You are on page 1of 16

EC201 Intermediate Macroeconomics EC201 Intermediate Macroeconomics


Lecture 35-36: Consumption

Lecture Outline: - Keynesian consumption function and the consumption puzzle; - Fishers two period model of consumption; - Life Cycle and Permanent Income theory of consumption; - Consumption as a random walk; Essential reading: Mankiw: Ch. 17 Keynesian Consumption Function When we discussed the Keynesian Cross model and the IS-LM model we did introduce the Keynesian Consumption Function: C t = C 0 + cYt The main features of that function are: a) the only determinant of current aggregate consumption is current income; b) the Marginal Propensity to Consume (MPC) is 0 < c < 1 ;
c) the Average Propensity to Consume (APC) income; The properties a), b) and c) are called the Keynes Conjectures. Property b) means that only a part of current income is consumed and so a proportion 1 c becomes saving. Property c) says that as income increases (meaning consumers become richer) consumers will save a larger fraction of income. Graphically the Keynes Consumption function looks like:


Ct C0 = + c is decreasing with Yt Yt


C0 slope=APC Y


The slope of any ray connecting zero with a point on the consumption function is the Average Propensity to Consume. As income increases a ray connecting zero with a point on the production function becomes flatter meaning that the slope is decreasing. Why the slope those rays is the APC? Consider the point C1 , Y1 on the consumption function. Take a ray starting from zero and passing through the point C1 , Y1 . Then you have a triangle (a right triangle). From basic trigonometry the slope of the ray is then the opposite side ( C1 ) divided by the adjacent side ( Y1 ).


C1 slope=C1/Y1 Y1 Y

Now we can ask: Is the Keynesian Consumption function a good representation of consumers behaviour? This is an empirical question. By looking at early empirical evidence (in the 30s and 40s) we had:

1) Cross-sectional evidence: this was evidence coming from surveys about households at a given point of time. For example a sample of 1000 consumers in 1934. The results from this evidence were: a) Richer households consumed more than poorer ones MPC > 0 b) Richer households saved more than poorer ones MPC < 1. c) Richer households saved larger fractions of their income APC asY . d). The correlation between current income and current consumption was found to be very strong (this was found during the Great Depression). Therefore according to this evidence it seemed that the Keynesian Consumption Function was a good representation of consumers behaviour. 2) Time series evidence: in 40s new pieces of evidence about aggregate consumptions were found by Simon Kuznets (a Nobel prize winner). He created a set of data from the US national accounts from 1869 to the 1940s on aggregate Y and C. According to the Keynes Consumption Function aggregate consumption should grow more slowly than income. This is because as Y increases, C also increases but proportionately less than income. Moreover as income increases APC should decrease. Kuznets found that the ratio C/Y was very stable in long time series data. This implies that C grew at the same rate as income and as income increased APC did not fall. Therefore we have two different pieces of evidence giving very different results. The difference between the two was that the first one was cross-sectional in detail (they looked at a snapshot of the economy at a point) whereas Kuznets study was of a time series nature (it looked at the economy over many points in time). So the evidence seemed to indicate that there were two consumption functions: a short-run consumption function which seemed to conform to Keyness conjectures and a longrun consumption function in which the APC was basically constant. This is known as the Consumption Puzzle. We can see how this looks with the following graph:

Consumption function from long time series data (constant APC )

Consumption function from crosssectional household data (falling APC )

Fishers Two Period Model of Consumption The Keynesian Consumption Function is not microfounded. It is not derived from a model of optimal behaviour of consumers. Here we look at microfoundations of aggregate consumption and see if the Keynesian Consumption Function is consistent with a microfounded analysis. The following analysis is due to Irving Fisher (the one of the Fisher Effect). The model is in practice very similar to the one we did for Ricardian Equivalence. Consider a representative consumer (there are many consumers but they are all equal) that lives for only two periods. There is no government. The consumer is rational and forward looking. Meaning: he maximises his own utility over his lifetime subject to his lifetime budget constraint and in deciding what to do today he takes into account what it will happen in the future. The consumer can lend and borrow at the real interest rate r . The budget constraint in period 1 is:

C1 + S1 = Y1


where C is consumption, S denotes saving and Y is income. In period 2 the budget constraint is:

C 2 = Y2 + (1 + r ) S1


Obviously in period 2 there is no saving since the consumer dies and so it will 4

consume all his income in period 2. The intertemporal budget constraint of our consumer is:
C1 + C2 Y = Y1 + 2 1+ r 1+ r 4)

