PRELIMINARY AND INCOMPLETE

New Theoretical Perspectives on the Distribution of Income and Wealth among

Individuals
Joseph E. Stiglitz1
A central question of economics has been how do we explain the distribution of income among factors
or production, and the distribution of income and wealth among individuals. Some fifty years ago,
theorists tried to develop explanations for what were then viewed to be the stylized facts, articulated,
for instance, by Nicholas Kaldor.2 Among the central facts was the constancy of the capital output ratio
and the relative shares.
Today, there seems to be a new set of stylized facts that have to be explained, many of them markedly
different from those that were the center of attention a half century ago. Among the empirical
observations are the following (some of these facts are more true for some countries than others; and
there are a few country exceptions):3
(a) Growing inequality in both wages and capital income (wealth)
(b) Wealth is more unequally distributed than wages
(c) Average wages have stagnated, even as productivity has increased, so the share of capital has
increased
(d) Significant increases in the wealth income ratio.4
(e) The return to capital has not declined, even as wealth-income ratio has increased

University Professor, Columbia University. The author is greatly indebted to the Institute for New Economic
Thinking for financial support. This is a revised version of a paper originally presented at an IEA/World Bank
Roundtable on Shared Prosperity, Jordan, June 10-11, 2014. I am grateful for the helpful comments of the
participants in the roundtable, and for the research assistant of Feiran Zhang and Eamon Kircher-Allen. The issues
discussed in Part I of this paper on the measurement of wealth and capital were discussed at a special session of
the IEA World Congress, Amman, sponsored by the OECD on the Measurement of Well-Being, and at a meeting
sponsored by the OECD High Level Expert Group on the Measurement of Economic Performance and Social
Progress, Rome, September 2014. I am indebted to Paul Schreyer, Martin Durand, Chief Statistician of the OECD,
and other participants at those meetings for their helpful comments and insights into the key issues of the
measurement of wealth and capital. I am also indebted to Martin Guzman for his comments.
2
Kaldor (1961). For a recent review of the attempts to explain these facts, see Jones and Romer, 2009.
3
These are not the only stylized facts that need to be explained. There is a large literature trying to explain the
shape of the income and wealth distribution, e.g. why the tails of the distribution are Pareto (fat-tailed), and why
at lower levels of income, the income distribution seems to be described by a lognormal distribution (Aitchison and
Brown, 1957; Lebergott, 1959.) In this paper, we will not address these issues. We note, however, that simple
extensions of several of the models presented here generate distributions which are consistent with these stylized
facts.
4
See Piketty (2014) and Piketty and Zucman (2014). For the U.S., U.K., Germany, and France, the wealth income
ratio rose from about 200-300% in 1970 to 400-600% in 2010..

These facts together help explain the growing inequality of overall income: inequality in each of the
components is increasing, and the relative importance of the more unequal component, capital, is also
increasing.
These facts put a new light on Kuznets hypothesis5 that, while in earlier stages of development,
inequality would grow, eventually inequality would fall. While that may have been true in the golden
age of capitalism, between the end of World War II and, say, 1980, the period in which Kuznets was
writing, such a conclusion no longer seems warranted. In particular, Piketty (2014) has argued that the
decades following World War II were an historical anomaly, the one period in which capitalism was not
characterized by a high level of inequality. He argues that not only has there been a large increase in
inequality since 1980, but that inequality will continue to grow.
There is a fundamental problem we confront in assessing the validity of that prediction. We are in the
midst of history: we cant be sure that inequality will continue to increase. There is an alternative
possibilitythat we are simply in the midst of a transition from one equilibrium to another.
Pikettys analysis, based on the hypothesis that the rate of return, r, is greater than the rate of growth,
g, has not been totally persuasive, at least to some commentators. He has, for instance, assumed that
capitalists save all of their income; but if they do not, the neat condition under which there is ever
increasing inequality, r > g, would no longer seem operative. He has assumed moreover that in the long
run, in spite of all the wealth accumulation of the rich, the return to capital does not diminish
seemingly contradicting a basic law of economics, that of diminishing returns. Under conventional
assumptions, as wealth accumulates, r falls, and the condition r > g would no longer hold.6
Anecdotes arent proofs, but they sometimes can alert us to factors that might have escaped attention
in a simple model. John D. Rockefeller was Americas first billionaire. At death, in 1937, his assets
amounted to 1.5 percent of GDP. Had his assets grown at the rate g (the rate of growth of the
economy) they would be worth $340 Billion. If r (the relevant rate of return) were just 1% more than g,
their family wealth should have grown to $680 billion. If, using numbers that Piketty might say are still
conservative, but more realistic, the disparity between g and r is 2%, their wealth would have been $1.3
trillion. Instead, the total value of the family assets is estimated to be $10 billionless than 1% of the
predicted amount-- divided among almost 300 members.7
Theory, together with empirical studies, can shed some light on these alternative hypotheses. It can
ascertain whether there are plausible conditions under which inequality (in some sense) increases
without bound. Alternatively, it can point to changes in technology or the economic environment that
might result in our moving from an equilibrium with one wealth distribution to one with much more

Kuznets (1955).
It is worth noting that in standard models, the condition r > g must be satisfied if the economy is intertemporally
efficient. If Pikettys analysis were correct, it would imply that any efficient economy would be characterized by
ever increasing inequality.
7
Roberts (2014). It appears that Piketty's analysis seems to have overestimated "r" and underestimated the
importance of the division of wealth among one's heirs.
6

inequality. Theory can also help identify policy interventions that might lead to an equilibrium
distribution manifesting less inequality.
In particular, consistent with our earlier work (Stiglitz, 1969a; Bevan and Stiglitz, 1979) we show that
there is some presumption that there exists an equilibrium wealth distribution (i.e. inequality does not
increase without bound.) We identify the parameters that determine the equilibrium inequality. We
identify key centrifugal and centripetal forcesforces leading to greater or less inequality. We suggest
that it is plausible that the factors contributing to growing inequalitymany of which are policy
relatedhave grown relative to those that might work to diminish inequality. This analysis is consistent
with observations on the marked increase in income and wealth inequality in recent decades in most
(but not all) advanced countries.
Overall wealth inequality is related both to the transmission mechanisms for human and financial capital
and to life cycle savings. There appears to be a significant increase in inherited inequality relative to life
cycle inequality8. In the models explored here, there is an equilibrium distribution between inherited
and life-cycle savings; but changes in key parameters can change that equilibrium.
While we believe that the most plausible models generate an equilibrium wealth distribution, we also
identify circumstances in which wealth inequality can steadily increase even taking into account the long
run macro-economic equilibrium conditions that Piketty seemingly ignored. This is true even if the rich
continue to save a significant fraction of their income.
Piketty suggests that significant capital taxation can ensure that this adverse outcome does not occur,
but if the equilibrium rate of interest is endogenous (as presumably it would be) there can be tax
shifting; the before tax rate can increase, so much so that the effect of the capital taxation, if not
carefully designed, could increase inequality. But we show that there are policies which can decrease
inequality.
We noted in the beginning of this introduction some puzzles posed by the new stylized facts: the fact
that there has been a marked increase in the wealth income ratio, but not the expected decrease in the
return to capital or the increase to the return to labor. There are further anomalies: most (but not all)
studies of the elasticity of substitution suggest that it is less than unity. This would imply an increasing
share of laborcontrary to the new stylized facts.9
The analysis presented here provides an answer to each of these anomalies. A key insight is that the
observed increase in wealth actually corresponds to a decrease in capital, as usually understood in
8

Bowles and Gintis (2002) and Piketty (2014). But there is not unanimity about this conclusion. See, in particular,
for the US, Wolff and Gittleman (2011).
9
There are still other anomalies about which we will have only a little to say in this paper. Globalization was
supposed to increase societal welfare for all countries; even if there were distributional effects within countries,
the gainers could more than compensate the losers. There is increasing evidence that there are indeed losers
(Acemoglou et al, 2014); but the losers are being told that they must accept further cutbacks in wages and
government services in order for the country to compete, seemingly suggesting that globalization requires them to
accept a lower standard of living.

neoclassical models. Some of the increase in wealth is, as Piketty himself recognizes in the value of
land. 10 Some of the increase in wealth is the increase in the capitalized value of what might be called
exploitationassociated with monopoly rents and other deviations from the standard competitive
paradigm. 11 An analysis of the forces giving rise to the increase in land values and exploitation provides
some insights into why there has been such a marked increase in wealth (and income) inequality,
enables us to assess whether such increases are likely to continue, and to identify policies that might
militate against these increases.
The paper is organized as follows. Part I focused on the distribution of income among factor shares
(capital and labor). Section I provides a quick overview of two key anomalies to which Pikettys work
has called attention. Section II, focuses on the fact that a large proportion of the increase in wealth is
related to the increase in the price of real estate.12 It was understandable why land was ignored in
earlier neoclassical models (including Solows, and those, like my own, trying to explain inequality): in a
modern economy, land is not a central input into production. The section develops several models
explaining the determination of the price of land, demonstrating why much (and in some instances,
more than 100%) of the increase in wealth would go into the value of land., It suggests that the
increases in wealth income ratios observed in recent years may in fact be only temporary: there may be
(and there evidently have been) bubbles.
The section argues further that the increase in land prices is closely linked with the provision of credit.
Changes in the rules under which credit is provided may lead to large changes in the price of landand
large increases in inequality.
Finally, the section sets up a simple model in which we can ascertain the equilibrium division of wealth
between workers (life cycle saving) and capitalists. In that model, we again show that a change in the
structural parameters of the economy that leads to capital deepening also leads to an increase in the
value of land, so that wealth increases more than the value of the capital stock.
Part II of the paper focuses on the distribution of income and wealth among individuals. Section 3
presents the basic model, in which it is established that there is a presumption that there is an
equilibrium wealth distribution. Section 4 includes a broad discussion of the centrigual and centripetal
10

Included in the increase in the value of land is the value of artificially created scarcity, e.g. through zoning
requirements. Land rents are likely to go up significantly with increasing urban agglomerations--it is not (as Piketty
(2014) seems to suggest that rents some places go up, and others go down. For instance, in a simple model of the
city, Arnott and Stiglitz (1979) show that land rents go up with aggregate transport costs. Not surprisingly, the
importance of agglomerations increase with the size of local public goods. Thus aggregate transport costs increase
with the size of public goods. (In their highly idealized model, they obtain the result that with cities of optimal size,
differential land rents are equal to the expenditures on local public goods, and are one half the value of aggregate
transport costs.)
11
This paper does not deal with the rents associated with intellectual property rights, which have taken on an
increasingly important role in the economy. This is an important omission, which we hope to address on another
occasion.
12
Of course, showing that capital has increased, but not as much as suggested by wealth data, resolves the earlier
noted anomalies only to the extent that the "capital" (correctly measured) output ratio has not increased. This
explanation should be viewed as complementary to the other explanations presented in Section I.

forces at play in the economy. Section 5. Identifies the relative role of life cycle savings vs. inherited
savings, and why that might have changed in recent years. Finally, section 6 describes in greater detail
the relationship between credit (and the institutions associated with credit creation) and wealth
inequality. The way our credit system functions (or mal-functions) has played an important role both in
the increase in the wealth income ratio and in the increase in wealth inequality.

Part I
1. Distribution of Income among Factor Shares: Key anomalies
There are real challenges in reconciling several of the observations in Pikettys book with standard
neoclassical theory (if we interpret wealth, W, in the usual way as capital, K). The most obvious is that
normally, we would expect that as the capital labor (capital-output) ratio increases, wages would
increase and the rate of return would decrease. For the United States, there is ample evidence on wage
stagnation , and there is considerable evidence that the return to capital has not decreased
significantly, including that presented by Piketty. .13
There is a second puzzle. It has been observed that the share of labor is decreasing. There is a wealth of
evidence arguing that the elasticity of substitution is less than unity.14 If wealth is "capital" in the usual
sense, then an increase in the Wealth/income ratio is associated with an increase in the capital effective
labor supply ratio, and if that is increasing, then the share of labor should be increasing. (The effective
labor ratio increases at a rate equal to the increase in the labor force plus the rate of labor augmenting
technological progress. )
The magnitude of the increase in the wealth income ratio itself presents a puzzle. Take a simplified
version of the model that seems to underlay Pikettys analysis, a Kaldorian savings model (Kaldor, 1957),
where capitalists save a fraction sp of their income and workers nothing.15 For simplicity, assume the
after tax rate of return on capital is 5%, sp = .4 Then16 capital would increase at the rate of .05 x .4 = .02.
If the growth rate were greater than 2%, the private capital output ratio would be declining. (Note that
if the share of capital is around .2 this generates a national savings rate of 8%, just slightly higher than
the actual savings rate.) Even if the savings rates were slightly higher, or the return to capital (and what
matters is after tax returns) slightly higher, it is hard to generate plausible increases in the real capital

13

Real U.S. wages have stagnated for decades, see Shierholz and Mishel (2013). Adjusted for inflation, average
hourly earnings of production and non-supervisory employees have decreased some 30 percent since 1990. See St.
Louis Fed data at <http://research.stlouisfed.org/fred2/series/AHETPI/>.
14
See Arrow et al. (1961); Young (2013). It should be noted that some authors have recently argued otherwise. See
e.g. Mallick (2007). But the assumption of an elasticity of substitution is greater than unity has one very disturbing
consequence: in models where the factor bias (i.e. whether technological change is capital or labor augmenting) is
endogenous, the steady state equilibrium is a saddle point. The economy is unstable. Stiglitz (2014)
15
Piketty implicitly seems to assume sp = 1, but the overwhelming evidence is that even the very rich save a much
smaller fraction of their income than that. Saez and Zucman (2014) estimate that the average saving rate for the
wealthiest 1 percent of Americans was 36 percent from 1986 to 20120.
16
Assuming that workers save nothing.

stock that could account for the observed increases in the wealth income ratios in recent decades. In
the case of the United States and many other countries, short run fluctuations in the wealth income
ratio are dominated by capital gains (and occasionally, recessions, we depress the denominator.) The
marked decline in the wealth income ratio in the US after 2008 highlights the importance of capital gains
and losses--gains before 2008 and losses since.
But what about the long run? Is there any reason to believe that the economy might be moving to a
new long run equilibrium with a higher wealth (capital) output ratio. Over the past sixty years, a wide
variety of models describing the growth of the economy have been formulated. In each, in the long run
(steady state) there is a particular capital output ratio. In each, changes in the underlying parameters
(the rate of growth of the labor force, the rate of growth of labor augmenting technological progress,
and savings behavior) can explain a change in the long run capital output ratio. The question is, have
there been any changes in these parameters sufficient to explain/account for changes in the capital
output ratio and distribution of income of the magnitude observed? (In the appendix, we briefly
present these models.)
For instance, in the Solow growth model, the long run capital output ratio is given by s/g*, where g* is
the long run growth rate, equal to the rate of growth of labor supply plus labor augmenting
technological change, and s is the savings rate. The rate of growth g* has varied, for instance increasing
in the 90's amd the first part of this century , while the savings rate (in the US) has decreased, which
would suggest a decrease in the long run capital output ratio, not an increaselet alone an increase of
the magnitude asserted.17 18
In the Kaldorian model, the long run capital output ratio is given by spSK/g, where SK is the share of
capital. While there has been an increase in SK, it has not been large enough to account for the increase
in the capital/output ratio.
This perspective contrasts markedly with that of Piketty (2014) and Piketty and Zucman (2013). Just as a
matter of national accounting, if s is the fraction of national income saved,
(1.1)

d ln K/dt sY/K

and, treating for the moment K and W as identical,

17

For instance, between 1960 and 2000, the savings rate fell from 8% to 2% while the rate of growth increased
from 2.3% to 4.1%. If these were permanent changes, then the long run capital output ratio would have fallen by
a factor of almost 8. (Actually observed growth rates will be higher than g* if there has been capital deepening,
less than g* if the reverse has been happening. In the period beginning in the 90s, the net national savings rate
was probably too low to sustain capital deepening. See the discussion in the next footnote.)
18

Matters are no better if we view the savings rate as endogenous, determined by intertemporal utility
maximization. Then, the critical variable is the intertemporal discount rate, and again, it is hard to see
changes in that variable of the magnitude that would account for changes in the observed capital-output
ratio.

(1.2)

dln (K/Y)/dt = sY/K g.

