The World’s Largest Open Access Agricultural & Applied Economics Digital Library

This document is discoverable and free to researchers across the

globe due to the work of AgEcon Search.

Help ensure our sustainability.

Give to AgEcon Search

AgEcon Search
http://ageconsearch.umn.edu
aesearch@umn.edu

Papers downloaded from AgEcon Search may be used for non-commercial purposes and personal study only.
No other use, including posting to another Internet site, is permitted without permission from the copyright
owner (not AgEcon Search), or as allowed under the provisions of Fair Use, U.S. Copyright Act, Title 17 U.S.C.
 For Restricted Circulation
to Seminar Participants only

BACKGROUND PAPERS FOR DISCUSSION

AT

THE INTERNATIONAL SEMINAR

ON

"COMPARATIVE EXPERIENCE OF AGRICULTURAL

DEVELOPMENT IN DEVELOPING COUNTRIES
SINCE WORLD WAR II"

NEW DELHI

25th, 26th, 27th and 28th OCTOBER. 1971

THE INDIAN SOCIETY OF AGRICULTURAL ECONOMICS

46-48, Esplanade Mansions, Mahatma Gandhi Road, Fort,
BOMBAY-I. .
J

AGRICULTURAL PRODUCTIVITY DIFFERENCES

AMONG COUNTRIES

Yujiro Hayami and V.W. Ruttan*

The source of productivity growth over time, and of

productivity differences ambng countries and regions have
emerged as a central unifying theme of growth theory and deve-
lopment economics. 1 In recent years a concensus seems to have
emerged to the effect that productivity growth in the agri-
cultural sector is essential if agricultural output is to
grow at a sufficiently rauid rate to meet the demands for
food and raw materials that typically accompany urbanization
and industrialization. 2 Failure to achieve rapid growth in
agricultural productivity can ·result either in the drain of
foreign exchange or in shifts in the internal terms of trade
against industry, and thus seriously impede the growth of
industrial production. Failure to achieve ranid growth in
labor productivity in agriculture can also raise the cost of
transferring labor, and other resources, from the agricul-
tural to the nonagricultural sector as development proceeds.
-- - - - - - - - -
* Extract. The American Economic Review, Vol. LX, No.5,
December 1970.
l J.R. Hicks has suggested that growth theory and deve-
lopment economics have no connection. This view wculd
seem to be invalid in view of Hicks' own criteria. See
A.O. Krueger.
2 See articles by Irma Adelman and Cynthia T. Morris,
Dale W. Jorgenson, Gustav Ranis and J.C.H. Fei, and
v.w. Ruttan.
 - 2 -

Extremely wide differences in agricultural producti-

vity exist among countri-es. Agricultural output per worker
in India is appro~imately bne-fiftieth of that in the United
States. Relatively few underdeveloped countries have achieved
levels of output per .wo'rker one-fifth as high as in the United
States. FurtheTmore, these differences have widened during
the last decade. 3 This l a g in the rate of productivity
grow~h irl; agrie1,1.lture represents a serious constraint·· on
econolfic growth in many developing economies. ' 1
Recent empirical research supports a classification of
the sources of productivity differences; or of produ~tivity
growth, into three broad categories, (a) resource
. ,. · e~dowments,
(b) technology, as embodied in fixed or working capit~l, and
(c) human canital, broadly conceived to include the .~du?ation:
' I

skill, knowledge and capacity embodied in a country's

population. Al though this 'i s clearly an oversimplification it
does represent a substantial advance · over the earlier emphasi s
on a single key or strategic factor. 4
Our analysis indicates that the three broad Cc;¼._tegories
,· outlined above account for approximately 95 per cent of the
differences in labor productivity in agriculture between a
reP,r~s.e nt,ati v.e group · of Less Level oped Countrias (LDC' s) '
and of Developed Countries (DC's). In this compari~on the
th_ree factors are of roughly equal importance. When. co~pared
to th.e DC' sr·· of recent settlement ( Australia_, Canada,. N53w
Zealand, and the United States) favorable resource endowments
account for sbmewhat more than one-third of the diffe-Tences •.

3 See Hayami and associates.

4 See studies by Zvi Griliches, A.O. Krueger, R.R~
Nelson, and T.W. Schultz.
J
 .>.

27
r

- 3 -

Resource endowment is the major factor accounting for differ-

ences in labor productivity between the DCts of recent
settlement and the older DC's. Nevertheless it seems apparent
that the LDC's could, over time, achieve labor productivity
levels in agriculture well over half as high as in the more
recently settled DC's, roughly comparable to the levels
achieved in the older DC's, through increased use of technical
inputs supplied from the industrial sector and improvements
in the auality of the labor force, even in the absence of sub-
stantial changes in man-land ratios.

!. The Method and the Data

The approach used in this study involves the esti-

mation of a cross-country production function of the Cobb-
Douglas type for thirty-eight developed and underdeveloped
countri~s. 5 Differences in agricultural output per worker
are accounted for by differences in the level of conventional
and nonconventional inputs per woTke-r, classified as ( a) interna:_
resource accumulation, (•) technical inputs supplied by the

5 Countries included are: Argentina, Austria, Australia,

Belgium, Brazil, Canada, Ceylon, Chile, Colombia, Denmark,
Finland, France, Germany, Greece, India, Ireland, Israel,
Italy, Japan, Mauritius, Mexico, Netherlands, New
Zealand, Norway, Peru, Philippines, South Africa, Spain.
Surina~, Sweden, Switzerland, Syria, Taiwan, Turkey, '
U.A.R., U.K., U.S.A., and Venezuela.

