PUBLISHED VERSION

Pomfret, Richard William Thomas

State-directed diffusion of technology: The mechanization of cotton harvesting in Soviet
Central Asia Journal of Economic History, 2002; 62(1):170-188

Copyright © 2002 The Economic History Association

Originally Published at:

http://journals.cambridge.org/action/displayJournal?jid=ISH

PERMISSIONS

http://journals.cambridge.org/action/displaySpecialPage?pageId=4676

Institutional repositories

2.4. The author may post the VoR version of the article (in PDF or HTML form) in the Institutional
Repository of the institution in which the author worked at the time the article was first submitted, or
(for appropriate journals) in PubMed Central or UK PubMed Central or arXiv, no sooner than one
year after first publication of the article in the Journal, subject to file availability and provided the
posting includes a prominent statement of the full bibliographical details, a copyright notice in the
name of the copyright holder (Cambridge University Press or the sponsoring Society, as appropriate),
and a link to the online edition of the Journal at Cambridge Journals Online.

23 April 2014

http://hdl.handle.net/2440/2196
 State-Directed Diffusion of Technology:
The Mechanization of Cotton Harvesting
in Soviet Central Asia
RICHARD POMFRET
When Soviet central planners began to mechanize the cotton harvest in earnest in
1958, they expected more rapid diffusion than the market-driven process that had
begun in the United States a decade earlier. But despite high output of cotton-picking
machines, the share of the crop harvested mechanically grew more slowly than in the
United States. The factor proportions in Central Asia did not justify mechanization:
although planners could enforce introduction of the new technology, investment in
cotton-harvesting machines was largely a waste of resources. The costs of premature
introduction are estimated at over $1 billion in 1960s prices.

T he relative merits of market-based and state-directed approaches is a

recurring theme in the literature on the diffusion of new technology.
This article addresses the issue in the context of technology embodied in a
machine, the mechanical cotton harvester. It contrasts the market-driven
diffusion process in the United States with state-directed diffusion in the
Soviet Union.
In market economies the speed of diffusion depends on the technical
characteristics of the machine (including improvements and variations on
the basic design), relative factor prices, and, in some cases, scale effects due
to indivisibilities or complementarities.1 Competition can be effective in
prompting continuous improvements, inducing institutional innovation to
overcome obstacles to diffusion, and filling missing markets. Competition
may, however, lead to inefficient use of resources, pursuing too many tech-

The Journal of Economic History, Vol. 62, No. 1 (March 2002). © The Economic History
Association. All rights reserved. ISSN 0022-0507.
Address correspondence to Richard Pomfret, School of Economics, Adelaide University, Adelaide
SA 5005, Australia. E-mail: richard.pomfret@adelaide.edu.au.
Most of the research behind this article was done during a sabbatical leave spent at the University
of California, San Diego, and I am grateful to the Graduate School of International Relations and
Pacific Studies for hospitality. Helpful comments on earlier drafts were provided by Michelle Barnes,
Alan Olmstead, Jonathan Pincus, John Siegfried, the editor of this JOURNAL and three anonymous
referees.
1
On the importance of technical improvements (the D in R&D) and of informational gaps and
interdependencies, see Rosenberg, Inside the Black Box. The classic account of the S-shaped diffusion
path of new technology is Griliches, “Hybrid Corn.” The threshold model introduces indivisibilities,
and combined with the size distribution of potential adopters generates an S-shaped diffusion path; for
application to the diffusion of mechanical reapers, see David, “Mechanization”; and Pomfret, “Mecha-
nization.” On tractor adoption, see Ankli, “Horses”; and Lew, “Diffusion.” The threshold model has
been criticized by Olmstead, “Mechanization” and “Diffusion”; and by Olmstead and Rohde, “Beyond
the Threshold,” who also emphasize the uncertainty inherent in the market and the nature of credit
institutions and farm organization. On the diffusion of tractors among U.S. cotton farmers see Musoke,
“Mechanizing”; and Whatley, “History.”

170
 State-Directed Diffusion 171

nological variants or producing too many models in suboptimally scaled

factories. Public policy may be superior to market forces in providing such
public goods as technical standards and secure intellectual property rights.
If indivisibilities matter, the state can encourage consolidation and coopera-
tion among both producers and consumers of the capital goods that embody
the new technology. In the Soviet Union the state played all of these roles,
consolidating both production of the capital equipment and potential adopt-
ers of the new technology.
Technical problems in the development of economical cotton-harvesting
machines were overcome in the 1930s and 1940s. The machine-picked share
of the cotton harvest in the United States increased in a classic S-shaped
diffusion pattern between 1949 and the late 1960s, starting in high-wage
California and occurring later in low-wage southeastern states, especially the
less geographically suited coastal areas of South Carolina.2 By the end of the
1960s, 96 percent of the U.S. cotton crop was harvested mechanically.3
Mechanization of the Soviet cotton harvest began in earnest about a decade
later, but with the government giving high priority to the process after 1958
it was expected that centrally planned diffusion would be rapid.4
The outcome was, however, not a steep diffusion path to more or less
complete mechanization. Diffusion was slower in the Soviet Union than in
the United States, and even reversed itself after 1981 (Figure 1). Three-
quarters of the cotton crop was mechanically harvested by 1959 in Califor-
nia, by 1962 in Texas, by 1965 in Mississippi, and by 1968 in South
Carolina, but no major cotton-producing Soviet republic ever reached this
share.5 Despite rapid build-up of machine production in the early 1960s and
an increasing share of machine-harvested cotton in the 1960s and 1970s,
machine use was below capacity from the start; and by the 1980s, the share
of machine-picked cotton was declining in the Soviet Union.
How to explain this phenomenon? A picture is often painted of bungling
politicians and planners failing to maintain production of mechanical har-
vesters and complementary inputs, or failing to appreciate the importance
of model variations to suit differing conditions or of repair and maintenance
services. Other authors see a plot to delay diffusion in order to keep Central

2
See Musoke and Olmstead, “Rise”; and Grove, “Farmers.”
3
Whatley, “New Estimates,” p. 199.
4
Priority was given to replacement of back-breaking work by modern machines in part because it
provided good publicity when, at the peak of Soviet prestige in the Third World, Soviet Central Asia
was held up as a model, contrasted to southern neighbors both in the satisfaction of basic needs and in
economic dynamism. Western economists also accepted this image; see Nove and Newth, Soviet
Middle East; Conolly, Beyond the Urals, pp. 55–241; and Wilber, Soviet Model.
5
The U.S. data are from Musoke and Olmstead, “Rise,” pp. 405–06. The minor Soviet exception was
Kazakhstan, where the reported share of the harvest picked by machine exceeded 75 percent in 1985
and 1988, although these spikes look implausible in the long-term context (Table 6). Less than 5
percent of Soviet land under cotton in the late 1980s was in Kazakhstan (Table 2).
172 Pomfret

100

90
United States
USSR
80

70
Percent of harvest mechanized

60

50

40

30

20

10

0
1949 1954 1959 1964 1969 1974 1979 1984 1989

FIGURE 1
THE MECHANIZATION OF COTTON-HARVESTING IN THE UNITED STATES AND THE
USSR, 1949–1988

Sources: See the text.