Piketty and Zucman present data showing that the average net national savings rate of the US over the
period 1970-2010 is 5.2%19, and that the average growth rate of the economy was 2.8%. The
wealth/income ratio varied, beginning the period at just over 4 and ending at 4.3. Thus, (1.2) would
have predicted a decline in the capital (wealth) income ratio, at an average annual rate of somewhat
more than 1.5%, in contrast to the observed increase. If these numbers were accurate, the observed
increase in wealth income ratios must come from somewhere else than the steady accumulation of
capital goods. Capital gains, and the increase in the value of land in particular, is the obvious answer.20
(We will note later in this section that an increase in the degree of exploitation also provides part of the
answer.)
We can reframe (1.2) to ask what is the critical net savings rate such that there is an increase in the
"real" capital output ratio? Let k be the effective capital labor ratio, g* be the "natural" rate of growth
of the economy, the sum of the rate of growth of population (work force) and the rate of labor
augmenting technological progress, = W/Y, and = K/W, the ratio of the value of produced capital to
wealth (which includes land); then
(1.3) dln k/ dt = sY/K - g* = s/ - g*,
so that capital deepening (an increase in the effective capital labor ratio) occurs if and only if
(1.4) s > g* .
If it were assumed that the US growth over the last forty years was close to its natural rate, 2.8%, = 4,
and = 1 (land is an unimportant), then s would have to be greater than 11.2%, more than twice the net
savings rate for the US. More realistic, even if = .8, s would have to be greater than 8.9%. Given the
US savings rate of 5.2%, only if < .46 will there be capital deepening.
All of these are different ways of making the same point: one cannot understand what is happening to
macro-economic variables like wealth income, capital income, and capital labor ratios without taking
into account land (or other forms of non-produced wealth). These non produced assets are of first
order importance. It was the omission of these that represents the most important lacuna in my 1969
theory of the equilibrium distribution of wealth and income, which this paper attempts to rectify.

1.2. Can wages fall, the capital output ratio increase, and the return to capital not fall as k increases

19

Earlier, in the discussion of the Kaldor model, we focused on private wealth. The net private savings rate for the
US over the period 1970-2010 has been 7.7% (Zucman and Piketty, 2014) As they point out, most of the variability
in wealth income ratios (at least as conventionally measured) can be attributed to the private sector.
20
This can be expressed in another way. The average annual increase in the capital stock for the US they estimate
to be 3.0%, of which the average "real" savings accounts (by the calculation above) to about 1.5%, or just half.
(Piketty and Zucman (2014) suggest that savings accounts for 72% of the increase in the wealth income ratio.)

The previous section argued that in none of the standard models of economic growth can one plausibly
obtain an increase in the equilibrium value of the capital output ratio of the magnitude observed if we
interpret wealth as capital. If one interprets W as capital, then there has been not only an increase in
the capital output ratio, but also in the capital labor ratio. Our ultimate objective is to understand the
distribution of income, both among individuals and among factor shares. We now ask, can wages fall (as
they have been) as k (the capital labor ratio) increases, within the standard neoclassical model.
Some have suggested that some forms of capital are like robots, and compete directly with workers,
lowering their wages. But highly skilled workers still need to manage the robots, and even if the
increased capital lowers the return to unskilled workers, it increases the return to the skilled workers.
We show here that under standard assumptions, an appropriately weighted average wage must
increase.
Assume Y = F(K, L 1, L2. ) is constant returns to scale. In the following discussion, we will simplify and
assume only two types of labor. In a two factor production function, an increase in capital must increase
the marginal productivity of labor. In this model, an increase in capital may not increase the marginal
productivity of both types of labor. Constant returns to scale (CRTS) implies that
FL1L1 + FL2L2 + FKK= F,
so
FL1KL1 + FL2KL2+ FKKK= 0,
Diminishing returns implies FKK < 0, which is why if there is only one type of labor FLK > 0: an increase in
capital (relative to labor) must increase the wage. Here, it is clear that the wage of one of the two types
of labor could go down.
But consider the average wage,
w(K) = (FL1L1 + FL2L2)/L
where L = L1 + L2.
w = (FL1KL1 + FL2KL2)/L = - FKKK/L > 0.
The weighted average wage must increase when capital (the capital labor ratio) is increased.
Data for the United States, for instance, shows otherwise: a stagnating or declining average wage rate
during the past four decades, during which the capital output ratio has increasedif we interpret
wealth as capital.21

Technological change
21

. For wage data see Shierholz and Mishel (2013).

There is a related hypothesis: that technological change has diminished the returns to unskilled labor. It
is skill biased.22 While the timing of the changes in the share of labor and the decrease even in wages of
relatively skilled labor in more recent years argues against skill biased technological change as the major
or at least sole explanation of changes in distribution23, here we focus on the analytics.
If there were a single type of labor, then labor augmenting technological change increases the effective
labor supply, and, everything else being the same, would reduce the effective capital labor ratio, and
hence the wage per effective labor unit. But each worker would represent a larger number of effective
labor units, so whether the wage per worker increases or decreases would depend on the elasticity of
substitution.24 Only if the elasticity of substitution is substantially below unity would wages fall. (As we
noted earlier, interpreting wealth as K implies an elasticity of substitution greater than unity, which
would imply an increase in wages. Similar results hold in the longer run, when there is an adjustment
in the capital stock.25)
Assume now there are two types of labor, skilled and unskilled, and technology is skilled bias, say
increasing the productivity of the skilled workers, while leaving that of unskilled workers unchanged.
Whatever the factor bias of technological change, it must move the factor price frontier outwards,
which means that if the return to capital doesnt change, then the return to at least one of the two types
of labor must increase, and if the stock of capital were fixed (or increasing), the average wage would
have to increase.26 Again, it is not easy to reconcile observed patterns of changes in factor prices with
the theory.

1.3. The resolution of the seeming paradox: There is more to wealth than capital
The previous section argued that it is hard to reconcile the new stylized facts with virtually any form of
the standard growth model under the assumption that the increase in wealth corresponds to an increase
in productive capital. What then is going on? The most plausible hypothesis is that wealth and capital
are markedly different objects (as Piketty himself recognizes, but the full implications of which he does
not take on board), and that wealth can be going up even as capital (as conventionally understood) goes
down.
22

The first to propose the idea of skill-biased technological change was Griliches (1969). See also Krusell et al.
(2000).
23
See Card and DiNardo (2002); Autor (2002); and Autor, Katz, and Kearney (2008).
24
Nave interpretations of Pikettys work, which confuse the increase of wealth with an increase in capital, argue
that there must be an elasticity of substitution greater than unityhow else could one explain the rising share of
capital. But if the elasticity of substitution is greater than unity, then labor augmenting technological change
would lead to an increase in wages at a fixed capital stock, and an even larger increase in wages were the capital
stock to increase. (The elasticity of substitution has to be substantially below unity for the wage to decrease. If
there are different kinds of labor, similar results hold for the average wage.)
25
Labor augmenting technological change leads to a higher return to capital, and the presumption is that that
would lead to higher investment. This would lead to a still higher wage.
26
Obviously, in the real world there can be changes in technology and changes in capital stock occurring
simultaneously; still, we can decompose the effects, into what would happen with a change in technology, given K,
and what would happen with a given technology as we increase K.

If capital is not going up much (or even going down) in tandem with the increase in the effective labor
supply, it would explain why the interest rate has not gone down, and possibly even why real wages
might not have gone up.27
Index number problems
Actually, if land, T, is treated as a factor of production,28 then wages will be related to inputs of both K
and T. If T is fixed, then the increase in K has to be even greater (e.g. to ensure that wages increase) to
offset the failure of T to rise. We need to add up K and T somehow to ascertain what is happening to
the aggregate input, which we will refer to as C. And how we add the two together matters a great deal.
And what makes sense for one purpose may not for another.
If T and K were additive in the production function (Y = F(K + T, L)) then we would simply add them up
linearly. But then there shouldn't be any changes in the relative price of T and K, since there are perfect
substitutes. In the case of France, this aggregate "C" has been going up more slowly than GDP, even
though K has been going up slightly faster than GDP. (See Figure 1)
On the other hand, we could have a production function of the form
Y = F(C,L)
where now
C = KT.
Then, since T is fixed,
dln C/dt = dln K.
Now, C is increasing if K is increasing, but whether it is increasing faster or slower than GDP depends on
the relative weights assigned to the two factors, . With even a relatively high value of , C/Y appears to
be declining for France.
It is obvious that wealth and capital are two markedly different concepts. There are many forms of
wealth that are not productive assets. Much of the increase in wealth in recent years is associated with
an increase in the value of land. The increase in the value of land does not, however mean that there is
more land, and that therefore the productivity of labor should go up.29 And an increase in the value of
27

though if the capital effective ratio were simply stagnant, real wages should increase at the rate of labor
augmenting technological progress, which has been significant. This suggests that an increase in the degree of
exploitation has also played an important role.
28
Although land is not very important in most industrial processes (certainly not as important as it is in
agriculture), housing services represent an important component of GDP, and land is an important input into real
estate.
29
Piketty (2014) suggests that though there has been a marked increase in the value of real estate, it is not the
changes in land value which matter (but rather the buildings that have been constructed on them). As he puts it
forcefully, the increase in the value of pure land does not seem to explain much of the historic rebound of the
capital/income ration (sic) in the rich countries The construction of capital indices involves, of course,

10

land does not mean that the marginal productivity of capital should decrease. Once we sever the
relationship between K (and more broadly "C," the value of the aggregate inputs) and W (where W
refers to wealth), all the paradoxes described in the previous section disappear.
Not only are the concepts different, but there are difficult measurement and aggregation problems
involved in each. The "volume" of capital goods resulting from saving out of national income (letting
consumption goods be the numeraire) will be affected by changes in the price of capital goods relative
to consumption goods. And the effective increase in "K" will also be affected by (capital augmenting)
technological change. (Indeed, the two issues are closely related; because there are constant changes in
the design of capital goods, one has to establish a "hedonic" index of equivalency.) If the only capital
good were computers, the increase in the "volume" of K from a given amount of savings has increased
enormously over time. And in calculating aggregate "K," we have to add up capital of different types,
whose relative prices and productivities are changing over time.
The standard wealth income measure, constructed by adding up the money value of wealth and dividing
it by the money value of income, and tracing how that ratio, and ownership of that wealth, evolves over
time captures something that is important in our economy: control over resources. But changes in the
wealth distribution, so measured, does not even necessarily reflect well the distribution of "well-being."
For the bundles of goods bought by those at different income/wealth levels may differ--indeed, in some
of the models below, the increase in wealth is closely linked to the increase in the price of a good which
is consumed only by the rich, so that the increase in inequality in well-being is markedly lower than the
increase in money-wealth.30
But what is clear is that the measure of wealth so constructed is not a good measure of the relevant
inputs into the production process--wealth could be going up, and yet any reasonable measure of inputs
(whether the production measure we identified as C or the narrower measure K) could be moving in the
opposite direction. The theoretical models we present below are sufficiently simple that there is no
need to form an aggregate measure of capital inputs, "C." We can trace out the evolution over time of
K/Y and W/Y and of the distribution of wealth, and the implications of policy on each of those variables,
without ever constructing such an aggregate measure.

complicated adjustments for changes in relative prices. Paul Schreyer concludes (personal note to author) by
observing the distinction between the wealth and production aspects of capital is indeed important and a story
about W does not immediately translate into a story about K. Associated with the two perspectives are different
measures that evolve quite differently. However, the key aspect in the analysis of capital in production and its link
to income shares seems to be the treatment of non-produced assets, in particular land.
30
These problems are similar to those that have arisen in the measurement of poverty, with Pogge and Reddy
(2010 ) arguing that standard estimates do not adequately reflect differences in prices faced by the poor--a claim
that Martin Ravallon has disputed, illustrating that these index number problems are both difficult and
contentious.

11

There are, in fact, three reasons that W can increase without a concomitant increase in K, besides an
increase in the value of land or other inelastically supplied factors31, described below. Some of these, as
we shall see, have as much to do with our accounting frameworks as with anything else.
Changes in market power and exploitation
Underlying the Solow model is the assumption of competition. But there is an increasing consensus that
much of observed inequalityespecially at the topis associated with rent seeking, including the
exercise of monopoly power. 32 If monopoly power of firms increases, it will show up as an increase in
the income of capital, and the present discounted value of that will show up as an increase in wealth
(since claims on the rents associated with that market power can be bought and sold.)33
Note that such increases in wealth are associated with a decrease in the economys effective
productivity, because they are associated with an increase in market distortions. Moreover, it is an
implication of such exploitation that even though W is increasing, wages are decreasing.
Presumably, there is a limit to the ability to increase market power, and therefore a limit to the extent
to which the wealth/income ratio increases (and the share of wages decrease.) But this provides little
comfort: there may be marked increases in inequality before we reach this limit.
While monopoly rents are the most obvious example of an increase in wealth unassociated with an
increase in the productive capacity of the economy, there are many other forms of exploitation, the
capitalized value of any change in which would show up in a change in wealth. If the financial sector
improved its ability to exploit the poor through predatory and discriminatory lending practices, and

31

In the short run, there can be capital gains on producible assets as well, but such increases cannot be sustained
in the long run, since they will elicit a supply response. Some of the increase in "seeming" wealth that occurred in
the US prior to the 2008 crisis was may have attributable to capital gains on buildings (though it is difficult to parse
out such capital gains from capital gains on land). But the "correction" brought now the price of real estate to or
below the reproduction cost.(If we take consumption goods as our numeraire, the price of capital goods could
increase or decrease; one of the most important capital goods, that associated with "technology," has, in these
terms, been experiencing a capital loss.
32
Piketty, Saez, and Stantcheva (2014) provide an interesting empirical test, pointing out that increases in tax rates
at the very top are not associated with slower rates of growth. See Stiglitz (2012, 2014) for a broader discussion,
including of the many forms that rent-seeking takes in a modern economy, and other evidence that rents have
become an important source of income at the very top. Galbraith (2012) emphasizes the role of the financial
sector in the generation of inequalities Much of the return to the financial sector is associated with rent seeking,
e.g. that associated with market manipulation, insider trading, predatory lending, abusive and anti-competitive
practices in credit and debit cards. There is an extensive literature discussing why we might expect an increase in
monopoly power in a modern economy, e.g. as a result of network externalities (Katz and Shapiro 1994) and the
fixed costs associated with research (Dasgupta and Stiglitz 1980). (Many of these arguments, however, are
inconsistent with the assumption of a constant returns to scale production function.)
33
The timing of increases in the share of capital are perhaps more consistent with those being explained by rapid
changes in the degree of exploitation than by sudden changes in the effective capital labor ratio. A permanent
increase in the share of capital by just 1% would, when capitalized at a real discount rate of 1.5%, imply an
increase of the wealth income ratio of .67; an increase of market exploitation leading to an increase in the share of
capital by 5% would lead to an increase in the wealth income ratio by more than 3.

12

abusive credit card practices (and the resulting profits were not bid away because of imperfections of
competition) then there would be an increase in standard metrics of wealth.
Successful corporate rent-seeking: transfers from the public sector to the private
There are more subtle forms of "exploitation." Government allows too-big-to fail banks. The value of
those banks is higher than they otherwise would be, because of government risk-absorption. But the
contingent-liability of the government is not capitalized, and because it doesnt show up in the national
balance sheet, it appears as if the wealth of the economy has increased. But with appropriate metrics
(where the decreased wealth of wage-earning citizens, as a result of the increase in the expected
present discounted value of the higher taxes that they will have to pay to bail out the banks, just the
opposite would have happened: we would have recognized that because of the distortions associated
with too big to fail banks, the productive capacity of the economy has been diminished; that the bailouts are Pareto-inefficient, and that the wealth of the economy has been diminished.34
In each of these situations, a change in the flow of resources that accrues to capital gets capitalized in
wealth, and the present discounted value of the decreased flow to the rest of the economy is not
reflected in our wealth metrics. We dont, for instance, value the change in the stream of tax revenues
to the government or the expenditures by the government or the reduced wages accruing to workers as
a result of increased market exploitation.
Changes in discount rates
There is a further reason for an increase in the value of wealth without a concomitant increase in the
physical productive capital stock, and that is that the rate of discount may fall, e.g. because of a
decrease in the interest rate (or an increase in the marginal tax rate), and this may induce large changes
in the relative price of different goods (and in the price of capital goods relative to consumption). This
was the essential issue in the Cambridge-Cambridge controversy some half a century ago, where it was
observed that the value of capital and the choice of technique may be non-monotonic in the interest
rate. 35
Other data problems
This section has explained why data on wealth do not reflect capital. Several of the stylized facts
involved inequality metrics. Some question the magnitude of some of the increase in inequality, say the
share of income at the top for the US, because of changes in the tax law in 1986 which may have led to

34

This discussion raises similar issues as those the Commission on the Measurement of Economic Performance and
Social Progress discussed in moving economic activities from the public to the private sector
35
See Sraffa (1960) and Stiglitz (1974). Thus, in models with the production of commodities by means of
commodities, the economy at a low interest rate and a high interest rate may look the same (the same
technologies are employed), while at an intermediate interest rate a different technology is employed. Even if the
value of wealth has changed in going from the low to the high interest rate, there has not been capital deepening,
in any meaningful real sense. There are a variety of other reasons that there can be changes in intertemporal
pricing, with large consequences to the valuation of assets. See the discussion below.

