Anda di halaman 1dari 13

Professor Mark Gertler Macroeconomic Theory I Fall 2000 Lecture 2

Innite Horizon Consumption-Saving Decision under Certainty


We abstract here from labor supply. The representative consumer chooses {Ct+i } to i=0 max
X i=0

i u(Ct+i ), (0, 1)

(1)

subject to:

At+1 = Rt At + Wt Ct ,
i

(2) (3)

lim

Rt+i At+i
j=1 i Q

=0

Rt+j

with At given.

The subjective utility function u() is a concave function with the usual properties. The impatience parameter ensures that the subjective utility function does not blow up. At is the stock of nancial assets at time t, while Rt is the gross interest rate (principal plus net return on nancial assets) at time t. Wt is labor income (real wage). At is a predetermined endogenous variable: it is taken as given at the beginning of period t but it may evolve over time. The consumer can aect its path over time. Condition (3) is meant to rule out Ponzi schemes. The individual has to satisfy an intertemporal budget constraint: it eventually has to pay o any debt, it cannot continuously play with Ponzi schemes. The paths of Rt and Wt over time are taken as given. They are a certain and given sequence. They are exogenous from the point of view of the representative consumer. They may not be constant over time but they are stationary, or, in other words, bounded: lim Rt+i = R, lim Wt+i = W .
i

The dynamic budget constraint (2) characterizes the behavior of wealth over time: we can rearrange it: Rt At = Ct Wt + Rt+1 At+1 . Rt+1 (4)

This is a rst-order dierence equation in Rt At . Shifting (4) forward one period: Rt+1 At+1 = Ct+1 Wt+1 + Dividing (5) by Rt+1 : Rt+1 At+1 Ct+1 Wt+1 = + Rt+1 Rt+1 Rt+1 We can thus plug (6) into (4) and rearrange to get: Ct + Rt+2 At+2 Rt+2 Rt+1
{z

Rt+2 At+2 Rt+2

(5)

(6)

< 0 if the individual is borrowing

Ct+1 Wt+1 Rt+2 At+2 = Wt + + Rt At Rt+1 Rt+1 Rt+2 Rt+1

(7)

Then, shifting (6) forward one period and dividing by Rt+1 : Ct+2 Wt+2 Rt+3 At+3 Rt+2 At+2 = + Rt+2 Rt+1 Rt+2 Rt+1 Rt+2 Rt+1 Rt+3 Rt+2 Rt+1 We can then plug (8) into (7) to get: Ct + Ct+1 Ct+2 Wt+1 Wt+2 Rt+3 At+3 + = Wt + + + Rt At Rt+1 Rt+2 Rt+1 Rt+1 Rt+2 Rt+1 Rt+3 Rt+2 Rt+1 (9) (8)

By repeated substitution, iterating the procedure k times:


k X i=0

Ct+i
j=1 i Q

Rt+j

k X i=0

Wt+i
j=1 i Q

Rt+j

+ Rt At

Rt+k+1 At+k+1
k+1 Q j=1

(10)

Rt+j

The last term on the RHS of (10) represents discounted future wealth. Letting k and imposing the No Ponzi Game/transversality condition (3) we get an innite horizon intertemporal budget constraint:
X i=0

Ct+i
j=1 i Q

= Rt At + Ht ,

(11)

Rt+j 2

with: Ht = where Ht is discounted human wealth.


X i=0

Wt+i
j=1 i Q

Rt+j

The crucial assumption here is that there are perfect capital markets and that Rt is the same regardless of whether At is positive or negative. If we did not impose the terminal condition (3) the consumer would consume more than its wealth, we would have an unbounded consumption sequence. Hence the dynamic budget constraint plus the No-Ponzi game (NPG) condition yields the innite horizon intertemporal budget constraint. We may have an alternative way of rearranging the dynamic budget constraint: update (2) one period forward: At+2 = Rt+1 At+1 + Wt+1 Ct+1 , and then divide by Rt+1 and solve for At+1 : At+1 = Plug (13) into (2) and get: Ct+1 Wt+1 At+2 + Ct + = Rt At + Wt + . Rt+1 Rt+1 Rt+1 Iterating the procedure and imposing the solvency condition we eventually get to (11):
X i=0

(12)

At+2 Ct+1 Wt+1 + . Rt+1 Rt+1

(13)

(14)

Ct+i
j=1 i Q

= Rt At +

Rt+j

X i=0

Wt+i
j=1 i Q

Rt+j

Dynamic Programming 1. Solve recursively, working backwards. 2. Collapse the innite horizon problem into a sequence of two-periods problems.

