we can find he(O, 1) such that i<(t)+ Ak(t) 2 0 a.e. * k(t) 2 k,eeht 2 0 in W”; i.e., k(t)&iT for all HEW,. Next let b > 0. Since k(-) is continuous for all b > 0, we can find tO~[O,

b] such that jlk( to) 11= max[jIk(t)II: te[O, b]]. We claim that Ilk(t,)II < max[ Ilk,, 11, (I111, M] = y. Suppose not. Then llk( to) )I > y and because of the continuity of k(m), we can find 8 > 0 small enough so that y < Ilk(t) II for all te[ t, - 19, t,]. But then since kc*) is R y-valued from hypothesis (H4), we have k(t))< 0 a.e.

(k(t),

on [t, - 0

*%llk(t)ll”
Ilk(to)II < Ilk(to- 0) II

* contradicting for all te[O,

the choice

b]. Since

on

of toEIO, b]. So indeed llk( to) II < y a IIk( t) II < y was arbitrary we deduce that

beR+

Ilk llmGY Then from hypothesis

(3.1)

(H2), we have

Ilk.

llac.~~(Y)

whereas since 0 < c(t) < q for all t&d+, Euclidean norm, we have

(3.2)

from

the

monotonicity

of the

lb IImGII1 II GY So from

Equations

(3.1),

(3.2),

and

(3.3)

(3.3) we see

IIc llapIlk IlaD < MI, where M, = max(y, NY)). Now we are ready to state and prove our main existence of a weakly maximal program.

that

Ilk II_,

theorem

Ili( II,__,

about

the

N. S. PAPAGEORGIOU

182

THEOREM 3.1. Zf hypotheses weakly maximal program.

(Hl)-(

H8)

hold,

then

there

exists

a

Let (k, C)EF be the good (eligible) program postulated by (H6). Using Lemma 3.1 and the continuity of u(. , . , . ), we get

PROOF.

hypothesis

that there exists

M, > 0 such that for all b > 0 we have

I;/-bu(k(t), k(t), b(t)) 0

M,

d+

whereas clearly 11l/ b/,b k(t) dt 11< M,. Then from the Heine-Bore1 rem we know that we can find (A*, &)&I x Ry such that for sequence

theosome

b, -+ 00 we have

$Lbnu(k(t),

i(t),

b(t))

dt-+P

and

1

s h

k(t) dt+^k

as n+a.

b,O Furthermore,

3.1 we have

from Lemma

and

Il$Lbnd(t)dtl(=

Because

of the convexity

\I+%) -co\1

b,

< 2M,

as n-+oo.

.r+O

of the technology

set G, for every

have

k(t) dt, $

s”‘ k(t) dt, ; Jbn6(t) ” 0 n 0

dt

n > 1 we

183

Weakly Maximal Program Passing (HI)),

to the

limit

as n +OJ

and since

G is closed

(see

hypothesis

we get (R, 0, O)EG.

Recall that (k, C)E F is a good program. So by definition -y&j and b, > 0 such that for all b > b,, we have

where

b, + 00 with b, 2 b,. Passing

there

exists

to the limit as n+ 00, we get

Note that because of the positive homogeneity and concavity of the inequality, we utility function (see hypothesis (H8)) and using Jensen’s have

;J,4”ikit), i(t),

k(t)) dt

k(t)&, Exploiting

the continuity

of u( -, -, a), in the limit as n + 00 we get

P
K =

So

{(A, y)mxWm: A< u( x, y, z), (x, y, +G}

Clearly, because of hypothesis (Hl) and (H8), K is a convex set, whereas because of hypothesis (H5), we have (U, O)+ K. So we can apply

184 the

N. S. PAPAGEORGJOU Minkowski

R m \ { (0, 0))

separation

theorem

forall(X, we can

y)~K.Ifp=O,thenwehave(p, find XER”’ such that (x, hypothesis (H7). Thus p z 0. Divide X+( Then

of convex

sets

and get

(CL, ~)ER

x

such that

for this

y)
6, y)
~?EW” consider

y~R”forwhich

y, z)EG. But then by p z 0 to get

this

contradicts

fi = p/p.

the following

minimization

inf[J~[“-Ujr(t),ijt),~(t))-(B,i(t))]dt:

problem

(j,w)eF]=m”