13

a change in reported income, not actual incomes earned. 36 (We should note that the studies of
inequality looking at the increased inequality at the top have attempted to deal with this obvious
problem.37) But the pattern of increased inequality (an increased share of total income going to the top
1%) continued even after tax changes were partially reversed in 1993. Moreover, other countries
without corresponding changes in tax codes have seen similar increases in inequality. (Interestingly,
because in the US, the top is the only part of distribution that has done very well, if it were the case that
most of their seeming increase in income is just a change in reporting, it would imply that that the
overall performance of economy has been really dismal; one would have to explain how it is that, given
all of the increase in wealth, all of the improvements in economic policy, and all of the alleged gains
from globalization and technology, all of these together seem to have generated so little improvement
in standards of living to any group in our society, not even, allegedly, the very top.)

36
37

Feldstein (2014).
Piketty and Saez (2003).

14

2. Land

The most important source of the disparity between the growth of wealth and the growth of
productive capital is land: much of the increase in wealth (in some cases, more than 100%) is an
increase in the value of landnot associated with any increase in the amount of land and therefore
of the productivity of the economy.
In the following sections, we describe models that might account for much of the increase in the
value of wealth taking the form of an increase in the price of land. The final model is the most
explicit in linking the increase in wealth through an increase in the value of land to an increase in
inequality. These ideas will be developed further in sections 5 and 6 of the paper.

2.1. A simple model with land rents

The simplest model is one in which the rents associated with land are fixed and last in perpetuity,
while the production of industrial goods requires no land. Then a slight decrease in the (long term
real) interest rate can lead to a large increase in the value of land.38 Thus, national output is given
by
(2.1) Q = F(K,L) + R
where K is productive capital and L is labor, for the moment assumed fixed. Then the value of
wealth, W, is given by39
(2.2) W = K + R/r = K + R/FK,
so that
(2.3) dW/dK = 1 - RFKK/FK2 > 1
If F is, for instance, a unitary elasticity of substitution, with coefficient on capital of , then
(2.4) dW/dK = 1 + (R/Q) (1 )/ .
If, for instance, that R/Q = .3 and = .2, then dW/dK = 1 + 1.2 = 2.2: the increase in wealth is more
than twice the increase in the productive capital.

38

If R is the rent from the land, and r is the real interest rate, then the value of land V T = R/r, so that there is an
equiproportionate increase in the value of land from a decrease in the real interest rate.
39
This analysis applies to a comparison across steady states with different K.

15

2.B. Positional goods

Similarly, if land serves as a positional good, there can be an increase in the value of land, without any
increase in the productive potential of the economy. Rich individuals compete for houses in the Riviera.
As the rich get richer, they compete more vigorously for this real estate. The price of this fixed asset
increases, without any increase in the productive potential of the economy.
Assume there are some assets in fixed supply (positional goods) that do not affect production of
conventional goods. Assume all the wealth of the economy is held by the rich (an assumption which
does not depart too far from reality) and that the demand by rich for these goods is given by M(W,p)
with the equilibrium given by
(2.5) M(W,p) = pT
where p is price of land, T, which is fixed supply, W = K + pT. For simplicity, we choose units so T = 1.
(2.5) can be solved for p as a function of W, and K can then be solved for
(2.6)

K = W - p(W).

Then
(2.7) dK/dW = 1 - p = 1 - MW/[1 Mp] < 1
If the wealth elasticity of the demand for positional goods is large enough and the price elasticity is
small enough, then an increase in W may even be associated with a decrease in K.

2.C. Bubbles: the dynamic instability of the market economy

Bubbles are a pervasive and recurrent aspect of market economies. While the recession may have
represented a correction, the economy may not have fully corrected the price of real estate.40
Hahn and Shell-Stiglitz41 showed the dynamic instability of the economy with heterogeneous capital
goods in the absence of a full set of futures markets extending infinitely far into the future(or without
perfect foresight extending infinitely far into the future). The steady state was a saddle point.
The same result also holds for a model with capital and land (with two state variables, K, the stock of
capital, and p, the price of land). For simplicity, we assume that population is fixed, and there is no
technological change.42 The short run dynamics are described by

40

The recurrence of bubbles has been noted by Kindleberger (1978).

Hahn (1966), Shell and Stiglitz (1967).
42
Standard models with land in a conventional production function encounter a problem because of diminishing
returns if population grows. This can be offset by land augmenting technological progress, in a borderline case.
41

16

(2.8) dK/dt + K+ T dp/dt = s[F(K,L,T) + T dp/dt]

and
(2.9) dp/dt + FT = pFK
where L is the labor supply, the depreciation rate, F(K,L,T) the neoclassical production function43, and
where we have assumed individuals save out of full income including capital gains. (Shell, Sidrauski,
Stiglitz, 1969a).44 The RHS of (2.8) is gross savings. This goes into an increase in the value of land (land
savings) or gross capital accumulation. (2.9) simply says that the (short term) return on land and
capital are the same.
Without loss of generality, we choose units so T = 1. Substituting (2.9) into (2.8), we obtain a pair of
differential equations describing the dynamics of the economy:
(2.10) dK/dt = sF(K,L,T) K (1-s) (pFK FT) = sF(K,L,T) K (1-s) dp/dt
(2.11) dp/dt = pFK - FT
Figure 2 shows that there is a unique steady state, given by the solution to the loci along which dp/dt = 0
and dK/dt = 0, given respectively by
(2.12)

p = FT/FK

and
(2.13) p ={s[F(K,L,1) - K + (1-s)FT} /{(1-s)FK if FK not equal to zero.45
We define K** as the value of K at which the numerator of (2.13) is zero46
(2.14) sF(K**,L,1) + (1 s)FT (K**,L,1) = K**
and K* as the value of K at which
(2.15a) sF(K*,L,1)= K*,
i.e. the value of K at which dK/dt = 0 at which dp/dt = 0, that is the value of K at which the two curves
cross, and therefore the long run equilibrium value of K. The steady state value of p* is then given by
(2.15b)

p* = FT(K*, L, 1) /FK(K*, L, 1)

Since

43

For simplicity, we assume that FK approaches infinity as K approaches zero, and that the marginal product of
capital falls to zero only as K approaches infinity.
44
Similar results can be obtained with other savings functions.
45
We do not need to pay much attention to this case because of the assumptions made in footnote 35.
46
There may be more than one such value. If there is, K** is defined as the smallest such value.

17

(2.16) dp/dK|dK/dt = 0 = (sFK ) /(1-s) FK + dp/dK|dp/dt = 0

And
at {K*,p*}
(2.17) /s = F(K*,L,1)/K* > FK(K*, L, 1)
it follows that at the point of intersection
(2.17) dp/dK|dK/dt = 0 < dp/dK|dp/dt = 0,
that is the dK/dt = 0 cuts the dp/dt = 0 curve from above. This means that there can be only a single
intersection for K > 0, i.e. a single value of {p*,K*} solving (2.15a) and (2.15b).
Under natural restrictions47 along both loci, when K = 0, p = 0.
The dp/dt = 0 is upward sloping provided only that
FTK/FT > FKK/FK
which is clearly satisfied if T and K are complements. On the other hand, the dK/dt = 0 locus is never
monotonic, reaching 0 once again at K** > K*.
Note that from (2.10) if p is above the dp/dt = 0 locus, p increases, i.e. if land prices are too high, for
ownership of land to generate the same returns as capital, the price of land has to increase. On the
other hand, if p is above the dK/dt = 0 locus, it means that the price of land is increasing; the increase in
the value of land (savings in this sense) acts as a substitute for real capital accumulation, and K
accordingly diminishes. The result is that the steady state equilibrium is a saddle point, as depicted in
the figure.
With futures markets extending infinitely far into the future, p is set along the trajectory converging to
the steady state, i.e. there is a unique value of p(K) for each K such that the economy converges to the
steady state.
Without futures markets extending infinitely far into the future or infinite foresight, there is no reason
to believe that the transversality condition will be satisfied. But along the paths which satisfy the short
run arbitrage equation but do not converge to the long run equilibrium because the initial price is too
high, the price of land eventually increases superexponentially.48 As a result, in finite time, the bubble
will be corrected. But it can be a long time. And even when there is a correction, it may still be on
a bubble path. The prices falls, but to a level still above the convergent path.Figure 2a depicts the

47

E.g. that the Inada condition (the limit as K goes to zero of FK is infinity) and that the limit of FT as K goes to zero
is bounded.
48
When the price is too low, eventually, the price may shrink to zero. For the rest of the analysis, we ignore this
case.

18

limiting case where land and capital are perfect substitutes, that is the production function is of the
form F(K + T, L), in which case p in equilibrium must be unity. We now have
(2.11a) dp/dt = FK (p -1)
which equals zero if and only if p = 1, i.e. the dp/dt = 0 locus is a horizontal straight line. The rest of the
analysis follows as earlier.49
Note that on the trajectories in which p explodes, eventually, the increase in the value of land crowds
out capital accumulationthe capital stock declines, even though wealth continues to increase. The
differential equation describing wealth accumulation is given simply by
(2.18) dW/dt =s[F(K,L,T) K + (pFK FT)].
Above the dp/dt =0 locus, pFK > FT, and for K < K*, F > sF > K. Thus,on the left- diverging trajectories
eventually wealth increases even though capital decreases.
Does this explain the increase in Wealth/Income ratio? It seems that this is at least a better explanation
than the alternative theory of an ever increasing true effective capital labor ratioa theory for
which it is hard to specify in a coherent model.

2.D. Land in a life cycle model

There is a fourth model which illustrates equilibria in which increases in wealth do not correspond to
increases in capital stock. In the standard life cycle model, workers save out of wages, based on
expectations of returns in the future. Wages are a function of the capital stock today, and the rate of
return is a function of the capital stock next period. Assume, as before, that there are two assets, land
and capital, and as before, that the return to land must equal the return to capital. In this section, for
simplicity, we focus only on the steady state and continue to assume the labor supply is fixed.50 Because
the only variable of interest is then the capital stock, wages and the returns to capital (and other
relevant variables) can all be expressed as functions of K. Hence as before, in the steady state
p = FT/FK,
and
(2.19) s(w(K),r(K))w(K) = K + FT/FK,

49

Except that the steady state is totally unstable, rather than a saddle point. If p ever deviates below 1, the price
continues to fall, and conversely if it ever is above p = 1. The dK/dt = 0 equation is now given by p = 1 + [sF(K,L,T)
K]/ (1-s) FK, which at K = 0 is above unity, and equals unity at K = K**. There is a K*** at which this turns negative.
50
For a more complete analysis of this model, see Stiglitz (2010) .

19

in the obvious notation were wages and returns to capital are functions of the capital stock. Workers
save a fraction of their wage income, with the fraction depending on their wages and the rate of return
to capital. Savings are put either into capital goods or into land holdings.
It should be obvious that there may be more than one steady state (more than one value of K solving
2.19) for plausible savings and production functions. It is useful to rewrite (2.19) to focus on savings in
capital:
(2.19a) s(w(K),r(K))w(K) - FT/FK = K.
Any value of K solving (2.19a) is a steady state equilibrium.
There can be multiple equilibria, as illustrated in figure 3. As K increases, wages increase. The slope of
the LHS can be greater or less than unity, and can vary with K, so that the LHS can cross the 45 degree
line more than once. There is a natural sense in which stability requires that the savings curve cut the
45 degree locus from above, i.e. the increase in savings into capital from an increase in the capital stock
is less than the increase in the capital stock itself.
Looking across (steady state) equilibria, it is clear that
(2.20) dW/dK = d[K + FT/FK] /dK = 1 + FTK/FK FTFKK/FK2.
If
(2.21) FTK/FT - FKK/FK > 0,
then W, wealth, increases more than K. That will always be the case if T and K are complements (as in a
Cobb Douglas production function).
By the same token, we can ask what happens if there is an upward shift in the savings function, i.e. the
savings function is given by s(w(K),r(K)). Then
(2.22) dK/d = sw/ {1 + FTK/FK FTFKK/FK sw wds/dK},
while
(2.23) dW/d = dK/d ( 1 + FTK/FK FTFKK/FK2).
Again, we get the result that W can increase more than K.

2.E. Credit and the creation of land bubbles and inequality

How much rich individuals are willing to spend for positional goods depends, of course, on the cost of
capital. In this section, we provide a bare-bones model that we think may capture more accurately what
has been going on than any of the models presented so far: the banking system provides credit based
on collateral. When the price of land in the Riviera goes up, the banks are willing to lend more. If the
20

banks are willing to lend more, the price of land in the Riviera goes up. There is, essentially, an
indeterminacy: it is the decision of the banks (the central bank) concerning credit availability that drives
the price of land (real estate).
We modify the model of section 2.2 by assuming three distinct classes of individualsworkers who just
consume, capitalists who save out of profits, own enterprises and invest only in capital goods, but have
no access to credit, and rentiers, who own land51. Their demand for positional good (land in the Riviera)
is given by M(WT ,C,p), with the equilibrium condition now being given by
(2.24) M(WT ,C,p)= pT = WT + C,
where C is the amount of credit that is available and WT is the wealth of the rentier, which is just the
value of the land minus what they owe in credit: WT = pT - C We can solve for
(2.25) p = (C)
The wealth of the rentiers is entirely driven by the provision of credit
(2.26) WT = pT C = (C) - C
To close the model, we need an additional equation describing capital accumulation. We take the
simplest version, due to Kaldor52. Capitalists-entrepreneurs save a fraction of their income, sp, putting
their money into capital goods
(2.27) dK/dt = sprK - K,
where is the depreciation rate, so in steady state
(2.28) FK(K*,L) = /sp.
In this model, the provision of additional credit has no effect on the equilibrium capital stock. We thus
obtain from (2.26), letting W = WT + K, the sum of the wealth of the rentiers and the capitalists,
(2.29) dW/dC = C -1 = [M2 M3]/[1 M1 + M3] > 0,.
provided the price and wealth elasticities are not too large. An increase in credit increases wealth
through an increase in land prices, but has no effect on the capital stock. Since it is only the wealthy
51

The model is obviously stylized, but there are good reasons why land should serve better as collateral that
capital goodscapital goods tend to be constructed for specific purposes, and are less malleable, alterable to
other uses, with often large asymmetries of information concerning the prospects of returns not only in the
intended use, but also in alternative uses. There are other reasons that the provision of credit typically gets
reflected in land bubbles (or bubbles in other fixed assets): when the price of capital goods exceeds the
production costs, the supply will increase, and this limits the extent to which the price can rise or the duration of
any bubble associated with a produced good. (Nonetheless, bubbles of produced goods do occurthe tech
bubble in the nineties and the tulip bubble in the seventeenth century being the most famous instances.
52
And, as we have noted, underlying Pikettys analysis. For simplicity, here we assume that sp is the gross savings
rate, which is assumed to be fixed and based on gross income, where r is now the gross return to capital. We
could rewrite all of these equations based on net savings and net income, without changing any of the results.

21

who own the land, that get access to credit, all of the increase in wealth (capital gain) goes to the
wealthy. Monetary policy causes both the increase in (non-productive) wealth and the increase in
wealth-inequality. Note that in this (polar) model, since credit simply leads to asset price increases (and
an increase in the price only of the fixed asset land)but not commodity price increasesthere is no
reason that a monetary authority focusing on commodity price inflation would circumscribe credit
creation.
The model presented here is highly stylized, and can easily be generalized. We have assumed, in
particular, that capitalists-entrepreneurs are the only ones who do real savings, while
landowners/rentiers simply buy land, and that credit is only provided to the latter rather than the
former.
Alternatively, we could assume that land and capital goods are perfect substitutes for each other, that
there is no consumption value to land, and there are not two separate classes of entrepreneurs and
rentiers. Land and capital are simply alternative stores of value, and in equilibrium they must yield the
same return. Then, as before,
(2.30) dlnp/dt + FT/p = FK.
Moreover, the full income of capitalists is now FK(pT + K), so that capital accumulation is described by (as
before)
(2.31) dK/dt + T (pFK FT) = sp(FK(pT + K)) K.
As before, (2.30) and (2.31) describe the full dynamics of the economy in terms of {p,K}.
Now assume, however, that banking system53 only provides credit with land as collateral, but provides it
at zero interest rate, so that owners of land borrow as much as they can. The central bank limits the
amount of credit that is made available. As more credit is provided, the price of land will be bid up, and
in equilibrium
(2.32) C = pT.
where is the collateral requirement. Thus, there is a path of expansion of the credit supply which
ensures that (2.30) is satisfied. If the financial system expands the credit supply at a pace that is faster
than that implied by (2.30) and (2.32), the return to land will exceed the return to capital. In this polar
model, if this were anticipated, no one would want to hold capital. The price of capital goods would fall
below 1, and the production of capital would halt. K would decrease with depreciation. We then
replace (2.30) and (2.31) with
(2.30a) dlnp/dt + FT/p = FK/q

53

Because we do not want to address issues involving the banking system and the wealth of its owners, we will
simplify the analysis and assume that it is government owned.

