kuznets paper curve ushape

8/16/2019 Kuznets Paper Curve Ushape

1/31

Economic Growth and Income Inequality

Simon Kuznets

The American Economic Review, Vol. 45, No. 1. (Mar., 1955), pp. 1-28.

Stable URL:

http://links.jstor.org/sici?sici=0002-8282%28195503%2945%3A1%3C1%3AEGAII%3E2.0.CO%3B2-Y

The American Economic Review is currently published by American Economic Association.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/aea.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. Formore information regarding JSTOR, please contact [email protected].

http://www.jstor.orgWed May 16 07:01:51 2007

http://links.jstor.org/sici?sici=0002-8282%28195503%2945%3A1%3C1%3AEGAII%3E2.0.CO%3B2-Yhttp://www.jstor.org/about/terms.htmlhttp://www.jstor.org/journals/aea.htmlhttp://www.jstor.org/journals/aea.htmlhttp://www.jstor.org/about/terms.htmlhttp://links.jstor.org/sici?sici=0002-8282%28195503%2945%3A1%3C1%3AEGAII%3E2.0.CO%3B2-Y


2/31

The

American

Economic Review

VOLUME XLV

MARCH, 955

NUMBER ONE

E C O N O M I C G R O W T H

AND

I N C O M E I N EQ U A L I TY *

y SIMO NKU ZN ET S

T he cen tral theme of this pap er is the chara cte r an d causes of long-

term changes in the personal distribution of income. Does inequality

in the d istribution of income increase or decrease in th e course of a

country s economic gro w th? W hat factors de termin e the secular level

and tre nd s of income inequa lities?

These are broad questions in a field of stu dy th at h as been plagued

by looseness in definitions, unusu al scarcity of data , and p ressures of

strongly held opinions. While we ca nnot com pletely avoid th e resulting

difficulties, it m ay help to specify the ch arac teristic s of the size-of-

income distributions that we want to examine and the movements of

which we want to explain.

Five specifications may be listed. First, the units for which incomes

are recorded and grouped should be family-expenditure units, properly

adju sted for th e num ber of persons in each-rather tha n incom e re-

cipients for whom the relations between receipt and use of income can

be widely diverse. Second, the distribution should be complete, i.e.

should cover all units in a country rather than a segment either at the

upper or lower tail. Third, if possible we should segregate the units

whose main income earn ers are either still in the lea rning or alrea dy in

the retired stages of the ir life cycle-to avoid com plicating the pic tu re

by including incomes not associated w ith full-time, full-fledged parti cipa -

tion in economic activity. F ou rth , income should be defined a s it is now

for national income in this country, i.e. received by individuals, in-

cluding income in kind, be fore an d af te r direct taxes, excluding cap ital

gains. Fifth, the units should be grouped by secular levels of income,

free of cyclical and other transient disturbances.

Fo r such a distribution of m atur e expen diture units by s ecular levels

Presidential address delivered a t the Sixty-seventh An nual Meeting of the American

Economic Association, Detroit, Michigan, December 29,

1954.


3/31


4/31


5/31

THE

AMERICAN ECONOMIC REVIEW

of income per cap ita, we should m easure sha res of some fixed ordinal

groups-percentiles, deciles, quintiles , etc. I n the und erlying ar ra y th e

units sho uld be classified by averag e income levels for a sufficiently long

span so tha t they form income-status groups-say a generation or

about 25 years. Within such a period, even when classified by secular

income levels, units m ay shift from one ordinal group to another. I t

would, therefore, be necessary and useful to study separately the rela-

tive sha re of u nits th at , throug hou t th e gene ration period of refere nce,

were continuo usly within a specific ordinal g roup, and th e sha re of the

units that moved into that specific group; and this should be done for

the shares of residents and n~ igra nts within all ordinal groups.

Without such a long period of reference and the resulting separation

between resident an d migrant units a t different relative income

levels, the very distinctio n between low an d high income classes

loses its meaning, parti cul arly in a stu dy of long -term changes in shares

and in inequalities in the distribution. To say, for example, that the

lower income classes gained or lost during the last twenty ye ars in

th at the ir share of t ota l income increased or decreased has meaning only

if the un its hav e been classified as members of the lower classes

throughout those

2

years-and for those who have moved int o or ou t

of those classes recently such a s tate m ent ha s no significance.

Furth erm ore, if one ma y ad d a final touch to wh at is beginning to

look like a sta tistical economist's pipe drea m, we should be able to trac e

secular income levels not only through a single generation but at least

throug h two-connecting th e incomes of a given genera tion with those

of its immediate descendants. We could then distinguish units that.

throughout a given generation, remain within one ordinal group and

whose children-through their generation-are also within th at group ,

from units that remain within a group through their generation but

whose children move up or down on the relative economic scale in their

time. Th e numb er of possible combinations an d perm utation s becomes

large; bu t it should not obscure th e main design of the income structu re

called for-the classification by long-term incom e st at us of a given

generation an d of it s im me diate descen dants. If living members of

society-as produ cers, consu me rs, savers, decision-makers on secular

problems-react to long-term changes in income levels an d sha res, da ta

on such a n income struc ture a re essential. An economic society can then

be judged by t he secular level of the income shar e tha t it provides for

a given generation and for its children. T h e im portan t corollary is that

the stud y of long-term ch anges in th e income distribution m ust d istin-

guish between changes in the sha res of resid ent groups-resident within

either one or two generations-and changes in th e incom e share s of


6/31

UZNETS: ECONOMIC GROWTH ND INCOME

INEQUALITY

groups that, judged by their secular levels, migrate upward or down-

ward on the income scale.

Eve n if we had d ata to approxim ate the income structu re just o ut-

lined, the broad question posed a t the start-how income ineq uality

changes in th e process of a cou ntry s economic growth-could be

answered only for grow th unde r defined economic and social conditions.

And, in fact , we shall deal with this qu estion in terms of the ex perience

of the now developed countries which grew under the aegis of the busi-

ness enterprise. But even with this limitation, there are no statistics

that can be used directly for the purpose of measuring the

secular

income stru cture . Ind eed , have difficulty in visualizing how such

info rm atio n could practicably be collected-a difhculty th at m ay be

du e to lack of familiarity with th e studie s of our colleagues in de-

mography and sociology who have concerned themselves with prob-

lems of generation or intergeneration mobility and status. But although

we now lack data directly relevant to the secular income structure,

the se tting up of reasona bly clear an d ye t difficult specifications is

not merely an exercise in perfectionism. For if these specifications do

approximate, an d trus t that they do, the real core of our interest when

we talk ab ou t shares of economic classes or long-term changes in these

shares, then prop er disclosure of our meaning a nd intentions is vitally

useful. I t forces us to examine and evaluate cri tically the data tha t are

available; it prevents us from jumping to conclusions based on these

inadequate data; it reduces the loss and waste of time involved in

mechanical manipulations of the type represented by Pareto-curve-

fitting to groups of da ta whose meaning , in term s of income concep t,

un it of observ ation, an d proportion of th e tota l universe cove red, re-

mains distressingly vague; and most important of all, it propels us

toward a deliberate construction of testable bridges between the avail-

able data a nd the income struckure that is the real focus of our interest.

I

Trends n Income Inequal i ty

Forewarned of the difficulties, we turn now to the available data.

These data, even when relating to complete populations, invariably

classify units by income for a given year. From our standpoint, this is

their major limitation. Because the data often do not permit many

size-groupings, and because the dilference between annual income

incidence and longer-term incom e stat us has less effect if the n um ber of

classes is small a nd th e limits of each class ar e wide, we use a few wide

classes. T hi s does not resolve the difficulty; an d the re a re others d ue to

the scantines s of da ta for long periods, inade quacy of th e unit used-

which is, a t best, a fam ily and very often

a

reporting unit-errors in th e


7/31

4 THE AMERICAN ECONOMIC REVIEW

data, and so on through a long list. Consequently, the trends in the

income structur e can be discerned b ut dimly, and th e results considered

as preliminary informed guesses.

T he da ta are for the United States, England, and Germany-a scan t

sample, but at least a starting point for some inferences concerning

long-term changes in the presently developed countries. The general

conclusion suggested is tha t the relative distribution of income, as

measured by annual income incidence in rather broad classes, has been

moving tow ard equality-with these tren ds particularly noticeable

since the 1920's but beginning perhaps in the period before the first

world war.

