American Economic Association

Farm Size Distribution, Income Distribution, and the Efficiency of Agricultural Production:
Colombia
Author(s): R. Albert Berry
Source: The American Economic Review, Vol. 62, No. 1/2 (Mar. 1, 1972), pp. 403-408
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/1821574
Accessed: 13/12/2010 14:03

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the 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 at
http://www.jstor.org/action/showPublisher?publisherCode=aea.

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

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.

http://www.jstor.org
 Farm Size Distribution,
Distribution, and
Agricultural Production:
By R. ALBERT BIERRY*
The conditions of Colombia's agricul-
tural sector seem, at first glance, to sug-
gest the "conflict" version of the relation
between output growth and income dis-
tribution. If indeed this is the situation,
the current inequality of income distribu-
tion (see below) suggests that policy deci-
sions should be very difficult. And the
nature of change over the last decade in
the sector suggests that those decisions
have essentially favored production with
only sporadic and limited attention given
to problems of distribution and quality of
employment. But pertinent information
raises doubts as to whether improved dis-
tribution and rapid output growth are
really in conflict. This study considers that
question, at the same time analyzing the
determinants of income distribution and
its changes over time, and the relative
social efficiency of smaller and larger
farms. To set the stage the information on
income distribution and its changes over
time is summarized.
Table 1 presents a best estimate of the
personal distribution of income1 of the
agricultural sector in 1960; the average
income level of each decile is also pre-
sented.2 The (Gini) coefficient of concen-
tration is estimated at 0.58. Any such cal-
culation of income distribution is nat-
urally fraught with many statistical prob-
* Associate Professor of Economics, Yale University.
I National accounts concept; family helpers excluded.
2 In 1960 the official exchange rate was 6.7 pesos per
dollar.
403
Income
the Efficiency of
Colombia
lems.3 It should be emphasized espe-
cially that, due to the methodology em-
ployed, Table 1 describes the distribution
of income generated in agriculture, not the
distribution of all income (from agricul-
ture or other sources) of people involved
in agriculture.4 These problems notwith-
standing, relevant conclusions can be
drawn: (a) the great majority of the agri-
cultural labor force had an income from
agriculture of below 5,000 pesos (about 700
U.S. dollars) and the bottom half had less
than 37000 pesos (400 U.S. dollars); (b) the
top 15 percent had close to 60 percent of
the income (say 55-65) and the bottom
85 percent therefore had 35-45 percent.
Determinants of the Skewness of the
Distribution of Agricultural Income
At a first level of discussion, the ex-
planation of the inequality of incomes in
agriculture is the unequal distribution of
land. The high incomes correspond to
3 Although the methodology used here (see Table 1)
has many weaknesses, it seems more reli'ablethan direct
sample techniques. Details of the methodology are
available in a more detailed version of this paper,
Berry, 1971.
4 While the pressure of their very low agricultural
incomes pushes them to earn incomes from other
sources, the limited evidence does not suggest that this
is very important for people in the bottom half of the
"income from agriculture" distribution. Impressionistic
evidence does suggest that incomes from other sectors
may be a high share of total income for some of the
people toward the top of the agricultural distribution
(i.e. the absentee farmers and the partially absentee
''commercial farmers"). Also capital gains (not included
in the tables) are hard to guess at.
404 AMERICAN ECONOMIC ASSOCIATION
TABLE 1-PERSONAL DISTRIBUTION OF INCOME TABLE 2-AVERAGE INCOMEOF
FROM COLOMBIAN AGRICULTURE, 1960 PRODUCERSBY FARM SIZE
(1960 pesos)
Cumulative Average
Decile Percentof Percent of Annual Farm Size Average Income Number of Producers
Income Income Income (Pesos)
(hectares)

(1) (2) (3) 1-2 1,300 191,350

1 2.24 2.24 865 2-3 1,900 117,000
2 2.87 5.11 1,110 3-4 2,320 92,000
3 3.34 8.45 1,290 4-5 2,640 58,200
4 3.73 12.18 1,440 5-10 3,670 169,150
5 4.21 16.39 1,625 10-20 5,580 114,200
6 4.68 21.07 1,807 20-30 6,750 44,050
7 5.78 26.85 2,232 30-40 8,430 26,500
8 7.90 34.75 3,060 40-50 10,203 16,240
9 12.77 47.52 4,940 50-100 12,800 40,000
10 52.48 100.00 20,270 100-200 23,800 22,300
200-500 41,140 13,700
Source: The estimate presented here is an adjusted ver- 500-1000 102,500 4,140
sion of that presented in Berrvz and Alfonso Padilla, 1000-2500 189,800 1,975
Appendix Table A-I. > 2500 527,700 790
Total 6,145 911,595

people with large farms (see Table 2). The
distribution of income among salaried
laborers, while showing some range (much
of it due to wage differentials among dif-
ferent regions of the country), does not
contribute much to the skewness of the
overall distribution, since all these incomes
Soutrce:Berry and Padilla, Appendix Table A-2, revised
to take account of new information.
come toward the bottom of it. Table 3 pre-
sents estimates of the source of income by
decile, distinguishing the labor share and
the capital share, and within the labor