_)
28
- 4 -
J

nonagrict: __ture sector, and (c) human capita1. 6 All the

d ~tn us ed in this study are taken from a recent compila-
tion of international agricultural production statistics
by Yujiro Hayami and associates. 7
Production functions were estimated for three
different periods; 1955 (1952-56· averages), 1960 (1957-
62 averages), and 1965 (1962-66 average~). 8 The analysis
was conducted in gross output (net bf seeds and feed) terms
in order to include the effects of current intermediate .
inputs such as fertilizer. Individual agricultural commo-
dities were aggre~ated by the farm gate (or import) prices
of the United States, Japan and India, to produce three
different output series. The series were then averaged
geometrically into a single composite output seri e s wh ~ct.
-:; .r -:_ .-:- ~ - • ,·, ,. -. .... L ..:..:r .:.• , ;

6 For a report on a preliminary attempt see Hayami (l9G9,

1970). Maj or extensions from the previous study ~.11clude:
( a) a comprehensive revision of data; (b) introducti0n oi '
the livestock variable; (c) analysis on a per far.rr: basis
in addition to a national aggregat e basis; (d) test of
stability of the production function over tine· and (e) 1'~ -
finAments in the procedures used to account f or produ-
ctivity differences.
7 The basic data were collected from publications b~ the
United Nations organizations (FAO, ILO and UNESCO)~ tn e
Organization fer Economic Cooperation and De.,.rel.opment; ar.1.0
the governments of various c0untries. Th8se data ¼ere
processed by Hayami and associates to be consistent with
the definitions of variables and, also, to be comparable
among countries. Earlier estimates of agricultural c.,7.:t-
puts reported by Hayami and Inagi \01'3"-su hst;.;n..ti a:::.:i...y ·
revised fr:ir this study.
8 Averageswere taken for flow variables ( output and fer1, i 1 i , . ... .
input). Stock variables were in principle measur ed by l 0 5 -'S
1960, and 1965 levels. It would seem more consis ~cnt ~o
have averages of 1953-57, 1958-62, and 1963-67, b~t the
original estimates of agricultural output are of 1957-62 ·
averages ( see Hayami and associates) and, whe n we t1,iec to
extend the 1958-62 output series to 1955 -~d 106T th e
F-Aa ir.Zc.ex ,:ff a !7TI.."C.U-l : u:·hl rr').(]Uct-ion- Wa.S r,v:..·-Lla blc ···,:·11
until 1966.
r

,.
- 5 -

was used as the dependent variable. 9

The independent variables used in the study
include labor, land, livest9ck, fertilizer, machinery, ~du~
cation, and technical manpower. In sum~ing up the effects
of resource endowments, technology, and human capital on
productivity per worker, land and livestock ~0rve as proxy
variables for internal resource accumulation; machinery and
fertilizer for technical inputs; and general and technical
) education in agriculture for human capital.
Land(measured by hectares of agricultural Uand) used
for agricultural production cannot be regarded as a mere gift
of nature. It represents the result of previous investment
in land clearing, reclamation, drainage, fencing, and other
development measures. Similarly, livestock (as neasured by
livestock units) represents a form of internal capital
accumutation. T~us, in our perspective, land and livestock
represent a form of long-term capital formation embodying
inputs supplied primarily by the agricultural sector. 10
Both high inouts of land and of 11":'~ estock pe:- · wrirker t end t o OP
associated with low levels of labor and high levels of lanrt
per unit of output. In contrast, fertilizer (as measured by
the N + P2 o5 + 'K2o in commercial fertilizers) and maci'linery
(as measured by tractor horsepower) r epresent inputs supplied ··
by the industrial sector. Technical advances stemmin~ fro~

9 This procedure was applied for 1960 data. 1955 and 1S55
output estimates were extrapolated from the 1960 e~ti-
mates by using the FAO indexes of agricultural producticn
'h>y countries.
10 Perennial nlants belong to the same category of inputs
as livestock; but they are not included due to the lack
of data.
30 - 6 -

both public and private sector research and development are

embodied in or complementary to these modern industrial inputs.
Mechanical innovations are usually associated with larger
inputs of power and machinery. Biological improvements,
such as the inno·:ations embodied in high yeilding varieties,.
are typically associated with higher levels of fertilizer
u·~e. In this analysis these two industrial inputs represent ·
proxies for the whole range of inputs which carry modern l.'
mechanical and biological technologies.
The proxies for human capital include me~sures of
both the general educational level of the rural population
and specializea education in the agricultural sciences and
technology. Two alternative measures of the level of general
education were attempted: (a) the literacy ratio and (b) the
school enrollment ratio for the primary and secondary levels.
Both sets of data are deficient in that they apply to the
entire population and are not sensitive to differences in the
'
quality of rural and urban education. Education in the
agricultural sciences atld technology was measured by the
number of graduates per ten thousand farm workers fro m agri-
cultural faculties at above the secondary level. These
graduates represent the major source f technological and
scientific personnel for public sector agricultural
, r e search
and extension and for research development and marketing in -~·
the private agribusiness sector. 11
A critical assumption in this ap~roach is that the
technical possibilities available to agricultural produqers
in the different countries can be described by the sa~e
~ - ~ -------
11 In a sense this variable may be superior as the proxy
for the level of research and extension to the "state
average of public expenditure on research and extension
per farm" used by Griliches, because our variable reflect[
the research and extension activities in the private
sector as well as in the publi.c sector.
 I I

