This study examines the impact of agricultural policies and investment on productivity in Zimbabwe and South Africa, for which the most complete productivity-related data in sub-Saharan Africa are available. Of particular interest are comparisons of the effects of such policies and investments on commercial and smallholder agriculture. The commercial and smallholder sectors in the two countries exhibit potentially significant differences in resources, policy support, and market conditions and thus in the impacts of policy changes and investment on agricultural productivity. Existing research on commercial and smallholder agricultural productivity in Zimbabwe and South Africa is reviewed. Recognizing the extent to which historic policy measures influenced the structure of agriculture in the two countries today, the study also assesses the sensitivity of existing estimates of productivity to changes in investment, policy and assumptions about human capital.
Zimbabwe and South Africa differ from most other African countries in the degree of past and present European involvement in their economies. Nevertheless they share a similar institutional heritage with many other African countries, particularly in eastern and southern Africa, in terms of the evolution of agricultural policy over the course of the twentieth century. This evolution has been characterized in much of eastern and southern Africa by an early bias towards large-scale commercial agriculture, with land, labour and capital markets structured accordingly. This was followed around the period of independence by subsequent policy redirection in support of the smallholder agricultural sector, followed more recently still by market liberalization and structural adjustment.
TABLE 7.1
Selected Indicators of Agricultural Productivity
Indicator |
Zimbabwe |
South Africa |
Sub-Saharan Africa |
World |
Fertilizer use (kg/arable ha, 1980) |
61 |
87 |
16 |
86 |
Fertilizer use (kg/arable ha, 1995) |
62 |
51 |
13 |
94 |
Tractors (per 1 000 ha cropland, 1993) |
7 |
8 |
3 |
21 |
Ag output (percent of GDP, 1996) |
14 |
5 |
24 |
na |
Ag output growth (percent/yr, 1980-1990) |
2.4 |
2.9 |
1.8 |
2.8 |
Ag output growth (percent/yr, 1990-1996) |
4.5 |
1.4 |
2.1 |
1.7 |
Cereal yield (kg/ha, 1980) |
1 359 |
2 117 |
1 100 |
2 230 |
Cereal yield (kg/ha, 1995) |
1 163 |
1 918 |
1 041 |
2 561 |
Ag productivity (US$/ha, 1980) |
34 |
45 |
53 |
na |
Ag productivity (US$/ha, 1993) |
41 |
49 |
68 |
236 |
Ag productivity (US$/worker, 1980) |
294 |
2 361 |
458 |
na |
Ag productivity (US$/worker, 1995) |
266 |
2 870 |
392 |
na |
Note: figures are three-year moving averages.
Source: World Bank (1998); tractors from WRI/UNEP/UNDP/WB (1998).
In seeking to understand the impact of these policy changes on the productivity of agriculture, Zimbabwe and South Africa are uniquely valuable as case studies in the extent to which sector-specific agricultural productivity-related data and research are available. This allows intersectoral comparisons of agricultural productivity based on differences in inputs, land quality and policy variables while controlling for factors such as climate, as well as identification of potentially interesting intersectoral spillover effects that may also be significant. Identifying the sources of smallholder agricultural productivity growth in Zimbabwe and South Africa may thus provide important insights into agricultural development strategies elsewhere in sub-Saharan Africa (SSA).
Table 7.1 provides selected indicators of input use, output growth and agricultural productivity for Zimbabwe and South Africa in relation to the rest of SSA and the world. At 50 to 60 kilograms per arable hectare in 1995, fertilizer use in both countries is four to five times the level in the rest of SSA, but (with the exception of South Africa in 1980) remains 30 to 40 percent below world averages. A similar pattern is evident for use of tractors. Agricultural output grew at an average rate of 2.4 percent per year in Zimbabwe and 2.9 percent per year in South Africa in the 1980s, at roughly the same pace as world agricultural output growth, compared with 1.8 percent per year in the rest of SSA. Between 1990 and 1996, agricultural output growth slowed to 1.7 percent per year for the world as a whole, slowed to 1.4 percent per year in South Africa, accelerated to 2.1 percent per year in SSA and accelerated to 4.5 percent per year in Zimbabwe.
Changes in agricultural productivity depend on how rapidly agricultural output changes in comparison with changes in input levels. Partial indicators of agricultural productivity, such as yields per hectare or the value of output per worker are also reported in Table 7.1. For example, cereal yields fell from 1 359 kilograms per hectare to 1 163 kilograms per hectare in Zimbabwe between 1980 and 1995, a decrease of 1.0 percent per year. Over the same period, cereal yields in South Africa fell from 2 117 kilograms per hectare to 1 918 kilograms per hectare, a decrease of 0.7 percent per year. Meanwhile SSA-average cereal yields fell from 1 100 kilograms per hectare to 1 041 kilograms per hectare, and world-average cereal yields rose from 2 230 kilograms per hectare to 2 561 kilograms per hectare (an increase of 0.9 person annually).
Land productivity in agriculture, as measured by the value of output per hectare of arable land, grew in both countries and in SSA as a whole between 1980 and 1995. In Zimbabwe, agricultural land productivity grew from US$34 per hectare in 1980 to US$41 per hectare in 1995, an annual increase averaging 1.3 percent. Agricultural land productivity in South Africa grew from US$45 per hectare to US$49 per hectare over the same period, an average increase of 0.6 percent per year. Agricultural land productivity grew even more rapidly in SSA as a whole, starting from a higher base, from US$53 per hectare in 1980 to US$68 per hectare in 1995, an average increase of 1.7 percent per year. Even at US$68 per hectare, however, SSA land productivity levels remain less than one-third world-average levels.
Labour productivity is measured as the value of agricultural output per agricultural worker. Here patterns differ from those for land productivity. While agricultural labour productivity rose at an annual average rate of 1.3 percent in South Africa between 1980 and 1995, from US$2 361 per worker to US$2 870 per worker, it fell in Zimbabwe and in SSA as a whole. Labour productivity fell in Zimbabwe at an average rate of 0.7 percent per year, from US$294 per worker to US$266 per worker, and in SSA as a whole at an average rate of 1.0 percent per year, from US$458 per worker to US$392 per worker.
These mixed trends in input use, output growth and agricultural productivity raise questions about the performance of agriculture in Zimbabwe and South Africa. The remainder of this paper explores these trends in productivity and the factors that underlie them, with a focus on the role of investment in improving agricultural productivity. Special attention is given to investment in human capital and its impact on productivity.
Deininger and Binswanger (1995) argue that in Zimbabwe and South Africa, as in Kenya, farming on large operational holdings was made economically feasible in large part due to policy distortions designed to favour large-scale European farmers over African smallholders. Otherwise, they argue, given the relatively low capital intensity and high labour intensity of both large-scale and small-scale agricultural production in Africa, supervisory costs for labour would become an important factor favouring smallholders.