22

where q is the price of capital goods in terms of consumption goods; and54

(2.31a) dln K/dt = - .
Consistent with what has been observed, K decreases and p increases. If C increases fast enough, the
value of wealth increases:
(2.33) Tdp/dt K = (dC/dt)/ K > 0
provided only that the pace of credit increase is large enough:
(2.34) d ln C/dt > K/C.
It is clear that this condition can easily be satisfied (and plausibly has been).
Note that the ratio of the (full) income of capitalists to that of workers will, along such a trajectory, be
increasing.
(2.35) YC/YL = FK(pT + K)/FLL.
where YC is the (full) income of capital and YL that of labor. Note too that the return to capital will be
increasing and that to labor decreasing (consistent with what has been happening), while the value of
wealth is increasing.55 Hence trajectories where there is a rapid expansion of credit shift the income
distribution towards capitalists. Of course, on such trajectories, growth in output will be low, in spite of
the rapid increase in wealth.56 This simple model is consistent with all of the stylized facts described in
the beginning of this paper.
In more general models, where rentiers also may buy some capital goods, and where there is not a
linear production possibilities frontier, then an increase in credit leading to an increase in the value of
land can initially lead to more investment, but eventually an increasing proportion of savings is absorbed
by increases in the value of land, and, as hereand evidently as in many countriesreal capital
accumulation diminishes.
While in this and other models in this section, the increase in wealth may be largely (or entirely) due to
an increase in land values, one might ask: does this lead to real inequality. After all, the rich consume
the positional goods. The increase in land values affects them, and them only. Workers are only
affected to the extent that the increase in land values crowds out capital accumulation, so K decreases
(or does not increase as much as it otherwise would.)
54

For simplicity, we assume the labor supply is fixed at unity, so that K is equal to the capital labor ratio. Nothing
essential depends on this assumption.
55
Similar, but less dramatic results obtain in a two sector model, in which as q falls, the production of capital goods
(investment) decreases, but does not drop to zero.
56
Indeed, in this polar model, with no technological change and no increase in labor supply, the rate of growth of
the economy is negative. But the model can easily be extended to the case with technological change and labor
force growth, where there is growth, but lower growth than there would have been without the rapid expansion of
credit.)

23

While this conclusion is true in the simplified models we have constructed here, it is natural that there
be a spill over to workers (and in practice, such spillovers typically occur.) Assume, for instance,
landlords/capitalists rent out some of their land to workers, at a rental price of pFK. Then, policies and
behavior which lead to an increase in pFK disadvantage workers.
Still, the observation that the increase in land prices (or of other positional goods) disproportionately
affects the wealthy has several important implications. First, it reminds that in making comparisons
across different income groups, we have to take into account the different market baskets of goods that
they consume. The increase in the relative prices of positional goods means that there may not have
been as large an increase in inequality as would appear to be the case. Secondly, it helps explain
differences in savings behavior both over time and across income levels. To achieve success as
demonstrated by acquiring expensive positional goods may require more savings today than when the
price of such goods were lower. In effect, an increase in credit leading to an increased price of real
estate may induce a greater willingness to hold assetsgreater savings that might, in turn, provide
monetary authorities greater comfort in their policies of credit expansion (since there is relatively little
spillover to aggregate demand, to the demand for produced goods and services.) Earlier in this paper,
we have noted different hypotheses concerning macro-economic savings functions, differentiating, for
instance, between the savings rate out of wages and capital. As an empirical matter, it may be that
there is a difference between savings out of capital gains, especially those arising from the increase in
the value of real estate, and other returns to capital, precisely because of the consequences of those
price changes for acquiring the goods in the future that the rich seek to purchase. Thirdly, by the same
token, patterns of inheritances and life-time giving across generations too may be endogenous, affected
in particular by such changes. If increases in real estate prices make it difficult for even reasonably
successful workers to purchase a home that they and their parents believe is appropriate to their station
in life, wealthy parents will provide larger intra vivos transfers. Note that, in some sense, the direction
of causality has changed: greater wealth and wealth inequality arising from an increase in real estate
prices has led to greater inheritances and intra vivos transfers across generations among the top.

Part II
Distribution of wealth among individuals
There are forces in the economy which lead to increases and decreases in wealth inequality. We can
think of these are centrifugal and centripetal forces. There exists an equilibrium distribution of wealth
when the two sets of forces are balanced. My earlier work (Stiglitz, 1969a, Bevan-Stiglitz, 1979)
analyzed models in which there existed an equilibrium wealth distribution. In this section, we ask (a) if
there exists an equilibrium wealth distribution, what are the forces that might lead to greater or lesser
inequality (in equilibrium)? And (b) are there plausible circumstances in which the centrifugal forces are
so great that wealth inequality continually increases (as Piketty seems to have suggested)?
3. Basic Model

24

The basic model is a variant of the Solow growth model, where we think of the economy as consisting of
dynastic families, leaving equal bequests among their children. For simplicity, we ignore technical
change. The evolution of wealth per capita for the ith family is described by the differential equation
(3.1) dln ki /dt = si yi ni ,
where yi is the ith familys income (per capita)
(3.2.) yi = wi + ri ki,
where wi is the ith familys wage, ri is its (after tax) return on capital, and ki is its capital (per capita). We
assume that there is perfect inheritance of both labor market and capital market productivity. ni is the
ith families rate of reproduction.
An essential part of the analysis is macro- and micro- consistency: aggregate k (the aggregate capital
labor ratio) determines the average return on capital, r, and wages57 with
(3.3) K = Ki
where KI is the ith familys total capital stock, and K is the aggregate capital stock of the economy. We
assume a neoclassical production function where output per (effective) worker is f(k), and
(3.4) ri = i f(k)
(3.5) wi = i(f fk),
where i is the relative return to the ith familys investment (some families are able to obtain a higher
return from their investments than others, with f(k) being the average marginal return across all
families)58 and where i is the relative return to the ith familys labor (some families receive higher
wagespayments per unit labor--than others, with f - kf(k) being the average wage across all families)
A special case59
Consider a Solow Model, where w, s, r, and n are the same for all families. Then
(3.6) dln ki/dt dln kj/dt = w/(1/ki 1/kj),
So regardless of initial distribution of wealth, there will eventually be equality of wealth.
If s, r, and n are the same, but wi differs across families, then in steady state the wealth distribution
corresponds precisely to the wage distribution. Asymptotically,
(3.7) ki*/ k* j = wi/wj .

57

In a more general model, expectations concerning those variables may affect s.

That is, in the obvious notation, E =1, E =1
59
Stiglitz (1969a)
58

25

If wages are lognormally distributed, so will wealth.

Extension to stochastic model60
Assume wages for each family are determined by the same stochastic process, with regression towards
mean, that there is a lower bound on wealth (individuals cant borrow more than a certain amount), and
that families optimize intergenerational utility.
Then there exists an equilibrium wealth distribution which is related to the nature of the stochastic
process of wages and the intertemporal/intergenerational discount factor. It should be obvious that if
there is faster regression towards the mean, then there is less need for those who are lucky (have high
incomes) to save, to redistribute income from themselves to later generations; hence there will be less
wealth inequality. By the same token, if the current generation has less concern for future generations
(a higher intergenerational discount rate) savings and wealth inequality will be lower. Finally, if there is
higher cross-sectional wage inequality (greater wage inequality at any moment of time) there will be (for
any given degree of intergenerational discounting and any degree of regression towards the mean)
more savings and wealth inequality.
Kaldorian savings
In Kaldors model, a given fraction of profits, sp, are saved, and none of wages.61
(3.8) dln ki /dt = spi ri ni
Assume sp, r, and n are the same for all families.62 Then relative wealth of all families would remain the
same; any initial inequality of wealth would be perpetuated, a result which contrasts starkly with that of
the Solow model, where any initial differences in wealth asymptotically have no consequences:
(3.9) dln ki/dt dln kj/dt = 0.
Note that this is true regardless of the value of r, n, or spthat is, the relationship between the interest
rate and the growth rate has nothing to do with relative wealth inequality. In fact, however, in this case,
in the long run there is a simple relationship between the rate of interest and the rate of growth: In long
run equilibrium
(3.10) sp r = n
so r is greater than rate of growth, but in spite of this, there is no further concentration of wealth. This
is true even if sp =1.
Assume, on the other hand, that families differ in their savings rate, i.e. si for some family is greater than
for some other family. Then its relative wealth will grow without bound. There is ever increasing wealth
concentration at the top. This is a result which is consistent with Pikettys conclusions, but it arises not
60

The details of this model are set forth in Bevan and Stiglitz (1979)
Similar results are obtained in the Pasinetti two-class savings model. (Pasinetti 1962.)
62
It seems as if Piketty assumes sp = 1 for all families, consistent with this assumption.
61

26

from the relationship between the rate of interest and the rate of growth63 but from differences in the
rate of savings among different capitalists.
As we noted earlier, any coherent model must reconcile macro-variables with the micro-analysis, i.e.
(3.11) dK/dt = dKi/dt = sirKi = r siKi =r s* K
where
(3.12) s* siKi / K,
is the weighted average savings rate. There is ever increasing wealth concentration at the top, an ever
increasing average savings rate, an ever increasing capital labor and capital output ratio, and an ever
diminishing rate of return on capital. But K, r, K/Y and s all approach asymptotically limiting values
given by {K**, r**, K/Y** ,s**} . Then asymptotically all of wealth is in hands of wealthiest, and
(3.13) s** = max si
The long run equilibrium is dominated by the family (families) with highest value of si.
(3.14) r** = n/s**.
Savings and returns to capital among capitalists
This analysis has made it clear that what matters is the relation between sr and the growth rate, not r
and the growth rate. The savings rate for even the rich is less than unity (especially once one accounts
for consumption of housing). But to the extent that sp < 1, the equilibrium return will be that much
greater than the rate of growth.
So far, we have assume that the return to capital of all capitalists is the same. Assume that all families
have the same savings rates and reproduction rates, but some families obtain a higher return to their
capital. (There is a semantic question about whether one should view these higher returns as a return
to a particular type of labor servicethe ability to manage capital. In this view, excess returns to
capital should be viewed not really as a return to capital, but as a return to labor. Nothing hinges on this
semantic issue, other than the distribution of income among factors.64) Then
(3.15) dln ki/dt dln kj/dt = sp(ri rj),
Again, capital gets increasingly concentrated in the family with the highest return. On the other hand, if
there is regression towards the mean in the ability to manage capital (as in the Bevan-Stiglitz model),
then there will be an equilibrium wealth distribution, with greater equilibrium inequality the slower the
63

In this model, with no technological change, the rate of growth of population is equal to the rate of growth of
the economy.
64
But for purposes of taxation, the distinction is very important, as the dispute about compensation for equity
managers (covered interest) illustrates. Most of the seeming returns to capital can be viewed as (i) a return to the
management of capital; (ii) a return to risk bearing; and (iii) a return to market power and other forms of
exploitation.

27

speed of regression towards the mean and the greater the variance of abilities of the children, given the
ability of the parent.
Returns to investment are related to the investments one makes in acquiring information, and the
optimal investment is a function of the size of ones wealth (simply because information is a fixed cost.)
We thus postulate that the return to capital for any family is a function of its capital,65
(3.16) ri = r(ki), with r > 0.
In the formulation of (3.16) there are no innate differences in the ability to obtain returns; the only
differences are those that arise out of the optimal amount of resources to allocate to the management
of wealth. If that is the case, then (3.15) and (3.16) imply that families that are richer initially will
increase in wealth relative to other familiesand this is true regardless of the relationship between the
(average) rate of return on capital and the rate of growth. As before, the economy comes to be
dominated by the family (families) with the highest levels of wealth. The system is unstable, in that any
perturbation, in which one family gets more wealth than another, is not only perpetuated, but the
differences increase, to the point where a single family has all the wealth.66
On the other hand, if we combine a model with stochastic ability with regression towards the mean with
increasing returns to capital, it is still possible that there be an equilibrium wealth distribution. The
advantages of returns to scale are offset by the (eventual) disadvantage of relative incompetence
(including the incompetence arising from not knowing that one is incompetent and the inability to hire
competent managers). There is ample anecdotal evidence of this process at work. Still, the effects of
the advantages of scale reflected in (3.16) can give rise to large inequalities within the lifetime of an
individual.67
Inherited human capital
Finally, let us return to the Solow model, but with one modification: assume that all families are
identicalthe only reason they differ in productivity is differences in endowment of human capital. The
wage that the individual gets paid is related to the wealth of his familywealthier families can invest in
more human capital. We let the wage of the individual depend on the aggregate capital labor ratio and
his familys endowment:68
65

It is not necessarily the case that the increasing returns to capital ownership reflected in equation (3.16) is a
social return. It may only reflect the enhanced ability to grab rents, e.g. as a result of access to inside information.
Some of the higher average returns of the wealthier may be a result of their being better able to bear risk.
66
It is not inevitable that there be such a relationship: information is a fixed cost. One can imagine a financial
market which amortized these fixed costs uniformly, in which case those with large amounts of wealth would
obtain the same returns as those with lesser amounts of wealth (putting aside the slight differences that might
arise from non-linear transactions costs.) But, again, because of imperfections of information, there are costly
agency problems, and those with enough wealth may be able to mitigate these agency problems by running their
own investment fund. As we comment later, the markets for information may be an important driver in current
inequalities.
67
Even here, there may be a process of regression towards the mean, as the process of aging occurs.
68
In effect, h is the number of effective labor units provided by an individual.

28

(3.17) w(k,ki) = h(ki) [f(k) kf(k)],

where k is the aggregate effective capital labor ratio,
(3.18) k = ki h(ki) Li /L,
where Li is the total effective labor supply, L = h(ki)Li.
Then there can exist an equilibrium distribution of wealth. In this simple case, where there are no
differences in innate abilities, and there are just two wealth groups in the population
(3.19) [f(k*) k*f(k*)]h(kL) /kL =[f(k*) k*f(k*)]h(kH) /kH,
(3.20) s{ h(ki)[f(k*) k*f(k*)] + rki} = nk*i
(3.21) k* = kH + (1 )kL,
where is the fraction of the population in the upper wealth group, with capital (per capita) of kH. (3.203.21) are three equations in the three unknowns {kH, kL, k*}. In general, there exist many possible
equilibrium wealth distributions, e.g. corresponding to different values of .
If, now, families differ also in their inherited abilities, then the final distribution of wealth will reflect
both the underlying distribution of abilities as well as the fact that families that are lucky enough to have
large amounts of capital can and will invest more in their children.
Inherited human capital and progressive taxation
Assume now that the savings out of (after tax) wages may be less than out of profits, and that there are
separate taxes on wages and capital, with each being progressive, but that all families have the same
innate abilities, savings rates, etc. We denote the after tax income functions by Tw and TK respectively,
that is Tw[w] is the after tax wage income of someone with a wage of w. For the capital per capita of the
family to converge
(3.17) H(k) sw T w[ w(k)]/k + sp TK(r(k)k))/k = n.
where sw is the savings rate out of wages. If H(k) is monotonically declining, as in Figure 4a, wealth will
be equally distributed: families with k greater than k* will see their wealth (per capita) diminish. But
Figure 4b shows a case where, because of the increasing returns to investment, savings per capita
divided by k may, beyond some k, increase. Then, inequality can increase without bound. Finally, figure
4c shows a case where, because of the onset of strong progressive taxation eventually H(k) declines
below n. Then, the economy would wind up with a bi-modal wealth distribution, at k* and k**. (Of
course, in the more general case where there is inheritability of productivitynot related directly to k
there can be more realistic wealth distributions.)
Inequality of consumption versus inequality of wealth

29

Traditional social welfare focuses not on inequalities of wealth but of consumption, and in the models
that we have presented, these do not translate simply. Consider the limiting case of the Kaldorian
model where sp = 1. Then, though there can be greater wealth inequalities (and initial wealth
inequalities are perpetuated), because the capitalists (by assumption) consume nothing, there are no
consumption inequities. But if sp < 1, then wealth inequities translate directly into consumption
inequities.
Similarly, in the Bevan-Stiglitz model with regression of wages towards the mean, where savings serves
to smooth consumption intertemporally, reductions in these intertemporal transfers may lead to less
wealth inequality, but greater consumption inequality.69 In the concluding remarks, we comments on
why, even in this model (which does not capture key elements of inequality in our society today) we
should, nonetheless be concerned with wealth inequality.
Distribution of wealth among dynastic families vs. individuals
Most of this section has traced the evolution of the wealth of families over time. We have looked at per
capita wealth of different families, as their wealth increases as a result of savings and as per capita
wealth gets reduced, as a result of the growth of family size. The wealth of the family with the highest
savings rate may increase more rapidly than the economy as a whole and than that of other families.
But that does not necessarily mean that the wealth of even the richest individual in the family is
increasing faster that the economy.
The per capita wealth of the richest family is growing at the rate (in the Kaldorian model, with a tax rate
of t)
(3.18) spr(1 - t) - n
while the economy is growing at the rate of
(3.19) n +
where is the rate of labor augmenting technological progress. Thus, the size of the wealth of the
richest person in the economy relative to the size of the economy increases only if
(3.20) r > (2n + )/sp(1- t) = (n + g*)/sp(1 - t)
a condition which is much less likely to be satisfied that the condition r > g*. For instance, if n = 1%, =
1.8%, sp = .5, and t = . 2, then r must be greater than 9.5%, a number considerably in excess of the 2.8%
growth rate.