3
)

production function. Cross~section production functions,

using individual countries or regions as observations, have
been widely used. Cross-country aggregate production functions
for the agricultural sector were first estimated by Jyoti
Bhattacharjee in 1953. An aggregate agricultural production
function similar to that used in this study, using states
in the United States as observations was employed by Zv1
Griliches in an attempt to account for the impact of research
and education on agricultural output. Anne Kruegerls r ecent
efforts to estimate the contribution of factor endowment
differentials to variations in per capita income employs the
assumption that all countries are subject to a uniform pro-
duction function~
In a recent paper Richard Nelson has argued that the
assumptions of a common production function"··· get in the
way of understanding international differences in productivity--
particularly differences between advanced and underdeveloped
countrl,.es" (p.1229). Nelson's objections appear directed
primarily to the empirical results obtained from use of re-
latively primitive two-factor production functions, where
intercountry differencas in value-added per worker are related
to the capital-labor ratio. He insists, as a r~sult of
differential diffusion of new technolcgy, that".~. at any
given time one would exnect to find considerable variation
among firms with respect to the vintage o_f their technology,
certainl;rvbetween countries, but even within a country" (p.1230).
We share the Nelson perspective. -Agricultural producers
I

in different coun~ries, in different. regions of the same

country~ and on different farms in the same region are not all
oh the same micro-production function. This reflects differ-
ences among producers in their ab~lity to adopt new t e chnology.
More iI!lportantly, it is also the result of differential

-
 - 8 -

diffusion of agricultural technology, and, -to an even

greater degree, of differential diffusion of the scientific
and technical capacity to invent and develop new mechanical,
biological, and chemical technology specifically adapted
to the ~actor endowments and prices in a particular co'untry
or region.
We may call the envelope of all known and potentially
.. '
discoverable aotivities a s ecular or "meta-production
fynction". The full range of technological alte.rnatives
• I •
described by the meta-production function is only partially
available to individual producers in a particular country
or agricultural region during any part·icular _historical
"epoch 11 • 12 It is, however, potentially available to agri ...
cultural scientists and technicians.

12 In the short run, in which substitution between capital

and labor is circumscribed by the rigtdity of existing
capital and equipment, production relationships are best
des~ribed by an activity with relatively fixed factor- 7 ~. , - - ~
factor and factor-product ratios. In the 1ong run in
which the constraints exercised by existing capital dis-
appear and are renlaced by the fund of available technical
knowledge, including all alternative feas~ble factor-
factor and factor-product combinations, production re-
_. lationships can be adenuately described by the neoclassi-
cal production function. In the secular period ©f
production, in which the constraints given by the avail-
able fund of technical knowledge are further relaxed
to admit all potentially discoverable knowledge 1 pro-
duction relationships can be described by a meta-pro-
duction function which describes all potentially dis-
c0verable technical alternatives. The meta-production
function can be regarded as the envelope of neoclassical
production functions. Although the term is not employed,
the meta-production function concept is implicit in the
work of Murray Brown a.pd of W,E.G. Salter. We have dis~
cussed the rationale for the metaproduction function
concept in Japanese and U,S., agricultural development in
greater detail elsewhere (see Hayami and Ruttan)! The
elasticity of substitution among factors increases con~
tinuously as the time period increases from the short run
to the secular period.
 33
- 9 -
)

We view the common or cross-country production

function which we have estimated as a meta-production
function. It is assumed that the invention and diffusion
of a new "location specific" agricultural technolo gy through
the application of the concepts of physical, biological, :md
chemical science and of engineering, craft, and husbandry
skills, is capable of making the factor productivities impli-
cit in the cross-country production function available to
producers in less developed countries. It is also assumed
~hat the capacity of a country to engage in the necessary
research, development and extension is me asured by the two
proxy variables for human canital, namely general education
and technical education in agriculture. It appears to us
that this effort, and that of Griliches and Krueger, are not
incons.istent with the persnective presented by Nelson in hi s
criticism of the empirical results obtained from two fact or
cross-country production functions.
The production function employed in th1s study was of
the Cobb-Douglas type. It was used mainly because of its
ease in manipulation and internretation. A test presented
in the Appendix* indicates that the unitary elasticity of
substitution implicit in the Cobb-Douglas production function
Q is an acceptable assumption. The ordinary least squares
estimation procedure was used. The possibility of simul-
taneous equation bias seems small because all inputs, except
fertilizer, are measured in stock terms and can be treated as
predetermined. In a few cases, however, the method of in-
st1umental variables was tried to see if any different in-
ferences might be drawn. The assumption of a common produetion
function among countries is a testable hypothesis. However,

* Apnendix is not reproduced here.

