The Economic Journal, 86 (March 1976), 45-58

Printed in Great Britain

THE SUBSTITUTION OF LABOUR FOR

CAPITAL IN KENYAN MANUFACTURING1

A major concern of governments of LDCs has been the growing level of urban
unemployment. Particularly disappointing has been the failure of manufactur-
ing employment to grow nearly as quickly as industrial output. The most
widely accepted explanation of the rapid growth in labour productivity has
been that capital deepening has occurred, and most of the literature has dealt
with the proximate causes of this phenomenon. Some economists allege that
growth in capital-labour ratios is inevitable as virtually no efficient factor
substitution is possible in most industrial processes.2 Others have argued that
efficient factor substitution is, in fact, feasible, but that incorrect market prices
of primary factors have conveyed the wrong signals to entrepreneurs.3
Surprisingly, in view of the considerably different policy implications of
these arguments, there has been relatively limited empirical examination of
production relations at the firm level.4 Most studies have focused on sectoral
aggregates (all manufacturing) and have used indirect C.E.S. estimating
procedures. However, the strong assumptions required for this procedure to be
valid, and their probable violation in the LDC context, suggest the need for
more disaggregated analyses using more direct evidence.5 T'o obtain the latter,
a series of interviews and visits to plants in Kenya were undertaken. Before
presenting the evidence from these interviews, Section I analyses aggregate
data which are helpful in providing a framework within which the micro data
can be interpreted. Section II describes salient aspects of the observed produc-
tion processes. Section III suggests some generalisations on the evolution of
labour productivity based on the evidence of Section II. Section IV considers
the role of managers and Section V contains conclusions.

1 I have received helpful comments on earlier drafts from Frank Child, Nathaniel Leff, Janet
Rothenberg Pack and W. B. Reddaway. Portions of this research were financed by funds provided
by the Agency for International Development to the Economic Growth Center, Yale University, under
contract CSD 12492. However, the views expressed in this paper do not necessarily reflect those of
AID.
2 For a recent statement of this position see Helen Hughes, "Industrialization, Employment and
Urbanization", Finance and Development, 1971, no. I.
a For a careful exposition of this view see Gustav Ranis, "Output and Employment in the 70's:
Conflicts or Complements", in Ronald Ridker and Harold Lubell, Employment and Unemployment
Problems of the Near East and South Asia (Vihas Publications, Delhi, I97I).
4 A survey of the existing literature including some recent studies in the same spirit as ours is given
by David Morawetz, "Employment Implications of Industrialisation in Developing Countries: a
Survey", ECONOMIC JOURNAL, September 1974.
5 For an excellent survey of the problem see Shankar Acharya, "Fiscal/Financial Intervention,
Factor Prices and Factor Proportions: a Review of Issues", Staff Working Paper 183, I.B.R.D.,
Development Economics Department.

[ 45 ]

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 46 THE ECONOMIC JOURNAL [MARCH

I. AGGREGATE PRODUCTIVITY GROWTH

For the period I964-7I, data for the entire Kenyan manufacturing sector
suggest little, if any, growth in the capital-labour ratio. As there are no
estimates available of the capital stock in manufacturing, I have considered
the implication of a range of assumptions. Table I shows the estimated rate of
growth of the constant price, fixed capital stock (net of depreciation) when the
initial ratio of this stock to (gross) value added originating in manufacturing
(on national accounts definition) is alternately assumed to have been IPo, I 5
and 2o and the straight line rate of depreciation is assumed to be 0-05 or 0-07.

Table I

Calculation of Rate of Growth of Real Fixed Capital Stock* in Manufacturing

Initial capital-output ratio:

Depreciation rate Io0 1-5 2-0

Rate of growth (% per annum)

005 11-2 7 1 4 7
0o07 91I 4.9 2.3

* Net of straight-line depreciation applied to initial stock and to new investment. The underlying
value added and fixed price investment series were obtained from national accounts data in Republic
of Kenya, Statistical Abstract, various issues.