69

For a more extensive discussion of this point, see Stiglitz (1976b)

30

4. The centrifugal and centripetal forces in the economy

The above model provides a framework for understanding the forces that can lead to an equilibrium
wealth distribution that is more or less unequal. As we have already noted, differences in wi, si, ri , and ni
and the stochastic processes for these variables determine differences in relative wealth positions.
Indeed, the model presented in the beginning of section 3in which the wages of each generation are
the same can be thought of as one limiting case: there is no regression towards the mean. The other
extreme is that where there is no correlation across generations (effectively, equal opportunity.) There
will still be inequality.
More dispersion of returns to capital and wages and persistence of differences in returns and wages will
lead to more dispersion of wealth. Thus, one of the factors limiting the build-up of ever increasing
inequality is that the children and grandchildren of those who made a fortune typically have a greater
likelihood of squandering the family fortune that magnifying it, reflected in the adage, from rags, to
riches, and back to rags in three generations.70
So too, more dispersion in reproductive rates and savings rates will lead to greater dispersion in wealth.
If some families have a small number of children, inheritances will have to be divided among a smaller
number, and so wealth (per capita) increases. In particular, if richer families have smaller families, then
there will be more wealth inequality.71
Differences in norms and social custom can lead to other aspects of wealth dispersion: if some groups in
society save more, than they will have more wealth. But one has to be careful in interpreting observed
differences: as we noted earlier, it is possible that there is a poverty trap. Those with low
income/capital save little, but if one could somehow increase their income/wealth enough, they would
move out of this poverty trap. It is not that differences in savings caused the observed differences in
income; it is that the observed differences in income caused differences in savings. This may be
especially so if, as we have suggested, the returns to capital are higher for those with more wealth; for
then, their incentive to save may be higher.
Thus, even the basic model presented in section 3 can give rise to wealth inequality. Assume the savings
function of each family is a function of ki. Then, instead of (3.6), we have
(3.6) dln ki/dt dln kj/dt = w(s(ki)/ki s(kj)/kj) + r (s(ki) - s(kj)).
As before, there can clearly exist equilibrium wealth distributions.
70

We have already noted that if each generation cares a great deal about future generations (low discount factor)
then wealth will become more concentrated (but inequality of consumption across generations will be lower).
71
But the boundary value condition, where there are no children, has just the opposite effect, if those without
children give their estate away. Similar results hold if wealthy individuals decide not to give a large proportion of
their wealth to their children.

31

Beyond the basic model

The model helps frame a discussion of other forces that could lead to greater inequality, by focusing on
the underlying drivers, the level of dispersion of wages and returns to capital, the perpetuation of those
differences, and the transmission of advantages and disadvantages across generations.
For instance, if instead of dividing inheritances equally, the eldest son inherits all the wealth
(primogeniture), then there will be more inequality. Changes in norms of inheritancewhere it
becomes more the norm to divided wealth equallylead to more equality of wealth.
We noticed the potential role of demographics. There are other ways in which demographics, broadly
understood, affects wealth inequality. If, for some reason, there was an increase in assortive mating,
with those with high wages (productivities) marrying others with high wages, then arguably, the pace of
regression towards the mean might be slowed. If previously, alpha males chose mates based on looks,
rather than on characteristics that drove market returns, then the pace of regression towards the mean
in wages would presumably be faster. If one organizes tertiary education to increase the likelihood of
assortive mating on the basis of market productivity, then the pace of regression towards the mean will
be slower.
The OECD has called attention to the role of changing social patterns (e.g. in family structure and
household formation) on income inequality among households.72 These patterns translate, over time,
into wealth inequalities. For instance, if there are some households with two earners and only one
child, they are likely to pass on to their child more wealth than is the case for households with two or
more children and only one earner. Patterns, such as those emerging in the United States, where those
at the bottom of the distribution are less likely to get married are likely associated with less transmission
of human capital across generations, leading to more wealth and income inequality.73
In section I we suggested that some, perhaps much of the increase in wealth and wealth inequality is
associated with an increase in exploitation, broadly understood. If very rich can use their wealth, and
more broadly the position in society that that wealth gives to them, to get higher returns to their capital
and access to better jobs (rents in the labor market, above normal returns in the capital market), then
wealth will become more concentrated. This has, of course, always been trueconnections matter, and
connections are passed on across generations; but if the extent to which this is true changes, then there
will be a change in the equilibrium distribution of income and wealth.
One might have thought that in a meritocratic society these connections matter less, and that may
indeed by the case in those countries, like Scandinavia, who take meritocracy seriously. But in countries
like the US, there is little evidence that the importance of connections has significantly decreased.
Indeed, meritocracy may increase the importance of connections. For instance, increasingly to get a

72
73

OECD (2011).
Daly and Valletta (2006).

32

good job one needs an internship, which is often unpaid. Not only can the children of the less affluent
not afford these internships, but it often takes connections to even get this unpaid work.74
Connections matter in another sphere: politics. In many countries, those with connections are able to
extract rents from the public.75 This is true even in democracies, though it has to be done in a more
rule based way: the manner in which the banks first purchased deregulation, and then received
mega-bailouts, is a case in point.76
How wealth begets wealthand how those in poverty become trapped there-- is well understood.
Those near bankruptcy have to pay higher interest rates, making their descent towards the bottom even
more steep.77 Their attempts at survival occupy so much of their energies that they cannot think about
the long term; and accordingly, they do not make the long term investments that would increase their
incomes.78
Those without wealth cannot get access to credit markets. This becomes especially important in an era
of super low interest rates. But in an era in which interest rates as such are near zeroand even the
return to many risky assets is very lowhow can the inequality of income increase? In Pikettys
analysis, there should be an era of wealth convergence. But instead, there is wealth divergence. Our
usual models differentiate between labor and capital and it would seem that the return to capital,
with the savings glut has plummeted.79 Shouldnt that mean that the share of capital would have
plummeted too80, and so too income inequality? It hasnt, and the reasons that it hasnt are instructive.
The scarce factor in our economy would seem not to be capital, but knowledge.81 Capital flows
relatively freely across borders; yet differences in per capita income persist, and largely because of
impediments to the free flow of knowledge. The banks manipulation of the LIBOR and foreign
exchange markets as well as insider trading scandals exemplify the returns that can be obtained from
information asymmetries82even information asymmetries deliberately created by the market. While
these were outside the law, there are pervasive opportunities to do similar things (with perhaps slightly
74

Perlin (2011).
Much of inequality in many countries is related to privatizations and the sale of public assets at below market
prices (Indias spectrum auction is one of the most recent examples. But there are a wealth of others.) Sometimes
the transfers occur in a more indirect way: the government issues a banking license to someone that is politically
connected; the private bank lends money to favored parties to purchase the state assets that are being
privatized. Restrictions on who can bid ensure that the prices are below what they would be in a competitive
market. Much of the Russian oligarchy was created in this way.
To the extent that connections can be purchased, this just reinforces the increasing (private) returns associated
with wealth ownership.
76
See Johnson, Simon, and Kwak (2010) or Stiglitz (2010).
77
Battiston et al. (2007) refer to this as trend reinforcement.
78
Mani et al. (2013); Mullainathan and Shafir (2009).
79
Bernanke (2005).
80
Under the assumption of an elasticity of substitution less than unity. See the discussion of section I.
81
In further support of this view, we note the emphasis placed by businesses on confidentiality and secrecy.
They know that knowledge is both power and money.
82
It should be noted that only the most egregious examples of the use of inside information are illegal and get
prosecuted.
75

33

lower returns) within the law. It is the belief that there are returns to knowledge that motivates those
who manage capital to invest so much in the acquisition of knowledge.83 But not everyone has equal
access to knowledge; and in markets timing is critical: knowing something slightly before others can
yield large returns.84
Given the asymmetries of informationthose without access to special information know the equity
markets is a stacked gameand given that less well-off individuals are more risk averse85, it is natural
that the richest individuals own a disproportionate share of equities; and that means that when the
monetary authorities lower interest rates, it increases the well-being of those at the very top relative to
everyone else. (See Section 6 for a fuller discussion of this point.)

Earlier, we explained how, if richer individuals (high wage individuals) invest more in the human capital
of their children, so their children have higher wages, the pace of regression towards mean will be
slowed and there will be more wealth inequality. High quality public education can counter this force,
ensuring that everyone faces a more level playing field. But in a society, like the US, where there is a
reliance on local funding for school, if there is more economic segregation86 then there will be more
inequality in the transmission of human capital. So too if greater reliance is placed on tuition for
financing tertiary education, in the absence of adequate scholarships, and this is even true if debt
financing is made available, unless the debt repayments are income contingent, as in Australia. Higher
interest rates charged on student loans will lead to more inequality of human capital; so too would the
passage of a bankruptcy law that makes student debt not-dischargeable even in bankruptcy (as the US
has done with a series of laws dating to the 1970s, the most recent expansion of which was the
Bankruptcy Abuse Prevention and Consumer Protection Act of 2005).
Changes in markets may also lead to changes in the equilibrium wealth distribution. Better insurance
markets mean that individuals have to accumulate less precautionary savings. Those who die with large
amounts of precautionary savings leave more to their children. Better rental markets or reverse
mortgages mean that the elderly are less likely to hold large amounts of real estate wealth.87 Most
importantly, better annuity markets would mean that individuals would have to hold less wealth against
the risk that they live a long time. But annuity markets remain highly imperfect, and individuals are
living increasingly long and that means that upper middle class individuals typically hold significant

83

Interestingly, the efficient markets hypothesis suggested such investments yielded no return: information
disseminated perfectly and instantaneously throughout the economy. But why then would rational individuals
invest so much money in gathering information. . See Grossman and Stiglitz (1980). The evidence, however, is
that markets are not informationally efficient, and that means there are returns to investment in
information/knowledge. Shiller (2000
84
Especially if other market participants are overconfident or unaware of their informational disadvantage. Note
again these are private returns, not social returns. See Stiglitz (1982)).
85
It is a standard assumption that there is decreasing absolute risk aversion.
86
Evidence is that economic segregation has increased. See Bischoff and Reardon (2011).
87
It is worth noting that there are large differences across countries in the relative role of rental markets vs. home
ownership. In Germany, homeownership is relatively low.

34

amounts of wealth into and beyond their retirement. Again, this can give rise to large differences in the
amounts inherited, depending on the age at which individuals die. (Stiglitz 1978.) Thus an increase in
the difference between life expectancy and the age of retirement88 and an increase in the variance of
the age of death will lead to more wealth inequality.
Public policy can have, indirectly, a major impact on each aspect of the generation of wealth inequality
described above. Here, I note only four points. First, following on our discussion of annuities, we note
that the provision of public annuities reduces the need for individuals to save for retirement. But since
most countries only provide limited public annuities, the effect is to make savings for retirement
essentially zero for the bottom part of the population. Hence, overall, wealth inequality is probably
increased.
Taxation of capital and especially bequests has both an income and a substitution effect reducing
bequests, and thus the transmission of inequality. But it results in greater incentives to provide
bequests through the transfer of human capital, which may lead to more wage inequality. And, as we
note in the next section, there can be an effect on the returns to capital, which can increase wealth
inequality.
Earlier, we noted that monetary policy, whether intentionally or not, affects the distribution of income
and wealth. We described how quantitative easing transferred wealth to the wealthy individuals who
own the bulk of equities. Low interest rates encourage firms to use more capital intensive technologies,
reducing the demand especially for low skilled workers. If monetary authorities tighten whenever
wages start to rise, the effect will be a ratcheting down of the wage share.89
Third, we note that any change in markets or public policy which affects the distribution of wages will
(according to our basic model) affect the distribution of wealth. There is an extensive recent literature
on the determinants of wage dispersion, discussing, for instance how globalization and skill biased
technological change may have led to greater wage inequality. The extent to which this is true is not
just determined by market forces, but how those market forces are shaped by public policy, e.g. the
rules governing unionization.
Finally, we note that changes in policy affect not just the distribution of wages, but also the distribution
of factor incomes. For instance, asymmetric trade liberalization (where capital market and goods
market liberalization precede labor market liberalization) exacerbates downward wage pressures in
advanced countries. (Charlton and Stiglitz 2005). Going forward, changes in globalization, the rules
governing it, and the structure of the global economy may affect inequality for another reason: the
increasing share of services (Greenwald and Kahn, 2009) may increase the importance of local
monopolies.
Cyclical effects
88

This variable is critical in the life cycle model described in section 5. In the absence of annuity markets,
individuals care not just about the mean life expectancy; the variability in life expectancy will also affect savings
ratesand therefore the importance of life cycle savings.
89
See chapter 10 of Stiglitz (2012a) for a more extensive discussion of the distributional effects of monetary policy.

35

The models of this paper are concerned with the long run evolution of the wealth distribution. Yet, one
cannot separate the consequences of economic instability from the long run analysis, particularly in the
presence of asymmetries and hysteresis effects. It is those at the bottom that suffer the most from
economic fluctuations (see, e.g. Furman and Stiglitz, 1998)), and in the boom, they do not make up for
what they lose in the recession (especially if monetary authorities follow the kinds of policies described
earlier). Instability may thus contribute to income and wealth inequalitythe recent economic
downturn being a case in point.90
5. Relative role of life cycle savings vs. inherited savings
A key concern in the growing wealth inequality is the worry that we are giving rise to an inherited
plutocracy. Piketty emphasized that with sp =1 and the rate of interest were greater than the rate of
growth (in our notation, if r > n), inherited wealth would increase. On the other hand, the fact that
individuals are living longer and must save for their retirement means that life cycle savings is increasing,
reflected in part in the huge increase in pension funds.91 In this section, we construct a simple model
incorporating both inherited and life cycle savings.
We assume two groups: workers who live two periods, and save for their retirement. The magnitude of
their savings is life cycle savings. Then there are the capitalists, who save a fixed percentage of their
income, sp For simplicity, we use a discrete time model, and assume zero growth of labor and zero
technical progress.
The difference equations describing the evolution of the system are given by
(5.1) Kct+1 = sp f(Kt) Ktc
and
(5.2) Kwt+1 = s(Kt+1) w(Kt)
where KW and Kc are workers and capitalists capital, respectively, where we have allowed the savings
rate of workers to depend on the (rationally expected) interest rate92, and where
(5.3) Kt = Kwt + KCt .
These equations fully describe the dynamics, given an initial value of workers and capitalists capital. In
the steady state Kct+1= Ktc and similarly for Kw. Hence, from (5.1)

90

It should be pointed out, however, that these effects are not unambiguous, since many economic fluctuations
are associated with stock market crashes that especially adversely affect those at the top. Income and wealth
inequality fell after the stock market crash of 1929. The current crisis may have especially adversely affected
workers because of the disproportionate effect on housing wealth, and government policies which seem to have
restored stock market wealth more effectively than housing wealth.
91
data
92
We could have employed a more general savings function: s(Kt, Kt+1) where the savings rate depends not only
on the rate of return on capital but also on wages.

36

(5.4) 1= sp f(K*),
where K* is the steady state value of K and f(K*) is the steady state gross return on capital, equal to 1 +
r. Note that r here is the return over a generation, i.e. if a generation is 30 years, and the annual
interest rate is 2%, r 1. The steady state level of capital (and the equilibrium interest rate) is

determined simply by capitalists saving propensity.