Asian peasants on the land and prevent the social disruption associated with
rural–urban migration. On these views, it is not the state-directed approach
itself that was at fault, but rather poor implementation by planners. I, how-
ever, shall argue that the “failure” of mechanization was due to resistance
by cotton growers. The central authorities were pushing for too-rapid diffu-
sion, and by the 1980s they failed to maintain even the diffusion levels
achieved in the 1960s and 1970s due to resistance at the level of the kolkhoz
(collective farm). Planners saw the benefits of mechanization as self-evident,
ignoring its real and opportunity costs, as well as its distributional impact.
How do we know that the intended diffusion was premature? One indica-
tion is the phenomenon of idle machinery during the Soviet period. Even
when machines were available, kolkhoz managers chose to leave them idle
and have the cotton picked by hand. Despite the low valuation by the users,
real resources went into the manufacture of the machines. Another sign that
diffusion had been too rapid is that when the post-Soviet successor states
shifted to greater market orientation in the 1990s, demechanization contin-
ued apace.6 Although reliable data on the share of machine-picked cotton in
6
Of course, many things changed after the collapse of central planning which may have affected the
demand for machinery, but the fundamental factor proportions had not. Characterizing the successor
 State-Directed Diffusion 173

post-Soviet Central Asia are hard to come by, anecdotal evidence suggests
that hand-picking became even more prevalent in the 1990s.7 By the turn of
this century, machine production had collapsed.
I will first set the scene by providing technical background material and
discussing the role of political leaders and central planners at the innovation
stage, where some of the Soviet Union’s latecomer advantage was dissipated
by political intervention. A following section discusses arguments that neo-
colonial political leaders hampered the diffusion process. I go on to posit an
alternative hypothesis, that resistance to mechanization was at the kolkhoz
level where, for the majority of adult male members, the costs of mechaniza-
tion exceeded the benefits; I assess the outcome in terms of social costs and
benefits, and argue that although the interests of kolkhoz managers were
determined within the framework of collectivized agriculture, their actions
reflected the economic unsuitability of mechanical harvesting in labor-abun-
dant Soviet Central Asia. A lower-bound valuation of the resources wasted
in premature diffusion is $1 billion at 1960 prices. The final section draws
some conclusions, emphasizing that while public intervention can accelerate
the pace of technological diffusion, this may not be a desirable outcome.

POLITICS, PLANNING, AND INNOVATION

Cotton harvesting was one of the slowest major agricultural processes to

undergo mechanization, even in high-wage countries. Although the first
patent had been taken out in 1850 and cotton mechanization was henceforth
always “just around the corner,” general adoption of mechanical cotton
pickers began in California only in 1949, and in the southeastern United
States in the early 1960s.8 The problems were both economic and technical.

economies as more market-oriented is not to deny that the cotton sector remains highly distorted in the
successor states, most strikingly with state monopsonies in Uzbekistan and Turkmenistan. Neverthe-
less, the end of central planning did lead to changes. Especially significant in the present context is the
disconnection of the machinery industry from the cotton growers, who are now less pressured to accept
delivery of new machines. In the planned economy, the level of compulsion in the labor market may
have reduced the incentive to economize on labor, but the shift to market-based labor allocation should
have reversed this bias. Note that in analyzing factor proportions input markets are of greater signifi-
cance than the lack of competition in the output market, assuming that decision-makers place some
weight on cost reduction.
7
The World Bank’s poverty assessment of Tajikistan (20285-TJ, 29 June 2000, p. 64) mentions that
“cotton harvesting is almost exclusively manual.” Andrew Apostolou, an Oxford University historian,
refers to the Uzbekistan 1999 cotton crop, “an incredible 96.5 percent of which was collected by hand”
(“Uzbek Model”).
8
The quotation is from Hon, “Rust,” p. 384. Strippers, which snapped cotton bolls off at the top of
the stalk and failed to distinguish ripe bolls from unripe ones and trash, were used in Texas and
Oklahoma after 1926, but were less efficient than the later cotton pickers which picked only ripened
bolls. Production of cotton pickers appears to have begun in 1946, when 107 machines were produced;
by 1951, over 3,500 machines were being produced annually in the United States (Street, New Revolu-
tion, p. 133).
174 Pomfret