My own guess about the most accurate assumptions would be a capital-

output ratio of I15 and a depreciation rate bracketed by 0o05 and 0-07. This
implies a range of growth of the real capital stock between 4.9 and 7 I % per
annum compounded.' Comparing these figures with the manufacturing
employment growth rate of 6-5 % there was, at most, a small degree of capital
deepening and it is possible that capital shallowing occurred. Over the period
considered, real output grew at a compound rate of 8-2 % per annum, so that
average labour productivity was increasing at a rate of I -6 % per year, a smaller
rate than that observed in many other LDCs. The allocation of this growth
between "technical progress" and capital deepening depends upon which
capital growth rate is assumed to be correct and the output elasticity with
respect to each factor. If 7.I % is used for the capital growth rate, and the
elasticity of output per head with respect to capital per head is assumed to be
0-5, assumptions favourable to the capital deepening explanation, about I8 %
of the growth in labour productivity was due to such deepening.2 If the lower
1 A recent study by Raymond Powell builds up the capital stock for the entire non-agricultural -
non-housing sector from investment series extending back to I931. His estimate of the rate of growth
of the real net capital stock in the I964-71 period for this broadly defined sector is 5-8 %. See "The
Stock of Fixed Capital in Kenya, I964-71 ", mimeographed, Institute for Development Studies,
Nairobi.
2 Of course, it would be possible to alter these results by assuming a range of rates of capital-
embodied progress. However, as will be reported below, much of the equipment investment consisted
of used machines and most new factory structures are simple enclosures which do not embody new
designs such as those permitting easier internal movement of material. Although some embodied
technical progress undoubtedly occurred, it is unlikely that it was significant.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 1976] SUBSTITUTION OF LABOUR FOR CAPITAL IN KENYA 47

capital growth estimate of 4-9 % is used, labour productivity grew at the same
time that the capital-labour ratio declined. The sector-wide data in manu-
facturing thus suggest that, given the absence of dramatic shifts in the sectoral
composition of manufacturing, microeconomic evidence at the firm level
should be consistent with little, if any, capital deepening yet sustained labour
productivity growth.
The growth figures analysed do not, however, indicate anything about the
levels of capital intensity. Using the (net) capital stock data at the beginning of
I97I implied by the same set of plausible range assumptions as before (an
initial capital-output ratio of I *5 and depreciation rates of o05 and 0.07), the
capital-labour ratios for I97I are K?896 and KL772, respectively. (In I97I
the Kenyan pound was pegged at I * I 7 British pounds.) To obtain a benchmark
against which to measure these figures an aggregate capital-labour ratio for all
of Kenyan manufacturing was calculated by weighing the Kenyan distribution
of value added in I97i by U.S. I939 capital-labour ratios and adjusting this
to a I97I price base. The weighted average is KL6,I 75.1 Even recognising the
difficulties involved in such comparisons, it appears that the capital intensity
of Kenyan manufacturing in I97I was a fraction of that in the United States
in I939. This finding is consistent with direct evidence on the capital intensity
of individual Kenyan firms included in the survey.

II. CURRENT PRODUCTION METHODS

The industries in which I conducted interviews cover a considerable range

within the manufacturing sector, including inter alia all branches of food
processing, paints, shoe polish, soap, shoes, cement, textiles, metal and plastic
containers and home toiletries. In total, 42 plants were visited, of which 4I
had more than 5o employees. Firms of this size accounted for 82 % of manu-
facturing value added in I 967 and the industries represented in the sample did
not differ greatly from the overall pattern. Most of the following is applicable
to all branches: exceptions will be noted explicitly.
In contrast with much of the literature, visits to plants show that, at least
in this economy, many widely accepted views of the output-employment
relation are erroneous in a number of important dimensions. For the most part
these factories use rather simple labour intensive technologies; plant managers
were invariably able to indicate more capital intensive processes that were
available, but whose installation was not currently profitable given the level of
wages and volume of output.
The labour intensity of individual factories may be demonstrated in a
number of ways. For example, detailed data for individual plants are available
from a set of recent United Nations publications.2 A comparison of these with

1 The detailed U.S. capital-labour ratios were calculated from tables C-5 and C-6 in Anne P. Carter,
Structural Change in the American Economy (Cambridge, Harvard University Press, 1970). Capital excludes
all working capital. The investment price deflator since 1939 was obtained from U.S. Department of
Commerce, Survey of Current Business, various issues. Dollars have been converted to Kenyan pounds at
the official exchange rate of 2-8 dollars per pound.
2 UNIDO, Profiles of Manufacturing Establishments, three volumes. I have analysed the UNIDO

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 48 THE ECONOMIC JOURNAL [MARCH

the data I collected shows much lower fixed capital-labour ratios (capital being
measured exclusive of buildings), in Kenya in the early I970S than in Japan
and in many Indian plants a decade earlier. For example, the average capital-
labour ratio in three Kenyan plants producing textiles was K/73o1 - less than
several Indian ones, though the latter face lower wages. Similar results hold for
all other industries in which comparisons were made.
Apart from factor intensities it is possible to examine suggested methods of
production in engineering textbooks.2 The current descriptions of desirable
production techniques are of interest chiefly for their complete absence from
most Kenyan plants.