If workers save more, the economy does not become richer; income does not go up; wages do not
increase. All that happens is that they increase their share of total capital.
The steady state capital of workers (life cycle capital) given by
(5.5) Kw* = s(K*)w(K*)
Hence
(5.6) Kw*/K* = s(K*)w(K*)/K*
Using (5.4) this can be rewritten
(5.7) Kw*/K* = s(K*)w(K*)/ f-1(1/ sp )
It is clear that if the savings rate of workers increases, for instance because of increased expected
longevity93, workers wealth increases proportionately, while aggregate wealth remains unchanged. By
the same token, in this model, if the generosity of social security increases, so the savings rate of
workers decreases, workers wealth decreases proportionately, while aggregate wealth remains
unchanged.
The effect of taxation
If we impose tax on capital at the rate tc, we obtain instead of (5.4)
(5.4a) 1= sp (1 tc ) f(K*),
implying that the after tax return to capital is not affected by the tax. There is, in effect, full shifting.
As the tax rate increases, the equilibrium capital stock diminishes.94

93

As we have noted earlier, there are a number of other factors that could affect life cycle savingsthe adequacy
of provision of health care for old age, the efficiency of annuity markets and the extent to which they are affected
by asymmetries of information, and uncertainties both about retirement age, rates of return to capital, and life
expectancies. In practice, there are other institutional factors: most individuals save through retirement
programs, and the rules and regulations concerning those retirement programs can have first order effects on the
amount set aside.
94
We should emphasize that this result is not general. Later in this paper, we consider, for instance, a model in
which capitalists have a choice of assets to hold, and in equilibrium, they hold all of the risky assets. In a
generalization of that model, it is easy to show that a tax on the excess returns to capital over the safe interest rate

37

The effect of the tax on the relative importance of inherited versus life cycle savings is complicated, and
can be shown to depend on elasticity of substitution, on s (the sensitivity of workers savings rate to
interest rates); on how the proceeds of the tax are spent; and on the degree of progressivity of the tax.
Capital taxation with proceeds distributed to workers
To ascertain the effect on the relative importance of lifecycle savings, we have to specify what happens
to the tax revenue. Assume it is redistributed to workers. Then the transfer T is given by
(5.8) T = tcrK*.
Now, (5.6) can be rewritten as
(5.9) Kw*/K* = s(K*)[w(K*) + T]/K*.
Then, to ascertain the effect of an increase in the tax rate on the share of inherited wealth we
simply have to ascertain the sign of
(5.8) d{s(K*)[w(K*) + T] /K*}/dK*.
Normally, an increase in the tax rate will lower the wage, but at least for low tc increase the transfer.
Workers lifetime income w + T increase as tc increases if95
(5.9) [-Kf + tc (f + fK*)] dK*/dtc + rK where
( 5.10 ) dK*/dtc = f/(1-tc) f.
The sign of (5.9) is thus that of tcf2/f(1 tc) < 0 for tc > 0. Hence, the loss in wages is always greater
than the benefit from the transfer.
It is easy to show that an increase in the interest income tax may increase the relative importance of
inherited wealth. At tc = 0, an increase in tc leaves w + T unchanged, lowers K*, and hence, so long as s
0, the share of inherited wealth increases.
The tax also has an adverse effect on the distribution of consumption (well-being). Since the after tax
interest rate facing capitalists is the same, their flow of consumption (in steady state) is unaffected.
Workers life time utility is given by
(5.11) V(r(K), w(K) + T)

leads to more risk taking, i.e. a shift in their portfolio to higher return assets. (Domar and Musgrave, 1944; Stiglitz,
1969b). If these assets are complements to labor, that shift by itself may increase wages. We note later that
taxes on capital gains in land may redirect investment into forms that are more complementary with labor.
95
c
c
c
From (5.4) fK/f dln K = [t /1 t ] dln t

38

We have already shown the derivative of W + T with respect to tc is negative (except at tc where it is
zero). But because the return the worker gets on his savings is unchanged, workers are unambiguously
worse off.
Thus, in the case that would seem to be the most favorable to workerswhere all the proceeds are
redistributed to themtheir income is reduced, their welfare is reduced, and inequality is increased.
Inheritance tax with proceeds distributed to workers
Assume now that only the return on inherited wealth is taxed. Life cycle savings is exempted, e.g.
through IRA accounts. Now, we have a somewhat more complicated problem:
(5.12) T = tcr(K* - Kw*)
where
(5.13) Kw* = s(K*)[w(K*) + T].
Substituting, we obtain
(5.14) Kw* = s(K*)[w(K*) + tcfK*)]/(1 + tcf)
We have already shown that as tc increases w(K*) + tcfK* decreases. Similarly, as tc increases the
denominator increases. Hence, so long as s 0, Kw* decreases, as does the share of life-cycle wealth.
Now, however, the effect on relative consumption (well-being) is more ambiguous. In particular,
at tc = 0,
(5.15) d (w + T)/dtc = d(w/dK)dK/dtc + fK* - fK*w = - fK*w.
On the other hand, f = 1/ sp [1 tc], so df*/dtc = rf*/1 tc = f* at tc = 0. Hence
(5.16) dV/dtc = VI [ - fKw* + fKw*] = 0
That is, the loss in income is precisely offset by the increased return to capital. But for tc > 0, the
interest rate effect is larger, and (because K*w is smaller) the transfers are larger, and workers utility is
increased.
Public investment
The results of this section so far on the ability of the government to improve the wealth distribution
through capital taxation are somewhat disheartening. The problem is that the transfer of money from
capitalists to workers lowers average savings rates, and this leads to an increase in the return to capital,
with a shifting of the burden of taxation. But if, instead, government invests the tax proceeds in the
same proportion that capitalists would have saved, and distributes the rest, instead of equation (5.4a)
we have
(5.4b) 1= sp (1 tc ) f(K*) + sptc f(K*),
39

from which it follows that K* and r are not affected by the tax rate, but the share of overall capital
owned by the capitalists diminishes. Letting W*c represent capitalists wealth, then it is easy to show
that so long as sp < 1, workers total income (w + T) is increased, so that workers wealth (life cycle
savings) is increased, while capitalists wealth diminishes by a factor (1 tc). Moreover, the return
workers get on their savings is unchanged, so their welfare is unambiguously increased.
Extension to land/money
As I have noted elsewhere96, life cycle models without land and/or money, and only capital, have to be
treated with caution. For instance, the problem of oversaving noted by Diamond (1965) simply cannot
occur if there is land: the value of land would go off to infinity if (in the case here were n = 0) the return
to capital were to fall below zero .
We extend the production function in the straightforward way so that Y = F(K, L, T), where T is the
supply of land, T and L are fixed, and F is constant returns to scale.
Instead of (5.1) the capitalists wealth accumulation equation is described by
(5.1a) Kct+1 + pt + 1Tct+1= Wct+1 = sp f(Kt) [Ktc + pt Tct]= sp fK (Kt) Wct
where, in the obvious notation, Tct is the land holdings of the capitalists at time t. In steady state, the
return to capital and the return to land (the return to each of the assets) is the same. We focus on the
steady state where the price of land is fixed. From (5.1a) it is clear that (5.4) holds. We similarly rewrite
(5.2) as (continuing with the obvious notation)
(5.2a) Wwt+1 = s(Kt+1) w(Kt) .
Hence, the steady state equations for life cycle wealth relative to total wealth is now just
(5.17) Ww*/W* = s(K*)w(K*)/W*,
where W* = K* + TFT/FK, and where
since the price of land must be such that the return on land and the return on capital is the same,
(5.18) p = FT/FK,

where K* is, as before, the value of K at which the return to capital is equal to 1/ sp .
A tax on the return to wealth will, as before, lead to a decreased value of K. The reduction in
wealth will typically by high than the reduction in K, because the higher discount rate will
reduce the value of land, but if land and capital are strong substitutes, the return to land may
increase, offsetting this effect. But while in a two factor production function, the decrease in
the value of K necessarily leads to a lower wage, now it may not. Capital and labor may be
96

Stiglitz (2010b).

40

substitutes rather than complements. (Robots may be a substitute for unskilled labor.) If
capital and labor are substitutes, then capital and land have to be complements, and the tax on
capital unambiguously reduces wealth inequality (increases the share of wealth held in life cycle
savings.) But if capital and labor are complements, the opposite may happen.

Human capital
Of course, even if we reduce the capacity of the rich to advantage their children through financial capital
by imposing taxes on inherited wealth, the rich can advantage their children through passing on more
human capital. Here, the structure of the education system is crucial: even with the provision of public
education, a mixed system, such as that of the United States, can provide the children of the rich with an
elite education which passes on advantages from one generation to the next, not only through the
formal skills acquisition (including the ability to think creatively) but also through the informal networks
and social skills which are imparted.
Towards a more general model
This section has assumed that society is composed of two groups of individuals, workers who engage in
life cycle savings, and capitalists who pass on wealth from one generation to the other. We have
ignored the kinds of inequalities within each group that were the focus of sections 3 and 4. Obviously,
we could combine the analyses: wage inequalities will give rise to inequality in life cycle savings;
inherited inequalities will be passed on from one generation to the other, as described earlier. We can
also formulate models in which there can be transitions from one class to another. Assume, for
instance, that providing bequests is a luxury, and that when individuals wealth exceeds a certain level,
they begin to act like capitalists. On the other hand, with stochastic returns to wealth, the wealth of
dynastic families can fall below that critical threshold, in which case they stop providing inheritances:
the only savings is life-cycle savings.
6. Credit and wealth inequality
We have argued that much of the growth of wealth is associated with the increased value of land (and
other fixed assets); and hinted that much of the increase in that growth is related to an increase in
creditgiven a system of credit based on collateral. By the same token, the system by which credit is
provided may be one of the main sources of wealth and income inequality: if a favored few get access
to credit, then their wealth increases relative to those without such access. Nowhere was this clearer
than in the former Soviet Union, where bank licenses were granted to some politically connected
individuals. The access to funds that this provided enabled them to acquire state assets as they were
being privatized; the limited access to funds meant that competition was limited and they could acquire
the state assets at far below fair market value.
Our system of credit creation may perversely create not only inequality at the top, but also at the
bottom. It persuades the poor to borrow beyond their ability, and then charges them usurious interest
41

rates. Changes in bankruptcy laws making it ever harder to discharge debts creates a system of partial
indebted servitude. Struggling to survive, they have no ability to make investments that would help
them emerge from such poverty, and indeed, even to think long term.
To see more precisely how the rules of the game on credit creation can affect the distribution of
wealth, first consider the model of section 2.E, where credit is provided at a low rate against land as
collateral. The return to holding land T is then the capital gain on land, the yield on land, and the
option that owning land provides to get access to capital at a low rate:
(6.1) T = (1 + )FK = (1 + )( d ln p/dt +FT /p )
where here owning a dollars worth of land allows one to borrow enough to increase ones land holdings
to (1 +), on each unit of which one obtains a return equal to the return on capital. In equilibrium, the
return to land must equal the return to capital, and this means that along the equilibrium path, if there
is a change in the rules of the gamesay a lowering of the collateral required for a loanthen there will
be an increase in the price of land: those who are lucky enough to own land at that moment receive a
large capital gain. Such a change could be motivated by an improvement in the ability to manage risk, or
by political influence, with the financial industry persuading politicians that such a change would allow a
more efficient capital market. Of course, such a change in the regulations regarding lending does not
increase the amount of real resources available in the economy, even if it might allow banks to lend
more, and thereby might increase the profitability of banking.
A slight variation of the life cycle model of section 5 allows us to explore in more detail some of the
distributive consequences of such a change or similarly, of a change in monetary policy that resulted in
lower lending rates. Here, we investigate these issues in a highly stylized model, that provides insights
into the natural reasons that the ownership of land or other assets that might be used for collateral
should be concentrated to the top. The issues can be seen more clearly in the context of a model where
we assume only two factors of production, capital and labor, and that the ownership (equity in capital)
can be used for collateral.
Assume that workers are very risk averse, while the wealthier capitalists are (close to) risk neutral. We
augment the model of section 5 by assuming that the government issues a fixed number of bonds B;
each bond pays a fixed (real) interest rate, r*, which is controlled by the government (monetary
authority). We assume that the returns to capital are variable, so that all the capital is owned by the
capitalists (they are the owners of equity), and all government bonds are owned by workers. Again, for
simplicity, we assume that capitalists save and reinvest all of their gross income. The price of the bond
is q. Thus the real rate of return to holding a bond is r*/q. Because of risk aversion, r*/q can be
substantially below EFK, the expected return on capital, and workers will still hold their wealth in
government bonds. On the other hand, so long as r*/q is less than EFK no capitalist will hold a
government bond. The price of the bond adjusts so that all of workers savings is held in bonds97, i.e.
assuming a constant savings rate of s,

97

We again assume a constant labor supply and normalize the labor supply at unity.

42

(6.2) Bq = sw
Interest on government bonds is financed through taxation. Not surprisingly, the structure of taxation
matters.
Assume for simplicity that interest payments to workers are financed through a lump sum tax on wages,
i.e.
(6.3) r*B =
and
(6.4) Bq = s(w ) = s(w r*B)
so
(6.5) q = s(w/B r*).
It can be shown that equilibrium requires q =198, i.e.
(6. 6) B = sw/(1 + sr*)
Now, a change in r* financed by a tax on labor leaves the returns to capital unchanged, and that means
that K* is unchanged99 and w is unchanged; but it necessitates a change in B and . In particular, it can
be shown that an increase in r* leads to an increase in 100. It thus leads to decreased first period
consumption, but to increased second period consumption.101 Since across steady states,
(6.7) C1 + C2 = w,
the steady state utility of workers is maximized at r* = 0 (when in effect individuals face the same
constraint).102
In this model, the T bill rate is totally divorced from the rate of return on capital. We can, however, link
the two, by assuming that the government, while borrowing from workers (who are engaged in life-cycle
savings) is willing to lend to capitalists at a rate that is equal to or greater than that rate. For simplicity,
we assume that there is a single rate, but that the government rations the amount it is willing to lend to
capitalists, since so long as r* < FK, risk neutral capitalists will want to borrow as much as possible. The
98

With all of profits going into (gross) investment, aggregate consumption must equal wages. Second period
consumption is just B + r*B, i.e.
C1 + C2 = (1 s) (w ) + + B = w s(w ) + B = w, from which (6. 6 ) follows immediately
99
As in our earlier models, the long run equilibrium return, the aggregate capital labor ratioand hence the
wageare determined simply by capitalists saving behavior.
100
= rB = srw/(1 + sr*). dln /dn r* = 1 sr*/1+sr* = 1/1+ sr* > 0
101
C2 = (1 + r*) B = (1 + r*) sw/(1 + sr*). dln C2/dln r* = [r*/1 + r*] -[ sr*/1+ sr*] = r*{1 + sr* s sr*}/(1 +
r*)(1+sr*) = r*(1-s)/ (1 + r*)(1+sr*) > 0.
102
One could conduct a full dynamic analysis, rather than focusing on steady states, with much the same results.
Focusing on steady states greatly simplifies the calculations.

43

way it rations credit is to require collateral. Hence, if a unit of capital allows a firm to borrow , the
overall return to a dollar of accumulation is FK(1 + ) r*. Hence, the long run equilibrium condition is
(continuing with our simplifying assumption of sp =1)
(6.8 ) FK(1 + ) r*= 1
In the short run, a lowering of r* leads to an increase in the net income of capitalists by an amount
proportional to K* and a reduction of the income of life-cycle savers/workers by an amount
proportional to sw (the value of their savings). Note that in this model, the distributive consequences of
a lower of the interest rate are the opposite of that derived in conventional class analysis, where
workers are seen as debtors and capitalists as creditors. In that model, a lowering of the interest rate
hurts capitalists and helps workers. Here, workers and capitalists are both owners of capital, but
different kinds of capital. A lowering of the interest rate helps owners of equity and hurts those who
hold government debt. This model seems to be a better description of the modern economy, and in this
model, lowering interest rates unambiguously contributes to growing inequality. (This model,
however, abstracts from Keynesian aggregate demand effects, which are the central motivation in
lowering interest rates. We have assumed a full employment neoclassical economy.)
Over the long run, with fixed, a lowering of r* increases the return to investing, implying a higher
equilibrium value of K, and a higher wage rate, from which workers gain. Moreover, as r* is lowered,
they gain also from the lowering of . But once r* is lowered below zero, there is an offsetting distortion
in the intertemporal pattern of consumption. This means that there is (from workers long run welfare
perspective) an optimal r* < 0.103
Inequality in wealth is given by
(6.9 ) R = sw/K
where R is the share of wealth owned by workers104 105
( 6.10 ) dln R/dln r* ={ -Kf/f Kf - 1} dlnK/dln r* = {FK / - 1} dlnK/dln r*,
where is the elasticity of substitution.
For very large elasticities of substitution, the increase in K has little effect on w, so inequality increases;
while for small elasticities of demand, the increase in K increases wages significantly, and reduces
inequality. Note that in the steady state equilibrium, from (6.8)
( 6.11 ) FK = [1 + r*]/1 + ,

103

In our model, the rate of growth of the labor force is zero, and the rate of labor augmenting technical progress is
zero. Thus, the long run rate of growth of the economy is zero. The critical condition involves the relationship
between the rate of interest and the rate of growth.
104
The elasticity of substitution is equal to f(f-kf)/kf
105
We note that because we have normalized labor supply at unity, which is fixed, the capital labor ratio, usually
denoted by k, is the same as the level of capital stock, K.