V (Rt At ) = max

i u(Ct+i ) =
X i=0

(15)
i

= max u(Ct ) + = subject to:


Ct ,At+1

i=0

u(Ct+i+1 ) =

max {u(Ct ) + V (Rt+1 At+1 )}

At+1 = Rt At + Wt Ct , with At given. V () is the value function of the problem: it gives the value of the objective function evaluated at the optimum; it represents an indirect utility function. It gives the value of the discounted stream of consumption, assuming that the individual has already maximized, that she/he has received the maximum possible utility one can get. V (Rt+1 At+1 ) represents the maximized value of wealth tomorrow. Expressing the dynamic budget constraint in terms of Ct Ct = Rt At + Wt At+1 , and substituting in (15): V (Rt At ) =max {u(Rt At + Wt At+1 ) + V (Rt+1 At+1 )} .
At+1

(16)

Assuming V () is dierentiable we can write the FOC: u (Ct ) + Rt+1 V (Rt+1 At+1 ) = 0
0 0 0

(17)

The term u (Ct ) is the marginal cost of saving, i.e. the utility cost of transferring one unit of consumption to the future. Rt+1 is the return on saving, while V (Rt+1 At+1 ) is the discounted shadow value of wealth, the marginal gain in utility from an additional unit of nancial wealth. Rearranging: u (Ct ) = Rt+1 V (Rt+1 At+1 ) 4
0 0 0

(18)

The envelope condition: V (Rt At ) V (Rt At ) = + (Rt At ) (Rt At )


z
Direct Eect

}|

Indirect Eect

= 0 at the optimum

Rt At may aect the choice of At+1 through the dynamic budget constraint. By the envelope theorem this eect is zero. To evaluate the shadow value of an exogenous variable, one needs to look only at the direct eect: we can thus ignore the impact of Rt At on the control variable: V (Rt At ) = u (Ct ). Shifting (20) forward one period: V (Rt+1 At+1 ) = u (Ct+1 ).
0 0 0 0

V (Rt At ) At+1
{z

}|

At+1 At

(19)

(20)

(21)

What (21) tells is that the shadow value of wealth is equal to the marginal utility of consumption at the optimum. Plugging (21) into (18) we get the consumption Euler equation: u (Ct ) = Rt+1 u (Ct+1 ).
0 0

(22)

How many restrictions, if any, do we need to close the model ? An optimal consumption program has to satisfy an innite horizon budget constraint: thus the additional constraint we need is the budget constraint (11):
X i=0

Ct+i
j=1 i Q

= Rt At +

Rt+j

X i=0

Wt+i
j=1 i Q

Rt+j

We need either an initial or a terminal condition, the latter being the budget constraint. The solution has to satisfy two basic properties: 1. Consumption must depend on wealth, as opposed to income (Permanent Income Hypothesis). The assumption of perfect capital markets is crucial.

2. Consumption smoothing: because of the concavity of the utility function the consumer is better o with a smoother intertemporal path of consumption. Consumption smoothing does not mean that the intertemporal prole of consumption is at (Ct = Ct+1 ). This would occur when Rt+1 = 1. Lets now deal with a special utility function: u(Ct ) = 1 C 1 , 0 1 t = log Ct , = 1 (23)

This is the case of CRRA (constant relative risk aversion) or isoelastic preferences. This utility function has the property of homotheticity: consumption is a linear function of wealth, human and nonhuman, or in other words, the ratio between consumption and wealth is constant within each period. It is possible to obtain an analytical solution for the case of logarithmic utility ( = 1) or for the case of constant R. u (Ct ) = Ct > 0 u (Ct ) = Ct1 < 0
00 0

(24) (25)

We want to get a measure of the curvature of the utility function, by computing the elasticity of the slope of the utility function (marginal utility of consumption) with respect to Ct , which measure the percentage change in marginal utility over a given percentage change in consumption: u (Ct ) Ct u (Ct ) Ct u (Ct ) Ct1 0 = = 0 Ct = Ct = . u (Ct ) Ct Ct u0 (Ct ) u (Ct ) Ct
0 0 00

(26)

is the coecient of relative risk aversion: it is constant and independent of the level of consumption. It measures the curvature of the utility function: it tells how the slope of the utility function is changing as consumption varies. A higher means a more concave utility function, thus a higher curvature of the utility function. If = 0 the individual is risk neutral and she/he does not care about consumption smoothing, the utility function is 6

linear and we have a constant marginal utility of consumption. As increases, risk aversion increases and so does the willingness to smooth consumption. There is a direct link between the degree of risk aversion and the desire to smooth consumption over time: in face of more uncertainty, with a higher degree of risk aversion, the individual wishes to do more consumption smoothing. The Euler equation in case of CRRA utility function:
Ct = Rt+1 Ct+1 .