Processing by Stage
All plants can be characterised by five basic operations: material receiving,
processing, material handling among processes, packaging, and storage of the
finished product.3 Considerable emphasis in the literature has usually been
placed upon the possibilities of substitution in processing.4 However, in most
plants a relatively small percentage of the labour force is involved in such
operations (Table 2, column 7). The high ratio in food processing reflects
the presence of two companies with unusually high ratios of workers in pro-
cessing relative to other activities while in textiles the ratio is intrinsic to the
process. In others the ratio was below 40 %. Most employment occurs in the
auxiliary activities and it is there that the potential payoff to substitution of
labour for capital is of the greatest significance.
Where the processing operation involves mixing and/or heating (e.g. paints,
much of food processing, soap, shoe polish) the possibilities of economically
rational substitution of labour for capital appear to be limited, since reversion
to older methods of human, rather than automatic control, is likely to reduce
quality and uniformity while adding only a small number of workers to the
labour force.5 For example, a soap factory which recently introduced such
equipment to ensure better quality experienced a decline of only two workers

data in detail in " Capital-labour Substitution - A Microeconomic Approach ", Oxford Economic
Papers, November I974. The capital stock data are valued at replacement cost and capital-labour
ratios were corrected for differences in the number of shifts worked.
1 As the Kenyan data were obtained during 197I and 1972, the replacement estimates given in the
profiles were adjusted to a 1972 base using a variety of data sources.
2 For example, F. H. Slade, Food Processing Plant, vol. i (London, I967).
3 A similar scheme was used by Harry Jerome in his Mechanization in Industry (New York, National
Bureau of Economic Research, 1934). This monograph summarises extremely careful studies of the
process of substituting capital for labour in a number of industries and has considerable relevance for
the LDCs. I am indebted to Simon Kuznets for this reference.
4 This, despite the presence in the literature of discussion about the possibilities of substitution in
"ancillary" or non-processing activities. For example, David Granick, "Economic Development and
Productivity Analysis: the Case of Soviet Metalworking", The Quarterly Journal of Economics, May 1957;
Gustav Ranis, "Factor Proportions in Japanese Economic Development", American Economic Review,
May 1957, and W. B. Reddaway, The Development of the Indian Economy (Homewood, Richard D. Irwin,
I962), pp. 74-6-
5 There were some interesting exceptions to the general rule of a lack of substitution possibilities in
processing. One of the soap manufacturers, for example, used a detergent process which is completely
different from the conventional one. Instead of soap "noodles" (the solid, wet soap which is produced
by mixing) being dried in large metal towers by gusts of hot air, the alternative process utilises a

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
I976] SUBSTITUTION OF LABOUR FOR CAPITAL IN KENYA 49

out of a total labour force of 60,1 though the new machine replaced two of
four workers in processing. The absence of a larger percentage decline reflects
the small proportion of labour engaged in processing. However, in plants that
are prematurely mechanised or in which the production process intrinsically
allows relatively few workers in auxiliary activities (e.g. textiles), the instal-
lation of modern processing equipment may have much more severe effects on
total employment.
None of this is meant to convey the image that existing Kenyan plants use
only the most modern processing equipment. Almost all plants began produc-
tion with a large proportion of used machinery.2 Moreover, even when pur-
chasing new equipment, many firms buy older models where these are still
available.
Table 2

Technical Characteristics of Firms, by Industry

Employment in
Percentage of firms using automatic: processing as
-m - proportion of
Conveyors or Mean total employment
Processing Filling Packing piping of wage WA (Av. of firm
equipment machines machines raw materials (W) W means)
( I (2) (3) (4) (5) (6) (7)

Food processing IO 28 10 10 308 I-22 0o52

Bottling 100 100 0 100 355 I *00 o i8
Paints 0 33 33 0 354 104 o028
Other chemicals 20 30 0 20 38i 1-07 o026
Textiles, clothing
and misc. final
products 67 NA 0 0 300 III o-68
Plastic containers o NA o o 463 1-05 0o40

NA, not applicable.

Notes: WA is the average wage in those firms in which one or more activities have been mechanised.
If a firm used any automatic machinery in a given phase of production it was considered to be automated
in that stage.

simple bin-drying procedure. The equipment is a third as costly as spray drying. However, only three
additional workers (out of a plant labour force of 50) are required as a result of the difference in process.
Other cases of substitution can also be cited. For example, another soap factory used an old railroad
wood-fuelled steam-engine to obtain heat for the mixing of basic materials. This added two jobs in
the plant for wood cutters. Perhaps the most startling exception occurred in a factory manufacturing
plastic containers in which six men were assigned to manually insert the tops of plastic caps. Although
a machine exists for this operation, it is very fast and suitable for only a limited range of sizes. In the
absence of large production runs for a given size, installation of such equipment was not calculated to
be economically efficient.
1 The question of whether the added quality is warranted is of some importance. While it could be
argued that for domestic consumption it is a luxury and heavy excise taxes should be levied to dis-
courage such consumption, existing world export markets require the higher level of uniformity. This
suggests obvious areas of mutual interest among LDCs to encourage intra-LDC trade. For extensive
discussion of such issues see Frances Stewart, "Technology and Employment in Less Developed
Countries", mimeo, Quieen Elizabeth House, Oxford.
2 Moreover, the purchase of used equipment confers a substantial benefit in so far as its lower price
requires less of scarce investment funds (and foreign exchange) to undertake a given level of production.
This assutmes that the lower price of equipment does not solely reflect physical depreciation or else