Let me cite some figures, all for income before direct taxes, in sup-

port of this impression. I n the Un ited States, in the distribu tion of in-

come among families (excluding single individu als), th e shares of th e

two lowest quintiles rise from 13 per cent in 1929 to 18 per cent in the

years af ter the second world war (av era ge of 1944, 1946, 1947, an d

1950) whereas th e sha re of the to p quintile declines from 55 to 44 per

cent, and tha t of the top 5 per cent f rom 31 to 20 per cent. I n the

Un ited Kingdom , the sha re of the to p 5 per c ent of u nits declines from

46 per cen t in 1880 to 43 per cent in 1910 or 1913, to 33 per cent in

1929, to 31 per cent in 1938, and to 24 per cent in 1947; t he sh are of

the lower 85 per cent remains fairly constant between 1880 and 1913,

between 41 an d 43 per cent, bu t then rises to 46 per cent in 1929 and

55 per cent in 1947. In Prussia income inequality increases slightly

between 18 75 and 1913-the shares of th e to p quintile rising from 48

to 50 per cent, of the top 5 per cent from 26 to 30 per cen t; the share

of the lower 60 per cent, however, remains abou t the same. I n Saxony,

the change between 1880 an d 1913 is minor: the sh are of the two

lowest quintiles declines from 15 to 14 per cen t; th at of the third

quintile rises from 12 to 13 per cen t, o f- he fo urth quintile from 16

to abou t 18 per cen t; tha t of the top quintile declines from 56 to

54 per cent, an d of th e top 5 per cent fro m 34 to 33 per cent. I n

Germany as a whole, relative income inequality drops fairly sharply

from 1913 to the 192 03, app arently du e to decimation of large for-

tunes and property incomes during the war and inflation; but then

begins to re turn to prew ar levels dur ing th e depression of the 1930's.'

'The following sources were used in calculating the figures cited:

United States. For recent years we used Income Dis tr ibution b y Size 1944 1950 (Wash-

ington, 1953) and Selma Goldsmith and others, Size Distribution of Income Since the

Mid-Thirties,

ev Econ. Stat .

Feb. 1954, XXXVI, 1-32; for 1929, the Brookings Institu-

tion data as adjusted in Simon Kuznets,

Shares of Upper Groups in Income and Savings

(New York, 1953),

p.

220.

United Kingdom.

For 1938 and 1947 Dudley Seen, The

Levelling of Income Since

1038


8/31

UZNETS: ECONOMIC GROWTH AND INCO ME INEQUALITY

Even for what they a re assumed to represent, let alone as approxima-

tions to shares in distributions by secular income levels, the data are

such that differences of two or three percentage points cannot be as-

signed significance. One must judge by the general weight and consen-

sus of the evidence-which un for tun ate ly is limited to a few cou ntries.

I t justifies a ten tative impression of constancy in th e relative distribu-

tion of income before taxes , followed by so me narrow ing of relative

income inequ ality af ter the first world war-or earlier.

Three aspects of this finding should be stressed. First, the data are

for income before direct taxes and exclude contributions by govern-

ment e.g., relief and free assistance). I t is fair to argue that both th e

proportion an d progressivity of direct taxes and th e proportion of total

income of individuals accounted for by government assistance to the

less privileged economic groups have grown during recent decades. This

is certainly true of the United States and the United Kingdom, but in

the case of Germ any is subject to furth er examination. I t follows th at

the distribution of income after direct taxes an d including free contribu-

tions by government would show an even greater narrowing of in-

equality in developed countries with size distributions of pretax, ex-

government-benefits income similar to those for the United States and

the United Kingdom.

Second, such stab ility or reduction in the inequality of the percen tage

shares was accompanied by significant rises in real income per capita.

T h e countries now classified as developed h ave enjoyed rising per

capita incomes except during catastrophic periods such as years of

active world conflict. Hence, if the shares of groups classified by their

annual income position can be viewed as approximations to shares of

groups classified by their secular income levels, a constant percentage

share of a given group means that its per capita real income is rising

at the same rate as the average for all units in the country; and a re-

duction in inequality of the shares means that the per capita income

of the lower-income groups is rising a t a more rapid ra te tha n the per

capita income of the upper-income groups.

T he third point can be put in the form of a question. Do th e distribu-

(Oxford, 1951) p. 39; for 1929, Colin Clark, National Income and Outlay (Lond on, 1937)

Table 47, p 109; for 1880, 1910, and 1913,

A.

Bowley, The Change in the Distribution of

the National Income

1880 1913

(Oxford, 1920).

Germany. For the const i tuent areas (Prussia, Saxony and others) for years before the

first world war, based on S. Prokopovich,

iVational Income of We ste rn Europea n Co untries

(published in M oscow in the 1920's). Some summa ry results are given in Prokopo vich,

The Distribution of National Income,

Econ. Jour.

March 1926, XXXVI 69-82. See also,

Das Deutsche Volkseinkommen vor und nach dem Kriege, Einzelschrift zur Stat. des

Deutschen Reichs no. 24 (B erlin, 193 2), an d

W.

S and

E

S. Woytinsky, WorId Popula-

tion and Production (New York, 1953) Table 192, p.

709


9/31

6 T H E AMERICAN ECONOMIC REVIEW

tions by annual incomes properly reflect trends in distribution by

secular incomes? As technology and economic performance rise to

higher levels incomes are less subject to transien t disturbances not

necessarily of the cyclical order th at ca n be recognized a nd allowed

for by reference to business cycle chronology bu t of a m ore irregular

type. f in the earlier years the economic fortunes of units were sub-

ject to greater vicissitudes-poor crops for some farm ers na tur al

calam ity losses for some non farm business units-if the over-all pro por -

tion of individual entr epr ene urs whose incomes were subject to such

calamities more yesterday but some even today was larger in earlier

decades these earlier distributions of income would be more affected

by transient disturbances. I n these earlier distributions th e temporarily

unfortunate might crowd the lower quintiles and depress their shares

unduly and the temporarily fortunate might dominate the top quintile

and raise its share unduly-proportionately more than in th e distribu-

tions for later years. f so distributions by longer-term average in-

comes might show less reduction in inequality th an do the distribution s

by annual incomes; they might even show an opposite trend.

One may doubt whether this qualification would upset a narrowing

of inequality as marked as tha t for the United States and in as short

a period as twenty-five years. Nor is it likely to affect the persistent

downw ard drif t in th e spread of the distributions in the Un ited King-

dom. B ut m ust adm it a stro ng element of judgment in deciding how

f a r this qualification modifies the finding of long-term stab ility followed

by reduction in income inequality in the few developed countries for

which it is observed or is likely to be revealed by existing data. The

imp ortant point is tha t the qualification is relevan t; it suggests need for

furt her study if we are to learn much from the available dat a con-

cerning the secular income structure; and such study is likely to yield

results of interest in themselves in their bearing upon the problem

of trend s in temporal in stability of income flows to individual unit s

or to econom ically significant groups of units in different sectors of

the national economy.

11 n ttempt at Explanation

f

the above summ ary of trends in the secular income struc ture of

developed c o ~ ~ n tr i e somes perilously close to pure guesswork a n

attempt to explain these dimly discernible trends may surely seem

foolhardy. Yet it is necessary to do so if only to bring to th e sur fac e

some factors that m ay have been a t play; induce a search fo r data

bearing upon these factors; and thu s confirm or revise our impressions

of the trends themselves. Such preliminary speculations a re useful


10/31

UZNETS: EC ONOMIC GR OW TH AND INC OM E INEQUALITY

provided it is recognized that we are at a relatively early stage in a

long process of interpla y am ong tenta tive summ aries of evidence,

preliminary hypotheses, and search for additional evidence that might

lead to reform ulation an d revisions-as bases for new analy sis an d

further search.

The present instalment of initial speculation may be introduced by

saying th at a long-term constancy , let alone reduction, of inequality in

the secular income structure is

puzzle. F or there ar e a t least two

groups of forces in the long-term opera tion of developed countries th at

make for

incre sing

inequality in the distribution of income before

taxes and excluding contributions by governments. The first group

relates to the concentration of savings in the upper-income brackets.

According to all recen t studies of t he app ortionm ent of income between

consumption and savings, only the upper-income groups save;

the

total savings of groups below the top decile ar e fair ly close to zero. F or

example, the top

5

per cent of units in the United States appear to

account for alm ost two-thirds of individuals7 savings; an d the top

decile comes close to acco unting for all of it. Wh at is pa rticu larly im-

porta nt is tha t the inequality in distribution of savings is grea ter tha n

th at in the distribution of pro pe rty incom es, an d hence of assets.