Despite the labor-intensity of hand-picking, the economic incentive to

introduce expensive harvesting machines was insufficient, even in high-
wage California, before the late 1940s. Mechanized picking also impaired
the quantity and quality of cotton harvested. Machines missed some plants
and knocked the bolls off others; they also compressed the soil, destroying
nutrients and reducing future yields. Machine-harvested cotton was moister
and contained more impurities than hand-picked cotton, and was therefore
harder to process, which led to problems with the processing equipment.
Moreover, mechanization presupposed costly complementary adjustments,
including the development of simultaneously ripening strains, the straighten-
ing of furrows and planing of fields, the application of herbicides and defoli-
ants, and the precise timing of the harvest before rainfall. Failure to provide
for any of these would reduce drastically the efficacy of the equipment.
Ever since the Russian conquest of Central Asia in the 1860s, the central
government had promoted cotton production through technological borrowing
and heavy investment.9 In the 1870s the first Russian governor-general sent
two specialists to Texas to obtain new seed varieties. As with the mechanical
harvester, U.S. varieties required selection and adaptation to the climatic and
topographical conditions of Central Asia, for the purposes of which an experi-
mental station was established in Tashkent. Railway construction facilitated
transport to the textile mills of Russia. Irrigation projects increased the land
available for cotton sowing, and here again the Soviet government drew on
foreign expertise.10 By 1940 cotton output was triple that of 1913, and 1980
output was four-and-a-half times that of 1940 (Table 1). Hectares sown with
cotton increased correspondingly (Table 2). Almost all of the cotton was grown
in Central Asia, with over half of the harvest from the Uzbek republic and most
of the remainder from the neighboring Tajik and Turkmen republics.11
Soviet policymakers closely followed the development of practical cotton
harvesting machinery in the United States. They purchased two of the ten
machines built by the Rust Brothers in the United States in 1936, and invited
the inventors to the Soviet Union in the same year. The first mass-produced
Soviet cotton harvester, a vertical-spindle machine called the SkhM-48,
appeared in 1949.
9
Before the 1860s Central Asia supplied less than 10 percent of Russia’s cotton, but by the 1930s
the Soviet Union was self-sufficient in cotton, and it became an exporter in the 1950s.
10
Arthur Davis, a well-known U.S. irrigation engineer, surveyed the feasibility of irrigating the
Goldnaya Steppe in 1913. Fifteen years later his proposal was accepted, and in June 1929 Davis went
to the Soviet Union, where he was given engineering responsibility for the Central Asian irrigation
program. Other U.S., Japanese, and Korean technicians worked on the Central Asian cotton economy,
although most left in 1931; see Sutton, Western Technology, pp. 32–43.
11
The data on cotton production and acreage (Tables 1 and 2), machine output and stock (Tables
3–5), and share of mechanized harvesting (Table 6) are drawn from secondary sources, which are
approximately but not precisely consistent, even though they draw on a similar set of primary sources.
The broad output trends and dominant role of the Uzbek republic are plain, even if precise magnitudes
and weights in deriving Soviet Union–wide measures are not.
 State-Directed Diffusion 175
TABLE 1
COTTON PRODUCTION IN THE SOVIET REPUBLICS AND SUCCESSOR STATES,
1913–1999
(thousands of tons of raw cotton)
Azeria Kazakhb Kyrgyz Tajikc Turkmend Uzbeke USSR
f
1913 64 28 32 69 517 744
f
1940 154 95 172 211 1,386 2,237
f
1953 388 134 390 308 2,432 3,853
f
1956 352 151 415 334 2,857 4,332
f
1960 366 126 399 363 2,949 4,289
f
1965 355 167 609 553 3,904 5,662
f
1970 336 187 727 869 4,495 6,890
f
1975 450 202 836 1,079 5,330 7,864
f
1980 884 206 1,011 1,258 6,245 9,962
f
1985 788 58 935 1,287 5,382 8,755
f
1988 616 79 963 1,341 5,365 8,689

Total

1992 336 246 52 515 1,290 4,129 6,568

1993 284 198 49 524 1,341 4,235 6,631
1994 284 208 54 531 1,283 3,936 6,296
1995 274 223 75 412 1,293 3,934 6,211
1996 274 183 73 318 436 3,350 4,634
1997 125 198 62 353 635 3,700 5,073
1998 113 162 75 385 707 3,220 4,662
1999 101 249 87 316 1,300 3,680 5,733
a
Azerbaijan after 1991.
b
Kazakhstan after 1991.
c
Tajikistan after 1991.
d
Turkmenistan after 1991.
e
Uzbekistan after 1991.
f
Kazakh production not identified in the source; it accounts for most of the residual.
Note: Other countries producing over a million tons in 1999 were China (11,490), United States
(9,517), India (6,218), Pakistan (4,486), Turkey (2,093), Australia (1,728), Brazil (1,416), and Greece
(1,185).
Sources: 1913–1988: Gleason, “Marketization,” p. 67. 1992–1999: Food and Agricultural Organization
website (www.fao.org).

TABLE 2
COTTON SOWN IN SOVIET CENTRAL ASIA, 1940–1988
(thousands of hectares)
Republic 1940 1950 1960 1965 1970 1980 1985 1988

Uzbek 924 1,098 1,387 1,550 1,709 1,878 1,990 2,017

Turkmen 150 153 222 257 397 508 561 636
Tajik 106 126 172 228 254 309 311 320
S. Kazakh 102 97 106 112 118 127 131 128
Kyrgyz 64 65 71 73 75 76 28 32
Total 1,345 1,539 1,958 2,220 2,553 2,897 3,021 3,133
Source: Craumer, “Agricultural Change,” p. 144.
176 Pomfret

Political infighting hampered the refinement and improvement of har-

vester design during the 1950s. The Tajik leadership advocated planting in
narrow rows and square clusters to increase cotton yields. The Uzbek lead-
ership and cotton officials in Moscow opposed this “for respectable techni-
cal reasons and in order to maintain investment momentum behind new
irrigation construction.”12 Khrushchev resolved the debate in favor of the
Tajik position, primarily because the Tajiks were fulfilling their cotton
quotas while the Uzbeks were not, and because he was not keen on more
investment in irrigation. Khrushchev also supported the Tajiks in their
advocacy of horizontal-spindle machines; production of the SkhM-48
ceased in 1954. In a market economy there would have been technical
competition and, inmost cases, a convergence on the technology revealed
to be superior.13 In the Soviet planned economy, the decision was ultimately
taken by a poorly qualified autocrat, apparently on political grounds and
without serious cost– benefit analysis.14 In any case, the need to design new
horizontal-spindle machines for narrow rows set the machinery industry
back several years.
Once production problems had been resolved in the late 1950s, both
Soviet and outside observers expected mechanization to proceed rapidly.
Annual output increased from a few hundred machines in the late 1950s to
3,200 in 1960, to a peak of 8,000 in 1965 (Table 3). The central planners
favored mass production of a limited range of standardized machines, in
defiance of the kolkhozy’s obvious diverse needs. In particular, even as late
as 1970 no Soviet machines were suited to harvesting some varieties of
long-staple cotton grown in the Turkmen and Tajik republics.15
These problems of politicization and bias in central planning can explain
slight delays in the timing of the initial large-scale mechanization in the
1950s, and some geographical limitations on the spread of mechanical har-
vesters. However, they cannot explain the slow diffusion of mechanical
harvesting after 1960, let alone its reversal after 1980.

12
Hodnett, “Technology,” p. 77 (italics in original).
13
The ability to make such economic experiments is identified as a key advantage of capitalism over
socialism by Rosenberg, Exploring the Black Box, pp. 87–108.
14
Khrushchev’s experience of agriculture was based on grain-farming regions of Ukraine and
Russia. He had little sympathy for Central Asian farmers, whom he considered to be lazy and shiftless.
Procurement prices for cotton were held back after 1954 because he considered them already too high
relative to grain prices. After 1954 Khrushchev’s main goal in Central Asia was promotion of the
Virgin Lands scheme to increase grain output in northern Kazakhstan; disruption of cotton-machinery
production could free resources to produce tractors and grain harvesters. When cotton output failed to
meet targets, he blamed the political leaders, and purged the leadership of every cotton-growing
republic between December 1958 and May 1961. By contrast, under Brezhnev the political leadership
was remarkably stable in Central Asia.
15
Hodnett, “Technology,” p. 78.
 State-Directed Diffusion 177
TABLE 3
ANNUAL OUTPUT OF COTTON HARVESTING MACHINES IN THE USSR, 1950–1979
(thousands)
Year No. Year No. Year No.