In both Zimbabwe and South Africa, a series of interventions favoured European farmers over African smallholders in markets for outputs as well as for land, labour, credit and other inputs (Kassier and Groenewald, 1990; Deininger and Binswanger, 1995). In Zimbabwe, the Land Apportionment Act of 1930 limited African land purchases to selected marginal areas, and government assistance was provided to white farmers for virtually all crop and livestock products through subsidies and state monopolies. In South Africa, the Native Lands Act of 1913 and other aspects of a deliberate apartheid policy sharply restricted African access to land, in order to reduce competition with white farmers in labour and output markets and increase the supply of African labour to the expanding mining industry. Government transfers to white farmers for marketing assistance alone averaged over 30 million South African Rand annually between 1948 and 1966 (Deininger and Binswanger, 1995).
These policies contributed to the emergence of the dualistic structure of landholding and production that characterizes agriculture in Zimbabwe and South Africa today. In Zimbabwe, one million communal farm households occupy about half of the arable cropland, while the other half is farmed by 4 500 large-scale commercial farmers, most of whom are white (Atkins and Thirtle, 1995). Compounding this, roughly three-quarters of communal lands are located in agro-ecological regions averaging 500 millimetres or less of annual rainfall (Mudimu, 1990). In South Africa, about 70 000 commercial farms occupy approximately 85 million hectares, while a subsistence-oriented sector, with problems characteristic of much of sub-Saharan Africa, occupies approximately 15 million hectares (van Zyl, Vink and Fenyes, 1987; van Zyl and Groenewald, 1988). Farmland per caput in the homelands established by the Native Lands Act of 1913 averaged less than three hectares in 1990, compared with over 18 hectares in the rest of South Africa (Kassier and Groenewald, 1990).
In addition, policy through much of the twentieth century favoured investment in European commercial agriculture over smallholder production. Research on hybrid maize began in Zimbabwe in 1932, and by 1949 Zimbabwe had become the second country in the world (after the United States) to market hybrid maize seed (Eicher and Rukuni, 1990). Zimbabwe's "first Green Revolution" (Eicher, 1995) was launched in 1960, spearheaded by and for white commercial farmers, and based on large public and modest private investments in new technology, physical and biological capital (e.g. rural infrastructure and new seed varieties), human capital (e.g. education) and farmer support institutions.
While improving yields on white commercial farms, these policies resulted in a gradual decline in the productivity of African smallholder agriculture. In Zimbabwe's communal areas, per caput grain production (a crude measure of labour productivity) declined from about 300 kilograms to about 200 kilograms between 1925 and 1980 (Jayne and Jones, 1997).
In South Africa, tax concessions on capital improvements and subsidies for white commercial agriculture in general led to steady investment in machinery in the decades following World War II (van Zyl, Vink and Fenyes, 1987). Until about 1970, mechanization was focused on the substitution of machinery for draughtt animal power in cultivation, while harvesting operations remained labour-intensive. With yields increasing, this brought about increased demand for labour. After 1970, with the introduction of mechanized harvesting, capital became a substitute for labour, leading to reductions in farm employment in major maize-producing areas. Subsequent increases in agricultural employment have been associated with the disappearance of interest rate subsidies and the phasing out of tax concessions for machinery purchases (Kassier and Groenewald, 1990; Thirtle, Townsend and van Zyl, 1998).
Following independence in 1980, government policy in Zimbabwe sought to balance equity and growth-oriented objectives through provision of public services and infrastructure, including substantially expanded credit and market infrastructure for smallholders, to encourage rapid increases in smallholder agricultural productivity (Atkins and Thirtle, 1995; Jayne et al., 1994). The new government in Zimbabwe also embarked on a programme of land resettlement that by 1990 had succeeded in resettling 52 000 families, most of them from the communal areas, on 2.5 million hectares purchased from European commercial farmers (Eicher and Rukuni, 1990). Ongoing plans call for the acquisition and resettlement of an additional 5 million hectares.
Zimbabwe's "second Green Revolution" focused on smallholders following independence in 1980 (Eicher, 1995). Underlying this second revolution were the inheritance of a productive public agricultural research system, a quadrupling in the number of government-provided loans to smallholders (Eicher and Rukuni, 1990), a sharp increase in guaranteed producer prices and a 30-fold increase in the number of Government Marketing Board (GMB) grain-buying depots and collection points. Together, these factors resulted in the rapid adoption of hybrid maize varieties and fertilizer by smallholders. Sales of hybrid maize seed and fertilizer to the small holder sector increased nearly five-fold between 1979 and 1985 (Rohrbach, 1990). As a result, smallholder maize production doubled from 738 000 tonnes in 1980 to 1.3 million tonnes in 1986 (Eicher, 1995). In the decade following independence, maize yields rose at an average rate of 6.7 percent per year, while cotton area and yields increased by 25 percent and 1.3 percent per year, respectively (Deininger and Binswanger, 1995).
Despite policy changes, yield differences between commercial and communal areas continue to reflect marked differences in access to input and output markets, land quality and climate. For example, maize yields in Zimbabwe's commercial sector are 50 percent to 200 percent higher than maize yields on comparable land in the smallholder sector (Deininger and Binswanger, 1995). Even within the smallholder sector, wide variations in behaviour are apparent. The 25 percent of communal lands in the most favourable agro-ecological zones, with 20 percent of the total smallholder population, accounted for 68 percent of smallholder maize production, 80 percent of smallholder maize sales and 91 percent of smallholder fertilizer use in the late 1980s (Rohrbach, 1990).
In South Africa, the reorientation of policy toward smallholders has included the scrapping of the Land Acts in 1991, the application of labour legislation to farm workers and the establishment of the Farmer Support Programme (FSP) (van Zyl, Fenyes and Vink, 1992). The FSP is designed to ensure that black smallholders in the homelands are provided with improved access to input markets, output markets, research and extension services, training and educational opportunities, physical infrastructure and secure tenure. The hope is that these services will create the conditions for the commercialization of the smallholder agricultural sector.
Across eastern and southern Africa, government-led provision of services to smallholders has proven to be extremely expensive and ultimately unsustainable. Jayne and Jones (1997) argue that production gains in Zimbabwe and South Africa have been achieved only at a cost greater than the value of the increased output. By the mid-1980s, the costs of these services, combined with vast accumulating surpluses of maize, forced a reappraisal of government policies in Zimbabwe (Eicher, 1995).