 - 10 -

it appears that the data us.e d in this study are too crude
to be employed for such a test~3

II. Estimation, of the Production Function

. We conducted an especially detailed analysis for
1960 because of (a) better comparability of ,output data and
( b) availability of data for 't he number of farms in that
year. 14 • The estimation was made both on per farm data
(output and conventional inputs deflated by th~ number of
farms) and on national aggregate data. The results from these
two setsor data are not sufficiently different to lead to
different infer~nces regarding the agricultural production
struotu~e, among countries.

13 In order to test the assumption that farmers in different

countries face the same production function, the pro-
duction function was estimated separately for the two
different groups of countries (DC and LDC's). The esti-
mation was tried for various groupings of DC's and LDC's,
but the results are all implausible with most of the
coefficients statistically nonsignificant or negative
in sign. It seems that measurement errors in our obser-
vations (especially of nonconventional variables) are too
large to make it possible to estimate the influences of
variables for the groups of countries within which the .
ranges of data variations are relatively small. The basic
assumption is 7 therefore, not testable on the presently
qVailable data. All we can claim is that differences in
agricultural productivity among countries can be ex-
plained well -with this assumption.
14 The 1960 World Census of Agriculture provides the data of
the number of farms for a large number of c·o untries.
Comparable data are available for only a small number
of countries for 1955 and 1965. See also fn.9.
 35
) - 11 -

Considering the crudeness of data, the levels of

stltistical s1gn1f1canee of the est~~~ted c0efficients seem
satisfactory in most cases. The coefficients stay fairly stable
when nonconventional variables are added or subtracted, though
the coefficients for labor and livestock tend to move opposite
to the coefficient for machinery. The results of estimation by
the method of instrumental variables (denoted as IV) compared
with the least square estimates provide no prima facie
evidence against the use of least squares.

Attempts to include other variables, e.g., the ratio

of irrigation land to total land area and the ratio of
cropland to pasture land, were tried in an attempt ½O
adjUgt' for differences in the quality of land input; but
it turned out that the coefficients for such variables are
either negative or nonsignificant. 1 ~
Plausibility of the estimates may be checked by a
comparison with the results of earlier attempt~ to estimate
aggregate production fune.tions in various countries.
Bhattacharjee obtained aggregate productlon elasticities
for his cross-country production function (including only
conventional variables) centered on 1950 of around 0.3 for
labor; 0.3 to 0.4 for land; and 0.3 for fertilizer. The
coefficients for livestock and tractors were not significant
at commonly accepted levels. Th9 Bhattacharjee results
indicate higher production elastieities for land and fertili zer
than the results obtained in our study. It would appear th ~t

15 This does not necessarily mean that suoh variables

have no significant influence, but rather it means that
the presently available data are too crude to estimate
the influences of such variables.

_,
 , I

36 - 12 -

our model is somewhat better specified in that we obtained

statistically meaningful coefficients for livestock and
machinery as well as· for the two proxy variables' foT human
capital.
The aggregate produ~tion elasticities of U.S • .
agriculture were estimated by Griliches as 0.4 to 0.5 for
labor; 0.1 to· 0.2 for land, fertilizer and machinery; ~.3
to 0.5 for education; 0.04 to 0.1 for research and exten-
sion. It is rather surprising that the Griliches' e~ti-
mates, desnite , the completely different nature of the data
used, coincide so well with the ones in this study •
•
The production elasticities estimated for Japanese
agrieulture by Yasuhiko Yuize in value-added terms are in
the ranges of 0.4 to 0.6 for labor and 0.2 to 0.4 for land.
Such figures are consistent with the estimates in this
study since according to the social account study by the
Japanese Ministry of Agriculture and Forestry the ratio
of value-added to gro~s output was around 0.7 in Japanese
agricult~re in the period when Yuize 1 s study was made. In
the less developed countries we do not have comparable
estimates of the aggregate agricultural production function.
)
Theodore Schultz has, however, inferred from the impact
of the 1918-19 influenza epidemic that the production el~s-
ticity of labor in Indian agriculture was 0.4. This is con-
sistent with our estimates. Such consistency with other
studies gives suppoTt to the results of estimation in this
study.
Griliches has found that in U.S. agriculture, a given
percentage increase in education·., whic~ improves the quality
of labor, has the same o.u tput effect as an equal percentage
increase in labor itself. In order to test whether the
same assertion holds in the international dimension, we
have estimated the production function by combining labor
_)
 - 13 -

Land general education E in a ~ltiplicative form L x E;

this resulted in little change (compare regressions 2 with
4, 3 with 5, 7 with 9, and 8 with 10). Further~ore, the
analysis of variance provides evidence in support of the
equality in the coefficients of labor and general
education. 16 •
Judging from the sums of coefficients of conven-
tional inputs, compared with the standard errors of those
) sums (shown in parentheses below the sums of coefficients),
constant returns seem to prevail both on the farm firm level
and on the national aggregate level. Note, however, that
increasing returns prevail when both private and socially
controlled inputs are allowed to vary. The constant returns
at the farm firm level may explain the existence of farms
of extremely different sizes producing the same commodities ••
The constant returns at the national aggregate level might
be one of the distinctive characteristics of agricultural
production and, if so, would have important implications for
the intersectoral investment priorities for national eco-
nomic development.
The stability of the agricultural production function
over time is tested on the 1955, 1960, and 1965 cross-
country samples. Because comparable data on the number of
farms were not available for 1955 and 1965, we assumed the
linear ho~ogeneity in the Cobb-Douglas production function
and regressed output per capita (per male worker) on