The present value of lifetime consumption is equal to the present value of lifetime income. Graphically the intertemporal budget constraint looks like:


(1 + r )Y1 +Y 2

Consumption income periods in

= both


Y1 +Y 2 (1 + r )


On the vertical axis we put consumption in period 2 and on the horizontal axis the consumption in period 1. The intertemporal budget line in the graph shows all combinations of C1 and C2 that just exhaust the consumers resources. The point C1 = Y1 and C 2 = Y2 showed in the graph is always feasible and is on the budget line. Our consumer can move along the budget line by saving and borrowing. The intertemporal budget constraint implies a trade-off between consumption in

period 1 and in period 2. If our consumer wants to consume more in period 2

compared to C 2 = Y2 he must consume less in period 1 in order to save. By doing that he moves up on the budget line. If he wants to increase consumption in period 1 he must accept less consumption in period 2 and he has to borrow in period 1 so he moves down the line. The slope of the intertemporal budget constraint is (1 + r ) . This is the opportunity cost (the relative price) of C1 in terms of C2. For a given

stream of income if the consumer wants to decrease consumption in period 1 by C1 , then according to 4) he can increase consumption in period 2 by C 2 = (1 + r )C1 so

C 2 = (1 + r ) . C1

The optimal consumption choice is where the intertemporal budget constraint is tangent to an indifference curve. An indifference curve shows all combinations of C1 and C2 that make the consumer equally happy.




The optimal consumption choice in both periods (C*1,C*2) is where the budget line is tangent to the highest indifference curve. At point O we have that MRS = 1 + r . The Marginal Rate of Substitution (MRS) is
the amount of C2 the consumer would be willing to substitute for one unit of C1.

How Consumption Responds to Changes in Income

Now we can ask what happens to the consumption choice over the life time if there is an increase in current Y1 or future income Y2 . What happens to current consumption and to future consumption? An increase in Y1 or Y2 shifts the budget line outward. Provided the consumption in both periods is a normal good, then both C1 and C 2 increase, regardless whether the income increase occurs in period 1 or period 2.


Difference between the Keynesian Consumption Function and the Two-period model: According to Keynes current consumption depends only on current income. In the two period model current consumption depends only on the present value of lifetime income. The timing of income is irrelevant because the consumer can borrow or lend between periods. How Consumption Responds to Changes in Interest Rate The interest rate is a price so a change in the interest rate creates an income and substitution effect. An increase in r rotates the budget line around the point (Y1,Y2 ) that becomes steeper. The optimal consumption bundle before the change in the interest rate is point A. After the increase in the interest rate the consumption choice is point B. As depicted in the graph C1 falls and C2 rises. However, it could turn out differently depending if consumption in period 1 is a normal or an inferior good. Income effect: If consumer is a saver, the rise in r makes him better off since he becomes richer. This tends to increase consumption in both periods. Substitution effect: The rise in r increases the opportunity cost of current consumption, which tends to reduce C1 and increase C2. Both effects will result in an increase in C2. Whether C1 rises or falls depends on the relative size of the income and substitution effects.


Y2 C1


In the microfounded two period model the timing of income is irrelevant since by assumption consumer can borrow and lend across periods. If our consumer learns that his future income will increase, he can spread the extra consumption over both periods by borrowing in the current period. This means that future consumption can be an important determinant of current consumption. Therefore the Keynesian Consumption Function seems not be very consistent with this microfounded analysis. However suppose that our consumer faces borrowing constraints (= liquidity constraints), then he may not be able to increase current consumption and his consumption may behave as in the Keynesian theory even though he is rational and forward-looking. The borrowing constraint takes the form: C1 Y1 . Consumer cannot borrow and so it cannot consume more than his income in period 1. Graphically:


The budget line with a borrowing constraint




The borrowing constraint can be non binding: C1 < Y1 . In this case the borrowing constraint does not matter since our consumer is a saver anyway. He does not want to borrow. The solution of our consumer when the borrowing constraint is non binding:




When the borrowing constraint is binding: C1 = Y1 . If the consumer wants to consume more than his income in period 1 he cannot. When the borrowing constraint is binding then it affects the optimal consumption choice: 9





The optimal choice is at point D. But since the consumer cannot borrow, the best he can do is point E. From our analysis we can say that the Keynesian Consumption Function can be consistent with a microfounded two period model of consumption if consumers face liquidity constraints.