With isoelastic preferences the consumption Euler equation is log linear in consumption growth. Rearranging:
1 Ct+1 = (Rt+1 ) = (Rt+1 ) , Ct

(27)

where =

is the intertemporal elasticity of substitution: it gives the sensitivity of the

intertemporal pattern of consumption to the interest rate. If Rt+1 > 1/ the individual picks an upward sloping intertemporal prole of consumption. The LHS of (27) is the gross growth rate of consumption Xt+1 : Xt+1 = Ct+1 Ct+1 Ct = 1 + gc , with gc = . Ct Ct

We can write (27) in a log linear form: ct+1 ct = rt+1 + log , (28)

where ct = log Ct , and rt+1 ' log Rt+1 = log(1 + rt+1 ). The LHS of (28) is the percentage change in consumption: ct+1 ct = log Ct+1 ' Ct+1 Ct = Xt+1 Ct

What (28) says is that consumption growth varies with the interest rate. The intertemporal elasticity of substitution tells the impact of an increase in Rt+1 on the growth rate of consumption. If Rt+1 increases there is a redistribution of consumption from today to tomorrow according to . If the interest rate increases, consumption today, Ct , decreases, 7

other things being equal; consumption is shifted toward the future, since the relative price of future consumption decreases. As , the coecient of relative risk aversion, increases, decreases: more risk aversion implies a greater desire to smooth consumption, and a less steep intertemporal prole of consumption. If there is a high degree of risk aversion, we have a small intertemporal elasticity of substitution: consumption growth is not very responsive to changes in the interest rate. More risk aversion means a smaller gain in marginal utility from an intertemporal reallocation of consumption. Hence, there is less attention to monetary gains in terms of adjusting the intertemporal prole of consumption: if is small, then there is a high desire to smooth consumption. On the other hand if is low, is high and consumption growth is very sensitive to changes in the interest rate. If utility is nearly linear, the degree of risk aversion, as represented by is pretty low; consequently the elasticity of intertemporal substitution is pretty large: what happens is that u0 () is nearly constant, hence the elasticity of marginal utility, which is also the coecient of relative risk aversion is close to zero: there is a strong elasticity of intertemporal substitution: the consumers are very willing to shift consumption over time in response to changes in the interest rate.. The Euler equation (27) gives the gross rate of change of consumption as a function of variables known at the current moment. It could be interpreted as suggesting that the consumer need not form expectations of future variables in making their consumption/saving decisions. However, it is clear from the innite-horizon intertemporal budget constraint that the individual cannot plan without knowing the entire path of both the real wage and the interest rate. In terms of the Euler equation, equation (27) only determines the rate of change, not the level of consumption. Although it is dicult in general to solve explicitly for the level of consumption, this can be done easily when the utility function is of the CRRA family: consumption is proportional to wealth; the marginal propensity to consume out of wealth is a function of the expected path of interest rates. We conjecture the following solution: Ct = t ({Rt+1+i } )(Rt At + Ht ), i=0 8 (29)

where t , the marginal propensity to consume out of wealth, depends in general on {Rt+1+i } , i=0 the sequence of current and future gross interest rates. It may be time varying, but it is exogenous from the point of view of the consumer: it is independent of individual characteristics. If t was constant between individuals then we can easily aggregate. Now note: Ht =
X i=0

Wt+i
j=1 i Q

= Wt +

Rt+j

X i=1

Wt+i
j=1 i Q

= Wt +

Rt+j

X i=0

Wt+1+i
j=1 i Q

= Wt +

Rt+1+j

Ht+1 . Rt+1

(30)

Now plug (29) into the dynamic budget constraint (2): At+1 = Rt At + Wt t (Rt At + Ht ). Add and subtract Ht in the RHS of (31) and rearrange to get: At+1 = (1 t )(Rt At + Ht ) + Wt Ht Rearranging (30) we get: Wt Ht = Plugging (33) into (32): At+1 = (1 t )(Rt At + Ht ) Rearranging: (Rt+1 At+1 + Ht+1 ) = (1 t )(Rt At + Ht ). Rt+1 one period forward, rearrange it and write: Rt+1 At+1 + Ht+1 = Using the Euler equation (27) into (36) we get to: Rt+1 At+1 + Ht+1 = R Ct t+1 , t+1 (37) Ct+1 . t+1 (36) (35) Ht+1 . Rt+1 (34) Ht+1 . Rt+1 (33) (32) (31)