16 The F-statistics calculated for testing the equality of

the labor and education coefficients are: 0.22 for
Regression 2 vs-. Regression 4; 0.31 for Regression 3 vs.
Regression 5; 0.65 for Re 6ression 7 vs. Regression 9;
0.77 for Regression 8 vs. Regression 10.
38
- 14 -

conventional inputs per capita and on non~conventional

. ,::-., -: ~:l.2:: ~ ..__ - -, -.,. I I. _., • -, •

inputs. In order. to make ihe ~ di,lta pomna.rabl..e.

among years we restricted the countries included in the
sample to 36 (Mauritius and Surinam were dropped from the
sample for lack of. labor data).

This appe·ars to be caused by high i.nteroorrelation

between land area, per worker and livestock per wt>rker.
Differences in the two sets of estimates do not seem to
imply different conclusions. The production parameters seem
largely stable over time. The null hypothesis of the
equality of the production coefficients among 1955, 1960,
and 1965 is accepted according to the results of analysis
of variance (the F-statistic calculated from Regressions
12, 13, 14, and 17 is only 0.95).

III. Accounting for Productivity Differ~nces

The results obtained from estimation of the agricul~
tural production function in ,t he previous sections may be
used to account for intercountry differences in labor pro-
ducti~ity (output per male worker) in agriculture in 1960).
Since our production function is now assumed to be
linear homogeneous (with respect to conventional:.j_n:put.s) -1.n
the Cobb-Douglas form, the percentage difference in output
per worker can be expressed as the sum of percentage diffe~-
ences in conventional inputs and non-conventional inputs
per worker each weighted by the relevant production elastici-
ties. Only the school enrollment ratio was used as the
education variable in this accounting, but the results would
have been essentially the satne it the· ·11 teracy ratio had
been used.
 \

3~
- 15 -

Two alternative- sets ofl"&9"U.lt'! are presented. The

first set involves group comparisons between LDC 1 s and DCts.
The second set involves individual comparisons of selected
LDC1s and DC's with the United States.

Group Comparisons
The sources of differences in la~or productivity
between the eleven LDC's and different groups of DC's are
) presented in Table 3. Each· column compares for each group
the percentage difference in agricultural output per w:>rker
between LDC's and DC's with the percentage differences in
input variables weighted by _the specified production elasti-
cities. Inside of the parentheses is shown the index with
the output-per-worker difference set equal to 100. The
countries classified as LDC's, for the purposes of this
comparison, all had per ca~ita incomeof less than 350 U.S.
dollars and more than 35 per cent of their labor force
engaged in agriculture. The countries classified as DC 1 s
had per capita income higher than 700 U.S. dollars and less
than 30 per cent of the labor force engaged in.agriculture.
Countries falling between these criteria are not included
in the comparisons presented in Tahle 3. ·
The difference in average agricultural output per
worker between the eleven LDC's and the thirteen DC's of
group 1 was 88.8 per cent; the difference between the eleven
DC's and the nine older DC's of group 2 was 83.5 per cent;
and the difference between the eleven LDC's and the f0ur
DC's of recent settlement--group 3--was 93.6 per cent. The
six variables included in the production function accounted
for 95, 85, and 96 per cent of the difference in apricul-
tural output per worker between the LDC's and the three
DC 1 s groups.
 - 16 -

TABLE 3f: ACCOUNTING FOR DIFFERENCE IN LABOR PRODUCTIVITY IN

AGRICULTU ETWEEN DEVELOPED COUNTRIES DC AND
LESS DEVELOPED COUNTRIES LDC AS PERCE OF THE
LABOUR PRO CI ITY OF DC
-------------------------------------------------==~-----------
Group 1 Group 2 Group 3
( 13 DC 1 s) (9 DC 1 s) (4 DC 1 s)
---------------------------------------------------------------
Difference in output per . ,
male worker - per cent 88.8 (100)* 8~,5 (lQO) 93.6 (100)
' \II
Percent of difference I

explained: Total 84.2 (95) 71.1 (85) 90.0 (96)