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
50 THE ECONOMIC JOURNAL [MARCH

We now consider the auxiliary plant activities, namely receiving, material

movement, filling (packing) and storing. In all plants, to take the operations
in order, receiving is done by hand, trucks being manually unloaded and
material moved to storage places from which it is carried by hand on to
the production floor. In intra-factory material movements among primary
machines there is little further scope for increasing labour intensity. Although
we expected to find considerable reliance on the use of conveyors, there were
only a few in use (Table 2, column 4) and these were primarily in processes in
which a conveyor is intrinsic - for example, the movement of crushed pine-
apple for short distances. Even quite heavy jobs where automatic methods
might be profitable (if batches were large) are done manually. For example,
after paint ingredients are weighed, they have to be raised to a platform from
which they are emptied into a mixer. In all plants except one this was done by
hand, two men carrying the drum rather than using either an electric hoist or,
as is done in western countries, piping the ingredients to the platform.
Filling operations generate considerable employment in many of the plants
visited. One finds relatively primitive operations rather than the automatic
filling operations suggested in some of the development literature (Table 2,
column 2). In a number of paint plants the mixed paint was poured into cans
by tipping the vat and using a funnel. In pineapple plants, cans are hand-
filled. Labour-intensive methods also extend to labelling. For the most part
labels are attached by hand, though automatic labelling is the rule in equivalent
western plants.
Finally, in the storage process only Io % of the firms use fork-lift trucks to
stack the final product in warehouses. In most of these it was due to the heavy
weight of drums' or, more rarely, to a desire to increase the height of vertical
storage, thus postponing the need to acquire further warehouse space; this
suggests the possibility of a building (space)-labour trade-off.
Every manager interviewed suggested that automatic filling and labelling
were the technically easiest operations to introduce. The failure to do so was
attributed to output levels which are so small that such equipment, given its
productivity, would lie idle much of the time. Reinforcing this was the fact that
many products are produced in small batches and require different container
sizes, while most automatic machines have limited size-flexibility.2 At consider-
ably higher output levels, estimated between three and ten times current
levels, it was argued that automatic filling machines would be profitable to

the lower initial price would simply mirror the shorter expected physical life and there would be no
alteration in the capital-labour ratio from that in the advanced countries. Among the firms inter-
viewed, used equipment was said to be fully as productive as new equipment and expected life was not
different between the two. A useful analysis of many of the issues concerning the desirability of utilising
used equipment can be found in A. K. Sen, " On the Usefulness of Used Machines ", Review of Economics
and Statistics, August I962.
1 It might be possible to break down drum size and thus avoid the need for a fork-lift. In many
cases it is not possible because of international export specifications.
2 Two earlier studies that analysed the relation between volume and factor substitution possibilities
are Gerard Boon, Economic Choice of Human and Physical Factors in Production (Amsterdam, North Holland
Publishing Co., I964), and W. Paul Strassman, Technological Change and Economic Development (Ithaca,
Cornell University Press, I968).

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
I976] SUBSTITUTION OF LABOUR FOR CAPITAL IN KENYA 5I

introduce regardless of the level of wages.' In plants where filling is the major
operation (e.g. carbonated beverages) and relatively long production runs of
a limited number of items are typical, automated filling is the rule, though
there is adaptation to low wages, e.g. stacking of bottles into cartons is done by
hand (Table 2, row 2).

(b) Some Numerical Examples

It is instructive to consider a numerical illustration of some of the preceding

observations. A typical example of the economic rationality of continuing to
use labour-intensive peripheral processes is provided by the operation of stack-
ing toothpaste tubes into cartons. Although the following data apply to one
operation within one firm, many other examples yield similar results. At
present stacking is done by three men who take filled tubes from a table at the
end of the semi-automatic filling machine and manually place them in cartons.
An available machine could replace two of the men and the remaining man and
machine would be three times as productive per shift as the three men. The
price of a used machine of this type is KC2,5oo and that of a new one KC4,ooo.
The production characteristics of the two are the same and the expected life
with either is at least I O years - the firm's current planning horizon. Differential
maintenance costs are viewed as a minor consideration. Thus, the relevant ques-
tion for the firm is the choice between the current, completely manual operation,
and the adoption of the used machine. The company opted for the manual
operation and the correctness of its choice can be demonstrated. The relevant
data are shown in Table 3.
Table 3

Cost of Production (K Shillings)

Unit Cost

Capital* Labour Current I*25 375

cost per cost per Total output current current
annum annum cost (8,ooo) output output

Used equipment 9,960 3,960 I3,920 I.74 I.39 0-46

Manual process o i i,88o i i,88o I *49 I*19 I*9

* Derived from the annuity formula Z = A I-( d)n ], where Z is the purchase price, n the
expected life, d the discount rate and A the annual cost.