Granted t ha t this finding is based on distribution of ann ual income,

and that a distribution by secular levels would show less inequality in

income and correspondingly less conc entration of savings, the in-

equality in savings would still remain fair ly sha rp, perhaps more so

than in holdings of assets. Other conditions being equal, the cumulative

effect of such ine qua lity in savings would be the con centratio n of a n

incre sing

prop ortion of incom e-yielding assets in the han ds of the

upp er groups-a basis fo r larger income share s of these gro cp s an d

their descendants.

Th e second source of the puzzle lies in th e indu strial struc tur e of

the income distribution. An invariable accompaniment of growth in

developed co untries is the sh ift aw ay f ro m agriculture, a process usu-

ally referred to s industrialization and urbanization. The income dis-

tribution of the total po pulation, in th e simplest model, m ay th erefo re

be viewed as a combination of the income distributions of the rural

an d of th e urba n populations. Wh at little we know of the structures

of these two comp onent income distributions reveals tha t: ( a ) th e

average per cap ita income of the rur al population is usually lower tha n

that of the urban ; ( b ) inequality in the percentage shares within the

See Kuznets,

op cit.

particularly Chapters

2

and

6.

Th e lower level of per capita incom e of the agricultural or rural population compared

with th at of ur ban is fairly ulell established, fo r this country by states, and for m any


11/31

T H E MERIC N ECONOMIC REVIEW

distribution for the rural population is somewhat narrower than in

that for the urban po pu lat ion -ev en when based on annual income; and

this difference would probably be wider for distributions by secular

income

le v e k 4 Operating with th is simple m odel, wh at conclusions do

we reach? First, all other conditions being equal, the increasing weight

of urba n population means a n increasing share for the more unequ al of

the two component distributions. Second, the relative difference in per

capita income between the rural and urban populations does not

necessarily drif t dow nward in the process of economic growth: indeed,

there is some evidence to suggest that it is stable at best, and tends to

widen because per cap ita productivity in urb an pu rsuits increases more

rapidly than in agriculture. f this is so, inequality in the total in-

come distribution should increase.

Two questions then arise: First, why does the share of the top-

income groups show no rise over time if the concentration of savings

has a cumulative eff ec t? Second, why does income inequ ality decline

an d particularly why does the sh are of the lower-income groups rise if

both th e weight of the more unequal urb an income distribution a nd the

relative difference between per capita urban and per capita rural in-

comes increase?

The first question has been discussed elsewhere, although the re-

sults are still preliminary hypotheses,' an d it w ould be impossible to

do more here th an summarize the m briefly.

Factors C ounteracting th e Concen tration of

aving

One group of factors co unte racting the cum ulative effect of con-

other countries (see,

e.g.

a summary table of closely related measures of product and

workers engaged, for var ious divisions of the productive system , in Colin Clark,

Conditions

of Economic Progress 2nd ed. [Lon don 19511, pp. 316-18). T he same table suggests, for

the countries with sufficiently long records, a stable or increasing relative difference between

per-worker product in agriculture and per-worker product in other sectors of the economy.

'This i s

true of the U.

S

distributions prior to the second world war (see sources cited

in footnote 1 ) in the years after the second world w ar th e difference seems to have disap-

peared. It is true of the distributions for Prussia, ci ted b y Prokopo vich; and most conspicu-

ous for India today as shown in the rough distributions by

M .

Mukherjee and A. K. Ghosh

in The P attern of Income and Expenditures in the India n Union:

A

Tentative Study,

International Statistical Conferences December 1951, Calcutta, India, Part 111, pp. 49-68.

ome elements of the discussion appeared in Prop ortion of Cap ital Fo rm atio n to Na -

tional Product, a pape r submitted to th e ann ua l meeting of the American Econom ic Associ-

ation in 1951 and published in Am Econ. Rev. Proceedings, M ay 1952, X L II , 507-26. A

more elaborate statement is presented in International Differences in Capital Form ation

and Financing (particu larly Appendix C, Levels an d Tr end s in Inco me Shar es of Upper

Income Groups), a paper submitted to a Conference

on

Capital Formation and Economic

Growth held

in

1953 under the auspices

of

the Universities-National Bureau Committec

for Economic Research.

t

is now in

press

as pa rt of the volume of proceedings of th at

conference.


12/31

UZNETS: ECONOMIC GROWTH AND INCOME INEQUALITY

centration of savings upon upper-income shares is legislative interfer-

ence an d political decisions. The se m ay be aimed at limiting th e

cumulation of property directly through inheritance taxes and other

explicit capital levies. They may produce similar effects indirectly,

e.g.

by government-permitted or -induced inflation which reduces the

economic value of accumulated wealth stored in fixed-price securities

or other properties not fully responsive to price changes; or by legal

restriction of the

yie d

on accumu lated property, a s happened recently

in th e form of re nt co ntrols or of artificially low long-term in terest rat es

maintained by the government to protect the m arke t for its own bonds.

To discuss this complex of processes is beyond the competence of

this paper, but its existence and possible wide effect should be noted

and one point emphasized. All these interventions, even when not

directly aimed a t limiting the effects of accum ulation of pa st savings

in the ha nd s of the few, do reflect the view of society on th e long-term

utility of wide income inequalities. Th is view is a vita l force th at w ould

ope rate in democratic societies even if the re were no o ther cou nterac t-

ing factors. This should be borne in mind in connection with

changes

in

this view even in developed countries, which result from the process of

growth and constitute

a

re-evaluation of the need for income in-

equalities as a source of savings for economic growth. T h e result o f

such changes would be a n increas ing press ure of legal an d political

decisions on upper-income shares-increasing as a co un try mov es to

higher economic levels.

We tur n to three othe r, less obvious grou ps of factors cou nterva iling

the cum ulative effects of con cen tration of savings. T h e first is demo-

graphic. I n the presen tly developed countries there have been dif-

ferential rate s of increase between the rich an d the poor-family con-

trol having first spread to the former. Hence, even disregarding mi-

gration, one can argue th at the top

5

per cen t of 1870 an d its descend-

ants would account for

a

significantly smaller percen tage of th e popu-

lation in 1920. This is even more likely in a country like the United

States with its su bsta ntia l immigration-usually enterin g the income

distribution at the lower-income levels; and may be less likely in a

country fro m which the poor have emigrated. T h e top 5 per cent of

population in 1920 is, therefo re, comprised only par tly of the de-

scendants of the top

5 per cent of 1870; perhaps half or a larger

fraction must have originated in the lower-income brac ke ts of 1870.

T hi s means th at th e period durin g which effects of con centra tion of

savings can be assumed to have cumulated to raise the income share

of an y given fixed ordinal group (whether i t be the top 1 5, or 10 per

cent of the population) is much shorter than the fifty years in the

span ; and hence these effects are much w eaker than they would have


13/31

1 T H E AMERIC AN EC ONOMIC R EVIEW

been if the top 5 per cent of

187

had, through their descendants, filled

completely the ra nk s of the top

5

per c ent of th e populatio n of 1920

Although t he c um ulative effect of savings m ay be to raise th e relative

income of a progressively diminishing top proportion of total popula-

tion, the ir effect on th e relative sha re of a

fixe

top prop ortion of th e

population is much reduced.

T h e second group of forces resides in the very na tu re of a dy nam ic

economy with relative freedom of individual opportunity. I n such a

society technological change is rampant and property assets that

originated in older industries almost inevitably have a diminishing

proportional w eight in the tota l because of th e more rapid growth of

younge r indu stries. Unless the descen dan ts of a high-income gro up

manage to shift their accumulating assets into new fields and partici-

pate with new entreprene urs in th e growing sh are of the new and more

profitable industries, the long-range returns on their property holdings

ar e likely to be significantly lower tha n those of the mo re recent

en tran ts into th e class of s ubs tantia l asset holders. Fro m shirt-sleeves

to shirt-sleeves in three generations prob ably exaggerates the effects

of this dynamism of a growing economy: there a re, among th e upper-

income groups of today, m any descendan ts of the upper-income group s

of more than three or even four generations ago. But the adage is

realistic in the sense tha t a long unbroke n sequence of connection with

rising ind ustries and hence with m ajor sources of continued larg e

property incomes is exceedingly rare; that the successful great entre-

preneurs of tod ay ar e rarely sons of th e grea t and successful entre pre-

neurs of yesterday.