Government-provided credit to smallholders was scaled back, as were marketing subsidies and grain-buying depots in communal areas (Atkins and Thirtle, 1995). In 1990, the government launched a programme of structural adjustment with the backing of the International Monetary Fund and the World Bank, which deregulated markets and reduced public expenditures (Marquette, 1997). Implementation of this programme coincided with an unusually severe drought across southern Africa in 1991-1992, causing declines in agricultural production, employment and wages, as well as in other indicators of social welfare that had improved dramatically in the 1980s. In spite of increased population pressure on smallholder lands, total cropped area in Zimbabwe has decreased slightly since the mid-1980s (Jayne et al., 1994), and fertilizer purchases by smallholders have stagnated since 1993 (Jayne and Jones, 1997).
Throughout eastern and southern Africa state maize marketing agencies continue to play a significant role in most countries despite regulatory reforms in recent years. In South Africa, maize price controls and trade restrictions were lifted in 1995, following similar measures in Zimbabwe in 1993 (Jayne and Jones, 1997). However, the effects of these measures are difficult to assess because of the confounding affects of recent extreme weather events, because these measures may not yet have been fully implemented, and because data on factor productivity remain incomplete. Nonetheless, both Zimbabwe and South Africa, traditionally reliable producers of grain surpluses, have seen steady declines in their net exports in recent years (Jayne and Jones, 1997).
TABLE 7.2
Selected Indicators of Human Capital
Indicator |
Zimbabwe |
South Africa |
Sub-Saharan Africa |
World |
Population (millions, 1996) |
11 |
38 |
596 |
5 754 |
Population growth (percent/year, 1980-1996) |
3.0 |
2.0 |
2.8 |
1.6 |
Population (millions, 2010) |
14 |
46 |
844 |
6 788 |
GNP per caput (US$, 1996) |
610 |
3 520 |
490 |
5 130 |
GNP per cap growth (percent, 1965-1996) |
0.4 |
0.2 |
-0.2 |
1.2 |
Life expectancy (m,f; years, 1996) |
55, 57 |
62, 68 |
51, 54 |
65, 69 |
Adult literacy (m,f; percent, 1995) |
90, 80 |
82, 82 |
66, 47 |
79, 62 |
Percent labour force in ag (m,f; 1994) |
58, 81 |
16, 10 |
65, 75 |
48, 52 |
Percent population < US$1 per day (1990s) |
41 |
24 |
na |
na |
Source: World Bank (1998).
Despite concerns that structural adjustment might result in maize price increases detrimental to the food security of urban and rural consumers, Jayne et al. (1996) report that some adjustment measures have improved household food security by reducing costs. Specifically, easing of restrictions on private grain transportation, marketing and processing since 1993 have significantly reduced the cost of maize meal, a staple in the Zimbabwean diet. The proportion of maize meal acquired through informal channels has risen from eight percent to 50 percent, at cost savings of 30 to 40 percent, representing savings equivalent to about ten percent of income among the poorest urban consumers. Rubey and Lupi (1997) estimate that for all but the highest income group, easing of these restrictions has more than offset the welfare losses due to removal of consumer subsidies.
Agricultural performance in Zimbabwe and South Africa is influenced by their resource base as well as by their policy histories. Table 7.2 presents selected indicators of human resources in the two countries. Zimbabwe's population of 11 million (1996) is growing at a rate of three percent per year, and is projected to reach 14 million by the year 2010. South Africa's 1996 population of 38 million is growing more slowly, at two percent, and is expected to reach 46 million by 2010. The two countries together currently represent about eight percent of sub-Saharan Africa's population.
GNP per caput was nearly six times as great in South Africa in 1996 as it was in Zimbabwe, and more than seven times the average for sub-Saharan Africa as a whole, although these figures mask sharp distributional differences within each country. At 55 years and 57 years for men and women, respectively, life expectancies in Zimbabwe were slightly higher than for sub-Saharan Africa as a whole in 1996, while South Africa's life expectancies were closer to world averages, at 62 and 68 years respectively. Conversely, adult literacy rates were 90 percent and 80 percent for men and women, respectively, in Zimbabwe in 1995, and 82 percent for both men and women in South Africa, higher than both world and sub-Saharan African averages.
TABLE 7.3
Selected Indicators of Natural Capital
Indicator |
Zimbabwe |
South Africa |
Sub-Saharan Africa |
World |
Land area (1 000 km2, 1995) |
387 |
1 221 |
23 628 |
130 129 |
Cropland (percent land area, 1980) |
7 |
11 |
6 |
11 |
Cropland (percent land area, 1995) |
8 |
13 |
7 |
11 |
Pasture (percent land area, 1995) |
44 |
67 |
34 |
26 |
Forest (percent land area, 1995) |
22 |
7 |
17 |
25 |
Protected (percent land area, 1994) |
8 |
6 |
6 |
7 |
Percent cropland irrigated (1995) |
5 |
8 |
4 |
18 |
Ag land per worker (ha, 1980) |
8.4 |
50.3 |
7.9 |
3.9 |
Ag land per worker (ha, 1993) |
6.0 |
51.0 |
6.0 |
3.8 |
Freshwater (m3 per caput, 1996) |
1 254 |
1 190 |
7 821 |
7 342 |
Source: World Bank (1998).
In Zimbabwe, as in most of sub-Saharan Africa, most women who are active in the labour force (three-quarters to four-fifths) are employed in agriculture, while in South Africa the figure is much lower, at ten percent. A similar pattern holds true for men: 58 percent in Zimbabwe, 65 percent in sub-Saharan Africa as a whole and 16 percent in South Africa. The corresponding worldwide averages are about one-half for both men and women. As a final indicator of human resources, about 41 percent of Zimbabwe's population are estimated to live on less than US$1 per day, as is true for about 24 percent of South Africa's population. Comparable figures are not available at the regional and global levels.
Selected indicators of agriculture-related natural resources are presented in Table 7.3. Both Zimbabwe and South Africa had proportionately more land in cropland (eight percent and 13 percent, respectively) and in pasture (44 percent and 67 percent, respectively) than did sub-Saharan Africa as a whole. Five percent of Zimbabwe's cropland was irrigated in 1995, compared with eight percent in South Africa and 4 percent for sub-Saharan Africa as a whole. In terms of overall freshwater availability per caput, both Zimbabwe and South Africa are relatively dry, each with about 15 percent the freshwater available at the average regional and global levels. By the year 2050, Zimbabwe and South Africa are projected to have about 860 and 560 cubic metres of freshwater available per caput, respectively. One thousand cubic metres is considered the threshold below which water availability becomes a severe constraint on socio-economic development and environmental quality (Meinzen-Dick and Rosegrant, 1997).