A ten-year life and a I5 %0 cost of capital are assumed. The average monthly
wage of these workers is 330 shillings and there are no skill differentials as the
machine could be operated with very small training costs by one of the three
1 Hand operations also require much more floor space than automated processes of equivalent
capacity. This suggests that substitution between labour and buildings (and land) may also become
important at increased volume; indeed, the constraint of factory floor-space was repeatedly cited as a
strong potential incentive to adopt automated devices in activities other than filling. Interestingly, this
phenomenon was much less pronounced in areas outside of Nairobi where both land acquisition and
construction costs are lower. In these areas a number of plants have in fact duplicated existing
production lines rather than automate.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
52 THE ECONOMIC JOURNAL [MARCH

workers currently employed. Potential output with three workers employed on

the manual process is io,ooo units per annum while actual current output is
8,ooo. At the latter quantity, unit costs on the equipment process are I.74
compared with I 49 on the manual one. Up to an output level of Io,ooo, the
ratio of unit costs remains the same, though the absolute level declines on both
processes. If we assume that labour cost on the manual process increases in
proportion to output above io,ooo units of production, the quantity at which
unit costs are the same (i.I9) for both processes is II,7I7. This quantity is
46 % above current output and the firm's decision to continue the use of the
manual process is correct. With higher production, the machine process
becomes increasingly efficient, and at its one-shift capacity (30,000) its unit
costs are 39 % of the manual costs (col. 6).
Despite the critical importance of volume, high wage levels could weaken
such choice as exists; for example, at a wage level only 26 % above the current
one the unit costs of the two processes would be equal at current output levels.
On the other hand, even at an output of 30,000 (3-75 times current volume)
the manual operation would yield the same unit cost if the wage were I,245
shillings per year - about one-third the current wage but arguably the shadow
price of labour to the urban sector.'
In contrast to peripheral operations, similar calculations for processing
equipment almost invariably show little scope for efficient substitution of
labour for capital. A typical one, analysing a juice-extraction process, shows
that the wage at which the labour-intensive process would yield equal unit
costs, at current output levels, is 4 shillings per month - surely less than the
shadow price of labour. Though choice exists in the sense that less labour per
unit of output requires more capital per unit of output, the lower productivity
of the labour-intensive process is such that its adoption cannot be defended.2

III. THE EMPLOYMENT-OUTPUT RELATION

The above material raises two general questions in analysing the relationship
between employment and output growth. First, what are the factors which
influence the initial choice of technique? Secondly, once this is chosen, what
are the implications for future employment growth?

(a) The Initial Choice of Technique

From the descriptive material of Section II it should be clear that there
appears to be considerable ex ante choice of capital intensity in most manufac-

1 Arnold Harberger, among others, has suggested that the shadow price of labour should be the
wage in the "informal sector" in urban areas. This is roughly IOO shillings per month in Nairobi
according to recent surveys. Harberger's view is set forth in "On Measuring the Social Opportunity
cost of Labour", in Fiscal Measuresfor Employment Promotion in Developing Countries (International Labour
Office, Geneva, I972).
2 In "Industrial Sector Labor Absorption", Economic Development and Cultural Change, April 1973,
Gustav Ranis shows that a number of measures, such as increasing the speed at which machines are
run, can alter the realised capital-output ratio with older equipment and increase its economic viability.
However, most processing equipment in our example would require implausibly large increases in
productivity to become a rational alternative.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
I976] SUBSTITUTION OF LABOUR FOR CAPITAL IN KENYA 53

turing branches and most firms have taken advantage of such choice to adopt
very labour-intensive methods, particularly in auxiliary activities. Although
the low cost of labour is undoubtedly a major factor determining the initial
technique in peripheral operations, this is too mechanistic a view of the process
of factor substitution. The neoclassical textbook view assumes there is a deus ex
machina at work, translating factor prices into correct choice of technique,
whereas such translation depends critically on the abilities and perception of
a set of talented managers who may not exist in various factories.'
A useful typology in analysing the role of managers is to divide them into
two categories: those with technical training or a background in production
and those without such education or experience. The technically trained
understand why operations are performed the way they are and the possibility
of using other methods. It is possible for them to envisage a production flow
which takes the output from a high-speed processor and divides it among
several hand-filling operations, rather than one which simply directs it into a
high-speed filler and wrapper. Instead of copying a U.K. process, they are able
to make the small, but important, adaptations which allow a more labour-
intensive process to function properly. In contrast, managers with sales or
finance experience and training and those who have extended wholesale and
retail operations backwards toward manufacturing appear to lack this ability.
They often duplicate the western process in toto, following the advice of con-
sultants and machine salesmen. Technical adaptation is only part of the role
of "good" managers. Typically they are also responsible for searching the
international market for appropriate equipment, whether used or new. This
involves the ability to evaluate specifications in the listings of used-machinery
dealers as well as the catalogues of new-equipment producers.2
We would argue that in the absence of technical expertise, appropriate
relative factor prices may be of limited efficacy in achieving socially appropriate
factor proportions3 (or indeed, private cost minimising ones for the given firm).
The effect of such expertise may be envisaged as extending the effective isoquant
out of which a firm chooses techniques. In Fig. I the unit isoquant is drawn
in two sections - the solid one reflects the typical range of alternatives actually
in use in advanced countries. The dashed part indicates the technically
feasible range if older peripheral equipment and other adaptations are im-
plemented. The "existence" of this part of the isoquant depends not only on
the physical feasibility of the process, but also on the ability of managers to

