Title: Consumption
1Consumption
- Anthony Murphy
- Nuffield College
- anthony.murphy_at_nuffield.ox.ac.uk
2Outline
- Consumption the biggest component of
GDP/national expenditure a good deal smoother
than income. - The two period model.
- Friedmans permanent income hypothesis PIH -
infinitely lived representative agent etc. - Modligianis life cycle hypothesis LCH finite
life, saving for retirement, population dynamics.
- Halls consumption function uncertainty,
rational expectations and the consumption Euler
equation. - Euler equations versus (approx.) solved out
consumption functions pros and cons. - Example of solved out consumption function for
US.
3Basic Two Period Model (1)
- Diagram
- Axes - c1y1 on horizontal axis (the present) and
c2,y2 on vertical axis (the future). - Intertemporal preferences Regular shaped
indifference curves (as opposed to linear or L
shaped ones). - Less than perfect trade-off between c1 and c2 so
want to smooth consumption over time. - Intertemporal budget line
- c1c2/(1r) y1 y2/(1r)
- (You can add an initial endowment a0(1r) if you
want to the RHS of the budget.)
4Figure 6.2(a)
Fig. 6.02(a)
Indifference curves Normal case
Consumption tomorrow
0
Consumption today
5Figure 6.2(b)
Fig. 6.02(b)
Indifference curves Zero substitution
Consumption tomorrow
0
Consumption today
6Figure 6.2(c)
Fig. 6.02(c)
Indifference curves Constant substitution
Consumption tomorrow
0
Consumption today
7Two Period Model (2)
- Budget constraint is a straight line thru
(y1,y2) point with slope equal to minus 1/(1r). - No borrowing or lending restrictions.
- Borrowing and lending rates are the same.
- Intertemporal budget constraint got by combining
period 1 and period 2 budget constraints - c1 a1 y1
- c2 a1(1r) y2
8Equilibrium in Two Period Model
- Equilibrium where highest attainable indifference
curve is tangential to the budget line. - You may be a borrower (c1 gt y1) or lender (c1 lt
y1) in period 1. - First order condition (FOC)
- slope of indifference curve
- slope of budget line
- ie. marginal rate of substitution (MRS) between
c1 and c2 1/(1 r).
9Figure 6.3(a)
Fig. 6.03
Optimal consumption borrower
Consumption tomorrow
0
Consumption today
10Figure 6.3(b)
Fig. 6.03
Optimal consumption lender
D
Consumption tomorrow
Y2
A
IC3
IC2
IC1
Y1
B
0
Consumption today
11FOC and Euler Equation
- Suppose preferences are additive over time so
U(c1,c2) u(c1) ?u(c2) where 0 lt ? lt 1 is a
discount factor. - MRS -dc2/dc1 holding U constant u'(c1) /
(?u'(c2)), where u'(c1) is the marginal utility
of c1 etc. - Thus FOC may be re-written as
- u'(c1) (1r)?u'(c2)
12FOC and Euler Equation
- u'(c1) (1r)?u'(c2)
- This is just a non-stochastic Euler equation!
- Note intuition indifferent between shifting one
unit of consumption between the present and the
future. - Complete smoothing of consumption (c1 c2) when
? 1/ (1r).
13CRRA Preferences
- CRRA preferences appealing constant savings
rate fixed allocation of wealth across assets
when interest rates constant. - u(c) c1-?/(1-?) with ? positive u'(c) c-? so
Euler equation is - c1-? (1r)?c2-?
- Take natural logs and note that ln(1r) is
approx. equal to r so - ?lnc2 (ln ?)/? r/?
14CRRA Preferences (2)
- The elasticity of intertemporal substitution EIS
is the coeff. on r in the Euler equation. - The EIS is 1/?, the inverse of the constant
coeff. of relative risk aversion. - The Euler equation implies that a higher interest
rate increases savings (c1 falls and c2 rises). - However, need to examine this effect in more
detail. (Why? Only looking at slope of budget
line not position of line).
15Playing Around with the Basic Two Period Model
- Rise in permanent income (both y1 and y2 rise)
outward parallel shift in budget line. c1 and c2
both rise. - Rise in current or future income budget line
shifts out parallel but not by as much as above.
Ditto for c1 and c2. - Current consumption is higher if future income
rises even if current income is unchanged! - A transitory rise in income may be represented by
a small rise in y1 (and possibly a offsetting
small fall in y2?). c1 and c2 only rise by a
small amount.
16Real Interest Rate Effects (1)
- Suppose r rises.
- Budget line swivels around (y1,y2) and is
steeper. - Need to look at substitution and wealth effects.
- Substitution effect given by Euler equation.
- Substitution effect on c1 is negative.
17Real Interest Rate Effects (2)
- For borrower, wealth effect on c1 is also
negative. - For lender, wealth effect on c1 is positive.
- Overall, the effect of a rise in real interest
rate on current consumption is not clear cut. - Empirical consensus is that interest rate effect
is small and negative. - Size of effect depends on incidence of credit
constraints and initial wealth, inter alia.