1950 4,641 1960 3,184 1970 10,406

1951 9,810 1961 4,293 1971 6,706
1952 3,921 1962 6,090 1972 6,810
1953 3,346 1963 7,101 1973 6,725
1954 3,123 1964 7,000 1974 7,402
1955 555 1965 7,749 1975 7,572
1956 885 1966 7,204 1976 8,000
1957 102 1967 6,600 1977 8,650
1958 20 1968 7,731 1978 8,800
1959 486 1969 9,800 1979 9,100
Note: The Soviet figures reported by Gleason are in “standard units”: that is, two-rowed harvesters
introduced in the early 1960s counted as two units and four-rowed harvesters introduced in the mid-
1970s counted as four. Hodnett’s (“Technology,” p. 80) figures are similar to Gleason’s until 1966, but
are lower for 1966–1970, presumably for this reason.
Source: Gleason, Between Moscow, p. 148.

POLITICS AND DIFFUSION

Gregory Gleason has provided the most sustained analysis of the slow
mechanization of the Central Asian cotton harvest. He focuses on the labor-
market effects of mechanization, arguing that it would have displaced the
native population from the land and created a potential source of unrest in
the towns of the region, or in other parts of the Soviet Union as they relo-
cated in search of work. Moreover, the urbanization of Central Asia’s work-
force would be associated with inflows of skilled labor from elsewhere in
the Soviet Union. Neither politicians in Moscow nor the leadership in the
Central Asian republics welcomed these consequences, and in the Brezhnev
era they conspired to discourage mechanization in order to forestall its social
consequences.16 Gleason concludes that “Moscow officials failed to order
the agricultural machinery industry to produce that which was required for
full mechanization,” and he predicted this would change with perestroika,
when market forces would push kolkhoz managers to mechanize in order to
cut costs and be competitive.
The idea that cotton kolkhozy were starved of machinery is belied by the
production record. The stock of machines continued to increase into the
1980s,17 even after the share of cotton harvested by machine had begun to
16
Gleason sees evidence of this complicity in the longevity of Central Asian leaders between 1961
and 1982, when there was only one change in the top official of the five republics. He also sees local
kolkhoz managers condoning the anti-mechanization strategy, because their influence was related to
the number of workers on the collective. See Gleason, “Marketization”; the quotation in the next
sentence is from p. 80.
17
See Table 4.
178 Pomfret

decline in every cotton-growing Soviet republic.18 Moreover, from early in

the mechanization drive there is evidence that the share of cotton harvested
by machine was far below its potential. In the Uzbek republic 2 percent of
cotton was harvested by machine in 1958 and 11 percent in 1962, even
though the stock of machines should have been sufficient to harvest a third
of the crop.19 In 1977 two International Labor Organization economists
visiting five kolkhozy in the Uzbek and Tajik republics were surprised at
how little of the harvest was machine-picked, even on a showpiece collec-
tive that had 40 percent more mechanical harvesters per acre than the aver-
age for the Uzbek republic.20
In contrast to the public-relations pictures of massed cotton harvesters
bringing in the crop, there are frequent references in the less official litera-
ture to idle machinery. Over 800 machines (out of a total of 3,500) were
reported idle during the 1970 cotton harvest in the Turkmen republic.21 Peter
Craumer cites a figure of 19 percent of machines (i.e., around 2,000 units)
standing idle during the 1982 harvest in the Turkmen republic.22
The counterpart to unused machines is strong folk memories of masses of
people picking cotton by hand in the 1970s and 1980s. Few outside observ-
ers made it to rural Central Asia during the cotton harvest, but in 1977
Azizur Rahman and Dharam Ghai “witnessed a great deal of cotton picking
by hand,” even though the five kolkhozy they visited were showpiece collec-
tives in locations relatively well-suited to mechanization.23 Visiting Uzbeki-
stan in the 1990s, I came across many people who had picked cotton as
students in the 1970s and 1980s, and their recalled images too were of large
numbers of people picking cotton by hand. In view of the prevailing pro-
modernization ideology, the official data on the share of cotton picked by
machine are likely to overstate the actual share, and the diffusion path is
likely to have been even flatter than that presented in Figure 1.
Idle machinery is usually explained by the low priority given to repair
shops in Soviet Central Asia, or more generally to central planners’ focus on
production and neglect of maintenance. Cotton-harvesting machinery is
everywhere subject to breakdown and in need of constant maintenance.24
18
See Table 6.
19
Hodnett, “Technology,” p. 79.
20
Khan and Ghai, Collective Agriculture, p. 113 n6. The tone of Khan and Ghai’s footnote is clearly
that the “numbers do not add up.” Elsewhere in their report, they refer to being “told that there were
technical problems of designing harvesters that would pick high quality cotton without waste” (p. 110
n26). This is a polite excuse rather than an explanation, because such problems are an inherent feature
of mechanical cotton harvesters and conditions around Samarkand, where the comment was made, are
about as favorable as possible for mechanical harvesting in Central Asia.
21
Hodnett, “Technology,” p. 74.
22
Craumer, “Agricultural Change,” p. 161.
23
Khan and Ghai, Collective Agriculture, p. 110 n26.
24
Hodnett (“Technology,” p. 113 n69) compares an Australian cotton farm he visited in 1971, where
four specialized mechanics were employed full-time to look after 50 cotton harvesters, with the situa-
180 Pomfret