1 Harvey Leibenstein has defined one aspect of entrepreneurship as the activities required to carry
on an enterprise where "not all markets are well established or clearly defined and/or in which the
relevant parts of the production function are not completely known " (" Entrepreneurship and Economic
Development", American Economic Review, May I968). Leibenstein emphasises the need to search for
cost-reducing methods that eliminate X inefficiency. However, in the LDC context the effort is likely
to militate in favour of a search for techniques that utilise low-price labour.
2 As we will note below, multinational corporations are often extremely effective in these activities;
indeed, some maintain special staff for this function, usually in the home office.
3 Thus one reason, beside distortions in relative factor prices, for the failure of "hothouse" import
substitution to generate as much employment as had been expected, may be the absence of " technical "
entrepreneurs. One object of such programmes is the development of entrepreneurs and this often
utilises those in trade as the "raw material".

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 54 THE ECONOMIC JOURNAL [MARCH

discern and implement this part of the isoquant.1 This implies that the text-
book isoquant has little meaning - the range of real world options is nowhere
conveniently laid out. The benefit to the firm of such extension is the realisation
of cost level FF rather than F'F' (for the given output level), as the firm can
now produce at point A. If the isoquant of Fig. I is viewed as an aggregate one,
a labour force-capital endowment (point C) to the right of OZ2 which, in the
absence of adaptation, implies unemployment of BC, is now consistent with full
employment; moreover a gain in output also is achievable as C will lie on a
higher isoquant than B.

F'/
F

0
F F' L

Fig. I

The extension of the isoquants thus achieved is equivalent to an increase

in the elasticity of factor substitution. The importance of this for an economy
in which potential labour supply is growing more rapidly than capital is that
an increased elasticity is equivalent to a (Hicks) labour-using (capital-saving)
innovation and allows the growth rate of output to be higher than it would be
in the absence of such innovation.3 In a country in which research explicitly
designed to use labour (and save capital) is extremely difficult to implement,
such achievement of correctly biased innovation is a potential substitute for
investment in research capability.
Although I have emphasised the role of low wage levels combined with
managerial adaptive ability in searching for appropriate techniques, the
incentives for such adaptive behaviour may be reinforced by differences in the
cost of capital perceived by the two types of managers. Managers without
1 The implication of this for the analysis of the sources of intertemporal growth is pursued in my
"The Contribution of Education to Growth: an Alternative Approach", Yale Economic Growth Center
Discussion Paper, no. 113. A set of unpublished papers by Richard Nelson and Sidney Winter, Jr.,
consider a wide range of questions concerning the " objective" existence of isoquants and its implication
for the theory of the firm.
2 OZ may be thought of as a ridge line determined not by labour having zero marginal productivity
because of the inherent physical properties of the production function, but as a result of the absence of
the ability to search for and implement new processes. It represents the capital-labour ratio which
almost all western firms exceed.
3 See, for example, Murray Brown, On the Theory and Measurement of Technological Change (Cambridge,
Cambridge University Press, I966), pp. 22-3. The economic interpretation of the effect of increasing
the elasticity of substitution is that such a change reduces the rapidity with which diminishing returns
set in to the faster-growing factor.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
1976] SUBSTITUTION OF LABOUR FOR CAPITAL IN KENYA 55

technical experience were found primarily in locally owned companies while

the technically trained ones were primarily in foreign subsidiaries. While the
former relied on the local market for investment finance, borrowing at rates of
6-7 %0, the latter typically obtained funds from the parent company and were
instructed to use discount rates of Io-I5 %, or three- or four-year payback
periods in calculating the profitability of purchasing additional equipment.
Given the similarity in the wage rate paid by the two types of firms, the
differential in the cost of capital may have strengthened the incentives of
foreign subsidiaries to search for labour-intensive processes, but such activity
would have reduced unit costs for local companies as well; their wage/cost-of-
capital ratio, though higher than that of foreign subsidiaries, warranted further
search for labour using processes but their technical abilities limited this
endeavour.

