This paper is based on a long-term research program with Rachel Kranton on the implications of identity

23

She receives income of Y

1

 in period 1, income Y

2

 in period 2, and she can borrow and lend at the rate of

interest r.

24

The simple proof is that her utility maximizing consumption will depend upon the intercept and the slope

of the budget line.  The budget line states that the present discounted value of consumption is the present discounted

value of her future income, which is what Friedman calls her wealth.  The intercept of the budget line is her wealth. 

That is how much she could consume today if she consumed nothing tomorrow.  And the slope of the budget line is

determined by the rate of interest r: on the budget line for every unit of c

1

 she gives up (1 + r) units of c

2

.  Her

consumption will be on the highest attainable utility indifference curve.  That will be the indifference curve that is

just tangent to the budget line.  As a result we see that given the utility function c

1

 will be a function of W and r. 

Note that current income does not come into this expression. 

25

Laibson (1997) shows that consumption with forward-looking consumers with hyperbolic discounting

will balance the marginal utility of present consumption out of wealth against the marginal utility of future

consumption according to an Euler condition.  Such a condition is wealth-based.  It is the generalization of the

16

V. Excess Dependence of Consumption on Current Income

This takes us to the second neutrality, where a similar critique applies.  According to this

result, other than its contribution to a consumer’s wealth, current income has no independent

effect on the consumption of a utility-maximizing consumer. 

Milton Friedman (1957) derived such consumption-income neutrality in the two-period

model of Irving Fisher.  In this model the consumer chooses her consumption between two

periods.  She maximizes her intertemporal utility function, given by the function U(c

1

, c

2

).  c

1

denotes her current consumption in the first period; c

2

 denotes consumption in the second

period.

23

  If she maximizes U(c

1

, c

2

), a dollar of income earned today will have the same effect

on her current consumption as a discounted dollar earned in the next period.  Thus her

consumption will only depend on the discounted value of her current and future income and the

rate of interest.  This proposition is easy to prove.  It generalizes to many different commodities

and to many different time periods, and, with quadratic utility, to uncertain incomes.

24

  It even

generalizes, but in a slightly messy way, to the standard (hyperbolic discount) models of

consumption with present bias.

25

  In standard terminology, the value of her discounted income is

---

tangency of the utility indifference curve to the budget line in the two-period model of Irving Fisher.  Both Friedman

and Laibson obtain consumption that is solely determined by current income if there is a constraint on current

borrowing and consumers’ desires for current consumption exceed their current income.  There is nothing inherent in

the preferences in either case that cause current consumption to be based on current income.   

26

Formally, permanent income is the product of the rate of interest and wealth.

17

called her wealth

 

; the amount of that wealth that can be spent without its depletion is called

permanent income.

26

  An alternative expression of Friedman’s hypothesis is that consumption

depends on permanent rather than on current income.  

As simple and general as Friedman’s proposition may be, it contradicted prior thinking

about the consumption function.  Keynes, and his followers, believed that current income played

an especially important role in the determination of current consumption. 

The fundamental psychological law [italics added], upon which we are entitled to depend

with great confidence both a priori from our knowledge of human nature and from the

detailed facts of experience, is that men are disposed, as a rule and on the average, to

increase their consumption as income increases, but not by as much as the increase in

income (Keynes, The General Theory, 1936, p. 96). 

The General Theory also discussed many other factors that could affect consumption.  The list

was sufficiently rich that it not only included current income, but also all the other determinants

of wealth, such as expected future income and the rate of interest.  But that does not make

Keynes’ theory identical to Friedman’s.   In the Keynesian theory consumers are more sensitive

to current income than to other changes in income that have similar effect on the consumer’s

wealth.

It turns out that it is surprisingly easy to test the hypothesis that current income plays no

special role in the determination of current consumption.  Campbell and Mankiw (1989) have

conducted a test that nests both Friedman’s view that consumption depends solely on wealth and

the simplified Keynesian view, that consumption depends solely on income.  They suppose that a

---

27

See Dornbusch and Fischer (1987, p. 284).  

18

fraction of consumers 

8 are pure Keynesians, while a fraction (1 - 8) behave according to the

permanent income hypothesis; they estimate 

8 from the extent to which consumption overreacts

to changes that would be predictable from past changes in income and consumption. Usefully

then, 

8 gives a natural measure of the departure from the permanent income hypothesis.  The

estimates of 

8 are both significant statistically and also of significant magnitude economically:

between 40 and 50 percent (depending upon whether three or five periods are used to predict the

change in current income).  Many other studies corroborate such excess sensitivity:  Shea (1985)

for union members whose contracts specified their future wages; Wilcox (1989) for social

security recipients who had been earlier notified of changes in cost-of-living adjustments; Parker

(1999) for payers of social security taxes with predictable inter-year changes; Souleles (1999) for

changes in disposable income net of tax refunds; and Banks, Blundell and Tanner (1998), and

Bernheim, Skinner and Weinberg (2001) for retirees.  

Textbooks are a useful source of how economists view this excess sensitivity.  It is

noteworthy that standard textbooks do not explain the violation of the neutrality as due to

consumer motivation other than maximization of the standard utility function, U(c

1

, c

2

).  Instead

of looking for missing motivation in consumer preferences, they explain this violation of the

permanent income hypothesis by frictions.  Dornbusch and Fischer (1987) stand out only

because they are the most explicit:  “Given that the permanent income hypothesis is correct [sic],

there are two possible explanations.”

27

  Those two explanations are liquidity constraints for

consumers and myopia in their projections of future income.  Other textbooks discuss these same

two causes for the breakdown, but they are vague regarding the exact number of such

---