Lal (1998) and Tagwira (1992) estimate that 15 percent of the land area in Zimbabwe is very severely eroded, 13 percent is severely eroded, 19 percent is moderately eroded and 53 percent is not eroded. Soil erosion averages 25 tonnes per hectare per year under traditional farming practices, 37 tonnes per hectare per year on commercial farms under continuous cotton and 17 tonnes per hectare per year on commercial farms under continuous maize. Erosion is estimated to have reduced maize yields by two to five percent per tonne of soil lost (Lal, 1998; Hudson and Jackson, 1959).
Jayne et al. (1994) examine changes in total factor productivity in Zimbabwe's smallholder agricultural sector over the period 1975-1990. They use a profit function approach to estimating total factor productivity (TFP), which facilitates measurement of policy effects on prices and productivity. They find that smallholders are responsive to maize price incentives, and thus to policies that influence maize consumption. For example, a subsidy that reduces the consumer price of maize by ten percent is estimated to stimulate subsequent year's crop input use by 1.6 percent and maize output by 5.6 percent.
Jayne et al. (1994) found the effects of research and development (R&D) on TFP in the smallholder sector to be insignificant. The use of maize hybrids increased from 29 percent of smallholder maize area in 1979 to virtually 100 percent by 1985. They note, however, that these hybrids had been available for a full decade previously, and attribute their eventual near-universal adoption to the post-independence expansion of credit and market infrastructure for smallholders in the early and mid-1980s. They conclude that the full impact of R&D may not be realized without complementary investment in the physical and institutional infrastructure necessary for input and output markets to function efficiently. By contrast, Thirtle et al. (1993) estimated that the rate of return to R&D in Zimbabwe's commercial sector, characterized by relatively well-functioning markets for inputs, outputs and credit, was between 40 percent and 60 percent.
While Jayne et al. (1994) integrated R&D directly into their profit function, Atkins and Thirtle (1995) use a modified two-stage methodology to analyse the same data in order to measure and explain growth in TFP in Zimbabwe's communal sector over the period 1975-1990. First, they derive a measure of TFP as the ratio of an aggregate output to an index of aggregate inputs. Changes in TFP are then explained in terms of changes in R&D, extension and farmer education. Atkins and Thirtle estimate that TFP grew at an annual average rate of 1.7 percent over the entire period, but find that this average conceals significant variation. Specifically, they estimate that TFP declined at an average rate of 0.8 percent annually between 1975 and 1980, increased at an average rate of 8.1 percent annually between 1980 and its peak in 1985, and then decreased at an average rate of 2.7 percent annually between 1985 and 1990.
Atkins and Thirtle find that labour and land together account for the bulk of input costs. Growth in population combined with the purchase of commercial sector farmland for resettlement contributed to growth in the use of both these inputs by the communal sector throughout the period, although the pace of resettlement has slowed since 1985. The use of hybrid seeds, fertilizer and chemical inputs grew at an average rate of 22.6 percent per year between 1980 and 1985, contributing to partial productivity growth rates for land and labour of 9.2 percent and 9.5 percent per year over the period, exceeding the growth in TFP. While the value of own consumption of agricultural output increased at an average annual rate of only 1.2 percent between 1975 and 1990, the value of sales by the communal sector (primarily maize and cotton) increased at an average annual rate of 26.4 percent over the same period. Between 1985 and 1990, sales declined at an annual average rate of one percent. However, while the government invested in infrastructure, credit and technology for the communal sector between 1980 and 1985, sales grew at an average rate of 33.1 percent per year.
In the second stage of their analysis, Atkins and Thirtle estimate how TFP is affected by changes in R&D, extension, education and other policy indicators (reflecting the number of government input-supply and output-purchasing depots and the provision of credit to the communal sector), while controlling for weather. They found that weather and the number of depots explain 80 percent of the variation in TFP between 1979 and 1990. Dalton, Masters and Foster (1997) also find a close association between production costs and access to markets in Zimbabwe's smallholder sector, emphasizing the importance of rural infrastructure in influencing production decisions.
While R&D was found to be insignificant in explaining changes in TFP, Atkins and Thirtle argue that hybrid maize still played a crucial role in the growth of TFP in Zimbabwe's communal sector. As Eicher (1995) points out, however, virtually the entire commercial sector maize crop was already planted to hybrid varieties well before independence, so communal sector farmers were able to draw readily on this technology.
Thirtle et al. (1993) use both a two-stage decomposition approach and an integrated production function approach to analyse productivity in Zimbabwe's commercial agricultural sector for the period 1970-1990. They find that the growth in aggregate output slowed to 2.7 percent annually in the 1980s, from 4.4 percent in the 1970s, while aggregate input use actually declined in the 1980s by 1.2 percent annually after increasing in the 1970s by 0.5 percent annually. The combined effect of these two trends was that total factor productivity in Zimbabwe's commercial sector was relatively stable before and after independence, growing at an average annual rate of 3.9 percent in the 1970s and 4.0 percent in the 1980s. The authors find that investment in R&D and extension explain over 90 percent of the variation in TFP over the period 1970-1990, and had an internal rate of return of 40 to 60 percent.
Thirtle, Sartorius von Bach and van Zyl (1993) studied TFP growth in South Africa's commercial sector over the period 1947-1991, using the Tornqvist-Theil approximation of the Divisia index. They find that the output index rose at an average annual rate of three percent over the period, the input index grew at an average annual rate of 1.8 percent and TFP grew at an average annual rate of 1.3 percent.
These overall averages mask a 0.9 percent annual decline in the input index after 1979 (following annual growth of 2.5 percent for 1947-1979), and a subsequent acceleration in TFP growth since 1981 to 2.9 percent per year. Labour and land productivity each rose by more than three percent annually between 1947 and 1991, due to the increased use of intermediate inputs (such as fertilizer and high yield varieties [HYVs]) and capital inputs (such as machinery) relative to labour and land.
Mechanization in particular was supported by cheap credit and tax concessions during the 1970s. When these concessions declined in the early 1980s, labour became cheaper relative to capital, and labour use increased for several years before falling again later in the decade.
Thirtle, Sartorius von Bach and van Zyl (1993) then seek to explain observed changes in TFP in terms of changes in R&D, extension services and farmer education. They also add variables for international technology transfer (based on lagged United States TFP) and for weather. Coefficients on R&D, extension and education were positive and significant, although very sensitive to changes in specification.
Khatri, Thirtle and van Zyl (1996) revisit the results found by Thirtle, Sartorius von Bach and van Zyl (1993), using a profit function approach directly incorporating the conditioning effects of local public sector agricultural research and international research spillovers. They point out that South African agriculture remains subject to widespread policy distortions that affect its structure and performance. For example, subsidized farm credit resulted in negative real interest rates in the 1970s and into the early 1980s, and resulted in considerable over-capitalization.