Mechanization also reduced kolkhoz income because of the poorer-quality

output, for which the producers received lower procurement prices.27 Thus,
C should include the fixed and variable costs plus the indirect costs of wast-
age and lower quality.
Decision-makers clearly saw the labor costs of hand-picking (w) as being
sufficiently low relative to the costs of machines (C) that mechanization was
resisted despite high values on Ls and Q. In the United States the labor re-
quired for cotton harvesting was reduced by 95 to 98 percent by mechaniza-
tion, depending on climate and other field conditions, and this appears to
have been true of Central Asia too.28 The value of Q also varied by loca-
tion.29 In Soviet Central Asia the degree of resistance to mechanization
varied, depending on the technical conditions reflected in Ls and Q; for
example, machine design was such that Ls and Q were lower and diffusion
was slower in the Tajik republic. Nevertheless, the phenomenon of idle
machines suggests a fairly ubiquitous tendency to consider costs not worth
incurring in order to bring marginal machines into service.
The labor costs of hand-picking, as seen by kolkhoz decision-makers,
were depressed by the varying status of kolkhoz members and by the state’s
ability to organize levies of students and urban workers. The main benefit
from mechanization—the saving of back-breaking labor—was not shared
equally. Despite the mass mobilization at harvest time, many kolkhoz mem-
bers did not participate. Grey Hodnett cites a figure of 400,000 nonpartici-
pants in the 1962 harvest in the Uzbek republic, when the total kolkhoz
population was about a million. The kolkhoz members who did participate
in the cotton harvest were mainly women; a plausible presumption is that
kolkhoz managers perceived their female workers as having a lower opportu-
nity cost than most male members.
Soviet Central Asia was often described as a region of surplus labor, in
contrast to the European parts of the Soviet Union. Although there was by
definition no unemployment, Western observers found substantial underem-

with hand-picking was between 2.5 and 5 percent, rising over time as the average age of pickers
increased.
27
The difference was about 10 percent in 1963 (Hodnett, “Technology,” p. 115 n93).
28
On the United States, see Street, New Revolution, p. 170; and Whatley, “New Estimates,” p. 201.
In 1999 in Uzbekistan cotton-harvester drivers were paid 100–200 sums per ton picked, whereas hand
pickers were paid 8–9 sums per kilogram, so that the reduction in labor costs from mechanization was
at least 97.5 percent.
29
Annual use of cotton-harvesting machines in the United States in the 1950s and 1960s varied from
225 hours in the southeast, to 350 hours in the Mississippi Delta, to 500 in the west; see Whatley, “New
Estimates,” p. 215. The main reason for this discrepancy was the drier and more reliable climate in the
west compared to the mud and morning mist of the southeast coastal regions. Musoke and Olmstead
(“Rise”) emphasize larger farm size and climate as reasons why Californian cotton farms were best-
suited to mechanical harvesting in the United States. Similar features in the core cotton-growing areas
of Central Asia suggest that neither physical conditions nor kolkhoz fragmentation are plausible expla-
nations for the slow diffusion of machine-picking.
 State-Directed Diffusion 181

ployment.30 One complication is the seasonality of agricultural work, and

especially the heavy demands during the cotton harvest. In the mid-1980s
over half of Uzbek cotton and three-quarters of Tajik cotton was hand-
picked, for which students and industrial workers were drafted for six to
eight weeks.31 The main social cost of these levies was the disruption of
industry and education, but the direct cost to the kolkhoz was small: in 1972
these casual workers were paid five kopeks per kilogram picked.
Mechanization would have had further effects on kolkhoz organization. In
particular, mechanization would make more economic sense if greater acre-
age were devoted to cotton and planting patterns devised to ease mechanical
harvesting. Such changes in sowing patterns were resisted by kolkhoz mem-
bers because they threatened to upset the established division between collec-
tive land and private plots. The private plots provided insurance against
sudden policy changes and other negative shocks to living standards. Al-
though hard evidence on the value of private plots in Central Asia is difficult
to come by, Hodnett claims that they probably accounted for at least a third
of average family income in the 1960s; Nancy Lubin has estimated that
private plots accounted for a fifth of Central Asian agricultural output and a
quarter of kolkhoz families’ income circa 1980.32 Sheila Marnie shows a big
range by republic and by family income in 1989: in the Uzbek republic, for
instance, for the poorest families (with per capita monthly income below 75
rubles) private plots accounted for 19–20 percent of family income, whereas
for families with per capita income above 200 rubles private plots accounted
for 32 percent of income.33 Thus, in the biggest cotton-producing republic all
kolkhoz families might resist any threat to their private plots, but the wealth-
ier—and presumably more influential—families had most to lose.
In sum, many kolkhoz managers considered the costs of mechanization to
outweigh the benefits, even though their “soft” budget constraints rendered the

30
Dienes (Soviet Asia, p. 131) reviews some of this literature, concluding that in 1979 surplus labor
was about 1.1 million in the Kyrgyz, Tajik, Turkmen, and Uzbek republics combined, and around
315,000 in the Kazakh republic. Marnie (“Soviet Labor Market,” pp. 203–05) estimates that in 1984
over a million people were voluntarily or involuntarily unemployed in the Uzbek republic, the core of
the Soviet cotton economy, and the non-employed amounted to 14 percent of the working-age popula-
tion, compared to 5.5 percent in the Russian republic. Klugman (“Wages,” p. 34), however, points to
large regional variations in Uzbekistan between cotton-growing areas, such as the labor-scarce Dzijak
region, and the surplus-labor Ferghana Valley.
31
In 1986 and 1987, 650,000 to 700,000 schoolchildren, 140,000 college and vocational school
students, and an unreported number of urban workers harvested cotton in the Uzbek republic (Craumer,
“Agricultural Change,” p. 162). In the Tajik republic over 300,000 students and industrial workers were
drafted during harvest seasons in the mid-1980s for one-and-a-half to two months cotton picking
(Dienes, Soviet Asia, p. 130).
32
Hodnett, “Technology,” pp. 85–86; and Lubin, Labor and Nationality, pp. 181–82 and 188.
33
Marnie, “Soviet Labor Market,” p. 223. This progressivity was more pronounced in the Tajik
republic, rising from 16 percent to 45 percent. In the Kyrgyz and Turkmen republics, the share fell as
family income rose. The share of family income coming from private plots was much lower in the
Turkmen republic (8–16 percent) than in the other three republics covered by Marnie.
182 Pomfret

machines themselves essentially free. One reason might be that their objec-
tives focused on meeting output targets, so that wastage costs, while rela-
tively minor to a firm in a market economy, loomed large in the Soviet
context; mechanization would in effect require planting more hectares with
cotton. Also, managers’ loyalty to their male colleagues would have discour-
aged disruption of the private-plot system, and led to little weight being
placed on the arduous labor done by others. In the specific institutional con-
text of Soviet Central Asia, central planners and kolkhoz managers were
acting at cross-purposes with respect to mechanization, and the question
arises of whether the resistance to mechanization was desirable on any
grounds other than the local decision-makers’ self-interest within the Soviet
system.