44

so inequality is reduced if
(6.12 ) < [1 + r*]/1 +
For instance, at r* = 0106, inequality is reduced if
( 6.13 ) < 1/1 + .
Thus, if there is a collateral requirement of .2 ( = 4), then wealth inequality increases as r* is lowered so
long as > .2. Under plausible conditions, then, lowering of the interest rate not only increases
inequality in the short run, but also in the long run.
It should be obvious that the higher the interest rate charged capitalists, the higher the collateral
requirements, and the lower the elasticity of substitution, the more likely is it that lowering interest
rates will reduce wealth inequality.
Similarly, the share of national income going to workers is
( 6.14 ) w + r*B /F(K) = w (1 + 2sr*)/(1 + sr*)F(K)
so
(6.15) d ln {w + r*B /F(K)}/dln r*
= (dln w/dln K)(dln K/dln r*) + Wsr* [ 2(1 + sr*) (1 + 2sr*)]/(1 + 2sr*)(1 + sr*)
= (dln w/dln K)(dln K/dln r*) + wsr*/(1 + 2sr*)(1 + sr*)
As before, a lowering of the interest rate reduces inequality if the elasticity of substitution is low
enough, but increases inequality otherwise.
The essential insight of this analysis is that differences between life cycle savers and capitalists affect the
asset composition of their holdings, and this means that policy changes (tax, monetary, and regulatory
policies) have differential effects on the two groups. A natural question is, cant the process of credit
allocation be changed to ensure that the rents that are effectively being given the owners of capital
through credit creation are more fairly shared? Why not have an auction of credit, so there wont be
any rents?
Part of the answer is provided by the theory of information asymmetries: Stiglitz-Weiss (1981) and a
large subsequent literature has explained why the provision of credit cannot be auctioned. There has to
be an allocation process, entailing judgments about who is most likely to repay. But if that is the case,
then who controls the allocation process makes a difference. Because it is a difficult task, entailing
difficult judgments, it is natural that it be entrusted to those who are better educated, to the elites. But
the elites are better judging those that are similar to themselves; there is an additional element of risk in
judging those that are different. Moreover, there are shared judgments about risks and values. Not
106

Recall the remark of fn. 97.

45

surprisingly, then, they allocate capital to those that are similar to themselveseven when and where
connected lending is prohibited; and, of course, even more so when connected lending is allowed. In
this manner, inequality builds on itself.
But that doesnt mean that there arent excessive rents built into the financial system, and not just
through the abuses that have been especially well-documented in the aftermath of the 2008 crisis,
based on market exploitation (see, e.g. Stiglitz, 2010). Consider, for instance, the allocation of credit for
mortgages. Today, such allocation is not based on judgment so much as credit scoring. It is an
information intensive process, involving the processing of information about the incomes of the
borrower and the values of the properties being acquired. But government entities have the best data
basis, and the government is in the best position to enforce the debt contract: the government, through
the income tax system, has a complete history of income, and through property registries, of the
transaction prices. The incremental cost of collecting mortgage payments through the income tax
system in negligible. Indeed, it could easily construct a system of income contingent mortgage loans
that would be far better than the current system.107 Administrative costs for such a system are likely to
be very low, so that mortgages could be provided at an interest rate only slightly greater than that paid
on government debt. The huge rents (and the associated instability and inequality) of the private
mortgage system could be greatly reduced.

7. Concluding Remarks
The first objective of this paper has been to explore whether there is a long run equilibrium wealth
distribution, or whether, under plausible conditions, wealth inequality continually increases. While
recognizing that there are models in which inequality could increase without bound108, our analysis
reinforces the earlier conclusions of Stiglitz (1969a) that under plausible conditions there is an
equilibrium wealth distribution. The increase in wealth inequality that we are currently observing is
more likely the consequence of a shift from one equilibrium distribution to a more inegalitarian one.
Moreover, in the models explored here, even when inequality increases, it approaches an asymptotic
value, and the increase is not so much related to the condition emphasized by Piketty, that the rate of
interest is greater than the greater of growth, as to differences in the savings rates, rates of return, and
fertility of different families. Indeed, even when the rate of interest is substantially greater than the
rate of growth, there can be convergence.
The second objective of this paper has been to try to understand what changes in the key parameters
determining the equilibrium wealth distribution might account for the growing inequalityto refocus
attention on many ways that inequality is created. Economists have long sought to explain inequality,
107

For a discussion of the merits of income contingent loans, see Chapman et al 2014
One model that we have not discussed is that of Stiglitz (2014), which notes that if the elasticity of substitution
is greater than unity, and the factor bias of technological change is endogenous, then the steady state equilibrium
is unstable, and the economy could diverge towards an equilibrium in which the share of capital is unity.
108

46

with the dominant explanation being marginal productivity theory. Under such a theory, it is the
transmission of assets (financial assets, abilities, human capital, other advantages, e.g. associated with
connections) that is crucial, and much of our paper is devoted to studying these transmission processes.
We have identified both centrifugal and centripetal forcesforces that contribute to less equal, or more
equal, wealth distributions.
With increased longevity, some had thought that life cycle savings would become increasingly
important, relative to inherited wealth, and for a while, that appeared to be the case, as pension funds
increased. But more recent evidence suggests a resurgence in the importance of inherited wealth.109 In
section 5 of this paper, we attempted to formulate a general theory that would explain the relative
importance of life cycle vs. inherited wealth, and within which we could ascertain the effects of
increased capital taxation. While increased longevity might indeed increase the share of life cycle
savings, the development of social insurance programs has had just the opposite effect (though at the
same time, the effect on the distribution of well-being can be quite different than the impacts on wealth
distribution.) More surprising was the result that unless care was taken in the imposition of capital
taxation, there could be sufficient shifting that the actual effect would (in the long run) be exactly the
opposite of that intended.
The intergeneration transmission of advantage, whether accomplished through transfer of financial
wealth, human capital, or connections, is important because societies in which positions are in an
essential way based on inheritance are fundamentally different from those in which positions arise from
individuals own efforts and abilities. Such societies cannot claim that there is a level playing field, that
there is equality of opportunity. There is already ample evidence that this is true in the US, with the
children of the rich who perform poorly in school ending up with higher incomes than the children of the
poor who do well; and with a young Americans life prospects being heavily dependent on the income
and education of his parents.
For more than two centuries, there has been an attempt to break away from a feudal system in which a
childs position in society is pre-ordained by that of his parent, and move to a meritocratic system where
it is determined by the childs own ability. In many respects we have succeeded, but perhaps not as
much as we had hoped: the evidence is that even in a society like the United States avowedly
committed to meritocracy, inherited advantages play a key role, and more than a role than can be
explained by the process of transmission of genes. The models presented here help explain why that is
so.
So too, we should be concerned with wealth inequality, however it is generated, because societies in
which there are large wealth (and income) inequalities function differently from more equalitarian
societies. There are social and political consequences.110

109

See Piketty (2014); Bowles and Gintis (2002).

It is worth noting that the attack on monopolies and trusts in the Progressive era was more motivated by
concerns about their political and social consequences than the market distortions to which they gave rise.
110

47

No one could easily defend the inherited inequalities of the feudal period; one could, at best, suggest
that they were pre-ordained, and to try to change these natural orders would lead to social havoc, with
an unacceptable price to be paid. But as new inequities emerged in the 18th and 19th century, new
explanations had to be found. Two broad strands of thought emerged: one seeking to understand the
evolving distribution in terms of exploitation of market power, the other in terms of social contribution
(the marginal productivity theory.) Nassau Senior, the first holder of the Drummond Chair of Political
Economy at Oxford (which I held in the 1970s), argued that inequality arose from the greater abstinence
of the wealthy, who abstained from consuming to increase their wealth, with the increasing wealth
increasing productivity and wages.
The analysis of this paper (based partly on a reinterpretation of the insights of Piketty) show that that
theory may provide limited insight into what has been happening in recent years111.
First, the increase in wealth is related to the slow and steady setting aside of income for the creation of
more round about means of production (to use Bohm Bawerks term), to enhance the capital deepening
of the economy, only to a limited extent. Rather, it is related to the increase in the value of landnot
the creation of more land, but just to the increase in the price of existing land. Indeed, the amount of
capital goods may actually be decreasing, and societys future prospects may get worse even as the
value of its wealth increases, a result which is strikingly different from that of the standard model. The
fact that the value of land in the Riviera has soared does not mean that there is more land, or that
France is richer; but if the increase in the capital stock is not large enough to offset the increasing
population and to offset the fact that the land supply is fixed, it means that the country (at least on a per
capita basis) is poorer. And it also means that wages are lower than they otherwise would have been.
There is an easy way of addressing the inequality that arises through capital gains on land: to tax capital
gains, especially those associated with increases in land values. This, of course, has long been part of
the progressive tax agendaadvocated in the nineteenth century by Henry George.112
We have suggested that the increase in the value of land and the distribution of ownership claims may
be related to the provision of credit by our financial system. As Galbraith (2012) and Turner (2014)have
suggested, our financial system is at the heart of the creation of inequality in our modern economy. The
standard (nineteenth century) model of the financial system is that it intermediated between savers and
borrowers, between farmers who had more seed than they wanted to plant or consume, and between
those who wanted to plant more seed than they had. The financial system was thus essential in
translating the abstinence of the savers into productive investments. But as Greenwald and I (2003)
pointed out, this model of a seed financial system does not describe our modern financial system.
Credit, giving one party the ability to spend more than his income, is created out of thin air, under the
credibility of the banks and the governments that back up the banks, and limited only by the regulatory

111

And perhaps what happened in former years as well

Henry George (1879). In this paper, we have focused on long run equilibrium models, most of which have been
simplified to the point that there are no capital gains, in the long run. Hence, the models presented here do not
enable us to address the question of tax shifting of such a tax.
112

48

authorities of central banks and the incentive structures that they provide based on bank capital.
Indeed, net, the financial system does not even raise capital for the corporate sector. (Mason, 2014)
Indeed, we have suggested that the traditional division of society into the owners of capital and workers
may no longer be appropriate, given the large amount of life cycle savings. The relevant division is
between capitalists, who pass on their wealth from generation to generation, and workers.
The central issue of wealth inequality is the proportion of overall wealth owned by the capitalists.113 If
the two groups differ in their risk aversion, then their asset holdings may differ markedly, and policies
affecting the returns on these different assets have large distributional effects. For instance, if
government bonds are viewed as safe, or at least much safer than capital goods, then
disproportionately, life cycle savings will be in government bonds. The division between capital and
labor needs to be rethought. With capitalists disproportionately controlling the equity in the economy,
we have to ascertain how different policies affect bonds (T-bills) and equity differently. Thus, a lowering
of the tax on capital gains provides benefits to the owners of capitalto the capitalists, not the life-cycle
saversand this increases inequality.
The real difference in these perspectives is seen in recent changes in monetary policy: a lowering of
interest rates benefits holders of equityagain the capitalistsbut hurts holders of government bonds,
disproportionately life-cycle savers. Traditional analyses would see a lowering of interest rates as
adverse to the interests of the rich.
There can be large distributive effects of monetary and financial regulatory policy, that are quite
different from those reflected in traditional discussions. (Of course, central banks focus on the effects of
their policies are aggregate demand; and in all of the models examined here, the economy is at full
employment. Still, the distributive effects we have identified may arise.) In a more fully articulated long
run equilibrium model with workers with life cycle savings and capitalists, we have shown that a
lowering of interest rates paid on government bonds and charged to capitalists is likely to increase
inequality, especially if the elasticity of substitution is not too low, and if collateral requirements are not
too high. By the same token, we have shown how a lowering of collateral requirements does not result
in an increase in the overall efficiency of the economy (there is nothing to the argument that such a
change allows capital to work more efficiently), but it does lead to more inequality.
We have also suggested that it is not just differences in social contribution, as suggested by marginal
productivity theory, that explain inequality; and it is not just differences in asset ownership (as described
above) that accordingly explain changes in inequality. There is ample evidence of what might be
described as market distortions, interpreting that word broadly to describe any deviation from the
standard competitive paradigm, including exploitation (discrimination, wealth appropriation as a result
of rent seeking, taking advantage of imperfections in corporate governance laws, a variety of forms of
exploitation by the financial sector). While there have been some, partially successful efforts to reduce
113

To repeat: these characterizations, while highly useful, obviously simplify. Large wage inequalities translate
into large inequalities in life cycle wealth, and there is some movement between the two categories we have
defined.

49

discrimination at the bottom, there are good reasons to believe that there may be more exploitation at
the top, and this too may account both for the rise in inequality (both of income and wealth) and the
increase in the wealth income ratio. Indeed, some have suggested that a defining characteristic of at
least American style capitalism in the 21st century is the growth of a new form of corporate control,
different from the managerial capitalism that defined the mid 20th century, dominated by rent
extraction (engineered via the financial sector) out of the corporate sector.114
Some, perhaps much, of what has occurred is related to the political process. It affects not just the
redistribution that occurs at any moment through the tax system, but also the before tax-and-transfer
distribution of income; indeed, much of the rent seeking occurs within the political process, and the
scope for rent seeking within the private sector is affected by the legal framework (anti-trust and
corporate governance laws and how they are enforced); policies, adopted through political processes,
limitor reinforcethe intergenerational transmission of advantages. The increase in inequality in
income and wealth, at least in some countries, has translated into increased inequality in political
power, reinforcing the centrifugal forces already at play. The growing inequality in our society is thus a
reflection as much of democracy in the 21st century as it is of capitalism in the 21st century.
Because so much of the increase in inequality in income and wealth is related to changes in policies and
politics, changes in those policies may be able to ameliorate this growing inequality. There are changes
in legal frameworks, taxes, and expenditures which would lead to less inequality in both the short run
and the long.
Our analysis has emphasized the importance of taking a general equilibrium perspective. Some policies
that might seem to reduce inequality may, because of the shifting of taxes and expenditures, have a
more ambiguous effect. For example, we showed that a tax on the return to capital, with the proceeds
provided as payments to workersa policy which on the face of it would seem to unambiguously
reduce inequalitymay have the opposite effect because of tax shifting; but if the proceeds of the tax
are spent on public investment goods, there can be unambiguous reductions in inequality.
In the end, we dont have to answer the question whether inequality will continually increase. If we
believe that there are large costs to our society, our democracies, and our societies of this growing
inequality, then at the very least, we should ask are these changes in policy which will slow down this
increase in inequality, and perhaps reverse it. An understanding of the forces that may be contributing
to the growing inequality is a first step in constructing such a policy agenda.

114

See, for instance, J.W. Mason, 2014. He points out a striking number of ways in which this new form of
capitalism is different from either managerial capitalism or the capitalism reflected in the neoclassical model, and
describes the financial system more as a mechanism for distributing cash out of the corporate sector than of
raising funds for the corporate sector.

50

Appendix: Theoretical Models Explaining the Long Run Capital Output Ratio
The standard Solow model provides the simplest framework within which we can attempt to interpret
recent changes in factor shares and the capital output ratio. In that model, a fixed fraction s of national
income is saved, so the rate of growth of capital K is given by
d ln K/dt = sY/K
where Y = income, produced by a constant returns production function F
Y = F(K,L) = L f(k) = f(k)/k
where f is output per worker and k is the capital labor ratio. If we assume labor augmenting
technological progress, so k is now the capital-effective labor ratio,
dln k/dt = sf(k)/k - n +
where is the rate of labor augmenting technological change. In steady state equilibrium.
f(k*)/k* = (n + )/ s.
The standard equilibrium is depicted in figure A1.
The first question we ask, then, is Can we explain variations in wages, shares, returns to capital, and
the capital output ratio in terms of such a model?
The model provides a straightforward explanation of a decrease in Y/K: k increased. And it provides a
straightforward explanation of why k might increase: either s increased, or n or decreased (Figure 1).
In many advanced countries (such as the US) it doesnt seem to be case that s increased, but n has
decreased.115 It is ambiguous what has happened to (some evidence that it may have decreased from
early 70s to early 90s, increased for a while after that, and then decreased.)116
But even if there were changes in s, n, and that could explain a change in k (and it is not clear that
there have been such changes) this interpretation runs afoul of three problems. The first is that the
increase in k should have been accompanied by increasing wages and diminishing returns to capital. It
has not. The second is that it would be accompanied by an increased share of capital only if the
elasticity of substitution is greater than unity. While there is some debate about the value of the
elasticity of substitution, most of the econometric evidence, cited in the text, suggests an elasticity of
substitution less than unity. Thirdly, it cannot account for the magnitude of the increase in the wealth
(capital) income ratio or the timing of the changes.
1.2. Kaldorian models
115

Net national savings rates for the US have varied considerably, from say 10.5, say in 1980, to 1.6% in 2009.
World Bank data: http://data.worldbank.org/indicator/NY.GNS.ICTR.ZS.
116
For the US, annual population growth rate: 1.17% in 1970, 0.72% in 2013. World Bank data:
http://data.worldbank.org/indicator/SP.POP.GROW.