Khatri, Thirtle and van Zyl find that public research expenditures and the international stock of knowledge (in the form of agriculture-related chemical and mechanical patents registered in the United States) are significantly and positively related to productivity growth in South Africa's commercial agricultural sector, particularly with respect to field crops. The authors estimate that the rate of return to public-sector R&D is 44 percent, but argue that these returns would be much smaller, or even negative, if returns to R&D could be adjusted for the full social costs of high unemployment and poverty in rural areas. Extension is found to have a significant but very small effect on productivity, while farmer education is found to have a significant and negative effect. The authors suggest this may be due to over-mechanization and over-application of fertilizer on the part of better-educated farmers, driven by emphasis on maximization of production rather than profits. The authors conclude that policy and R&D together have distorted incentives that would otherwise lead to profit maximization reflective of South Africa's relatively abundant labour and relatively scarce capital.
Thirtle, Sartorius von Bach and van Zyl (1993) did a careful study for South African commercial agriculture that is the first attempt to measure South African TFP at the national aggregate level, and this paper draws extensively on this work. For comparison, Thirtle et al. (1993) develop an innovative set of comparable computations for the commercial and communal sectors in Zimbabwe. Atkins and Thirtle (1995) note that a lack of adequate time series data has so far prevented the measurement of TFP elsewhere in sub-Saharan African agriculture. The comparisons in section 7.5 use data from published estimates of TFP for South Africa's commercial agricultural sector and for Zimbabwe's commercial and communal agricultural sectors (Thirtle, Sartorius von Bach and van Zyl, 1993; Thirtle et al., 1993; Atkins and Thirtle, 1995; Khatri, Thirtle and van Zyl, 1996; Thirtle, Townsend and van Zyl, 1998). The TFP indices calculated by these authors are described below. Included in the discussion are the strengths and weaknesses of the available statistical sources, and accommodations that were made for shortcomings. The result is three comparable data sets that address the conceptual problems involved in transforming aggregate agricultural accounting data into production data in a way that is consistent with its theoretical basis; the Tornqvist-Theil approximation of a Divisia index.
Outputs
The numerator of a TFP index is aggregate agricultural output. The output shares are calculated using the currently priced series for the three main output categories, namely crops, horticulture and animals. These were adjusted for subsidies to give the prices actually faced by farmers. Laspeyres indices are calculated for the value of crops, horticultural products and animal products, using crop-years2. The year-on-year ratios of these indices are used to form the Tornqvist-Theil index.
Inputs
The denominator of the TFP index is an index of inputs used in agricultural production. This index has two components, each weighted by cost shares, and again the year-on-year ratios of these indices are used to form the Tornqvist-Theil index. There are two primary inputs, labour and land. In South Africa, data are available on the number of farm employees and domestic servants on farms, differentiating between males and females and between blacks and whites. For the physical units of labour, it was necessary to estimate the number of domestic servants so that they could be subtracted from the total farm labour force, and to estimate the proportion of a full- labour unit that a part-time employee represents (Thirtle, Sartorius von Bach and van Zyl, 1993). The index is not quality-adjusted for sex, wage or education.
The land index used by Thirtle et al. (1993) also fails to account for quality differences. Although van Schalkwyk and Groenewald (1992) have constructed a land quality index for South Africa, it provides only one observation for each country. Therefore, the index of Thirtle et al. tends to under-represent land when cultivated area is expanded on to poorer lands. However, some features of land quality are captured in the data. For example, land quality is related closely to the buildings and land improvements series included in the capital items below. The costs of irrigation are included in the intermediate inputs index. The share weights for land are calculated from estimates of the values of land rent and the percentage of the area rented. Thus it was possible to impute a rental value to the entire area of land to get a measure of land value.
Intermediate inputs, as they are collected for the national accounts, are packing material, fuel, fertilizer, feed, dips and sprays and other inputs. The values of all inputs were deflated with the appropriate price indices, also from national accounts, in order to get the input series. As with the outputs, subsidies on fertilizer, feed and transport were used to adjust the value figures in order to reflect the actual prices to farmers. Farm-produced intermediate inputs will have used up primary inputs in their creation and allowance was made for this, so that the off-farm and on-farm items are treated consistently.
The treatment of capital items in the national accounts raises the same on-farm or off-farm issue and adds to it a stock-flow problem. It is the flow of services from the capital stock that should enter as inputs in the denominator of TFP, along with the other flow variables. The most obvious cases are items such as machinery and vehicles, where the purchase price is much larger than the annual value of service flows. The aggregate agricultural statistics provide sufficient information for the service flow emanating from these capital stocks to be measured as the sum of running costs, interest and depreciation.
The three capital items included are land and fixed improvements, machinery and equipment and animal capital stocks. The values of these capital items were used to calculate shares. The shares were then converted into interest and depreciation flows that were deflated using the appropriate price indices. The interest components of the service flows on all three items were fixed at the low rate of two percent per year to reflect the considerable credit subsidies that were available to commercial farmers over most of the period3. The depreciation series mostly deflate machinery at ten percent per annum and fixed improvements on land at two percent per year. The running cost elements obviously differ for the different categories. The appropriate charges that were included in the land improvement category are maintenance and repairs to fencing and buildings. For the animal capital stock, the situation was slightly different. An interest element was calculated on the value of the herds. Depreciation is included in the sense that the animals are depreciated 100 percent in the year of slaughter and feed represents a large element of running costs. Rather than keeping feed separate it was included in the animal capital series to prevent this from turning negative in bad years when large numbers of animals were lost. This was done because it is necessary for each of the individual series to be positive, so that they could be transformed into logarithms.
Having attained independence over a decade before South Africa with the Lancaster House agreement in 1980, the new government's agricultural policy in Zimbabwe was based on an important policy document "Growth with Equity" (Republic of Zimbabwe, 1981). One by-product of this policy was that communal agriculture's contribution to the national income accounts appears in a Central Statistical Office publication (CSO, annual), which is available from 1975 onwards. Available from this source is a comprehensive list of outputs and intermediate inputs, not including land and labour, in current value terms that are used to calculate value added in communal agriculture.
The problem with measuring the productivity of smallholder agriculture in Zimbabwe, as in most of the rest of sub-Saharan Africa, is that land markets are generally not free and active in the communal areas. In addition, there is only a small amount of hired labour, and the wage that is paid is often a poor indication of the returns to widely used family labour (Atkins and Thirtle, 1995). In Kenya, for example, Carter and Wiebe (1990) found that the marginal product of labour applied on small farms was only a fraction of the market wage for labour, indicating significant imperfections in the market for labour. Thus, price information is lacking in Zimbabwe for the two basic factors of production that together account for the major part of the cost of inputs. Thirtle et al. (1993) used information on the price of land purchased by the government for resettlement schemes and the commercial wage as proxies in order to calculate weights for land and labour.