Resource accumulation: · 29.2 (33) 17.5 (21) 32.6 (35)
Land 9.2 (10) 1.8 ( 2) 9.7 (10)
Livestock 20.0 (23) 15,7 (19) 22.9 (25)
Technical inputs: ~.3 (27) 24.3 (29) 24.5 (26)
Fertilizer 14.5 (16) 14.5 (17) 14.6 ( 16)
Machinery 9.8 (11) 9.8 (12) 9.9 (10)
Human Capital: 30.7 (35) 29.4 (35) 32.9 (35)
General Education 18.2 (21) 17.6 (21) 19.5 (21)
Technical Education 12.s (14) 11.7 (14) 13 •.4 . (14)
~==================•s===============~======•====~===a=========~
* Inside of parentheses are percentages with output per worker
set equal to 100.
LDC: Brazil, Ceylon, Colombia, India, Mexico, Peru, Philippines ,
Syria, Taiwan, Turkey, UAR.
DC : Austrialia, Belgium, Canada, Denmark, France, Germany,
Netherlands, New Zealand, Norway, Sweden, Switzerland,
U.K., U.S.A.
Group 1 includes all DC's;
Group 2 sxcludes Australia, Canada, New Zealand, and the
United States ffrom DC's;
Group 3 includes . only the four DC's excluded from Group 2.
Accounting formula:
l'
/'Yd - Yl0 = 0.1~ ad - a1 ) + 0.25c·sd - s 1 )' + 0.15 / fd - r 1)
~ Yd ad sd ~ fd 7 .
+ 0.10 (.md - m1) + 0.40 / _Ed - ~1 )+ 0.15 ( ud - u1 }
. \. md ~ Ed ) \_: Ud .
where y, a, s, f, m, are respectively, output, land, livestock,
fertilizer, machinery per male worker; E and U are, respectively,
· 'the general education (school enrollment ratio) and the technic nl
education va:riable; lower case letter d denotes DC and 1 denote s
LDC.
$ Tables 1 ~d 2 are not reproduc ed here.
 I' .\:

- 17 -

In the comparison between the elevBn LDC 1 s and

the thirteen DC's - group l - each generalized category,
internal resource accumulation (land and livestock), technical
inputs from the industrial sector (fertilizer and machinery),
and human canital (general and technical education in agri-
culture), account for approximately one-third of the explained
difference in labor productivity.
The main difference between group 1 and the other
) two groups is the amount of the difference explained by land.
Difference in land accounts for only 2 percent of the
difference in labor productivity between the LDC's and the
older DC's, while it accounts for 19 per cent between the
LDC's and the new DC 1 s. This implies that it should be
feasible for the LDC's, even with the present land-labor
ratios to achieve levels of productivity per worker roughly
equivalent to the labor productivity levels achieved by
workers in the older DC's - that is, roughly four times as
high as present LDC levels and well over half the level
achieved by the DC's of recent settlement. The critic~l
elements in achieving such increases in labor productivity
are the supply of modern industrial inputs in which the new
technology is embodied ar.d the investment in general education
and in research and extension which raises the capacity to
develop and adopt a more productivity technology.

Comparison of group 2 and 3 results does indicate

that resource endowments, particularly land, do represent a
serious barrier to efforts of both that LDC's and the older
DC's to achieve levels of output per worker comparable to
the levels currently enjoy~d in the more recently settled
DC 1 s. This is the first time, to our knowledge, that the
economic advantage of the favourable resource endowments in
these countries has been demonstrated ouantitatively.
 - 18 -

Individual Comparisons

The individual country comparisons presented in

Table 4 were developed in order to provide somewhat deeper
insight into the sources of differences in labor productivity
between different "ideal type" DC's and LDC's and the United
States. Each now compares the percentage dtfference in agri-
cultural output per worker between each country and the
United States with the linear combinations of percentage )
differences in input vari ables weighted by the specified
production elasticities. Inside of the parentheses is the
index with the output-per-worker differences set equal to
100. In general, the results are consistent with the group
comparisons.
In the four underdeveloped countries - Indi a ,
Philippines, United Arab Republic, and Colombia - inte rnal
resource accumulation accounts for approximately one-third
and technical inputs roughly one-fourth of the differences.
Human capital accounts for more than one-third of the differ-
ence between the United States and India, the United Arab
Republic, and Colombia. In the Philippines, which has
achieved a relatively high level of schooling and produces
a relatively large number of agricultural college graduat e s,
human canital explains less than one-fourth of the pro-
ductivity difference. The contrast between India und the
Philippines in this respect is quite striking.
In the comparisons between the countries of Europe ~n~
the United States, differences in internal resource accu-
mulation represent the most significant source of difference
in labor productivity, The constraint of land on agricul-
tural productivity is relatively modest for the Unit ed
Kingdom which experienced the drastic agricultural transfor-
mation after the repeal of the Corn Law; it is stronge st for
)
- 19 -

TABLE 4: ACCOUNTING FOR LABOR PRODUCTIVITY DIFFERENCES FROM

THE UNITED STATES AS PERCENT OF U.S. LABOR PRODUCTI-
VITY, ll SELECTED COUNTRIES
===============-===- -----=e:============-====-====-=-========--==-
Difference Percentage of difference
in output ex:Qlained by::
per worker Total Resource Technical Human
from U.S.as accumu.la- inputs capital
percent of tion (land (ferti- ( gene-
u. s. and live... lizer ral and
stock and ma- techni-
) chinery) cal edu-
cation)
-----------------------------------------------------------------~
LDC
India 97.8 102.1 32.7 25.0 44.4
(100) a (104) (33) (26) (45)
Philippines 96.2 82.1 33.4 24.9 23.8
(100) ( 85) (34} (26) (25)
UAR 95.6 "'97.0 33.8 24.6 38.6
(100) (101) (35) (26) (40)
Colombia 89. 7 89.4 25.8 24.7 38.9
(100) (100) (29) (28) (43)
EUROPE
Den.mark 52.3 51.0 20.4 13.2 17.4
(100) (97) (3P) (25) (33)
Netherlands 56.6 51.7 25,0 15.0 ll.7
0 (100) (91) •(44) (26) (21) _
United Kingdom 55.8 50.2 18.2 13.4 18.6
_(100) (90) (33) (24) (33)
France 63.9 64.3 26.2 16. 5 21.6
(100 (101) (41) (26) (34)
JAPAN 89.2 66.0 34.1 22.4 9.5
(100) (74) ( 38) (25) ( 11)
PASTORAL FARMING
Argentina 60.0 45.9 -4.8 24.3 26.4
(100) (76) (-8) (40) (44)
New Zealand -42.4 -49.l -55.2 2.7 3.4
(100) (116) (130) (-6) (-8)