COST–BENEFIT ANALYSIS

The previous section investigated the mechanization decision from the

perspective of the most influential members of the kolkhoz. This section
tries to make the same calculation on the basis of shadow prices. Was
mechanization appropriate in Soviet Central Asia given the region’s re-
source endowments?
One approach to this question is to assume technical parameters and
capital costs from the much-studied U.S. case. According to Warren
Whatley, in 1964 the total expected cost of machine-harvesting cotton
ranged from 3.1 cents per pound in California to 8.9 cents in North
Carolina.34 If the technical characteristics of Soviet and U.S. machines (Ls
and Q) were similar, then the key issue is whether shadow labor costs were
below this range.35 The casual harvest-time wage rate of five kopeks per
kilogram was equivalent to around two U.S. cents per pound at official
exchange rates, and even less at other exchange rates preferred by Sovietolo-
gists in the 1960s.36 Thus, it seems probable that labor was sufficiently abun-
dant in Soviet Central Asia for hand-picking to be the most efficient tech-
nique, even if physical conditions were as favorable to mechanization as in
the least-cost U.S. region.
This conclusion is consistent with international evidence. In the southern
United States in the late 1950s, the labor costs of hand-picking cotton were
34
Whatley’s estimates draw on earlier work by Street and a 1969 University of Chicago Ph.D.
dissertation by Frank Maier; see Whatley, “New Estimates”; and Street, New Revolution.
35
This is not a very stringent test. Insofar as U.S. machine-makers offered a wider range of cotton
harvesters than Soviet producers, Soviet labor costs would need to have been significantly higher than
three U.S. cents per pound to justify widespread mechanization.
36
This was not a market-determined wage, but the low wage is consistent with the underemployment
evidence presented in the previous section. Thirty years later, labor was still abundant enough to make
mechanization uneconomic; in the 1998 harvest season in Uzbekistan cotton pickers were paid five sum
per kilogram, which was just over two U.S. cents per pound at the official exchange rate and less than
one cent at the market exchange rate; see Pomfret, “Agrarian Reform.”
 State-Directed Diffusion 183

over five cents per pound.37 Australia, the other high-wage cotton-growing
country, also mechanized harvesting. Major producers such as India, China,
and Pakistan, however, continue to harvest by hand because their labor costs
do not justify mechanization.38 Soviet Central Asia was between these ex-
tremes of labor costs, and closer to Egypt or Iran or Turkey, none of which
has mechanized the cotton harvest.
One cannot, of course, rule out externalities. The social disruption of
mechanization would have been considerable,39 but not necessarily negative
and, as emphasized above, there is little evidence of policymakers delaying
mechanization for this reason. More plausibly, Soviet planners may have
placed great weight on the propaganda benefits of modernization. An eco-
nomic price was, however, paid for any such psychic benefits.
How high was this price? The phenomenon of idle machines suggests that
their true economic value was close to zero, and this is supported by the
collapse of sales after 1991. In that case the cost of premature mechanization
was the resource cost of building the machines. Over 60,000 machines were
produced in the 1960s (Table 3) and, although I have no data on deprecia-
tion rates, the numbers in Table 4 suggest that this flow was perhaps main-
tained through the 1970s and declined in the 1980s, before collapsing to a
few hundred per year in the late 1990s (Table 5).40 Soviet prices tell us little
about the resources going into each machine, because input prices, transport
costs, and output prices for producer goods were all meaningless in the
centrally planned economy. During the 1960s the list prices for generic
standard-model cotton pickers in the United States ranged from $6,000 to
$10,000.41 Given the competitive nature of the U.S. industry, this could be
considered the “world price” of a standard machine and, if Soviet machines
were of the same quality, then it provides a shadow price. Thus, during the
1960s the opportunity cost of premature mechanization of cotton-picking
was in the range of $360–600 million. The amount may have been slightly

37
Labor costs included not just the minimum wage payments of about $3.50 per hundredweight, but
also recruitment and organizing costs (0.86 and 1.16 cents per pound); see Whatley, “New Estimates,”
p. 211.
38
In 1999 the countries with the largest cotton production were China, the United States, India,
Pakistan, Uzbekistan, Turkey, Australia, Brazil, Turkmenistan, and Greece (Table 1).
39
In the southern United States the mechanization of cotton and subsequent release of unskilled labor
was arguably the greatest single economic and social event of the twentieth century, transforming not
only the cotton sector but also the entire labor market and social structure, leading to large-scale
migration to the northern United States and reinforcing demands for civil rights.
40
In interviews at the Tashkent Institute of Engineers for Irrigation and Agricultural Mechanization
(TIEIAM) in 1997 I was told that 2,000 cotton-picking machines were produced per year in the 1980s.
Output dropped to very low levels in the early 1990s, although TIEIAM staff explained that this was
due to input supply disruptions (especially steel) and technical problems which were being resolved
by negotiating a joint venture with a U.S. producer. With a total academic staff of 422, and 514 stu-
dents registered in its Faculty of Agricultural Mechanization, TIEIAM clearly maintains the tradition
of central advocacy of mechanization.
41
Whatley, “New Estimates,” p. 209.
184 Pomfret
TABLE 4
STOCK OF COTTON-HARVESTING MACHINES, 1960–1986
(thousands)
Republic 1960 1965 1970 1975 1980 1986

Uzbek 8.30 21.80 26.10 28.70 36.60 37.90

Turkmen 0.68 3.30 3.50 6.50 9.30 10.90
Tajik 0.17 2.23 2.85 2.91 3.42 4.11
Kazakh 0.70 ? 2.80 ? 2.50 2.60
Kyrgyz 0.14 1.64 1.74 1.27 1.05 0.74

Total 9.99 ? 36.99 ? 52.87 56.25

Source: Craumer, “Agricultural Change,” p. 161.

less in the next two decades, but it is unlikely that the total (undiscounted)
cost at 1960s prices was less than $1 billion.42