The Life Cycle Hypothesis (LCH) of Consumption The Keynesian Consumption Function while able to explain aggregate consumption of an economy at a point in time was not able to explain it over time. This is what we called consumption puzzle. Therefore there was the need to have a theory to explain the behaviour over time of the aggregate consumption. Two theories were developed at about the same time, one theory was proposed by Franco Modigliani (Nobel Prize in economics) called the Life Cycle Hypothesis. The other theory was developed by Milton Friedman called Permanent Income Hypothesis. Both theories can explain why over time the Average Propensity to Consume remains stable. Extending the idea of the Fishers model the Life Cycle model says that it is not only income in the current period that affected peoples observed consumption choices, but also income they expected in the future. The Life Cycle Hypothesis is that the income of people varies in a known way (systematic way) over peoples lives and that people use savings to move income from high-income periods to low-income periods in order to smooth consumption over their lifetime.


Basic Assumptions of the Life Cycle Hypothesis a) Perfect Knowledge of Lifetime: individuals know with certainty how much they are going to live and when they are going to die; b) Uniform Consumption: individuals prefer to have a constant stream of consumption over their lifetime; c) Zero Bequests: individuals do not die with positive income. They consume all their accumulated saving; d) Zero Real Interest Rate: this is a simplifying assumption; e) Ability to Calculate Future Income: individuals are able to guess correctly what their future income is going to be; A Basic Life Cycle Model Consider a representative consumer (many consumers all equal) that is going to live until time T. Denote with: - W the initial wealth of the consumer (this includes financial and real assets). - Y the constant annual income that the consumer is going to earn until retirement. - R is the number years until retirement.

The Lifetime Resources of the consumer is:

W + RY


Equation 5) is the discounted lifetime resources (remember that r = 0 by

assumption). To achieve constant consumption over time our consumer divides lifetime resources equally over time:

C= Or written differently:

1 (W + RY ) T


C = W + Y


where =

1 R is the marginal propensity to consume out of wealth and = is the T T

marginal propensity to consume out of income.

Notice that both marginal propensity to consume above are positive and less than 1. The Average Propensity to Consume (APC) is:
C W = + Y Y


Since over time aggregate wealth and aggregate income tend to grow together (given


the assumption of consumption smoothing if income increases, consumption remains constant and so saving (and wealth) will grow proportionally to income) then APC should remain stable over time.

Graphically, an example of a consumption profile of an individual under the Life Cycle hypothesis looks like:

When very young the individual has low income and may borrow to sustain their consumption level (think at students for example). Over time the individual starts to accumulate saving and keeps constant consumption. When old he retired and so his labour income goes to zero and in order to consume the same amount as before he has to dissave.

The Permanent Income Hypothesis (PIH) of Consumption Due to Milton Friedman (1957). The basic idea is that peoples income has a random element to it and also a known element to it and that people try to smooth the random part. Similarly as in the Life Cycle Hypothesis under the Permanent Income Hypothesis consumption should depend on more than just current income and people want to avoid fluctuations in their consumption over their lives. However, on the contrary of the life cycle assumption that income has a regular path along the life of a person, the assumption of permanent income emphasises the fact that income can change from a year to another due to temporary and random shocks Define Current Income Y as: Y = Y P +YT 9)

Y P = Permanent Income: this represents the long run (average) income which people


expect to persist into the future. This is given by your lifetime resources as in the LCH (wealth plus income over the future) divided by he number of years you expect to live (average). In practice the right hand side of 6); Y T = Transitory Income: this represents temporary deviations from average income. This is the random part of income that is unexpected. It can be positive or negative; Example: suppose that you are working and receive an annual salary of 30000. Suppose that you expect to get that salary every year in the future. Then 30000 represents the permanent part of your income and you expect to get 30000 every year also in the future. However assume that this year, since you have been very productive, you receive a bonus of 5000. This bonus represents a transitory income since you do not expect to get it every year from now on.

According to PIH consumers use saving and borrowing to smooth consumption in response to transitory changes in income. In particular permanent changes in income lead to much larger changes in consumption than temporary income changes. Thus, permanent income changes are mostly consumed while temporary income changes are mostly saved. For example, if you get promoted and you get a salary increase, this change will be probably permanent and so your consumption over time will probably rise. If instead you win the lottery, this represents a transitory income and you will probably not consume all of this transitory income. In practice the consumption function according to the PIH is:
C = Y P


where is the fraction of permanent income that people consume per year. Equation 10) is in practice the same as equation 7), the only difference is the economic reasoning behind the two.