=========================-========================================
a Inside parentheses are percentages with output per worker
differences set equal to 100.
 - 20 -

France which preserved peasant farms by protective tariffs.

Increase in the use ~f technieal inputs ·and improvements in
the ~uality of human capital canbring labor productivi~y of
the several European countries closer to the U.S. level. Never-
theless it seems ap~arent that major advances in labor pro-
ductivity in European agriculture ·(especially in countries
like France) toward the U.S. level are dependent on the
absorption of a higher percentage of the agricultural labor
force into the nonagricultural sector. The Japanese case
is similar to the European, except_ that ,! ~pan, . characterized by
a stronger constraint of land, has moved further toward the
exhaustion· of productivity differential.a,; associated with
investm~nt in education and research. In our judgement the
model· underestimates the significance of the land ·constra int in
the Japanese case and, to a lesser degree, in the European
r,ase. Without a significant increase in land are a pe r worker
it would be impossible for Japanese agriculture to incr eas e
technical inputs (especi ally machinery) to the U.S. l evel.
The two pastoral farming cases are of p articular
interest. In spite of low levels of technical inputs, labor
productivity in Argentina is roughly . comparable to that in
Europe. -This is due almost entirely to a f avourable man-land
ratio comparable to that in the United S~ates. Argentina
has, as a result of under-inve stment in technology and human
capital, tailed to .f~lly exploit its favourable man-land
ratio. New Zealand, in contrast, has achieved a level of
labor productivity well above the U.S. level (the highest
in the world) by complementing its f avourable resource en-
dowments with high levels of technical inputs and investment
in education and research.
The results obtained in both group and individual
comparisons are somewhat different than those obtained by
Krueger. Using a different methodology, Krueger found that
 45
) - 21 -

human capital exvlained more than half the difference in

income levels between the United States and a group of. less
developed countries. This is in contrast to our studies in
which human capital explains approximately one-third of the
difference in labor productivity. Krueger's results apply
to the entire economy and ours to only the agricultural
sector. It seems reasonable to expect that resource endowments
would be of relatively greater significance in the agricul-
tural sector than in the total economy. We see, therefore,
no inconsistency between our results and those obtained by
Krueger, In general the consistency between the results pre-
sented in Tables 3 and 4, combined with our general know-
l edge of the economies being studied, strengthens our con-
fid ence in the methodology employed in this study.

IV. Implications for Agricultural

Development Strategy.

The implications of this analysis for agricultural

development strategy in the less developed countries have
both encouraging and discouraging aspects. It is clear that
output per worker in the several LDC's can be increased by
several multiples, while land area per worker remains con-
stant or even declines slightly. To achieve increases of this
magnitude ~;11 require substantial investment (a) in rural
education and (1) in the physic~l, biological, and.social
sciences. The latter is reauired for the technical and
institutional infrastructure needed for the invention-, dev -
lopment, and extension of a more efficient agricultura-1
technology. It will also reauire the allocation of substan-
tial resources to the production of the technical inputs
supplied by the industri al s ector, by which new technology is
carried into agriculture. By and large , these change s
)
.
 - 22 ...

achieve the higher levels of output per worker through

increase in output per unit area.

A more discouraging aspect of this analysis is that

in order to achieve levels of labor productivity comparable
to the levels achieved in the DC's of recent origin it will
be necessary to complement those technical changes de signed
· to increase output per unit area with technologies that
· reduce the labor input per unit area. Significamt r eduction
in labor input per unit area is likely to occur, however,
only in those economies in which urban-industrial development
is sufficiently advanced to absorb not only the growth in
the rural labor force but also to permit a continuous redu-
ction in employment in rural areas. 17 It should be noted
that this has occurred in Japan only since World War II. In
most LDC 1 s it seems likely that. the agricultural labor force
will continue to expand more rapidly than the nonagricultural
demand for labor from rural areas.

The i~plications for agricultural development

strategy for most less developed countries seem relatively
clear. An attempt must be made to close the gap in the level
of modern· industrial inputs and in education and research.
Agricultural surpluses generated by closing the gap, over and
above the amount necessary to maintain the growth of agri-
cultural productivity, must be used to fina~~e industrial

19 See the article by Folke Dovring.

 - 23 -

development t-8

Maintenance of the rate of growth of agricultural

productivity can be expected to impose a substantial drain
on the savings that can be generated from the agricultural
surpluses. Initially a substantial component of industri.a l
capacity must be designed to provide technical inputs for
the agricultural sector. Substantial investment will be
needed to create the institut+onal infrastructure to improve
) general education in rural areas and to produce the technical
~and scientific manpower needed to bring about technical
cho11ges in agriculture. Investment in land ~evelopment,
such as irrigation and drainage, will also be necessary in
a number of countries in order to obtain a full return from
the new biological and chemical technology.