CONCLUSIONS

Labor productivity in Soviet Central Asia was very low. Cotton output per
day of labor in 1953 was even lower in the Soviet Union than in Egypt.43
Labor productivity increased during the 1950s and 1960s, but it fell in all the
cotton-growing republics between 1970 and 1987.44 Despite the large expen-
diture on irrigation, machinery, and fertilizers, the returns to both land and
labor used in cotton production were diminishing in the final two decades
of the Soviet Union, and the cotton economy was imposing huge environ-
mental costs as chemical fertilizers drained into the rivers and irrigation
projects cut off the flow of water to the Aral Sea.45 Soviet planners recog-
nized the low level of labor productivity and saw the need for mechaniza-
tion, but low opportunity cost of labor is a reason for doubting the appropri-
ateness of labor-saving investment.
Politicization and central planning delayed the development of the cotton-
harvester industry in the 1950s; once it did take off, the system’s rigidity and
biases restricted the development of technical improvements and of repair
facilities. These are, however, minor features of the story of the anemic
42
Because some machines had positive value in relatively labor-scarce areas such as Dzijak this
could be an overestimate, but in other respects it is a minimum estimate. The U.S. prices are for
bottom-of-the-line one-row pickers, whereas Soviet output included two-row pickers (which were
twice as expensive as one-row pickers in the United States) and other non-standard models. As noted
in Table 3, data on machine production in the 1950s and 1960s are imprecise about quality. Moreover,
substantial costs were incurred in the design and retooling associated with the SkhM-48 and subsequent
shift to horizontal-spindle machines.
43
Economic Commission for Europe, “Regional Policy,” p. 60; and Hodnett, “Technology,” p. 73.
44
Craumer, “Agricultural Change,” p. 159.
45
In March 1989 Gorbachev was reported in Pravda as complaining that over the previous 20 years
in Uzbekistan 1.6 million new hectares of irrigated land had been created and productive forces in
agriculture had increased sixfold, but gross production was up only 78 percent; see Gleason, “Marketi-
zation,” p. 72.
 State-Directed Diffusion 185
TABLE 5
ANNUAL OUTPUT OF COTTON-HARVESTING MACHINES IN UZBEKISTAN, 1994–1999
Year No.

1994 651
1995 1,121
1996 863
1997 1,049
1998 351
1999 278
Note: So far as I am aware, no cotton harvesters were produced elsewhere in the former USSR; so these
figures are comparable with Table 3.
Source: Ministry of Macroeconomics and Statistics data reported in Uzbek Economic Trends,
January–March 2000, p. 104.

diffusion of mechanical harvesting in Soviet Central Asia. The main point

is that, given the technical parameters and abundance of labor in Soviet
Central Asia, hand-picking was more economical than machine-picking.
Even with machines essentially free (in the sense that there was a stock and
no second-hand market), kolkhoz managers preferred to leave them idle.
Planners and politicians had little idea of how their strategy for increasing
labor productivity was received at the kolkhoz level.46
With over a decade of post-Soviet experience, we have a huge advantage
over writers who tried to explain cotton mechanization during the Soviet
era. Nevertheless, it is surprising just how universally accepted was the
premise that mechanization was a source of improved efficiency. Observers
cited the difference in labor requirements between hand-picking and
machine-picking, without considering the capital costs and other factors.
Since the end of central planning in the new independent states, there ap-
pears to have been a continuing decline in mechanical harvesting; at the
very least, there has not been the dramatic shift to machine-picking which
Gleason predicted would be the inevitable outcome as market forces became
more important under perestroika.47 (See Table 6 for the year-by-year per-
centages of cotton actually harvested by machine.) The implication is that
kolkhoz members, who opposed mechanization out of a narrow self-interest
shaped by distorted Soviet prices and other incentives, actually produced the

46
Irrigation projects, the other huge capital investment in Central Asian agriculture, also failed to
distinguish between kolkhoz-level incentives and social welfare, but in this case kolkhozy and central
planners tended to be on the same side. Irrigation projects were generally supported by kolkhoz mem-
bers, who saw private benefits from free or heavily subsidized water, which helped them to achieve
cotton-output targets as well as improving their private plots, with little cost to them. Little attention
was paid to effective use of the water, a scarce resource in the Central Asian desert, or to the environ-
mental costs of existing and planned irrigation projects. Even as late as the mid-1980s official opinion
was still in denial as to the costs of desiccation of the Aral Sea, and Chernenko was still considering
diversion of Siberian rivers. The common feature was the grand commitment to modernization without
attention to complexities “on the ground.”
47
Gleason, “Marketization.”
186 Pomfret
TABLE 6
PERCENTAGE OF COTTON HARVESTED BY MACHINE IN THE MAIN COTTON-
PRODUCING SOVIET REPUBLICS, 1965–1988
Uzbek Turkmen Tajik Kazakh Kyrgyz

1965 23 ? 14 21 33
1970 33 32 22 41 39
1975 46 47 28 69 62
1980 63 52 36 63 51
1981 68 49 29 51 61
1982 54 41 22 45 63
1983 34 43 12 43 49
1984 31 26 11 39 60
1985 40 52 13 76 23
1986 42 47 22 ? 69
1987 45 ? 20 57 69
1988 47 65 24 78 ?
Note: The USSR points in Figure 1 are weighted by republican production levels, from Table 1.
Hodnett (“Technology,” p. 79) gives similar figures: at “the end of the 1960s” 33 percent of the entire
Soviet crop was machine-harvested, with variation across republics: Kyrgyz 39 percent (1970);
Turkmen 33 percent (1970); Tajik 21 percent (1969); Azeri 9 percent (1965); and Uzbek 1.7 percent
(1955), 2 percent (1958), 11 percent (1962), 24 percent (1965), 29 percent (1969), and 34 percent
(1970).
Source: Craumer, “Agricultural Change,” p. 161.

economically efficient outcome in view of true relative factor prices in

labor-abundant Central Asia. Cotton-picking remains back-breaking work,
but in the more market-oriented economies of the 1990s the demand for
cotton-harvesting machines virtually disappeared and the supply of labor
was plentiful even at the low wages.48
The role of central planners in pushing cotton mechanization was similar
to that prescribed by advocates of public policy to accelerate technical diffu-
sion. In the long run, as incomes rise, machine-picking displaces hand-pick-
ing. That this process did not occur as smoothly or rapidly as planned was
not primarily due to the incompetence or deviousness of politicians or plan-
ners, but rather due to resistance at the kolkhoz level. The true cost of state-
directed diffusion was not that the planners failed to speed up diffusion, but
that they did it too far ahead of its time. The costs in terms of wasted re-
sources in machine production were large, amounting to at least $1 billion
at 1960s prices, even under conservative assumptions.
Accelerating technical diffusion, especially when it appears to be ob-
structed by scale barriers associated with expensive equipment, is often
advanced as a reason for government intervention, and not just in centrally
planned economies. In newly independent states in the 1950s and 1960s,
48
In the 1998 harvest season the base rate in Uzbekistan was only five sum per kilogram, or about
$5 per day at the market exchange rate (Pomfret, “Agrarian Reform,” p. 281). Apostolou (“Uzbek
Model”) reports that 96.5 percent of Uzbekistan’s 1999 cotton harvest was picked by hand, and that
out of cotton export earnings of $884 million just $3 million was accounted for by labor costs.
 State-Directed Diffusion 187