The third group of factors is suggested by the importance, even in

the upper-income brac kets, of service income. At an y given time, only

a

limited pa rt of th e income differe ntial of a top group is accounted for

by the concentration of property yields: much of it comes from the

high level of service income (professional and entre pren euria l earnings

and the like) . T h e secular r ise in the upper incomes due to this source is

likely to be less m arke d th an in th e service incomes of lower b rack ets,

an d for two som ewh at differen t reasons. Fi rst, in so far a s high levels

of service incomes of given upp er u nits a re du e to individ ual excellence

(as is t rue of m any profess ional and entrepreneur ia l pursuits) , there

is much less incentive for an d possibility of keeping such incomes a t

con tinue d high relat ive levels. H en ce, th e service incom es of the de-

scendants of an

initially high

level unit are not likely to show as strong

an upward trend as the incomes for the large body of population

a t

lower-income levels. Second, a su bst an tia l p ar t of t he rising trend in

per capita income is due to interindustry shift, i.e.

a shift of workers

from lower-income to higher-income industries. T h e possibilities of rise


14/31

KUZNE'I'S:

E C O N O M I C G R O W T H

A N D

I N C O M E I N EQ U A LI TY

due to such interin dus try sh ifts in the service incomes of the initially

high-income groups a re much m ore limited than fo r the population a s a

whole: they are already in high-income occupations and industries and

the range for them toward higher paid occupations is more narrowly

circumscribed.

Th es e three g roups of facto rs, even disregarding such legislative and

political intervention as is indicated ab ove, are all characteristics of a

dynamic growing economy. T h e differentials in rat e of na tur al increase

between th e upper- an d the lower-income groups ar e true only of a

rapid ly growing population-with or with out immigration-but ac-

companied by declining death rates and declining birth rates, a demo-

graphic pattern associated in the past only with the growing Western

economies. T h e imp act of new indu stries o n obsolescence of a lre ad y

established wealth a s a source of p rope rty income is clearly a function

of ra pid grow th, an d the more rapid th e growth the greate r the impact

will be. T h e effect of i nte rind us try s hifts on the rise of p er c apita

income , pa rtic ula rly of lower-incom e grou ps, is also a fu nction of

growth since only in a growing economy is there much shift in the

relative importanc e of the several industr ial sectors. One can then s ay ,

in general, that the basic factor militating against the rise in upper-

income shares that would be produced by the cumulative effects of

conc entra tion of savin gs, is the dyn am ism of a growing an d free

economic society.

Yet while the discussion answers the original question, it yields

no determinate answer as to whether the trend in income shares of

upper groups is upward, downward, or constant. Even for the specific

question discussed. a determinate answer depends upon the relative

ba lance of factors-continuous con centra tion of savings mak ing for

an increasing share, and the offsetting forces tending to cancel this

effect. T o tell wh at the tr end of upper-income shares is likely to be , me

need to know much m ore ab ou t the weights of these conflicting pres-

sures. Moreover. the discussion has brought to th e surface fa ctors tha t,

in an d of themselves, m ay cause either an upward o r a downward trend

in the sha re of u pper-income group s and hence in income inequality-

in distributions of ann ual or o f secular income. Fo r example, the new

en tra nts into the uppe r groups-the upw ard migrants -who rise

either because of esceptional ability or attach me nt to new industries or

for a varie ty of other reasons-may be enter ing th e fixed uppe r grou p

of say the top per cen t with an income differential-either ann ual or

long-term-that ma y be relatively grea ter than tha t of en tra nts in the

preceding generation. Nothing in the argument so far excludes this

possibility-which would mean a rise in the sh are of upper-income

groups, even if the share of the old resident pa rt remains con stan t or


15/31

2

THE


even declines. Even disregarding other factors th at w ill be noted in the

next section, no firm conclusion as to tre nds of upper-income sha res

can be derived from the bare model discussed. Search for further data

might yield evidence that would permit a reasonably rough but deter-

minate conclusion; but I have no such evidence a t hand.

T h e Sh ift from gricultural to Nonagricultural Sectors

What about the trend toward greater inequality due to the shift

from the agricu ltural to the nonag ricultural sectors? In view of the im-

por tance of industrialization a nd urbanization in the process of eco-

nomic growth, their implications for trends in the income distribution

should be explored-even though we have neither the necessary data

nor a reasonably complete theoretical model.

T h e implications can be brough t out m ost clearly with th e help of a

numerical illustration (see Table I). In this illustration we deal with

two sectors: agriculture (A) and all others (B). For each sector we

assum e percentage d istribu tions of tota l sector income among sector

deciles: one distribution E) is of mo dera te inequality, with the share s

sta rtin g a t 5.5 p er ce nt for th e lowest decile and rising 1 percentage

point from decile to decile to reach 14.5 per cent for the top decile; the

other distribution

( U )

is much more unequal, the shares starting at 1

per c ent fo r th e lowest decile, and rising 2 percentage po ints from decile

to decile to reach 19 per cent for the top decile. We assign per capita

incomes to each sector: 50 units to A an d 100 units to B in case I (lines

1-10 in the illustration) 50 to and 200 to B in case

I1

(lines 11-20).

Finally, we allow the proportion of the num bers in sector A in the total

number to decline from 0.8 to 0.2.

T h e numerical illustration is only a pa rtial sum mary of the calcula-

tions, showing the sh ares of the lowest and highest quintiles in th e in-

come distribution for the total population under different

assumption^.^

The basic assumptions used throughout are that the per capita income

of sector

B

(nonag ricultural) is always higher than th at of sector A;

that the proportion of sector

A

in the total number declines; and that

the inequality of the income distribution within sector A m ay be as

wide as that within sector B but not wider. With the assumptions con-

The underlying calculations are quite simple. F or each case we distinguish

20

cells within

the total distribution-sets of ten deciles fo r each sector. Fo r each cell we comp ute th e

percentage shares of both number and income

in

the number and income of total popula-

tion, and hence also the relative per capita income of each cell. The cells are then arrayed

in increasing order of their relative p er capita income an d cumulated. I n the resulting

cumulative distributions of number and countrywide income we establish, by arithmetic

interpolation, if interpolation

is

needed, the percentage shares in total income of the

successive quintiles of the country s population.


16/31

KUZNETS: ECONOMIC GROWTH AND INCOME INEQUALITY

3

Proportion of Number in Sector

A

to Total Number

0 . 8 0 .7

0 . 6

0 . 5

0 . 4

( 1 ) ( 2 ) ( 3 ) ( 4 )

( 5 )

? ?

??

I Pe r Capita Income of Sector A = 5 0 ;

of S ecto r B

100

1.

Pe r capita income of tota l pop-

ulation

60

65

i 0 75 80 85

90

Distribution ( E ) for Both Sec-

tors

2.

Sh ar e of

1st

quintile

10.5 9 .9 9 . 6

9 .3 9 . 4

9 . 8

10.2

3 .

Shar e of 5 th quintile

34.2

35.8

35.7 34.7

3 3 .2 3 1 .9

30 .4

4 . Range (3 -2 )

23.7

25.9

26 .1

25.3 23 .9 2 2 .1

20.2

Distribution ( U ) for Both Sec-

tors

5. Sh are of 1st quintile 3 .8 3 .8

3 .7

3 . 7 3 . 8

3 . 8 3 .9

6 . Share of

5th

quintile

40.7 41.9 42 .9

42 .7

4 1 .5 40 .2

38.7

7 . Range (6-5) 36.8

38.1 39 .1

39.0

3 7 .8 36.4

3 4 .8

Distribution

( E )

for Sector

A ,

( U )

for Sector B

8. Sh are of 1st qu intil e 9 . 3

8 .3 7 . 4 6 .7

6 .0

5 .4 4 .9

9 . Sh are of 5 th quintile

37.7 4 1 .0

42 .9

42.7 41.5 40.2

38.7

10.

Range

(9-8) 28.3

3 2 .7 3 5 .4

36 .0

35.5

34 .8

3 3 .8

11 Pe r Capita Income of Sector A=50;

of S ecto r B 200

11.

Per ca pita income of to tal pop-

ulation

80 95

110

125 140

155

170

Distribution ( E ) for Both Sec-

tors

12. Sh are of 1st quintile

7 . 9

6 . 8 6 .1

5 .6

5 .4

5 .4 5 .9

13. Sh are of 5 th quintile

50.0

49.1 45.5

41.6

3 8 .0

35 .0

3 2 .2

14.

Range

(13-12) 4 2 .1 4 2 .3 3 9 .4 36.0 32 .6 29.6 26.3

Distribution

( U )

for Both Sec-

tors

15. Sh are of 1st quintile

3 . 1 2 .9

2 .7 2 .6

2 .6

2.7

3 . 1

16. Share of

5th

quintile

52.7

56 .0 54 .5

51.2

47 .4 44.1

40.9

17. Range (16-15)

49 .6 5 3 .1

51 .8 48.6

44.8 4 1 .4

37.9

Distribution E) for Sector

A ,

( U )

for Sector B

18. Sh are of 1st qui ntile

7 . 4

6 .2

5 .4 4.7

4 . 2

3 . 9

3 .8

19.