If successful, the effort would, over time, result

in a rate of growth in the nonagri~ultural labor force suffi-
cient to permit a reduction in the agricultural labor forc e
and a rise in labor productivity toward the levels of the
DC 1 s of recent settlement. Clearly the process outlined here
is inconsistent with the low cost route to agricultural
development that seemed to be opened up by the dual economy
models which have dominated much of the theoretical dis-
cussion of agricultural development during the last decade.

18 Shigeru Ishikawa has suggested that achievement of

national agricultural output and productivity objectives
may, in some developing countries, require a net flow
of savings from the nonagricultural to the agricultural
sector, The possibility has been such a shock to
some students of development economics that they
recommend a "development without agriculture" policy
( e .g., M.J. Fl~nders).

C
 - 24 ...

REFERENCES

I. Adelman and C.T. Morris, An Econometric Model of

11

Socio-Economic and Political Change in Under"

developed Countries", Amer. Econ. Rev., Dec. 1968,
58, 1184-218.
K.J. Arrow, H.B. Chenery, B.S. Minhas, and R.M. Solow?
"Capital-Labor Substitution and Economic Efficiency",
Rev. Econ. Statist., Aug. 1961, 43, 225-50.
J.P. Bhattacharjee, "Resource Use and Productivity in
World Agriculture'', J. Farm Econ., Feb. 1955, 37,
57-71.
M. Brown, On the Theory and Measurement of Technological
Change, Cambridge 1966, 43-62.
F. Dovring, "The Share of Agriculture in a Growing
Population", Mon. Bull. Agr. Econ. Statist., (FAO)
Aug-Sept. 1959, 8, 1-11.
M.J. Flanders, "Agriculture Versus Industry in Development
Policy: The Planner's Dilemma Re- examined",
J. Develop, Stud., Apr. 1968, 5, 171-89.
Z. _G riliches, "Research Expenditures, Education, and the
Aggregate Agricultural Production Function",
Amer. Econ, Rev., Dec. 1964, 54, 961-74.
Y. Hayami, "Sources of Agricultur'.11 Productivity Gap
.Among Selected Countries", Amer. J. Agr. Econ.,
Aug. 19€9, 51, 564-75.
--------, "On the Use of the Cobb-Douglas Production
Function in the Cross-country Analysis of Agri-
cultural Production", Amer, J. Agr. Econ~,
forthcoming.
-------- and K. Inagi, 11 International Comparison of
Agricultural Productivities", Farm Economist, 1969,
· 11 (No.10), 1-13.
-------- and V. W. Ruttan, "Factor Prices and Technical
Changes in Agricultural Development: ' The _Unit ed
States and Japan, 1880-1960", J. Polit, Econ.,
Nov. 1970, 74.
) - 25 -

Y. Hayami, in association with B. Miller, w. Wade and

S. Yamashita, An International Comparison of
Agricultural Production and Productivities, Univ.
Minn. Agr. Exp. Sta. Tech. Bull., No.277, St.
Paul, 1970.
J.R. Hicks, Capital and Growth, Oxford 1965.
S. Ishikawa, Economic Development in Asian Perspective,
Tokyo 1968.
D.W. Jorgenson, "The Development of a Dual Economy", Econ.
l•, June 1961, 71, 309-34.
H. Kaneda, "Substitution of Labor and Nonlabor Inputs and
Technical Change in Japanese Agriculture", Rev.
Econ. Statist, Dec. 1968, 47, 163-71.
A.O. Krueger, "Factor Endowments and Per Capita Income
Differences among Countries", Econ. J., Sept. 1968,
?e-, 641-59.
R.R. Nelson, "A Diffusion Model of International Producti-
vity Differances in Manufacturing Industry",~-
Econ. Rev., Dec. 1968, 58, 1219-48.
G. Ranis and J.C.H. Fei, "A Theory of Economic Develop-
ment", Amer, Econ, Rev., Sept. 1961, 51, 533-64.
V. W. Ruttan, "Growth Stage Theories, Dual Economy Models,
and Agricultural Development Policy", Publication
No.AE 1968/2, Univ. Guelph 1968.
W.E.G. Salter, Productivity and Technical Change,
Cambridge 1960.
T.W. Schultz, Transforming Traditional Agriculture,
New Haven 1964.
Y. Yuize, "Nogyo ni okeru Kyoshiteki Seisankansu no
Keisoku" (The Aggregate Production_ Function in
Agriculture), NogYo Sago Kenkyu, Oct. 1964, 18,1-54.
Japan, Ministry of Agriculture and Forestry, Nogyo oyobi
Nakano Shakai Kanjo 1967, (Social Accounts of
Agriculture and Farm Households) Tokyo 1968.
United Nations, 1960 World Census of Agriculture, FAO,
New York.