SHARES,BY DECILESOF
TABLE3-LABOR ANDCAPITAL
THEPERSONAL DISTRIBUTIoNa
INCOME

Average
Income Hired Imputed Total
Decile (National Labor Pure Labor Labor Capital
Accounts Share Share Share' ShareL
Concept)

1 865 72.3 13.9 86.2 13.8

2 1110 70.2 14.9 85.1 14.9
3 1290 80.8 9.6 90.4 9.6
4 1440 80.6 9.7 90.3 9.7
5 1626 81.9 9.0 90.9 9.1
6 1807 57.7 21.1 78.8 21.2
7 2232 30.3 34.8 65.1 34.9
8 3060 7.5 40.4 47.9 52.1
9 4940 1.8 25.3 27.1 72.9
10 20270 1.2 5.7 6.9 93.1

Soutrce: Calculations by the author.

a The issue of uhether the income of the producer can meaningfully be distributed

among the factors he supplies is discussed elsewhere (see BerrV). The conclusion there is
that no plausible definition of the labor share could make its general relation with income
dramaticallv different from that shown in these figures.
b M\ore
precisely, hired labor share plus imputed pure labor share.
c More precisely, capital share including human capital income of operators.
 PROBLEMS OF DEVELOPMENT
share attempting to separate the paid
labor share and imputed pure labor share.
Although the figures presented in Table
3 are "best estimates" rather than precise
information, it is clear that (a) approxi-
mately the bottom half of income earners
receive by far the largest part of their in-
come as paid blue-collar workers on other
farms; (b) the capital share is very high on
the largest farms, possibly around 90 per-
cent but almost certainly above 75 percent.
Detailed information to permit calcula-
tions parallel to those presented above is
unavailable for other years, but the nature
of the 1960 distribution suggests that
knowledge of the path of the labor share
over time would give some feel for changes
in overall skewness. Real daily wage rates
appear to have been about the same in the
latter part of the 1960's as they were in the
mid 1930's; they underwent a decline in
the late depression years and the early
1940's, then rebounded and continued to
increase till the early 1960's, and have
since levelled off. Over the same period
average income per person engaged in
agriculture appears to have risen at an
average rate of 2-3 percent per year.
Since both land and capital have risen
faster thain labor over this period, the
labor slhare must lhave falleni substantially
(see Table 4).
It is not possible to make a neat deliinea-
tion between agricultural laborers and
farm operators in Colombia; all combina-
TABLE 4-"PURF" LABOR SIHARF IN AGRICULTURAL
VALIUE ADDFD, SELECTED GROLUP OF YEARS
1935-39 (66-84(( )a
1940-44 (56-79%,)a
1945-49 46-57%c
1950-54 40--47c%
1955-59 34-42>-(-
1960-64 35-43(%o
a Figures in parentheses appear to he ul)ward b)iased,
though not so nmuchas to inmlv that thev could be
below the 1945-49 figure. See the (liscussion in Berry,
p. 25.
POLICY 405
tions of operator-laborer (in terms of the
share of income from each) exist. As a re-
sult, it is unclear for what proportion of
income earners absence of a positive trend
in the real wage rate over this period im-
plies a failure of total income to rise.
Since in 1960 about 80 percent of the
labor force earned the majority of its in-
come from labor (paid or imputed), this
proportion must have been substantial.
The Relationship Between Factor Produc-
tivity and Income Distribution-A First
Level of Discussion
As just discussed, the majority of low
income people in Colombia's agricultural
sector earn most of their income working
on other people's farms; it follows ob-
viously that their income level, as well as
the quality and security of their employ-
ment, depends on the demand for labor
generated by the operators of those farms.
'I'he other major determinant of equality
of income distributioni is clearly equality
of land distribution. If in fact there is an
output-distribution conflict, it must be
true that either the heavy use of labor
(i.e., the applicationi of technologies which
require large labor inputs) and/or rela-
tively eclual distribution of land mutst he
inimical)le to hligllpro(ltictivity al ()utipult
growth. Oine piece of pertinent evidence is
the relative productivities of factors across
farm sizes.
'I'he figures of 'ITable5 show that value
added per effective hectare and per peso
of capital (including land) decreases with
size while output per person increases.
Output per worker is about ten times as
high for the largest size category used here
as for the smallest, while output per
hectare is only about 10 percent as high.
Most of the difference in output per
hectare appears to be dlue to the lower
average (uality of land on the larger
farms; the ratio of value added to either
value of land or value of land arid capital
406 AMERICAN ECONOMIC ASSOCIATION