51

While the Solow model is the simplest model of capital accumulation, it leaves out two key issues upon
which we focused in the paper: the fact that those at the top, many of whom derive much of their
income from capital, save at a much higher rate. In the 1960s, these ideas were extensively explored,
e.g. by Nicholas Kaldor and Luigi Passinetti. Moreover, one of the key concerns today is the relative
importance of inherited wealth versus life cycle savingsanother topic which was extensively analyzed
earlier.
Here, we show that the results of these models are (for our purposes) roughly consistent with those of
the Solow model. These models too have difficulty explaining the new stylized facts.
The essential feature of Kaldors model is that savings depends on distribution of income. Here, we
explore only the simplest version where only capitalists save. Using the same notation as in the
previous section,
dK/dt = sP rK
where r is interest rate, and sp is savings rate out of profits.117
Equilibrium in the Kaldor model with competition (where r = f) is given by
f(k*) = (n + ) /sp
The analysis much as before, replacing average product of capital with marginal product of capital.
Typically the two move together. Similarly, the analysis of the share of capital is basically unchanged.
(Figure A2)
1.3. Optimizing model
Much of modern macro-economics eschews these simplistic models because the level of savings is not
chosen optimally, on the basis of an intertemporal optimization model. There is considerable evidence
that in fact individuals do not behave according to that standard model, including studies in behavioral
economics.118 For our purposes, however, this debate is of limited relevance. The standard optimizing
models yield results that, in the long run (upon which we are focusing) are similar to those obtained in
the Solow and Kaldorian model, even though the savings rate is endogenous.
The standard model119 yields the equilibrium condition
n + = r = f(k*)
where is the pure rate of time preference and is the elasticity of marginal utility. An increase in is
an increase in impatience, and has an effect that is similar to a decrease in sp in the Kaldorian model.
(See figure A3).
117

Piketty in his most recent book (2014) seems to assume that sp = 1. The evidence is sp < 1.
Carroll (1998); Banerjee and Mullainathan (2010).
119
Where the representative individual maximizes the integral of discounted utility U(C). Note that the standard
model is highly restrictive (i.e. assuming intertemporal separability, and time independence of utility).
118

52

1.4. Generalized savings model

In the standard overlapping generations model, workers save a fraction of their wage income:
w(k) = f(k) kf(k),
so
s(kt+1, kt ) w(kt ) = (n + ) kt+1
where we have postulated that the workers savings rate can be a function of the (expected) return to
capital, as well as the wage, the former depending on kt+1, the latter on kt. It is easy to show that there
can exist multiple equilibrium steady states, i.e. solutions to120
s(k*, k* ) w(k* ) = (n + ) k*.
Shocks can move the economy from one equilibrium to another, or shift the equilibrium value of k.
(figure A4). But again, within the confines of this model, it is hard to generate changes in the capital
(wealth) output ratio of the kind observed (if we interpret wealth to be capital.)

Figures: see separate file.

120

53

See Stiglitz, 2010.

References
Acemoglu, Daron, D. Autor, D. Dorn, G. Hanson, and B. Price, Import Competition and theGreat U.S. Employment
Sag of the 2000s NBER working paper, August 2014
Aitchison, John and J.A.C Brown (1957). The Lognormal Distribution. The Economic Journal, 67(268): 713-715.
Arnott, R. and J. E. Stiglitz (1979) Aggregate Land Rents, Expenditure on Public Goods and Optimal City Size,
Quarterly Journal of Economics, 93(4), November 1979, pp. 471-500.
Arrow, Kenneth J., Hollis B. Chenery, Bagicha S. Minhas, & Robert M. Solow (1961). Capital-Labor Substitution and
Economic Efficiency. The Review of Economics and Statistics, 43(3):225-250.
Autor, David (2002). Skill Biased Technical Change and Rising Inequality: What is the Evidence? What are the
Alternatives? Working paper, MIT and NBER.
Autor, D. H., Katz, L. F., & Kearney, M. S. (2008). Trends in U.S. Wage Inequality: Revising the Revisionists, Review
of Economics and Statistics, XC(2), 300-323.
Banerjee, Abhijit and Sendhil Mullainathan (2010). The Shape of Temptation: Implications for the Economic Lives
of the Poor, NBER Working Paper 15973.
Battiston, Stefano, Domenico Delli Gatti, Mauro Gallegati, Bruce Greenwald, & Joseph E. Stiglitz (2007). "Credit
chains and bankruptcy propagation in production networks," Journal of Economic Dynamics and Control,
31(6): 2061-2084.
Battiston, S., D. DElli Gatti, M. Gallegati, B. Greenwald, and J.E. Stiglitz (2012). Liaisons dangereuses: Increasing
connectivity, risk sharing, and systemic risk. Journal of Economic Dynamics and Control, 36(8): 11211141.
Bernanke, Ben S. (2005). The Global Saving Glut and the U.S. Current Account Deficit, The Sandridge Lecture,
Virginia Association of Economists, Richmond.
Bevan, D. and J.E. Stiglitz (1979). Intergenerational Transfers and Inequality,, The Greek Economic Review, 1979.
Bischoff, Kendra and Sean F, Reardon (2011b). Income Inequality and Income Segregation, American Journal of
Sociology, 116: 1092-1153.
Bowles, Samuel and Herbert Gintis (2002). The Inheritance of Inequality. Journal of Economic Perspectives, 16(3):
3-30.
Braverman, A. and J.E. Stiglitz (1989). Credit Rationing, Tenancy, Productivity and the Dynamics of Inequality, in
P. Bardhan (ed.), The Economic Theory of Agrarian Institutions, Oxford: Clarendon Press, 1989, pp. 185201.
Card, David and John E. DiNardo (2002). Skill-Biased Technological Change and Rising Wage Inequality: Some
Problems and Puzzles. Journal of Labor Economics, 20(4): 733-783.
Carroll, Christopher D. (1998). "Why Do the Rich Save So Much?," NBER Working Papers 6549.
B. Chapman, T. Higgins, and J. E. Stiglitz (eds.), Income Contingent Loans: Theory, Practice and Prospects,
Houndmills, UK and New York: Palgrave Macmillan, 2014.
Daly, Mary C. and Robert F. Valletta (2006). Inequality and Poverty in United States: The Effects of Rising
Dispersion of Mens Earnings and Changing Family Behaviour, Economica, 75(289): 75-98.
Dasgupta, Partha and Joseph Stiglitz (1980). Industrial Structure and the Nature of Innovative Activity. The
Economic Journal, 90(358): 266-293.
Diamond, Peter A. (1965). National Debt in a Neoclassical Growth Model, The American Economic Review,55(5):
1126-1150.

54

Domar, Evsey D., and Richard A. Musgrave, Proportional Income Taxation and Risk-taking, Quarterly Journal of
Economics 58(3):388-422
Feldstein, Martin (2014). Pikettys Numbers Dont Add Up. Wall Street Journal, May 14. Available at
<http://online.wsj.com/articles/SB10001424052702304081804579557664176917086>.
Furman, Jason and Joseph E. Stiglitz (1998) Economic Consequences of Income Inequality, with Jason Furman, in
Symposium Proceedings Income Inequality: Issues and Policy Options, Jackson Hole, Wyoming: Federal
Reserve Bank of Kansas City, pp. 221-263.
Galbraith, James Kenneth (2012), Inequality and Instability: A Study of the World Economy Just Before the Great
Crisis, Oxford University Press
George, Henry (1879) Progress and Poverty: An Inquiry into the Cause of Industrial Depressions and of Increase of
Want with Increase of Wealth VI. New York: Robert Schalkenbach Foundation.
Griliches, Zvi (1969). "Capital-Skill Complementarity." The Review of Economics and Statistics, 51(4): 465-68.
Greenwald, Bruce and Judd Kahn, 2009, Globalization: n. the irrational fear that someone in China will take your
job, Hoboken, N.J. : John Wiley & Sons.
Greenwald, Bruce, and J. E. Stiglitz, Towards a New Paradigm in Monetary Economics, Cambridge: Cambridge
University Press, 2003.
Hahn F. (1966). Equilibrium Dynamics with Heterogeneous Capital Goods, Quarterly Journal of Economics, 80:
633-646.
Johnson, Simon and James Kwak (2010). 13 Bankers: The Wall Street Takeover and the Next Financial Meltdown,
Pantheon.
Jones, C., and Paul Romer, The New Kaldor Facts: Ideas, Institutions, Population, and Human Capital, NBER
Working Paper 15094, June 2009.
Kaldor, Nicholas (1957). A Model of Economic Growth. The Economic Journal, 67(268): 591-624.
(1961) Capital Accumulation and Economic Growth, in F. A. Lutz and D. C. Hague, eds. The Theory of
Capital, St. Martins Press, 1961, pp. 177-222.
Katz, Michael L. and Carl Shapiro (1994). Systems Competition and Network Effects. Journal of Economic
Perspectives, 8(2): 93-115.
Kindleberger, Charles P. and Robert Aliber. (1978). Manias, Panics, and Crashes: A History of Financial Crisis, New
York: John Wiley & Sons.
Krusell, Per, Lee E. Ohanian, Jos-Victor Ros-Rull, and Giovanni L. Violante (2000). "Capital-Skill Complementarity
and Inequality: A Macroeconomic Analysis." Econometrica, 68(5): 1029-1054.
Kuznets, Simon (1955). Economic Growth and Income Inequality. The American Economic Review, 45(1): 1-28.
Lebergott, Stanley (1959). The Shape of the Income Distribution. The American Economic Review, 49(3) 328-347.
Mallick, Debdulal (2007). The Role of the Elasticity of Substitution in Economic Growth: A Cross-Country Test of the
de La Grandville Hypothesis. Economic series, Deakin University, Faculty of Business and Law, School of
Accounting, Economics and Finance.
Mani, Anandi, Sendhil Mullainathan, Eldar Shafir, and Jiaying Zhao (2013) Poverty Impedes Cognitive Function,
Science 341(6149): 976-980.
Mullainathan, Sendhil, and Eldar Shafir (2009). Savings Policy and Decisionmaking in Low-Income Households,
Insufficient Funds: Savings, Assets, Credit and Banking Among Low-Income Households, 121-145.
OECD (2011). Divided We Stand: Why Inequality Keeps Rising, OECD.
Perlin, Ross (2011). Intern Nation, London: Verso, 2011.

55

Pasinetti, L. (1962). The Rate of Profit and Income Distribution in Relation to the Rate of Economic Growth,
Review of Economic Studies, 29(4): 267-279.
Piketty, Thomas (2014). Capital in the Twenty-First Century, Cambridge Massachusetts : The Belknap Press of
Harvard University Press.
Piketty, Thomas and Gabriel Zucman (2013) "Capital is Back: Wealth-Income Ratios in Rich Countries 1700-2010,"
working paper July 26, 2013.
(2014a) "Capital is Back: Wealth-Income Ratios in Rich Countries 1700-2010," Quarterly Journal of Economics.
(2014b). Wealth and Inheritance in the Long Run. Working Paper.
Piketty, Thomas and Emmanuel Saez (2003). Income Inequality in the United States, 1913-1998, Quarterly
Journal of Economics, 118(1): 1-39 (Longer updated version published in A.B. Atkinson and T. Piketty eds.,
Oxford University Press, 2007) (Tables and Figures Updated to 2012 in Excel format, September 2013).
Piketty, Thomas, Emmanuel Saez, and Stefanie Stantcheva (2014). Optimal Taxation of Top Labor Incomes: A Tale
of Three Elasticities. American Economic Journal: Economic Policy, 6(1): 230-271.
Pogge, Thomas and Sanjay Reddy (2010), in Debates in the Measurement of Poverty, ed. Joseph Stiglitz, Sudhir
Anand and Paul Segal Oxford University Press.
Roberts, Sam (2014)Why are Rockefellers Moving from 30 Rock? We Got a Great Deal, New York Times,
November 24, p. A1.
Saez, Emmanuel and Gabriel Zucman (2013); Capital is back: Wealth-Income Ratios in Rich Countries 1700-2010;
Quarterly Journal of Economics.
(2014). Wealth Inequality in the Unites States Since 1913: Evidence from Capitalized Income Tax Data, NBER
Working Paper 20625h
Shell, Karl, and Joseph E. Stiglitz (1967). Allocation of Investment in a Dynamic Economy, Quarterly Journal of
Economics, 81: 592-609.
, M. Sidrauski, and J.E. Stiglitz (1969). Capital Gains, Income and Savings, with K. Shell and M. Sidrauski,
Review of Economic Studies, 36(1): 15-26.
Shierholz, Heidi and Lawrence Mishel (2013). A Decade of Flat Wages, Economic Policy Insitute Report, available
at <http://www.epi.org/publication/a-decade-of-flat-wages-the-key-barrier-to-shared-prosperity-and-arising-middle-class/>.
Shiller, Robert (2002). Irrational Exuberance, Princeton, NJ: Princeton University Press.
Spence, Michael (1984). Cost Reduction, Competition, and Industry Performance. Econometrica, 52(1): 101-122.
Sraffa, Piero (1960). Production of Commodities by Means of Commodities: Prelude to a Critique of Economic
Theory, Cambridge University Press
Stiglitz, Joseph E. (1969a). Distribution of Income and Wealth Among Individuals, Econometrica, 37(3): 382-397.
(Presented at the December 1966 meetings of the Econometric Society, San Francisco.)
(1969b), A Re-Examination of the Modigliani-Miller Theorem, American Economic Review, 59(5), December
1969, pp. 784-793.
(1973). Education and Inequality, Annals of the American Academy of Political and Social Sciences, 409: 135145.
(1975): The Theory of Screening, Education and the Distribution of Income, American Economic Review,
65(3): 283-300. Reprinted in Selected Works of Joseph E. Stiglitz, Volume I: Information and Economic
Analysis, Oxford: Oxford University Press, 2009, pp. 99-121.

56

 (1976a). Estate Taxes, Growth and Redistribution, in Essays in Honor of W. Vickrey, R. Grieson (ed.),
Lexington: Lexington Publishing Company, pp. 225-232.
(1976b). Notes on Estate Taxes, Redistribution, and the Concept of Balanced Growth Path Incidence,
Journal of Political Economy, 86(2), part 2, pp. 137-150
(1978). Equality, Taxation and Inheritance, in Personal Income Distribution, W. Krelle and A.F. Shorrocks
(eds.), North-Holland Publishing Company, pp. 271-303. (Proceedings of IEA Conference, Noordwijk aan
Zee, Netherlands, April 1977.)
(1982). Information and Capital Markets, in Financial Economics: Essays in Honor of Paul Cootner, William F.
Sharpe and Cathryn Cootner (eds.), Prentice Hall, New Jersey, pp. 118-158. Also NBER Working Paper 678.
Reprinted in The Selected Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis:
Applications to Capital, Labor, and Product Markets, Oxford: Oxford University Press, 2013, pp. 55-84.
(1985). Equilibrium Wage Distributions, Economic Journal, 95(379), pp. 595-618. Reprinted in The Selected
Works of Joseph E. Stiglitz, Volume II: Information and Economic Analysis: Applications to Capital, Labor,
and Product Markets, Oxford: Oxford University Press, 2013, pp. 559-580.
(1993). Remarks on Inequality, Agency Costs, and Economic Efficiency, prepared for a workshop in
Economic Theories of Inequality, Stanford Institute for Theoretical Economics, Stanford University,
March 11-13.
(1998). Inequality and Growth: Implications for Public Finance and Lessons from Experience in the US, in
Beyond Tradeoffs: Market Reform and Equitable Growth in Latin America, N. Birdsall, C. Graham, R.H.
Sabot (ed.), Inter-American Development Bank, 1998, pp. 305-319.
(2010a). Freefall: America, Free Markets, and the Sinking of the World Economy, New York: WW Norton.
(2010b). The Wobbly Economy, mimeo, Columbia University.
(2012a). The Price of Inequality: How Todays Divided Society Endangers Our Future, New York: W.W. Norton.
Also published in the U.K. by Penguin/ Allen Lane. Translated in simplified Chinese by China.
(2012b). Macroeconomic Fluctuations, Inequality, and Human Development, Journal of Human
Development and Capabilities, 13(1): 31-58. Reprinted in Macroeconomics and Human Development,
Deepak Nayyar (ed.), Taylor and Francis.
(2014), Unemployment and Innovation, NBER Working Paper 20670.
and Andrew Charlton (2005). Fair Trade for All, New York: Oxford University Press.
Turner, Adair (2014). Wealth, Debt, Inequality And Low Interest Rates: Four Big Trends And Some Implications,
Institute for New Economic Thinking.
Young, Andrew T. (2013). US Elasticity of Substitution and Factor-Augmentation at the Industry Level.
Macroeconomic Dynamics, 17(4):861-897.
Wolff, Edward N. and Maury Gittleman, Inheritances and the Distribution of Wealth Or Whatever Happened to
the Great Inheritance Boom? BLS Working Paper 445, January 2011

57

Figure 1: Capitaloutput ratio including and excluding land, France

200.0

180.0

GDP

160.0
140.0
120.0
100.0
80.0

K+T/ GDP

60.0
40.0
20.0
0.0
19851987198919911993199519971999200120032005200720092011

Figure 2

**

Figure 2A

s(w, r)w

45

Figure 3

Ft

Fk

H ( )

Figure 4A

H ( )

Figure 4B

'

**