(b) Implications for the Growth of Productivity

The fact that the initial process chosen is relatively labour-intensive has
interesting implications for future growth. In most of the existing literature,
the analysis of the dynamic effects of choice of technique has been limited to
the question of the impact of such choices on the national saving rate.' Little,
if any, attention has been directed to the productivity implications of such
decisions, yet it is apparent from my survey that these effects may be quite
significant.
In the production function literature it is assumed that "technical" progress
is the same for all regions of the isoquant - no distinction being made about
the probable type or rate of progress at different levels of capital intensity.2
However, if one considers the potential for disembodied progress, it appears
plausible that a large number of such changes are possible in the simple
processes described above, whereas in more mechanised operations there is
greater difficulty in realising "disembodied" productivity gains, relative to
those obtainable from embodied changes. Indeed, much of the production
function literature seems quite arbitrary (though convenient) in this regard:
long time series are analysed and yet the nature of the technical change is
assumed to be uniformly embodied or disembodied over the entire period.3
It is quite clear in the Kenyan context that substantial productivity growth
has taken place simply as a matter of reorganisation of production and better
training and supervision. A typical change involved a simple rearrangement of
the position of two processes within the same plant. A worker who had formerly
been idle half of the time (evenly spaced over the day) was more fully employed
when two processes converged on him. In addition to these types of completely
disembodied change, there were some that could be called slightly embodied

1 See the excellent discussion and synthesis in A. K. Sen, Choice of Technique, 3rd ed. (Oxford, I967).
2 Two exceptions are A. Atkinson and J. Stiglitz, "A New View of Technological Change",
ECONOMICJOURNAL, September I969, and Harvey Leibenstein, "Technical Progress, the Production
Function and Dualism", Banca Nazionale Del Lavoro Quarterly Review, December I960.
3 For a carefully executed study of this type see M. Intriligator, "Embodied Technical Change and
Productivity in the United States, 1929-1958 ", The Review of Economics and Statistics, vol. XLVII,
February I965.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 56 THE ECONOMIC JOURNAL [MARCH

in so far as they were implemented with internally produced "equipment"

containing material and labour worth less than K?5o. The archetype of such
innovation occurred in a fruit-processing plant in which cans had formerly
been packed one by one. The manager had the machine-shop bore I2 holes,
each the size of a can top, into a sheet of metal. This was installed at the end
of the processing line. As the crushed fruit came down on to this metal, one
worker shoved the fruit through the holes until the cans were full. The manager
estimated that this increased output per worker at this point fourfold. Most
plants had benefited from a large number of simple reorganisations and
inexpensive, internally generated "inventions".
The initial employment and output benefits from the choice of labour-
intensive techniques are thus augmented by total factor productivity gains
which accrue from reorganisation and simple inventions, these being greatest
at low capital-labour ratios, and becoming progressively more difficult as firms
reorganise and become mechanised or, if no capital deepening occurs, more
difficult with the passage of time as innovation possibilities become exhausted.
These developments imply that real value added per worker will increase,
and if wages were constant the level of profits would grow rapidly unless prices
were sufficiently reduced. However, in a highly unionised sector, also charac-
terised by racial differences between workers and management, it is likely that
the former will insist on some increase in remuneration. Thus, rather than
growing wage levels providing the impetus to capital deepening and thence to
productivity growth, as is often assumed, the causal relation is, as delineated
here, more complex.
Finally, it should be emphasised that our investigations produced no evidence
that companies using labour-intensive methods suffer from lower efficiency or
increased costs. Although some do exhibit domestic prices which are not
competitive with c.i.f. prices of imported goods, this was most often attributable
to a low volume of production and the inability to exploit increasing returns.
In the presence of excess capacity and yet-to-be-exploited scale economies,
measured rates of effective protection are at best a weak guide to potential
efficiency. Nevertheless, many of these firms are competitive, as indicated by
the low rates of effective protection (in their industry) according to the Phelps-
Wasowl computation and the existence of considerable exports to areas other
than the East African Community.

IV. MANAGERIAL ABILITY

We noted above that the innovations making productivity growth possible

were the result of decisions by managers who possessed a firm understanding
of the technical aspects of the production process. Perhaps surprisingly, in view
of the conventional wisdom that foreign-owned firms (or those employing
western trained technicians) will duplicate western methods, it was typically a
subsidiary of a foreign firm which carried out labour-intensive adaptations and