TABLE 5-FACTOR PRODUCTIVITYAND FARM SIZE IN COLOMBIA,1960

(Value Figures in Thousands of 1960 Pesos)

Value of
Value Yield Crop Out- Percent of
Value of Value Value
Farm Size Output Added Added/ Value Added/ Overall Index of put/Hec- Arable and
(Hectares) per per Effective Added/ Value of Yield Cultivated tare of Pasture
Worker Worker Hectare Hectare Land and Indexc & Fallow Cropped& Land in
Capital Land Fallow Crops
Land

0-3 1.83 1.67 .75 1.37 .35 94.2a 80.5a 1.05a .87a
3-5 2.37 2.08 .79 .86 .36 96.8b 81.6b 1.03b 77b
5-10 3.15 2.71 .50 .73 .33 96.7 79.4 1.04 .66
10-50 4.15 3.47 .57 .44 .25 98.4 72.3 .96 .47
50-200 6.65 5.35 .38 .25 .17 117.8 68.8 .87 .28
200-500 10.76 8.61 .35 .21 .15 140.3 70.7 .90 .18
>500 17.16 15.07 .35 .13 .14 147.4 67.3 .89 .06
Total 4.44 3.71 .46 .28 .20

a 0-2 hectares rather than 0-3.

b 2-5 hectares rather than 3-5.
c Index of value of product per cultivated hectare assuming for each size category the distribution of land among
crops characterizing the crop sector as a whole.

Sources antdMet1todology:Figures in the first five columns are "best estimates" (Berry); alternative estimates and the
methodology of the various calculations are presented in Tables A-4, A-5 and A-7 of the cited study.