The Average Propensity to Consume (APC) is:

C YP YP = T Y Y Y +Y P


Since in the long run we should expect that most of income variations are due to permanent income the APC should be stable.
Y because Y P while Y T does not change. Therefore the numerator and the


denominator increase by the same amount and so

C should remain constant. Y

The Random Walk Hypothesis of Consumption This is due to Robert Hall (1978). The idea is to consider the Permanent Income/Life Cycle Hypothesis under uncertainty once we include the idea of rational expectations (people use all available information to forecast future variables like income) in the analysis. If PIH-LCH is correct and consumers have rational expectations, then consumption should follow a random walk: changes in consumption should be unpredictable. A change in income or wealth that was anticipated has already been factored into expected permanent income, so it will not change consumption. Only unanticipated changes in income or wealth that alter expected permanent income will change consumption. Consider first a basic model of PIH-LCH under certainty. Assume that r = 0 (real interest rate is zero for simplicity) and there is no
discounting. Assume that the representative consumer lives for T periods. A consumer chooses consumption in each period to maximise the sum of his per period utility over his lifetime given by:

U (C1 ) + U (C 2 ) + ... + U (CT )

subject to his lifetime budget constraint: C1 + C 2 + ... + CT = W0 + Y1 + Y2 + ... + YT where W0 is the initial wealth and Y is income.



Assume that: per period utility function is: U (C ) = C

a 2 C where a > 0 is a 2

constant (this is a quadratic utility function). From the two period model we know that optimal choice of consumption over two different period of time is where: MRS = 1 + r , where MRS is Marginal Rate of Substitution between consumption in the two different periods. Since r = 0 the condition here becomes: MRS = 1 . The MRS between consumption in two different periods, like C1 and C 2 for example, is given by:

dU (C1 ) / dC1 dU (C 2 ) / dC 2
The numerator is the marginal utility of C1 and the denominator is marginal utility of


C 2 . Using the specific quadratic utility function above, the MRS between C1 and C 2

1 aC1 . 1 aC 2 1 aC1 = 1. 1 aC 2

The optimal condition MRS = 1 is therefore:

This implies C1 = C 2 . Extending our reasoning to T periods, we must have that is optimal for our consumer to keep consumption constant over time: C1 = C 2 = ... = CT . In practice it is optimal to have consumption smoothing. Using C to denote the per period consumption, the constraint 13) implies that:


T 1 (W0 + Yt ) T t =1


The right-hand side of 14) is exactly the permanent income and the right hand side of equation 6) in the Life Cycle model . Now introduce uncertainty. The main difference is that now you must form an expectation about future variables. Suppose you are at time 1, you know the variables at time 1 but you must form expectations about variables at time 2, 3, etc. etc. The optimal condition calculated before becomes:

1 aC1 =1 E1 (1 aC 2 )


where E1 means the expectation formed in period 1. Given that 1 and a are constants we have E1 (1 aC 2 ) = 1 aE1 (C 2 ) .so equation 15) implies:

C1 = E1 (C 2 )
periods t and t 1 : C t 1 = E t 1 (C t ) Now we introduce the idea of rational expectations1: C t E t 1 (C t ) = t


Extending our reasoning to T periods, equation 16) must hold for any two different



where t ~ WN (0, 2 ) . Equation 18) says that the forecasting error is a random
variable in average zero (no systematic error in expectations). Substituting equation 17) into 18)

Remember lecture note 17/18 on the New Classical Macroeconomics where we introduced the idea of

rational expectations.


C t C t 1 = t Or C t = C t 1 + 19)

Equation 19) is the equation of a Random Walk. It says that if we add rational expectation to the permanent income life cycle hypothesis aggregate consumption will follow a random walk. This implies that the best prediction for tomorrow consumption C t is today consumption C t 1 and no other variable can be helpful in predicting future aggregate consumption. If we take expectations of 19) we have: Et 1[C t ] = Et 1 [C t 1 ] + Et 1 [ t ] Et 1 [C t ] = C t 1 Because Et 1[C t 1 ] = C t 1 since that variable is known at time t-1 and Et 1[ t ] = 0 since t by definition has a zero expected value. Moreover equation 19) says that changes in consumption are totally unpredictable since C t C t 1 = t . Changes in consumption from one period to another are totally random.

Does Aggregate Consumption Follow a Random Walk? The answer is normally no. Empirical tests show that some variables, like for example movements in stock market prices, can help in predicting variations in aggregate consumption.