18Figure 6.9
Fig. 6.09
Effect of an increase in the interest rate
negative income effect for borrowers, positive
for lenders
D
D
A
Consumption tomorrow
Consumption tomorrow
R
R
A
B
B
Consumption today
Consumption today
(a) Student Crusoe(borrower)
(b) Professional athlete(lender)
19Credit Constraints (1)
- Assume that representative agent cannot borrow in
period 1. - Budget line is now discontinuous at (y1,y2).
- Budget line same as before in lending region i.e.
to left of (y1,y2). - Budget line drops down to horizontal axis in
borrowing region i.e. to right of (y1,y2).
20Credit Constraints (2)
- Now a corner solution at (y1,y2) is a distinct
possibility. - A rise in future income y2 has no effect on
current consumption if credit constrained. - A permanent or transitory rise in current income
has a large effect if credit constrained
(marginal propensity to consume is one). - Interest rate effects smaller or zero if credit
constrained.
21Figure 6.11
Fig. 6.11
With a credit constraint, the choice set is
reduced.
C
Consumption tomorrow
A
R
D
0
Consumption today
22Permanent Income Life Cycle Hypotheses (1)
- Can generalize analysis from two periods to many
or an infinite number of periods. - Standard model often called PIHLCH model.
- Original permanent income model of consumption
uses a rational, infinitely lived, representative
consumer. - Emphasis on different response of consumption to
permanent and transitory changes in income etc.
23Figure 6.5
Fig. 6.05
Temporary vs. permanent income change
D
Temporary R to R Permanent R to R
D
Consumption tomorrow
Y2
R
Y2
A
AR
Y1
Y1
B
B
B
0
Consumption today
24PIH and LCH (2)
- In the life cycle model, aggregate consumption
derived from behaviour of individual consumers
(of different ages) with finite lifespans. - Consumption smoothing and the life cycle pattern
of income mean that the young borrow, the middle
aged save and the retired dis-save. - Obviously, aggregate consumption depends
positively on population and income growth. - The level of savings also depends on length of
retirement relative to length of working life.
25Stochastic Income Interest Rates
- Solved-out consumption functions useful e.g.
- c1 k(r)W
- where wealth W a0(1r) y1 y2/(1 r) .
and k(.) is a known function of the real interest
rate r. - Difficult to derive exact results in PIH-LCH
model when income and interest rates are random. - Interest rates often assumed constant and point
expectations of future income used. - Halls (1978) insight look at Euler equation.
26Halls Consumption Equation
- The stochastic Euler equation for the infinitely
lived representative consumer is - u'(c1) E1(1r1)?u'(c2)
- where Et is the conditional expectation at time
t given the information set It. - Aside Can rearrange Euler equation to get
pricing kernel / stochastic discount factor. - Rational expectations assumed.
27When Does Consumption Follow A Random Walk?
- Under very special and unrealistic assumptions,
Euler equation implies that consumption is a
random walk. - When u(c) is quadratic and ? 1/(1r), then ?ct
ut with Et(utIt) 0 so both ?ct and ut are
innovations (unpredictable). - Since Et(?ctIt) 0, Et(ctIt) ct-1.
28Stochastic Euler Equations (1)
- Halls Euler equation is only a FOC, as noted
already. - It does not tell you anything about the effects
of income shocks, uncertainty etc. - To examine these sorts of issues, you need to
embed it in a bigger model! - This is one reason why some argue that
approximate solved out consumption functions are
more useful.
29Stochastic Euler Equations (2)
- Assuming CRRA preferences, joint normality of rt
and ct etc., the best we can do is - Et?lnct (ln ?)/? rt/?½(sct)2/?
- which shows that uncertainty increases savings
(since ct rises and ct-1 falls as the variance of
ct rises). - Long list of assumptions rational expectations,
representative agent, no credit constraints etc. -
30Testing Consumption Euler Equations
- Consumption Euler equations do not fare very well
empirically. - For example, if the basic model is correct, then
variables in the information set at time t-1
should not help in predicting ?lnct. - A natural test of this hypothesis is to include
the prediction of ?lnyt, using variables dated
t-1. or even t-2, in the regression of ?lnct on a
constant, rt and a proxy for (sct)2. - Predicted income growth is always highly
significant.
31Solved Out Consumption Function for US
- See separate note for example of solved out
consumption function for US. - Source Muellbauer (1994), Consumers Expenditure,
Oxford Review of Economic Policy.
32Summary (1)
- Rational consumers attempt to smooth consumption
over time, by borrowing in bad times (or when
young) and saving in good times (or in middle
age) . - Consumption is primarily driven by the present
discounted value of current and future non-labour
income and initial assets. - Financial market imperfections generate credit
constraints. Current income matters more for
credit constrained consumers.
33Summary (2)
- The effect of a change in the real interest rate
is ambiguous since wealth effects differ for
lenders borrowers. Overall the effect is
probably small and negative. - Over the life cycle, consumption is smoothed by
borrowing when young, saving in middle age and
dis-saving when retired. - Temporary changes in income (or other
disturbances) have small effects. Permanent
changes or shocks have large effects.