Share of 5 th quintile

51.6

56.0 54.6

51.2

47 .4

44.1 4 0 .9

20.

Range

(19-18)

44.2 49.8

49.2

4 6 .5 43.2 4 0 .2

37.2

For m ethods of ca lculating the shares

of quintiles, see text p.

12and fn.

6 .

Some

differences

will not check because of rounding.

cerning th ree sets of factors-intersector differences in pe r ca pit a in-

come, intrasecto r distributio ns, an d sector weights-varying within the

limitations just indicated , the following conclusions ar e suggested :


17/31

14

T H E A M E R IC A N E C O N O M IC R E V I EW

Fir st, if th e per ca pita incom e differential incre ases, or if t h e income

distribution is more unequa l for sector B tha n for sector A, or if both

conditions ar e prese nt, t he rise over time in the relative weight of secto r

B causes a marked increase in inequality in the countrywide income

distribution. W e have her e a d em onstration of the effects upon tre nds

in income inequality of interin dus try sh ifts aw ay from agriculture dis-

cussed above pp . 7-8).

Second, if the intrasector inconie distribution is the same for both

sectors, an d th e widening inequality in the coun trywide income distribu-

tion is due only to th e increa sing per capita income differential in favor

of sector B, such widening is greater when the intrasector income dis-

tributions are characterized by moderate rather than wide inequali ty.

Th us, if the intrase ctor distributions a re of the

E

type, the range in th e

cou ntryw ide distribution w idens from 23.7 to 26.3 as prop ortion of A

dro ps from 0.8 to 0.2 and a s th e ratio of per ca pita income of sec tor

B

to th at of sector A changes from 2 to 4 see line 4, col. 1, an d line 14,

col.

7 ) .

If the distributions ar e used, the range, under identical con-

ditions, widens only from 36.8 to 37.9 see line 7, col.

1,

and line 17,

col. 7). This difference is revealed more clearly by the change in the

sh are of the 1s t quintile, which bears th e br un t of widening ineq uality:

for the E distribution, the share drop s from 10.5 line 2 , col. 1) to 5.9

line 12, col. 7) for the distribution, from 3.8 line 5, col.

1

to 3.1

line 15, col. 7).

Th ird , if the per cap ita income differen tial between sectors is con-

sta nt, bu t th e intrasector distribution of B is mo re unequal th an th at of

A, the widening inequality in the countrywide distribution is the

greater, the lower the assumed per capita income differential. Thus for

a differential of 2 to 1 , the ra nge widens from 28.3 when t he prop ortion

of A is 0.8 line 10 , col. 1 ) to 36.0 a t the pea k when th e prop ortion of

A

is 0.5 line 1 0, col. 4 ) a nd is still 33.8 when th e prop ortion of A drop s

to 0.2 line 1 0, col. 7 ). F or a per c ap ita incom e differential of 4 to 1, th e

widening of th e rang e a t the maxim um is only fro m 44.2 line 20, col.

1 ) to 49.8 line 20, col. 2 an d then the range declines to 37.2 line 20,

col. 7 ) , well below th e initial level.

Fo urth , th e assu mp tions utilized in the num erical illustration-of a

rise in prop ortion s of to ta l num ber in section B ,

of g rea ter inequ ality

in the distribu tion w ithin sector B , and of the growing excess of per

ca pita income in B over tha t in A-yield a decline in th e sh ar e of th e 1s t

quintile th at is much m ore conspicuous than the rise in the sha re of the

th

quintile. Th us th e shar e of the 1st quintile, with the proportion of A

at 0.8, distribution in B illore unequal than in A, and a per capita in-

come differential of 2 to 1, is

9.3

line 8, col. 1 ) . As we shift to a pro-

portion of A o f 0.2, and a per cap ita income differential of 4 to 1, the


18/31

KUZNETS: ECONOMIC

GROWTH

AND INC OME INEQUALITY

5

share of the 1st quintile drops to 3.8 l ine 18, col. 7). Under the same

conditions, the share of the 5th qu intil e changes from 37.7 line 9, col.

1 to 40.9 line 19, col. 7 ).

F if th , even if the differential in per cap ita income between th e two

sectors remains constant and the intrasector distributions are identical

for the two sectors, the mere shift in the proportions of numbers pro-

duces slight but significant changes in the distribution for the country

as a whole. In general, as the proportion of A drifts from 0.8 down-

wards, the rang e tends first to widen an d then to diminish. W hen the per

capita income differential is low 2 to l ) , the widening of the range

reaches a peak close to middle of the series, i.e. a t a proportion o f

A

equal to 0.6 lines 4 an d 7) an d the movemeilts in the range tend to be

rather limited. When the per capita income differential is large 4 to I ) ,

the range contracts as soon as the proportion of A passes the level of

0.7,

and the decline in the range is quite substantial l ines 14 and 17 ).

Sixth, of pa rticu lar bea ring upon th e share s of up per-income grou ps

is the finding th at the sha re of t he top qu intile declines as the propo r-

tion of A falls below a certa in, rathe r high fraction of tota l num bers.

Th ere is not a single case in th e illustration in which the share of the

5th quintile fails to decline, either throughout or through a substantial

segmen t of the sequence in the down ward movem ent of th e prop ortion

of A from 0.8 to 0.2. In l ines 6 and 9, th e share of the 5 th quin tile de-

clines beyond the point at which the proportion of A is 0.6; and in all

other relevant lines the downward tren d in the share of the 5th qu intile

sets in earlier. Th e reason lies, of course, in the fact th at with increasing

indus trialization, the growing weight of the nona gricultu ral sector,

with its higher per capita income, raises the per capita income for the

whole economy; and yet per capita income within each sector and the

intrasector distributions are kept constant. Under such conditions, the

upper share s would fail to decline only if there were either a grea ter

rise in per ca pita incom e of s ector B th an in tha t of sector A or increas-

ing inequality in the intrasector distribution of sector

B.

Several other conjectural conclusions could be drawn with ad ditional

variation s in assum ptions , an d multiplication of s ectors beyond th e

two distinguished in the numerical illustration. But even in the simple

inodel illustrated the variety of possible patterns is impressive; and

one is forced to the view that much more empirical information is

needed to perm it a prop er choice of specific assump tions an d con stants.

Granted that several of the conclusions could be generalized in formal

mathem atical terms, u seful inference s would b e within our reach only if

we knew more abo ut t he specific sector distributions a nd th e levels an d

trends in per cap ita income differentials amon g the sectors.

If then we limit ourselves to what is known or can be plausibly as-


19/31

6 T H E A ME RICA N E CO N O MI C RE V IE W

sumed, the following inferences can be suggested. We know that per

capita income is greater in sector B than in sector A; tha t, a t best , the

per capita income differential between se ctors A an d

B

has been fairly

constant e.g., in the United States) and has perhaps more often in-

creased; th at th e proportion of sector A in tota l num bers has dimin-

ished. The n, if we sta rt with intrase ctor distribution of B m ore unequal

than for A we would expect results suggested by either lines 8-10 or

18-20. I n th e former case, the range widens as th e proportion of A dro ps

from 0.8 to 0.5, a nd then narrows. I n th e latter case, the range declines

beyond the point a t which th e proportion of A is 0.7. B ut in both cases,

the sh are of the 1st quintile declines, an d fairly app reciab ly an d con-

tinuously (see l ines an d 18) . T h e magnitude and continuity of the

decline are partly th e result of the specific assump tions mad e; bu t one

would b e justified in arguing th at within the b road limits suggested by

the illustration, the assumption of greater inequality in the intrasector

distribution for sector

B

than for sector A, yields a downward trend

in the share of the lower-income groups. Yet we find no such tren d ir.

the empirical evidence that we have. Can we assume that in the earlier

periods th e intern al distribution for sector B was not more unequa l tha n

for sector A, despite the more recent indications th at u rban income dis-

tribution is more unequal than the rural?

Th ere is, obviously, room fo r conjecture. I t seems most plausible

to assume that in earlier periods of industrialization, even when the

nonag ricultural population w as still relatively sm all in th e total, its in-

come distribution was more unequal tha n tha t of th e agricultur al popu-

lation. This would be particularly so du ring the periods when indu stri-

alization an d urbanization were proceeding ap ace and t he urba n popula-

tion was being swelled, an d fairly rapidly , by immigrants-either fr o m

the country s agricultural areas or from abroad . Under these condi-

tions, the urban population would run the full gamut from low-income

positions of recent ent ran ts to th e economic peak s of the established

top-income groups. T h e urban income inequalit ies might be assumed to

be f a r wider tha n those for t he agricultural population which was or-

ganized in relatively small individual enterprises (large-scale units were

rarer then than now).

