Model Undergraduate Economics Exam Answer

How does the inter-temporal optimization model contribute to the understanding

of consumption at the aggregate level?

Inter-temporal optimization models argue that current consumption is

determined by a life-time optimization problem, and thus is determined by life-time or
permanent income rather than current income, and thus tends to be smoothed over an
individual’s life. In this essay I will consider the two main variants of this approach:
the life cycle model and Hall’s Rational Expectation Permanent Income Hypothesis
(hereafter REPIH). Both assume a representative individual framework, and thus,
though their conclusions are about individual behavior, they have simple aggregate
interpretations for an economy of n representative individuals. I will argue that the
models contribute to our understanding of consumption in a number of important
areas, but do leave a number of observed aggregate effects unexplained, due to limits
imposed on predictive power by the assumption of a perfectly rational representative
agent.
The central tenet of inter-temporal optimization models is that agents attempt
to keep the marginal utility of expenditure constant across time (in discounted terms).
Because the marginal utility of expenditure and expenditure itself are assumed to be
monotonically related this leads to consumption smoothing. The major assumptions
are: that agents have inter-temporally additive preferences; face perfect capital
markets (to facilitate smoothing); and form rational expectations. The key differences
between the life cycle model and the REPIH are the assumptions made about the
functional form if utility functions and issues of certainty. The general form of the life
cycle model assumes certainty over future income and tends to assume either
quadratic preferences or preferences of a constant elasticity of substitution (CES)
form. Individuals are assumed to maximize utility: U = U(Ct) = (1 + )-t U(Ct, Zt)
subject to the lifetime budget constraint: tt=1Ct/(1 + r)t = A0(1 + r) + tt=1Yt/(1 + r)t .
Here: Ct = consumption in period t; = rate of time preference; A0 = initial
non-labour wealth; Yt = labour income at period t; and r = real rate of interest
(assumed constant across all t). It is also assumed that = r for all t.
The Zt are variables affecting the desirability of consumption at different
points in the life cycle. The Euler equation from this optimization problem takes the
form: λ(Ct, Zt) = µ((1 + r)/(1 + t. µ is merely the Lagrange multiplier and thus is
constant, so the marginal utility of consumption is constant across time, subject to the
Zt variables being constant. This implies a life-time pattern of consumption with high
savings during adult working life, and then dissaving during retirement, so as to
ensure constant consumption across all periods. This has two key implications for
aggregate consumption saving behavior. Firstly, an increase in population growth
should increase the aggregate rate of saving as the % of the population in young, high
saving periods should increase. Secondly younger cohorts have higher life-time
resources, and thus their increased saving will outweigh the dissaving of the older
generation, which was based on a lower projection of life time income. If we allow
the Zt variables to vary we can allow the marginal utility of consumption to be greater
in some periods than others e.g. periods with a young family, but the broad shape of
the life time consumption pattern, and thus the aggregate conclusions, are not altered.
The life cycle theory can also shed light on the effects of asset price
movements on aggregate consumption. If we additionally assume quadratic
preferences, the solved out form of the consumption function can be written Ct =
Wt/Kt, where Wt = wealth = the right hand side of the budget constraint, and Kt = 1 +
1/(1 + r), which is the inverse marginal propensity to consume out of assets. Any
perceived permanent increase in asset prices (e.g. a permanent rise in house prices)
will increase current consumption and thus aggregate consumption. Similarly any
permanent change in r will change the propensity to consume out of assets and thus
affect current and aggregate consumption ( as an increase in r leads to decrease in K,
thus increasing W/K), though it should be noted that assuming a constant elastic form
of utility preferences with σ > 0 makes the effect of permanent changes in r
ambiguous.
Under uncertainty concerning expected future income streams consumers are
assumed to equate the expected marginal utility of income across all periods. Here to
make the Euler equation tractable we must assume quadratic preferences. Again
assuming r = and is constant across all periods, we have Ct = E(Ct+1). Given that
Ct+1 = E(Ct+1) + εt+1 where εt is an unexpected shock distributed iid ~ N(0, σ2), Ct+1 =
Ct + εt+1. Moving back one time period then gives: Ct = Ct-1 + εt.. This is the Hall
stochastic Euler equation, indicating that consumption in the current period is equal to
last period consumption bar any shocks to consumption. This implies the same
consumption smoothing broad conclusion as the life cycle model, but in a stricter
sense as it cannot allow for the possible variance due to Zt style variables. The solved
out consumption function is Ct = EtWt/Kt = A0(1 + r)/Kt + Etyp, yp = permanent
income, and thus it has the same asset price effect predictions as the life cycle model.
But one key explanatory failure of these models is that, though they predict
individually smooth consumption, aggregate consumption seems to be smoother than
the theories would suggest; the Deaton paradox. Statistical tests indicate income has a
unit root. Under the REPIH, the consumption shock will equal a shock to permanent
income. Thus the variance of aggregate consumption innovations should be similar to
that of current income innovation, if income has a persistent unit root process. But
while cross-country data on income seems to follow a random walk, data on
consumption is significantly smoother (see Muellbauer 1994), yet the key prediction
of the models are the smoothness of consumption.
There are two main explanations of this excess smoothness that shed light on
the usefulness of the models. Firstly it is argued that the paradox is explained by
credit constraints or precautionary savings behavior (Muellbauer 1994). The former is
assumed away in both approaches. Importantly, the assumption of quadratic
preferences precludes precautionary savings behavior, and thus if this explanation is
correct then the life cycle model is adjustable to include such (e.g. Caroll 97) but the
REPIH is not as it must assume quadratic preferences to draw its solution. Secondly it
is argued that the solutions to both models are too demanding to be on the rational
capacity of consumers to be practically implemented. Thaler (1994) further argues
that life-cycle saving patterns are not repeatable experiments – and thus there is little
chance for the development of accurate decision heuristics either. The assumption of
the rational consumer is key to both models, but it seems bounded rationality is an
important concern that questions their conclusions. Finally, to move successfully from
a representative consumer to aggregate conclusions, the REPIH must assume either
infinitely lived consumers or those whose bequest motive is strong enough for them to
be modeled as such (Barro 74). Otherwise existing households break the
inter-temporal link between Ct and Ct+1. Muellbauer argues that bequest motives are
rarely strong enough to make this a reasonable assumption.
Inter-temporal optimization models contribute to the understanding of
aggregate consumption as, partly due to their representative individual nature, they
have a number of conclusions for individual behavior that have corresponding
aggregate conclusions. However the assumption of non-boundedly rational consumers
and the necessary assumption of quadratic preferences under uncertainty are short
comings of such models that limit their predictive scope for aggregate behavior.