This approach is a reasonable one for the commercial sector of Zimbabwe, but when input prices vary as widely as they do in the communal sector, input costs need to be compared with the value of output for an accurate measure of TFP (Atkins and Thirtle, 1995). When this is done, it becomes apparent that the value of most communal land is well below the price paid by the government for resettlement land purchased from commercial farmers, and that the return to family labour in the communal sector is far below the commercial wage rate. Rather than imposing prices that are too high for the two main inputs in order to obtain weights for aggregation, the value of the intermediate inputs was subtracted from the value of output. The resulting residual, which could be considered value added, was divided between land and labour using share weights calculated from extensive gross margin calculations from University of Zimbabwe/Michigan State University (1992).
The effect of this new assumption (Atkins and Thirtle, 1995) is that all profits are attributed to labour and land inputs, rather than allowing a "return to managerial ability" to remain as part of the TFP index for Zimbabwe as it does for South Africa. TFP estimates for both South Africa and Zimbabwe, irrespective of this last assumption, are both Tornqvist-Theil approximations of Divisia TFP indices, and are therefore comparable. Given the degree of historic distortion in input markets in the communal and commercial sectors of these two countries, the sensitivity analysis in the following section explores the impact of assumptions about human capital on estimated TFP.
The studies analysed in section 7.3 provide a number of insights into the relationship between investment and agricultural productivity in Zimbabwe and South Africa. In this section the data described in the previous section are drawn on to look more closely at trends in total factor productivity in the three sectors to better understand how those trends have been influenced by patterns of investment in human capital and by government policies. The meta-analysis conducted here focuses on the labour component, distorted as it was by deliberate policy interventions over the course of the twentieth century.
Comparable TFP calculations for South Africa's commercial agricultural sector and for Zimbabwe's commercial and communal agricultural sectors are illustrated in Figures 7.1, 7.2 and 7.3. Moderate fluctuations and a rising trend since the mid-1960s characterize TFP in South Africa's commercial agricultural sector (Figure 7.1). Investment in R&D and extension services played a positive and significant role in explaining this trend (Thirtle, Sartorius von Bach and van Zyl, 1993; Khatri, Thirtle and van Zyl, 1996). Results for investment in education were mixed. During the 1970s, policy-generated market distortions in the form of subsidies led to over-investment in capital equipment by farmers. Subsequent reduction in these subsidies contributed to the observed acceleration in TFP growth during the 1980s.
FIGURE 7.1
TFP in South Africa's Commercial Agricultural Sector
FIGURE 7.2
TFP in Zimbabwe's Commercial Agricultural Sector
TFP in Zimbabwe's commercial agricultural sector has also been characterized by a rising trend since 1970, but has fluctuated considerably less than South Africa's commercial agricultural sector (figure 7.2). As in the South African case, investment in R&D and extension services explained most of the growth in TFP (Thirtle et al., 1993). While this sector was affected by the changes in policy favouring the communal sector brought about in the 1980s by the newly independent government, these changes produced similar drops in both input use and output in the commercial sector. As a result, TFP itself was largely unaffected.
FIGURE 7.3
TFP in Zimbabwe's Communal Agricultural Sector
FIGURE 7.4
TFP in South Africa's Commercial Agricultural Sector
TFP in Zimbabwe's communal agricultural sector was also characterized by a rising trend, but this growth trend was slower than in either of the two commercial sectors considered. It was also much more volatile, varying by as much as 100 percent from year to year. In contrast to findings with regard to the role of R&D in the two commercial sectors, Jayne et al. (1994) and Atkins and Thirtle (1995) find that investment in R&D was not significant in explaining changes in TFP in Zimbabwe's communal sector. It is important to recall that communal sector farmers in Zimbabwe were able to draw on a readily available stock of hybrid maize varieties that had already been nearly universally adopted by Zimbabwe's commercial farmers. Much more important than additional R&D, then, was the role of investment in infrastructure. Specifically, improvement in access to markets, including government-supported input-supply and output-procurement depots in the 1980s, were found to be very important by Jayne et al. (1994), Atkins and Thirtle (1995) and Dalton, Masters and Foster (1997).
The analysis involves testing the sensitivity of TFP to different assumptions about the quality of labour, particularly in light of the low wages earned by black farm employees. As an indicator of the contribution of these employees to agricultural productivity, one can ask what agricultural productivity would have been if these employees had not had very low reservation wages and had been able to command a higher wage. Investments in human capital through the rural education system, particularly in South Africa, may have had significant impacts on agricultural productivity through their effects on the quality of labour available from farm workers. This in turn could have allowed workers to command higher wages. Thus, adjusting these wage figures gives an informative counter-factual example of labour costs that can then be used to generate a new estimate of TFP.
In the mining sector, non-blacks earned more than four times as much as blacks. In the agricultural sector black regular-agricultural employees earned less than a quarter the wages paid to black mining employees, and black casual-agricultural employees earned even less. These figures are not strictly comparable, however, as agricultural workers generally received a greater share of their total compensation in forms other than financial remuneration, including clothing, household supplies, food and other goods and services. Agricultural workers in some cases were also given access to plots of land to cultivate, including via share tenancy arrangements (van Onselen, 1996).
As a result of prior discriminatory labour policies, there existed the potential for labour unions and other changes associated with the end of apartheid to increase the bargaining power of black agricultural workers in South Africa for higher wages. As a result, commercial farmers accustomed to an abundant supply of inexpensive labour are finding it increasingly difficult to compete. Some farmers report that while they had previously kept an excess supply of labour on the farm to meet seasonal periods of high activity, they are no longer able to do so (Kirsten, 1998). Using the same data and methods from the previous analyses of agricultural productivity in South Africa and Zimbabwe, the linkages between labour quality and agricultural productivity can be explored. In particular, how sensitive productivity estimates may be to changes in human capital investment and restrictive labour policies is considered. A Cobb-Douglas production function, which embodies unitary elasticity of substitution, is assumed. The labour aggregate is adjusted for the numbers of domestic servants and part-time farm employees, and the salaries for the different racial and gender categories are then multiplied by the numbers of each, in order to calculate a total money wage. The estimated value of in-kind wages is then added as a lump sum.
For this exercise, it is assumed that agricultural wage rates double with human capital investments, minimum wage legislation, labour union activities and the loosening of other restrictions on labour associated with independence and the end of the apartheid era. In keeping with the assumptions of the Cobb-Douglas model, the quantity of labour employed would be halved, keeping the overall wage bill constant. Working with the original spreadsheets, the effect of such a change on estimated agricultural productivity in the commercial and communal agricultural sectors of Zimbabwe and the commercial agricultural sector of South Africa is examined.