If we gran t the assump tion of wider inequality o f distribution in sec-

tor

B,

the shares of th e lower-income brack ets should have shown a

downward trend. Yet the earlier summary of empirical evidence indi-

cates tha t during the last 50 to 75 years there has been no widening in

income inequality in th e developed countries bu t, on the con trary, some

narrowing w ithin the last two to f ou r decades. I t follows th at th e intra-

sector distribution-either for sec tor A or fo r sec tor B-must have

shown sufficient narro wing of inequ ality to offset th e increase called


20/31

KUZNETS: ECONOMIC

GROWTH

AND INCOME INEQUALITY

7

fo r by the factor s discussed, Specifically, the sh ares of th e lower income

groups in sectors A and/or B must have increased sufficiently to offset

the decline th at would otherwise have been produced by a com bination

of the elem ents shown in the numerical illustration .

This narrowing in inequality, the offsetting rise in the shares of the

lower brackets, most likely occurred in the income distribution for the

urban groups, in sector B. W hile it m ay also have been present in sector

A i t would have had a m ore limited effect on the ineq uality in th e coun-

tryw ide incom e distribu tion because of the ra pidly diminishing weight

of sector

A

in the total. Nor was such a narrowing of income inequ ality

in agriculture likely: with indu strializatio n, a higher level of technology

permitted larger-scale units and, in the United States for example,

sharpened the contrast between the large and successful business

farmers and the subsistence sharecroppers of the South. Furthermore,

since we accept th e assumption of initially narrow er inequ ality in th e

interna l distribution of income in sector A than in sector B, any signifi-

cant reduction in ineq uality in the former is less likely than in t he latter .

He nce we may conc lude th a t the major offset to th e widening of in-

come inequality associated with the shift from agriculture and the

countryside to industry a nd th e city mus t have been a rise in th e income

sh are of th e lower groups within the nonagricultural sector of the

population. This provides a lead for exploration in what seems to me

a most promising direction: consideration of the pace an d chara cter of

the economic growth of t he urb an population, with partic ula r reference

to th e relative position of lower-income groups. Mu ch is to be said for

the notion tha t once the early turb ulen t phases of industrialization an d

urbanization ha d passed, a varie ty of forces converged to bolster the

economic position of the lower-income grou ps within t he ur ba n popula-

tion. The very fact that after a while, an increasing proportion of the

urba n population was native, i.e., born in cities rat he r tha n in the

rural area s, an d hence more able to tak e adva ntag e of the possibilities

of city life in prepara tion for the economic strugg le, meant a be tter

chance for organization and adaptation, a better basis for securing

greater income shares than was possible for the newly immigrant

population coming from the countryside or from abroad. The increas-

ing efficiency of the older, established urban population should also be

taken into account. Furthermore, in democratic societies the growing

political power of the u rb an low er-income group s led to a var ie ty of

protective and supporting legislation, much of it aimed to counteract

the worst eff ec ts of ra pid industrialization and urb aniza tion and to

support the claims of the broa d masses for more a deq uate shares of the

growing income of t he c oun try. Space does not pe rm it t he discussion of

demographic, political, and social considerations that could be brought


21/31

8 THE

MERIC N ECONOMIC

REVIEW

to be ar to explain th e offsets to a ny declines in the sha res of the lower

groups, declines otherwise deducible from the trends suggested in the

numerical illustration.

111

Other Trends Related to Those in nconze nequali ty

One aspect of the conjectural conclusion just reached deserves em-

phasis because of its possible interrelation with other important ele-

ments in th e process a nd the ory of economic growth. T h e sca nty em-

pirical evidence suggests that the narrowing of income inequality in

the developed countries is relatively recent and probably did not char-

acterize the earlier stages of their growth. Likewise, th e various facto rs

tha t have been suggested above would explain stability and narrowing

in income inequality in the later rather than in the earlier phases of

industrialization and urbanization. Indeed, they would suggest widen-

ing inequality in these e arly phases of economic grow th, especially in th e

older countries where the emergence of the new indu strial system had

sha tterin g effects on long-established pre-indu strial economic an d social

institutions. T hi s timing characteristic is particularly applicable to fac-

tors bearing upon the lower-income groups: the dislocating effects of

the agricultural an d industrial revolutions, combined with the swarm-

ing of pop ulation incident upon a rapid decline in dea th rates and the

maintenance or even rise of birth rates, would be unfavorable to the

relative economic position of lower-income groups. Furthermore)

the re ma y also have been a prepo nderan ce in th e earlier periods of fac-

tors favo ring ma intenance or increase in the share s of top-income

group s: in so far as their position was bolstered b y gains arising out of

new industries, by an unusually rapid rate of creation of new fortunes,

we would expect these forces to be relatively stronger in the early

phases of in dustrialization th an in th e later when the pace of ind ustrial

growth slackens.

One might thus assume a long swing in the inequality characterizing

th e secular income str uc tu re: widening in the early phases of economic

growth when the transition from the pre-industrial to the industrial

civilization was most rapid; becoming stabilized for a while; and then

narrowing in the later phases. This long secular swing would be most

pronounced fo r older cou ntries where th e dislocation effects of the

earlier phases of modern economic growth were most conspicuous; but

it might be found also in th e younger countries like the United States,

if the period preceding ma rked in dustrialization could be compared with

the early phases of industrialization, and if the latter could be com-

pared with th e subsequent phases of greater maturity.

f there is some evidence for assuming this long swing in relative

inequality in the distribution of income before direct taxes and exclud-


22/31

KUZNETS: ECONOMIC GROWTH

A N D

INCOME TNEQUhLITY

9

ing free benefits from government, there is surely a stronger case for

assuming a long swing in inequa lity of income net of direct taxes an d

including government benefits. Progressivity of income taxes and, in-

deed, their very im portanc e characterize only the m ore recent phases of

development of the presently developed countries; in narrowing in-

come inequality they must have accentuated the downward phase of

the long swing, contributing to t he reversal of trend in th e secular

widening and n arrowin g of income inequality.

No adequate empirical evidence is available for checking this con-

jecture of a long secular swing in income ine qua lity ; nor can the phases

be dated precisely. Ho wev er, to mak e it m ore specific, I would place the

early phase in which income inequality might have been widening, from

about 1780 to 1850 in England; from about 1840 to 1890, and particu-

larly from 1870 on in the U nited Sta tes; and , from the 1840 s to the

1890 s in Germ any. I would put th e phase of n arrowin g income in-

equality somewhat later in the United States and Germany than in

England-perhaps beginning with th e first world war in th e forme r an d

in the last quarte r of the 19th century in the latter.

I s there a possible relation between this secular swing in income

inequality and the long swing in other im portan t compon ents of the

growth process? For the older countries a long swing is observed in the

ra te of growth of population-the upw ard pha se repre sented by accel-

eration in the rate of growth reflecting the early reduction in the death

rate which was not offset by a decline in the birth rate (and in some

cases was accompanied by a r ise in the birth rate ) and th e downward

phase rep resented b y a shrinkin g in the rat e of growth reflecting th e

more pronounced downward trend in the birth rate. Again, in the older

countries, and also perhaps in the younger, there may have been a

secular swing in th e ra te of urbanization, in the sense tha t th e propor-

tional additions to urban population a nd the m easures of internal mi-

gration that produced this shift of population probably increased for

a while-from the earlier much lower levels; but then tended to diminish

as urban population came to dominate the country and as the rural

reservoirs of migration became proportionally much smaller. For old,

and perhaps fo r young countries also, there m ust have been a secular

swing in the prop ortions of s avings or ca pita l formation to to tal eco-

nomic produc t. Per capita product in pre-industrial times was not large

enough to permit as high a nationwide rat e of saving or capital fo rma-

tion s was attained in the course of industrial development: this is

7Prokopovich s data on Prussia. from t he source cited in footno te 1, indicate a sub-

tantial widening in income inequality in the early period. The share of the lower

9

per

cent of the population declines from 73 per cent in

1854

to 65 per cent in 1875; the share

of the top 5 per cent rises from 2 1 to 25 per cent. But I do not know enough about the

dat a for the early years to eva luate the reliability of the finding.