0

Let {(f,, w,,)}, > 1 E F be a minimizing sequence for the preceding variational problem. From Lemma 2.1, we know that {f,),~i{f,},~i { ri~~}“,i E Lz(R+) are all relatively w*-compact sets. Since L’,(R+) (the on the previous predual of Lz(R + )) is separable, the relative w*-topology sequences is metrizable; see [6], Theorem 1, page 426. So by passing to a subsequence

if necessary,

we may assume

f,, “-‘f, f,, sf and f, zf, f, $:f and ti,,

that

L”,(W+). So for every b > 0, we have L’,[O, b] (recall L”,[O, b] - Li[O, b]). Thus by using Mazur’s

in

[6] Corollary

14, page 422) and by passing

and since G is convex, b> 0 was arbitrary, we w)EF.Also because of know that the integral w*-1.s.c.

to a subsequence

ti,, 2 ti 2 2i, in

lemma (see if necessary

we get (f(t), j(t), ti~(t))~G a.e. on [0, b]. Since get that (f(t), f(t), ti(t))~G a.e. in R+. Hence (f, hypothesis (H8) and [5], Theorem 21, page 59, we functional (a’, b’, c’) -+ Joau( a’(t), b’(t), c'(t)) dt is

on L”,(W+)x

Lz(W+)x

L”,(R+).

Hence

m,=hn/m[u-u(f”(t),~“(t),ti”(t))-(c,f;”(t))]dt 0

~S~[ii-~(f(t),i(t),~(t))-(li,i(t),]dt 0

-m,=Srn[ii-u(f(t),i(t),~(t))-(B,i(t))]dt 0

185

Weakly Maximal Program In particular there exists b,

this implies

that for every

(g,

T)EF

and for every

E > 0,

> 0 such that

+eJb[u(g(t), tip)> i(t))-~(r(t)~f(t),~(t))]dt 0

(6 f(b)- g(b))

P-4) for all b > b,. Let ( g, r) = (k, C)E F be the good program is postulated by hypothesis (H6). Then we have

whose existence

/b[u(f(t),f(t),ti(t))--ii]dt+E 0 ZSb[~(kjt),L(t),djt))-ii]dt-(li,f(b)-k(b)) 0

for all b 2 b,. But since ) ( 6, f(b) - k(b)) 1 < IIfiI(. 2 M, and (k, C)E F is a good program, then so is (f, W)E F. We now claim that (f, ZD)E F is in fact the desired weakly maximal program. Suppose not. Then there exists &a > 0, b, 2 0 and (h, Z)E F such that

for all b 2 b,. Then since (f, W)EF is a good program, from inequality (3.4) above we have

(6, f(t)-

h(t))+

so is (h, Z)EF and

E

a/)(+)>+)> i(s))ds-~~(~(s).f(s),~(s))ds

N. S. PAPAGEORGIOU

186

for t 2 max(bO, b,) = b,.

Hence (fiJ-(t)-h(q)+E>GJ

*

-

‘(e,f(+qs))ds

s

0

““(e,f(s)-h(s))ds+E(t--bz)~Eg(t-h2) s 0

*tfi>

1

--

i!

l ~;(s)ds-;y),,

7

s b,

t-b,

(~,f(s)--h(s))ds+~EBt~O

t-b,

0

Recall that since (f, w), (h, Z)EF are good programs, we have +L as t-w. So in the limit as t+oo, we get l/t/;_/+)& 1/t/,“q+s E > Ed. But E > 0 was arbitrary. So &o = 0 a contradiction. Therefore we conclude that indeed (f, W)EF is the desired weakly maximal program. REFERENCES B. Peleg,

A weakly maximal

W. Brock over

an

golden-rule

Economic Reu., 14:574-581

International

and A. Haurie, infinite

time

program

On the existence

horizon,

Math.

for a multi-sector

economy,

(1973). of overtaking

of Operations

optimal

trajectories

Research,

1:337-346

(1976). D. Carlson Berlin,

and A. Haurie,

A. Takayama, bridge,

Znfinite Horizon Optimal Control, Springer-Verlag,

1987.

Mathematical

Economics,

Cambridge

University

Press,

Cam-

1985.

T. Rockafellar, Conjugate Duality and Optimization, Regional Conference Series in Applied Mathematics, Vol. 16, SIAM Publications, Philadelphia, 1976. N. Dunford and J. Schwartz, Linear Operators, Wiley, New York, 1958.