Based on the original labour data in South Africa's commercial agricultural sector, estimated TFP grew from a base of 100 in the late 1940s to more than 140 in the late 1980s, before falling back to about 130 in the early 1990s (Figure 7.1). By imposing the experimental change described above, doubling agricultural wages and halving agricultural labour employed, it is found that estimated TFP rises by more than 15 percent by the end of the period (Figure 7.4). This result can be explained by the substantial distortions that characterized South Africa's commercial agricultural sector throughout much of this period, especially those that resulted in the maintenance of excess supplies of labour in agriculture. In particular, if unskilled labour was employed in commercial agriculture to the point where its marginal product was not only low but actually negative, then reducing the amount of labour employed could increase efficiency and help explain the experimental result. However, it is important here to draw a careful distinction between estimates of productivity and efficiency, on the one hand, and implications for equity and social welfare on the other.
The same experiment can be performed on the data from Zimbabwe. As depicted in Figure 7.2, estimated TFP in Zimbabwe's commercial agricultural sector grew from a base of 100 in 1970 to about 220 in 1989. When wages are doubled and agricultural employment halved, estimated TFP is unchanged over the first decade, until about the time of independence in 1980, but then falls by about five percent relative to the original estimate over the course of the subsequent decade (Figure 7.5). Several factors are probably at work here. Agricultural inputs other than labour, like tractors, seeds and fertilizer, were far less readily available in Zimbabwe before 1980 than in South Africa. This was partially due to macro-economic policies to restrict imports of foreign technology. The concerns at the time were with maintaining a balance of payments, but the effect would have been to increase reliance on agricultural labour in Zimbabwe. Whereas in South Africa, agricultural labour would have been busy only during peak seasons, in Zimbabwe this condition would have persisted most of the time. With independence came a new set of policies designed to promote the adoption of modern technology, especially in maize production, and to increase opportunities for black farmers in the communal sector. It is at this time that increased demand for agricultural labour in the commercial sector would have manifested itself in Zimbabwe. In such circumstances, reducing labour employed while increasing wages, as in this experiment, would result in a decrease in productivity, particularly after independence in 1980.
FIGURE 7.5
TFP in Zimbabwe's Commercial Agricultural Sector
The improved opportunities for blacks in Zimbabwe following independence resulted in part from specific government policies to improve conditions in communal agriculture. Original estimates of TFP in Zimbabwe's communal agricultural sector depicted in Figure 7.3 reveal a slight upward trend that is dominated by wide variability over the 15-year period. In contrast to findings for the commercial sectors in Zimbabwe and South Africa, repeating the same experiment of doubling wages and halving agricultural employment in this sector results in no systematic change in estimated TFP (Figure 7.6). In terms of the earlier analysis of results for other sectors and countries, this could imply the absence of labour market distortions in the communal sector. Alternatively, and perhaps more likely, is the possibility that the communal sector relies on non-marketed (and thus unmeasured) inputs and outputs to such an extent that neither the original nor the experimental estimates of TFP fully capture what is happening on many African smallholdings. Furthermore, white labour in Zimbabwe's communal sector is virtually non-existent; if the experiment is interpreted in terms of a change in the quality of the labour input as proxied by racial differences, no change in TFP would be expected. Since the labour input is undifferentiated, changes that hold labour's share constant have no effect. Investments in educational infrastructure might be expected to increase the quality of the unskilled agricultural labour input, but as no substantial investments of this kind were made in the communal sector, no effect on TFP from the experiment are apparent.
FIGURE 7.6
TFP in Zimbabwe's Communal Agricultural Sector
Results of this experiment in the three agricultural sectors suggest that when data are scarce, estimates of TFP may be sensitive to alternative assumptions about differences in human capital investment, particularly in the context of policy distortions. This sensitivity may be reduced in cases, such as Zimbabwe's communal agricultural sector, where technology is relatively homogeneous.
Indicators of agricultural productivity show mixed trends in Zimbabwe and South Africa. Such partial measures as cereal yields are higher in the two countries than they are, on average, in sub-Saharan Africa as a whole. But land productivity in both Zimbabwe and South Africa is lower in terms of the value of agricultural output per hectare, and is growing more slowly, than it is in sub-Saharan Africa as a whole. Labour productivity is somewhat lower in Zimbabwe than it is in sub-Saharan Africa as a whole, and declined in both cases from 1980 to 1995. By contrast, labour productivity in South African agriculture is much higher, and grew over the same period. In terms of welfare implications, however, wage data in South Africa reveal sharp disparities in payments by race and sector, with black agricultural labour at the bottom of the scale.
As for more complete measures of agricultural productivity, Atkins and Thirtle (1995) estimated that total factor productivity (TFP) in Zimbabwe's communal sector grew by 8.1 percent annually in the early 1980s. The subsequent fall by 2.7 percent per year for the remainder of the decade was driven by reduced spending for costly post-independence policies supporting smallholder production, and also influenced by weather variation. In contrast, TFP growth in Zimbabwe's commercial sector was relatively stable, at about 4.0 percent annually in the 1970s and 1980s (Thirtle et al., 1993). TFP grew at an average annual rate of 1.3 percent in South Africa's commercial agricultural sector between 1947 and 1991, accelerating to 2.9 percent per year in the final decade leading up to independence (Thirtle, Sartorius von Bach and van Zyl, 1993). These rates compare with an average annual TFP growth rate of 1.3 percent for sub-Saharan Africa as a whole (with wide variation among countries) over the past three decades.
Of critical importance in facilitating increased efficiency in the use of conventional inputs is improved infrastructure. Jayne and Jones (1997) argue that policy in some areas must address two fundamental constraints, the first of which is infrastructure, and specifically weak transportation and communication systems. Poor infrastructure results in high transactions costs, reducing the ability of farmers to compete in both input and output markets.
The second fundamental constraint they identify is the need for increased investment in agricultural research to improve productivity growth and increase the stability of regional food production and prices. In this regard, productivity analysis that fails to account for policy changes and differences in the quality of infrastructure may overestimate payoffs to R&D, when in fact R&D may not generate the returns expected without complementary investments in physical and institutional infrastructure.
Similar biases are introduced by the lack of information on the quality of both conventional and non-conventional inputs to agricultural production. Lack of data on changes in resource quality over time, and on the physical relationships between such changes and agricultural production, remain formidable obstacles to projections of productivity trends into the future. The lack of information on non-market inputs and non-marketed outputs also poses challenges for measurement and analysis of agricultural productivity in sub-Saharan Africa, particularly in the smallholder sector. These limitations are evident in section 7.5, for example, in the observed lack of sensitivity of TFP estimates to experimental changes in wage and employment levels in Zimbabwe's communal agricultural sector.