23/31

2

THE


suggested by present comparisons between net capital formation rates

of 3 to 5 per cent of national prod uct in underdeveloped countries and

rates of 10 to 15 per cent in developed countries. If then, at least in the

older coun tries, an d perhap s even in th e younger ones-prior to initia-

tion of the process of mode rn development-we begin with low secular

levels in the savings proportions, there would be a rise in the early

phases to appreciably higher levels. We also know that during recent

periods the net capital formation proportion and even the gross, failed

to rise an d perha ps even declined.

Other trends might be suggested that would possibly trace rong

swings similar to those for inequality in income struc ture , rate of growth

of po pulation, ra te of urban ization an d inte rna l migration, an d the pro-

portion of savings or cap ital formation to nation al produ ct. Fo r ex-

ample, such swings might be found in the ratio of foreign tra de to

domestic activities; in the aspects, if we could only measure them prop-

erly, of government activity tha t bear upon marke t forces

there

mu st have been a phase of increasing freedom of m ark et forces,

giving way to grea ter intervention by g overnm ent). B ut the suggestions

alre ady m ade suffice to indic ate tha t the long swing in income inequa lity

must be viewed as part of

a

wider process of economic grow th, and

interrelated with similar movements in other elements. T h e long alter -

nation in the r ate of growth of pop ulation can be seen partly as a cause,

partly as an effect of the long swing in income inequality which was

associated with a secular rise in real per capita income levels. The long

swing in income inequality is also probably closely associated with the

swing in capital formation proportions-in so far as wider inequ ality

makes fo r higher, and narrower inequ ality for lower, country-wide sav-

ings proportions.

IV.

ompa rison of Developed and Underdeveloped ountrie s

W ha t is the bearing of the experience of the developed cou ntries upon

the economic growth of underdeveloped countries? Let us examine

briefly th e d ata on income distribution in th e latter, an d speculate upon

some of th e implications.

As might have been expected, such da ta for underdeveloped cou ntries

are scan ty. Fo r the p resent purpose distributions of family income for

Ind ia in 1949-50, fo r Ceylon in 1950, and for Pue rto R ico in 1948 were

used. While the coverage is narrow an d the margin of erro r wide, the

data show that income distribution in these underdeveloped countries

is somewhat

more

unequal than in the developed countries during the

period af ter the second world war. T h us the shares of th e lower quin-

tiles are 28 per cent in Ind ia, 3 per cent in Ceylon, an d 24 per cent in

Puer to Rico-compared wi th 34 per cent in the United States and 36


24/31

KUZNETS: ECONOMIC

GROWTH

AND INCOME INEQUALITY

per cent in the United Kingdom. T he shares of the t o p quintile are 55

per cent in India, 50 per cen t in Ceylon, an d

56

per cent in Puerto Rico,

compared with 44 per cent in the United States and 45 per cent in the

United Kingdom.'

This comparison is for income before direct taxes and excluding free

benefits fro m governments. Since the burden an d progressivity of direct

taxes are much greater in developed countries, and since it is in the

latte r th at su bs tan tial volumes of free economic assistance are extended

to the lower-income groups,

a

com parison in terms of incom e ne t of

direct taxes and including government benefits would only accentuate

the wider ineq uality of income distributions in the underdeveloped co un-

tries. I s this difference a re liable reflection of wider ine quality also

in the distribu tion of secula r income levels in underdeveloped coun tries?

Even disregarding the margins of error in the data, the possibility

raised earlier in this paper that transient disturbances in income levels

may be more conspicuous under conditions of primitive material and

economic technology would affect the comparison just made. Since the

distributions cited reflect the annual income levels, a greater allowance

should perhaps be made for transient disturbances in the distributions

fo r the und erdeveloped than in those for the developed countries.

Whether such a correction would obliterate the difference is a matter

on which I

have no relevan t evidence.

Another consideration might tend to support this qualification. Un-

derdeveloped coun tries are ch aracterized by low av erage levels of in-

come per capita, low enough to raise th e question how the po pulations

manage to survive. Let us assume that these countries represent fairly

unified population groups, and exclude, for the moment, areas that

combine large native populations with small enclaves of nonnative,

privileged m inorities, e.g., Ke ny a an d R hodesia, where income inequa l-

ity, because of the excessively high income sha res of the privileged

minority, is appreciably wider than even in the underdeveloped coun-

tries cited above.' On this assum ption, one ma y infer th at in countries

or sources of these da ta see Regional Economic Tren ds an d Levels of Living, sub-

mitted a t the Korm an Waite Harris Fou ndation Insti tu te of th e University of Chicago in

November 195'4 (in press in th e volume of proceedings). T his paper, an d an earlier one,

Underdeveloped Countries an d the Pre-industrial Phases in the Advanced C ountries: An

Attempt a t Comparison, prepared for the World Population Meetings in Rome held in

Septemb er 1954 (in press) discuss issues raised in this section.

'In one year since the second world war, the non-African group in Southern Rhodesia,

which accounted for only

5

per cent of total population, received 57 per cent of total in-

come; in Kenya, the minority of only 2.9 per cent of total population, received 51 per cent

of total income; in No rther n Rhodesia, the minority of only 1.4 per cent of tota l population,

received 45 per cent of t ota l income. See United Natio ns,

Natioturl Income and Its k

trihution in Underdeveloped Countries

Statistical Paper, Ser.

E

no.

3

1951, Tab le 12, p.

19.


25/31

THE AMERICAN ECONOMIC REVIEW

with low averag e income, the se cular level of income in the lower brac-

kets could no t be below a fairly sizable proportion of averag e income-

otherwise, the groups could not survive. This means, to use a purely

hypoth etical figure, th at the sec ular level of the sh are of the lowest

decile could not fall far short of

6

or 7 per cen t, i.e., the lowest decile

could not have a per cap ita income less than six- or seven-ten ths of the

countrywide average. I n m ore advanced coun tries, with higher average

per cap ita incomes, even th e secula r share of the lowest bra cke t could

easily be a smaller fraction of the countrywide average, say as small as

or per ce nt for the lowest decile, i.e., from a fifth to a third of the

countryw ide average-without implying a ma terially impossible eco-

nomic position for tha t group. T o be sure , there is in all countries con-

tinuous pressure to raise the re lative position of the bottom-income

groups; but the fact remains that the lower limit of the proportional

share in the secular income structure is higher when the real country-

wide per c apit a income is low than when i t is high.

f

the long-term s ha re of the lower-income groups is larger in th e

underdeveloped than in the average countries, income inequality in the

former should be narrower, not wider as we have found. However, if

the lower brackets receive larger share s, an d a t the same time the very

top bra cke ts also receive larger shares-which would me an th at the

intermediate income classes would not show as gre at a progression from

the bottom-the net effect may well be wider inequa lity. T o illus trate ,

let us compare the distributions for India and the United States. The

first quintile in In dia receives per cen t of tota l income, more tha n th e

6

per cen t sha re of the first quintile in the United States. Bu t the second

quintile in India receives only

9

per cent, the third 11, and the fourth

16; whereas in the United S tates, the shares of these quintiles are 12,

16, and

respectively. T h is is a rough statistic al reflection of a fa irly

common observation relating to income distributions in underdeveloped

compared with developed countries. T h e former have no middle

classes: there is a sharp contrast between the preponderant proportion

of population whose average income is well below the generally low

countrywide average, and a small top group with a very large relative

income excess. Th e developed countries, on the other ha nd, a re charac-

terized by a much more gradual rise from low to high shares, with sub-

stantial groups receiving more than the high countrywide income

average, an d the top groups securing smaller shares than the comp arable

ordinal groups in underdeveloped countries.

I t is, therefore , possible th at even the distribution s of secula r income

levels would be more unequal in underdeveloped than in developed

countries-not in th e sense th at the shares of the lower bra cke ts would

be lower in the former th an in th e latter , but in the sense th at the sha res


26/31

KUZNETS: ECONOMIC GROW TH AND INCOM E INEQUAL ITY

23

of the very top g roups would be higher a nd th at those of th e groups

below the top would all be significantly lower than a low countrywide

income average. T hi s is even more likely to be tru e of th e distribution

of income ne t of direct taxe s and inclusive of free govern me nt ben efits.

Bu t whether a high probability weight can be a ttache d to this conjecture

is a ma tter for further study .

In the absence of evidence to the contrary, assume tha t i t is tru e:

that the secular income structure is somewhat more unequal in under-

developed countries tha n in th e more advanced-particularly in those

of Western an d N orthe rn E uro pe an d their economically developed

descendants in the New World the United States, Canad a, Australia,

and New Zealand). This conclusion has a variety of important impli-

cations and leads t o some pregnant questions, of which only a fe w can

be stated here.