appears to be about twice as high on the
farms of 0-5 hectares as on those with more
than 500 hectares.' Total social factor
productivity emerges higher for smaller
farms for almost all plausible combina-
tions of assumptions about the social op-
portunity cost of factors. Even when
labor's annual opportunity cost is based
on the recorded average wage rate (250
days is taken as a typical working year)
and capital is accounted at its average
rate of return in agriculture, the larger
farms' productivity is only just equal to
that of the smaller ones; with those as-
sumptions the smallest farms (0-3 hec-
tares) have low measured efficiency, but
the other groups are all close to the aver-
age, with those in the 5-50 range appear-
ing slightly more efficient than the others.
The use of other assumptions about the
opportunity cost of labor gives the smaller
farms a clear-cut advantage, the relation
I These results, which are based on national aggre-
gate data and therefore subject to the various weak-
nesses of such data, are corroborated blvxmicro data col-
lected by James Grunig.
being monotonic above the size category
5-10 hectares; that group stands out as
the most efficient over the full range of
plausible assumptions. Unfortunately,
while they are suggestive, these relative
efficiency measures cannot (even ab-
stracting from the statistical deficiencies)
be accepted as definitive due to the in-
validity of the implicit assumption that
the capital and labor on various farm-
sizes are homogeneous; they do provide a
set of "benchmark" estimates, and while
substantial error may be present, it seems
unlikely that the qualitative relation be-
tween size and efficiency could be re-
versed.
Explanations of Differing Social Efficiency
of Farms by Size
The most plausible explanations of a
higher total social factor productivity on
small farms are (a) their facing a lower
price of labor and a higher price of capital,
and (b) the greater incentive associated
with low income levels which shows up,
according to some observers, in a higher
 PROBLEMS OF DEVELOPMENT POLICY
average quality of entrepreneurship on
smaller farms.6 Factors most frequently
hypothesized as working in the opposite
direction are economies of scale and the
greater ease of adoption of improved
technology by the better educated and
financed large farmers.
Colombian evidence suggests that most
of these factors are at work, at least in
parts of the agricultural sector. Tech-
nological advantage of the large farm is
suggested by the greater average yields
for specific crops, a differential which in
1966 appareintly reached 2:1 or more for
some crops like wheat, barley, sugar, and
potatoes. For other crops there is little or
no differential, but in no case does there
appear to be a negative relation between
farm size and yield; an overall yield index
is a little over 50 percent higher for farms
of over 500 hectares than for those under 2
hectares (see Table 5). Value added per
hectare is undoubtedly a less positive
(more negative) function of farm size,
since associated with the more modern
technologies which produce the higher
yields on the larger farms is a high "pur-
chased input/value of output" ratio;
figures are not available to quantify this
difference.
The major "proximate" factors asso-
ciated with the smaller farms' higher value
added per unit of scarce resources (land
and capital) appear to be their smaller
use of fallow (see Table 5) and the greater
proportion of their land directed to crops
relative to livestock (more precisely to
cattle).
The second question pertinent to the
possibility of an output maximization-
distribution improvement conflict is
whether those forms of production (tech-
nologies, sizes of farm, etc.) which achieve
6 Note also that the "quality" of relevance is not a
measure of the potential different in(lividuals have
(which would more likely he high on the larger farms)
but the quality of management actually applied.
407
high factor productivity imply low labor
utilization and hence unequal distribution
of income. No general answer can be
given on the basis of existing evidence, but
the available categorization by farm size
is again suggestive with respect to the
labor utilization scarce factor produc-
tivity relation; as seen in Table 5, the
small farms are much more labor inten-
sive.
Policy Relevance of the Above Conclusions
The existence of different factor returns
on farms which differ in some way (e.g.,
size), while on the one hand implying that
groups may differ systematically in social
efficiency, also implies imperfect factor
markets.7 Imperfect markets complicate
the concept of social efficiency and may
make impossible simple policy conclusions
about which group of farms should be
favored by public policy. It is possible that
the existing situation is "efficient" given
the imperfections in the system.
Broadly speaking, the information pre-
sented above might be relevant to two
sorts of policy questions: (a) decisions af-
fecting the size of farm units, e.g., the
size chosen for the colonization of public
or other newly settled lands, and the na-
ture and extent of redistribution via
agrarian reform; (b) decisions involving
the distribution (and possibly the pricing)
of factors whose supply the government
affects directly or indirectly, e.g., credit.
With respect to question (a), it might ap-
pear that the higher factor productivity of
the smaller farms implies the desirability
of land redistribution; but this conclusion
does not necessarily follow. There may,
for example, be transitional costs asso-
ciated with the movement of people who
previously farmed smaller plots (or were
I Unless the difference in measured returns is just off-
set hN-unrecorded nonmonetarv-l)enefits or costs, and /or
the cluality of the factor varies in proportion to the
returns.
408 AMERICAN ECONOMIC ASSOCIATION
landless) to the larger units and with the
large operators moving to smaller units or
out of agriculture entirely. Further, the
measures of effliciency used here refer to
relative average factor productivities
across farm size; they do not represent,
nor allow one to deduce, the relative mar-
ginal productivities of any factor or of all
together. Finally, it must be remembered
that relative social factor productivities
reflect relative product prices. At present
the composition of output on small farms
is qjuite different from that of large ones,
with each specializing in those products in
which they have a "comparative advan-
tage." The more land redistributed from
large units to small ones, the greater the
output of the typically "small farm"
products and the less of the typically
"large farm" products; the resulting
change in relative product prices would
tend to diminish the present differences
in relative efficiency.8
Given the political restraints on land re-
distribution, the central issue may more
likely be the distribution of other factors
ancl the key datum the relative marginal
productivity of the "mobile" factors on
differeint farm sizes. In the case of the most
mol)ile credit policy should be aimed at
directing it to those farms where the
marginial product of capital is highest.9
Here agaill, since there is I1o necessary
8 Other factors might, of course, wvorkin the opposite
direction; for example, if there wverefewer large farms,
rural services might he better, the pop)ulation would
prolhal)lYbecome imioreedlucatedl,etc. Still, the general
exl)ectation wvoulldbe that the efficiencv differential
would (liminiishas redlistrib)utionoccurred, especially in
the short runi.
I If it may lie assumed that the impact of credit is to
increase the capital stock (rather than to increase labor
emp1}loved).
relation between a high average produc-
tivity of capital and a high marginal
productivity, detailed and different in-
formation from that provided above would
be necessary to provide -a base for policy.
Despite the impossibility at this time of
any very satisfactory interpretation of the
interrelationship between size of farm and
factor productivities, the data suggest
strongly that unless a solution is found
largely outside the sector10there is no quick
solution for the bad distribution of income
in agriculture which does not involve land
redistribution as an important component.
It seems unlikely that large farms will be-
come more labor intensive over time (the
opposite is more likely), and while there
is no doubt that very small farms can be
made more productive by improved tech-
nology and additional capital, it seems
doubtful that over the short run farms of
less than five hectares can provide what
might reasonably be considered a mini-
mum acceptable income level in Colombia.
REFERENCES
A. Berry, "Land Distribution, Income Distri-
bution and the Productive Efficiency of
Colombian Agriculture," Yale Econz.
Growth Centcr, Discussion Paper No. 108,
Mar. 1971.
and A. Padilla, "The Distribution of
Agriculturally Basecl Income in Colombia,
1960," mimeo, 1969.
L. Currie, Accelerating Developmnenlt:The
Necessity an1ldthe Mlcanis, New York 1966.
J. Grunig, 'Some Comparisons of Produc-
tivity and Efficiency of Large and Small
Farms," mimeo, 1969.
10There are many cogent arguments in favor of this
solution. See Lauchlin Currie.