1 M. Phelps and B. Wasow, " Measuring Protection and its effects in Kenya ", Institute for Develop-
ment Studies (Nairobi), Working Paper no. 37.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 I976] SUBSTITUTION OF LABOUR FOR CAPITAL IN KENYA 57

was more willing to use older equipment. One of many examples was a local
subsidiary of a United States producer which had almost no new equipment;
indeed much of it was 25-40 years old, though it had been reconditioned.
Although the machinery was obsolete by United States standards, it was more
than serviceable. Strikingly, it was the parent company which had a depart-
ment specially devoted to searching the used-machinery market, arranging
delivery, etc. It is unlikely that a purely local company would have been as
successful.'
The adaptive behaviour of technically orientated managers stands in contrast
with that reported elsewhere.2 Such discrepancies are explainable in terms of
specific characteristics of the Kenyan firms. First, the technically oriented
management in Kenya has had limited formal education in engineering. Most
have been trained in the production operation within a firm. Though quite
conversant with the most modern technology, they have observed the work-
ability of older techniques. Thus they are less likely to think that the only
possible method of production is the latest - the charge most often directed
against graduate engineers. Moreover, since the "engineer" is usually the
managing director he is forced to consider the financial implications of choice
of technique: the artificial separation of the engineering and the economic
aspects of such choice is effectively overcome. Even if the "engineer" prefers
the modern, the non-schizophrenic managerial half is likely to impose financial
discipline.3 The combining of the two roles is, in my view, a critical one in
inducing adaptive performance. The relatively small size of Kenyan companies
is also of importance. Most references to the advocacy of indiscriminate adop-
tion of automated methods involve staff engineers who are divorced from the
management side. So far there has been little need for a separate engineering
staff in Kenyan companies: the director's own engineering time is sufficient.
To the extent that indigenous owners or managers had technical training,
there was little difference between foreign and domestic companies. One of the
most innovative plants had a local owner with a United States Ph.D. in
engineering. But to the extent that foreign plants are more likely to be directed
by those with considerable technical expertise, while locally owned ones are
more likely to be run by owners who have extended their selling operations
backward, foreign ownership, in this dimension, may well be desirable. To
the extent to which such abilities are not available locally or through expatriates
employed by local firms, the loss in adaptive ability should be weighed against
the presumed costs of their presence.
1 This point is also made in Ian Little, Tibor Scitovsky and Maurice Scott, Industry and Trade in
Some Developing Countries (London, Oxford University Press, I970), p. 57.
2 W. Paul Strassman and others have emphasised that foreign-trained engineers are likely to imitate
the advanced techniques of the developed countries.
3 Wells's managers have succeeded less well in integrating their engineering and economic selves,
though Wells does indicate that an economic rationale does exist for what initially appears to be a
proclivity for the modern, regardless of production costs incurred. Louis Wells, Jr., "Economic Man
and Engineering Man: Choice of Technology in a Low Wage Country", Public Policy, spring I973.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms
 58 THE ECONOMIC JOURNAL [MARCH I976]

-V. CONCLUSIONS

The two major findings of this study are: (I) there is considerable variation in
feasible efficient production methods, particularly in peripheral operations;
(2) substantial gains in labour productivity without capital deepening occur
due to the existence of considerable (disembodied) productivity gains and the
gradual elimination of excess capital capacity with consequent better utilisation
of the labour attached to that capacity. The rapid growth of labour productivity
results in reduced employment requirements for a given rise in output. This
phenomenon is likely to be transitional as realisation of the productivity gains
of the type described here becomes more difficult as industries mature.
These findings suggest two observations. First, it is important to obtain
additional direct evidence on whether companies in LDCs use excessively
capital-intensive methods. To infer capital deepening from the growth of labour
productivity is, at best, a questionable procedure unless supported by strong
microeconomic evidence. The factory visits reported here as well as earlier
studies relying on plant visits do not lend support to such views.1 Secondly, the
growth of labour productivity reflects efficiency gains which are more easily
achievable in labour-intensive regions of the isoquant. This implies that the
static factor choice decision may be more complicated than is usually realised.
Where shadow prices indicate that a capital-intensive technique is socially
optimal, given the initial levels of labour and capital productivity associated
with two techniques, it may nevertheless be easier to realise disembodied
productivity gains with the labour-intensive one whereas further productivity
gains with the capital intensive technique are likely to require embodiment and
thus use scarce investment funds (and foreign exchange as well). Analysis of
factor choice should take account of probable productivity growth associated
with each method.2 Moreover, to the extent that the adoption of a labour-
intensive process may also lead to machine embodied innovations for this
technique, further enhancement of the social profitability of such processes
results.

HOWARD PACK

Swarthmore College

Date of receipt offinal typescript: June 1975

1 See, for example, the references to Granick, Ranis and Reddaway in footnote 4, p. 48, above.
2 This is a point made by Atkinson and Stiglitz though they assume changing productivity is machine-
embodied. If machine-embodied productivity growth of the appropriate "bias" is to be generated,
it is likely that the LDCs will need their own capital goods industries. For a discussion of their role in
fostering such a development see H. Pack and M. Todaro, "Technological Change, Labour Absorption
and Economic Development", Oxford Economic Papers, November I969.

This content downloaded from 130.126.162.126 on Sun, 26 Jun 2016 09:34:01 UTC
All use subject to http://about.jstor.org/terms