I n the first place, the wider inequality in th e secular income stru ctu re

of underdeveloped cou ntries is associated with a much lower level of

average income per cap ita. Tw o corollaries follow-and th ey would

follow even if the income inequalities were of the same relative range

in the two groups of countries. Firs t, the impact is far sharp er in the

underdeveloped countries, where the failure to reach an already low

countrywide average spells much greater material and psychological

misery than similar proportional deviations from the average in the

richer, more advanced countries. Second, positive savings a re obviously

possible only a t much h igher relativ e income levels in th e underdeveloped

countries: if in the more adva nced co untries some savings ar e possible

in the fourth quintile, in t he underdeveloped countries savings could be

realized only a t the very peak of the income pyramid, say b y the top

5 or per cent.

f

so, th e conce ntration of savin gs an d of a ssets is even

more pronounced than in th e developed countries; and the effects of

such concentration in the pa st m ay serve to explain the peculiar charac-

teristics of th e secular income stru ct ur e in underdeveloped coun tries

today.

The second implication is that this unequal income structure pre-

sumab ly coexisted w ith a low rate of growth of income per c apita. T h e

underdeveloped countries today have not always lagged behind the

presently developed areas in level of economic performance; indeed,

some of th e form er ma y have been th e economic leaders of th e world in

the centuries preceding the last two. The countries of Latin America,

Africa, and particularly those of Asia, are underdeveloped today be-

cause in th e las t two cen turies, and even in recent d ecades, their ra te of

economic growth ha s been far lower than t ha t in the Wes tern World-

and low indeed, if any growth there was, on a per capita basis. The

underlying sh ifts in industrial structure, the oppo rtunities for internal


27/31

24 THE AMERICAN ECONOMIC REVIEW

mobility and for economic improvement, were far more limited than

in the more rapidly growing countries now in the developed category.

Th ere was no hope, within th e lifetime of a generation, of a significantly

perceptible rise in the level of real income, or even th at th e next genera-

tion might fare m uch better. I t was this hope that served as an impor-

tant and realistic compensation for the wide inequality in income dis-

tribution that characterized the presently developed countries during

the earlier phases of their growth.

T h e third implication follows from the preceding two. I t is quite

possible th a t income inequa lity has not narrowed in the uilderdeveloped

countries within recent decades. T he re is no empirical evidence to check

this conjectural implication, bu t i t is suggested by the absence, in these

area s, of the dy nam ic forces associated with rapid growth th a t in the

developed coun tries checked the upw ard trend of th e upper-income

shares that was due to the cumulative effect of continuous concentra-

tion of pa st sa vings ; a nd it is also indica ted by the failu re of the politi-

cal and social systems of underdeveloped c oun tries to initiate the govern-

mental or political practices that effectively bolster the weak positions

of th e lower-income classes. Indeed, th ere is a possibility tha t inequa lity

in th e secular income structu re of underdeveloped countries may have

widened in recent decades-the only qualification being th at where

there has been a recent shift from colonial to independent status, a

privileged, nonn tive minority may have been eliminated. But the im-

plication, in term s of the income distributio n among the

n tive

popula-

tion prope r, still remains plausible.

T he somber picture just presented m ay be a n oversimplified one. B ut

believe that it is sufficiently realistic to lend weight to the questions

it poses--questions a s to the bearing of th e recent levels an d tren ds in

income inequality, and the factors tha t determine them, upon the future

prospect of underdeveloped cou ntries within th e orbit of the free world.

The questions are difficult, but they must be faced unless we are

willing completely to disregard past experience or to extrapolate me-

chanically oversimplified impressions of pa st de velopme nt. T h e first

question is: I s the pa ttern of the older developed countries likely to be

repeated in the sense that in the early phases of industrialization in the

underdeveloped countries income inequalities will tend to widen before

the leveling forces become strong enough first to stabilize and then re-

duce income inequalities? While the future c annot be a n exact repetition

of the p ast, there ar e alre ady certain elements in the present conditions

of underdeveloped societies, e.g. swarming of population due to sh arp

cuts in death rates unaccompanied by declines in birth rates-that

threate n t o widen ineq uality by de pressing the relative position of lower-

income groups even fu rther. F urthe rm ore, if an d when industrialization


28/31

KUZNETS ECONOMIC GROWTH AND INCOME INEQUALITY

5

begins the d islocating effects on these societies in which the re is often

an old hardened crus t of economic an d social institutions ar e likely to

tc? be qu ite sharp-so sh ar p a s to destroy the positions of some of th e

lower groups more rapidly than opportunities elsewhere in the economy

may be created for them.

The next question follows from an affirmative answer to the first.

Can the political framework of the underdeveloped societies withstand

the s trai n which furth er w idening of income inequality is likely to gen-

erate ? Th is que ry is pertinent if it is realized that the real per ca pita

income level of many underdeveloped societies today is lower than the

per capita income level of the pre sently developed societies before

th ir

initial phases of industrialization. A nd ye t the stresse s of the disloca-

tions incident to early phases of indus trialization in the developed coun-

tries were sufficiently acute to strain the political and social fabric of

society force major political reforms an d sometimes result in civil war.

T he answer to the second question may be negative even granted

that industrialization may be accompanied by a rise in real per capita

product. If for many groups in society the rise is even pa rtly offset by

a decline in their proportional share in total produc t; if consequently

it is accomp anied by widening of income ineq uality the res ulting pres-

sure s and conflicts ma y necessitate d rastic chang es in social and political

organization. This gives rise to the next and crucial question: How

can e ither the institutiona l and political framewo rk of th e unde rde-

veloped societies or the processes of economic growth an d indu strializa-

tion be modified to favo r a s ustain ed rise to higher levels of economic

performance an d yet avoid the fatally simple remedy of an au thorita rian

regime th at would use the population as cannon-fodder in the fight for

economic achievement? How to minimize the cost of transition and

avoid paying the heavy price-in in te rnal tensions in long-run ineffi-

ciency in providing means for satisfying wants of human beings as

individuals-which th e inflation of political power rep resen ted by

authoritarian regimes requires?

Fac ing these acu te problems one is cognizant of th e dang ers of ta k-

ing a n extreme position. One extreme-particularly tem pting to us-

is to favor repetition of past patterns of the now developed countries

patte rns th at under the m arkedly different conditions of th e presently

underdeveloped countries ar e almost bound to pu t strain on the exist-

ing social and economic institutions an d eventuate in revo lutionary ex-

plosions and a utho ritarian regimes. Th ere is danger in simple analogies;

in arguing that because an unequal income distribution in Western

Eu rope in th e past led to accu mulation of savings and financing of basic

capital formation the preservation or accentuation of present income

inequalities in the underdeveloped countries is necessary to secure the


29/31

26

TH


same result. Even disregarding the implications for the lower-income

groups, we may find that in at least some of these countries today the

consumption propensities of upper-income groups are far higher and

savings prop ensities fa r lower th an we re those of the more puritanical

upper-income gro ups of the prese ntly developed countries. Because the y

may have proved favorable in the past , i t is dangerous to argue that

completely free markets, lack of penalties implicit in progressive taxa-

tion, a nd the like ar e indispensable for the economic growth of th e now

underdeveloped countries. Under present conditions the results may be

qu ite th e opposite-withdrawal of accu mu lated assets to relatively

safe channels, either by flight abro ad or into real est ate ; and the

inability of governments to serve as basic agents in the kind of capital

formation th at is indispensable to economic growth. I t is dangerous to

argue that, because in the past foreign investment provided capital

resources to spa rk satisfacto ry economic growth in some of the smaller

European countries or in Europe's descendants across the seas, similar

effects can be expec ted tod ay if o nly the underdeveloped cou ntries can

be convinced of th e need of a favo rable climate. Yet, it is equ ally

dangerous to take the opposite posit ion and claim t ha t the prese nt prob-

lems are entirely new and that we must devise solutions that are the

product of imagination unrestrained by knowledge of the past, and

therefore full of romantic violence. What we need, and I am afraid i t

is but a truism, is a clear perception of past trends and of conditions

under which they occu rred, as well as knowledge of the conditions th at

characterize the underdeveloped countries today. With this as a begin-

ning, we can then at tem pt to translate th e elements of a prop erly under-

stood past into the conditions of an a deq uately understood present.

V. oncluding Remarks

In concluding this paper,

I

am ac utely conscious of the meag erness

of

reliable information presented. The paper is perhaps

5

per cent

empirical information and 95 per cent speculation, some of it possibly

tainted by wishful thinking. Th e excuse for building an elabo rate struc-

ture on such a shaky foundation is a dee

kuznets paper curve ushape

Documents