The LHS of (35) is the discounted value of next periods wealth. Now we can update (29)

and by using (29): Rt+1 At+1 + Ht+1 = Now we are ready to plug (38) into (35):
Rt+1 t (Rt At + Ht ) = (1 t )(Rt At + Ht ). t+1 Rt+1 Rt+1 t (Rt At + Ht ). t+1

(38)

(39)

Simplifying we get a rst-order non linear dierence equation in t as a function of Rt+1 :


1 Rt+1

t = (1 t ). t+1

(40)

In the special case of log utility ( = 1), the marginal propensity to consume out of wealth is constant, since there is nothing driving t : thus we can solve for t (t = t+1 = ): =1 = 1 . (41)

When = 1 there is no intertemporal substitution, so is a constant: we have log utility and consumption is equal to a constant fraction of wealth. Income and substitution eects cancel out. For the case of constant interest rate, R does not change, and once again there is nothing driving t : = 1 R1 . Note: Ct = (1 R1 )(Rt At + Ht ) < 0 (the substitution eect dominates) R > 0 (the income eect dominates) < 1 R > 1 (43) (42)

In the rst case, we have a relatively high intertemporal elasticity of substitution and a relatively low risk aversion, hence the substitution eect dominates, and an increase in the interest rate has the eect of increasing savings, shifting consumption from today to 10

tomorrow. In the second case we have a relatively high coecient of relative risk aversion and the income eect dominates: an higher interest rate translates into higher wealth and higher consumption today. The time pattern of consumption is smoother. There is more consumption smoothing. In the case of log utility these two eects oset each other. Now we deal with the log linear approximation of (40): the LHS is already log linear while for the RHS we want to take a rst-order Taylor expansion around the steady state: we can write (40) as follows: ( 1) log Rt+1 + log + log t log t+1 = log(1 elog t ) (44)

We can take a rst-order Taylor expansion of the RHS around the steady state value of log t as follows: elog ( t ), (1 elog )

(45)

where is steady-state value of t and it is given by (42). Then we can write (45) as: 1 R1 1 1 ( t ) = 1 1 ( t ). R R Now we can write (44) in terms of deviations from steady-state values: ( 1)t+1 + t t+1 r 1 = 1 1 t , R
! !

(46)

(47)

where a hat variable denotes its log deviation from its steady state value: xt = log Xt log X = xt x Then, by rearranging (47) we can write: 1 t = (1 )t+1 + t+1 r R1

(48)

which is a forward-looking rst-order linear dierence equation in t . r t = 1 [(1 )t+1 + t+1 ] R 11 (49)

Updating one period forward: t+1 = Plugging (50) into (49): r r t = 1 (1 )t+1 + 1 [(1 )t+2 + t+2 ] . R R Iterating:
X t = 1 (1 ) 1 R i=0 R X i=1

[(1 )t+2 + t+2 ] r R1


)

(50)

(51)

!i

rt+1+i

(52)

or:

t = (1 ) Note that, other things being equal:

R1

!i

rt+i .

(53)

if > 1, rt+i > 0 = t < 0 (the substitution eect dominates) if < 1, rt+i > 0 = t > 0 (the income eect dominates).

(54)

Note the analogy between (54) and (43). As one can see an increase in the interest rate over its steady state value, given wealth, has two eects. The rst is to make consumption later more attractive by raising the opportunity cost of current consumption: this is the substitution eect. The second is to allow for higher consumption now and later: this is the income eect. In general, the net eect on the marginal propensity to consume is ambiguous. For the logarithmic utility function, however = 1, and the two eects cancel, as one can see from (54). Then the marginal propensity to consume does not deviate from its longrun level ( t = 0): it is then constant, independent of the path of interest rate and exactly equal to (1 ). If the utility function is of a more general CRRA formulation, the marginal propensity to consume out of wealth depends on the expected sequence of rates of return on nancial assets. The sign of the eect of changes in the expected rate of interest depends on whether the degree of risk aversion is less or greater than one, or equivalently on whether the elasticity of substitution is greater or less than one. 12

References
[1] Blanchard, Olivier J. and Stanley Fischer (1989). Lectures on Macroeconomics. Cambridge: MIT Press. [2] Romer, David (1996). Advanced Macroeconomics. New York: McGraw-Hill.

13

Anda mungkin juga menyukai