technological catch-up was one of the arguments for state-supported indus-

trialization. Even a well-managed economy like South Korea’s fell into this
trap with its heavy-industry drive of the late 1970s, aimed at speeding the
ascent up the technical ladder, but leading instead to an embarrassing climb-
down by policymakers. In high-income market economies a costly example
was the French and British governments’ success in providing their national
airlines with a supersonic aircraft when the market failed to do so. British
Airways and Air France, each forced to take seven Concordes, were un-
happy even when the aircraft were sold to them for one franc; the premium
they could charge supersonic passengers was inadequate to cover even the
operating costs.49 Supersonic travel will surely become economical some-
day, just as Central Asian cotton will eventually be picked by machine. But
anticipating the future can be a costly folly.
49
By the end of 1978 the British and French governments had spent $4.28 billion to build two
prototypes and 14 commercial planes. The two national airlines each made operating losses on their
seven Concordes, so that in 1981 both national governments were subsidizing Concorde operations;
see Feldman, Concorde and Dissent, pp. 83–120. Public spending on Concorde is a lower-bound cost
estimate because it understates the opportunity cost of many valuable resources.

REFERENCES

Ankli, Robert. “Horses vs. Tractors on the Corn Belt.” Agricultural History 54, (January
1980): 134–48.
Apostolou, Andrew. “The Uzbek Model confronts a Crisis.” Eurasianet, posted 5 October
2000 at http://www.eurasianet.org.
Conolly, Violet. Beyond the Urals: Economic Developments in Soviet Asia. London: Ox-
ford University Press, 1967.
Craumer, Peter. “Agricultural Change, Labor Supply, and Rural Out-Migration in Soviet
Central Asia.” In Geographic Perspectives on Soviet Central Asia, edited by Robert
Lewis, 132–80. London: Routledge, 1992.
David, Paul. “The Mechanization of Reaping in the Ante-Bellum Midwest.” In Industrial-
ization in Two Systems: Essays in Honor of Alexander Gerschenkron, edited by Henry
Rosovsky, 3–39. New York: John Wiley, 1966.
Dienes, Leslie. Soviet Asia: Economic Development and National Policy Choices. Boulder:
Westview Press, 1987.
Economic Commission for Europe. “Regional Economic Policy in the Soviet Union: The
Case of Central Asia.” Economic Bulletin for Europe 9, no. 3 (1957): 49–75.
Feldman, Elliott. Concorde and Dissent. Cambridge: Cambridge University Press, 1985.
Gleason, Gregory. “Between Moscow and Tashkent: The Politics of the Uzbek Cotton
Complex.” Ph.D. dissertation, University of California, Davis, 1984.
______. “Marketization and Migration: The Politics of Cotton in Central Asia.” Journal of
Soviet Nationalities 1 (Summer 1990): 66–98.
Griliches, Zvi. “Hybrid Corn: An Exploration in the Economics of Technological Change.”
Econometrica 25 (October 1957): 501–22.
Grove, Wayne. “Farmers at the Technological Crossroads: Post-WWII Diffusion of the
Mechanical Cotton Harvester.” Unpublished paper, January 1997 (drawn from a PhD
dissertation at the University of Illinois).
188 Pomfret

Hodnett, Grey. “Technology and Social Change in Soviet Central Asia: The Politics of
Cotton Growing.” In Soviet Politics and Society in the 1970s, edited by Henry W.
Morton and Rudolf L. Tokes, 60–117. New York: Free Press, 1974.
Hon, Ralph. “The Rust Cotton Picker.” Southern Economic Journal 3 (April 1937):
381–92.
Khan, Azizur Rahman, and Dharam Ghai. Collective Agriculture and Rural Development
in Soviet Central Asia. Basingstoke: Macmillan, 1979.
Klugman, Jeni. “Wages in Transition: The Case of Uzbekistan.” Ph.D. Thesis, Australian
National University, 1998.
Lew, Byron. “The Diffusion of Tractors on the Canadian Prairies: The Threshold Model
and the Problem of Uncertainty.” Explorations in Economic History 37 (April 2000):
189–216.
Lubin, Nancy. Labor and Nationality in Soviet Central Asia: An Uneasy Compromise.
Princeton, NJ: Princeton University Press, 1984.
Marnie, Sheila. “The Soviet Labor Market in Transition.” Ph.D. Thesis. European Univer-
sity Institute, Badia Fiesolana, Italy, 1992.
Musoke, Moses. “Mechanizing Cotton Production in the American South: The Tractor,
1915–1960.” Explorations in Economic History 18 (October 1981): 347–75.
Musoke, Moses, and Alan Olmstead. “The Rise of the Cotton Industry in California: A
Comparative Perspective.” This JOURNAL 42 (June 1982): 385–412.
Nove, Alec, and J. A. Newth. The Soviet Middle East: A Communist Model for Develop-
ment. London: Allen & Unwin, 1966.
Olmstead, Alan. “The Mechanization of Reaping and Mowing.” This JOURNAL 35 (June
1975): 327–52.
______. “The Diffusion of the Reaper: One More Time!” This JOURNAL 39 (June 1979):
475–76.
Olmstead, Alan, and Paul Rohde. “Beyond the Threshold: An Analysis of the Characteris-
tics and Behavior of Early Reaper Adopters.” This JOURNAL 55 (March 1995): 27–57.
Pomfret, Richard. “The Mechanization of Reaping in Nineteenth Century Ontario: A Case
Study of the Diffusion of Embodied Technical Change.” This JOURNAL 36 (June
1976): 399–415.
______. “Agrarian Reform in Uzbekistan: Why has the Chinese Model Failed to Deliver?”
Economic Development and Cultural Change 48 (January 2000): 269–84.
Rosenberg, Nathan. Inside the Black Box. Cambridge: Cambridge University Press, 1982.
______. Exploring the Black Box. Cambridge: Cambridge University Press, 1994.
Street, James. The New Revolution in the Cotton Economy: Mechanization and its Conse-
quences. Chapel Hill: University of North Carolina Press, 1957.
Sutton, Antony. Western Technology and Soviet Economic Development, 1930 to 1945.
Stanford, CA: Hoover Institution, 1971.
Whatley, Warren. “A History of Mechanization in the Cotton South: The Institutional
Hypothesis.” Quarterly Journal of Economics 100 (November 1985): 1191–215.
______. “New Estimates of the Costs of Harvesting Cotton: 1949–1964.” Research in
Economic History 13 (1999): 199–225.
Wilber, Charles K. The Soviet Model and Underdeveloped Countries. Chapel Hill: Univer-
sity of North Carolina Press, 1969.