Observed productivity trends, in conjunction with projected population growth, continue to warrant concerns about the prospects for sustainable production, income generation and food security in commercial and smallholder agriculture, both in Zimbabwe and South Africa in particular and in sub-Saharan Africa more generally. Commercial agriculture in South Africa demonstrates the potential benefits of investments in infrastructure, human capital and research, which are simultaneously needed to facilitate efficient use of conventional agricultural inputs and marketing of agricultural outputs. As Jayne et al. (1994) suggest, the challenge for sub-Saharan African countries facing budget pressures is how to redesign rather than abandon policy efforts to raise agricultural productivity in a sustainable fashion. At least as challenging is the need for further attention on the part of researchers and policy-makers to the equity implications of the alternative policy and investment strategies.
Atkins, J. & Thirtle, C. 1995. The productivity of communal agriculture in Zimbabwe, 1975-90. Oxford Agrarian Studies, 23(2): 99-115.
Carter, M.R. & Wiebe, K.D. 1990. Access to capital and its impact on agrarian structure and productivity in Kenya. American Journal of Agricultural Economics, 72(5): 1156-1150.
Central Statistical Office, Government of Zimbabwe. Annual. Production account, communal farming lands. Harare.
Dalton, T.J., Masters, W.A. & Foster, K.A. 1997. Production costs and input substitution in Zimbabwe's smallholder agriculture. Agricultural Economics, 17(2-3): 201-209.
Deininger, K. & Binswanger, H.P. 1995. Rent seeking and the development of large-scale agriculture in Kenya, South Africa and Zimbabwe. Economic Development and Cultural Change, 43: 493-522.
Eicher, C.K. 1995. Zimbabwe's maize-based green revolution: preconditions for replication. World Development, 23(5): 805-818.
Eicher, C.K. & Rukuni, M. 1990. Zimbabwe's experience in restructuring institutions to support communal farmers: lessons for Namibia. In C. Csaki, T. Dams, D. Metzger & J. van Zyl, eds. Agricultural restructuring in southern Africa. Windhoek, Namibia: International Association of Agricultural Economists.
Hudson, N.W. & Jackson, D.C. 1959. Results achieved in the measurement of erosion and runoff in southern Rhodesia. Inter-African Soils Conference. Dalaba, Central African Republic.
Jayne, T.S. & Jones, S. 1997. Food marketing and pricing policy in eastern and southern Africa: a survey. World Development, 25(9): 1505-1527.
Jayne, T.S., Khatri, Y., Thirtle, C. & Reardon, T. 1994. Determinants of productivity change using a profit function: smallholder agriculture in Zimbabwe. American Journal of Agricultural Economics, 76: 613-618.
Jayne, T.S., Rubey, L., Chisvo, M. & Weber, M.T. 1996. Zimbabwe food security success story: maize market reforms improve access to food even while government eliminates food subsidies. Policy Synthesis No. 18, Office of Sustainable Development and Office of Agriculture and Food Security, Global Bureau, USAID. April.
Kassier, W.E. & Groenewald, J.A. 1990. The agricultural economy of South Africa. In C. Csaki, T. Dams, D. Metzger & J. van Zyl, eds. Agricultural restructuring in southern Africa. Windhoek, Namibia: International Association of Agricultural Economists.
Khatri, Y., Thirtle, C. & van Zyl, J. 1996. Public research and development as a source of productivity change in South African agriculture. South African Journal of Science, 92: 1-9 (galley proof).
Kirsten, J. 1998. Personal communication. (Chair, Department of Agricultural Economics, University of Pretoria, Gauteng, South Africa).
Lal, R. 1998. Soil erosion impact on agronomic productivity and environment quality. Critical Reviews in Plant Sciences, 17(4): 319-464.
Marquette, C.M. 1997. Current poverty, structural adjustment and drought in Zimbabwe. World Development, 25(7): 1141-1149.
Meinzen-Dick, R.S. & Rosegrant, M.W. 1997. Managing water supply and demand in southern Africa. In L. Haddad, ed. Achieving food security in southern Africa. Washington, DC: International Food Policy Research Institute.
Mudimu, G.D. 1990. Achieving and maintaining national and household food security: Zimbabwe's experience and issues for the 1990s. In C. Csaki, T. Dams, D. Metzger & J. van Zyl, eds. Agricultural restructuring in southern Africa. Windhoek, Namibia: International Association of Agricultural Economists.
Paul, P. & Abey, R. 1984. Conceptual issues in measuring productivity growth. Paper presented at the 28th Annual Conference of the Australian Agricultural Economics Society, Sydney, February.
Republic of Zimbabwe. 1981. Growth with equity. Harare.
Rohrbach, D.D. 1990. Strengthening crop production and marketing systems for semi-arid areas in the SADCC region. In C. Csaki, T. Dams, D. Metzger & J. van Zyl, eds. Agricultural restructuring in southern Africa. Windhoek, Namibia: International Association of Agricultural Economists.
Rubey, L. & Lupi, F. 1997. Predicting the effects of market reform in Zimbabwe: a stated preference approach. American Journal of Agricultural Economics, 79(1): 89-99.
Tagwira, F. 1992. Soil erosion and conservation techniques for sustainable crop production in Zimbabwe. Journal of Soil and Water Conservation, 47: 370-374.
Thirtle, C., Atkins, J., Bottomley, P., Gonese, N., Govereh, J. & Khatri, Y. 1993. Agricultural productivity in Zimbabwe, 1970-90. Economic Journal, 103: 474-480.
Thirtle, C., Sartorius von Bach, H. & van Zyl, J. 1993. Total factor productivity in South African agriculture, 1947-91. Development Southern Africa, 10(3): 301-318.
Thirtle, C., Townsend, R. & van Zyl, J. 1998. Testing the induced innovation hypothesis: an error correction model of South African agriculture. Agricultural Economics, 19(1-2): 145-157.
United States Department of Agriculture. 1980. Measurement of US agricultural research productivity: a review of current statistics and proposals for change. Technical Bulletin No.1614, Economics, Statistics and Cooperatives Service, Washington, D.C.
University of Zimbabwe & Michigan State University. 1992. Documentation system summary: a guide to the Zimbabwe food security and marketing 1990-92: survey data. Harare.
van Onselen, C. 1996. The seed is mine: the life of Kas Maine, a South African sharecropper, 1984-1985. Hill & Wang Publishers, ISBN 080909603X.
van Schalkwyk, H. & Groenewald, J. 1992. Regional analysis of South African agricultural resource use and productivity. Agrekon, 31(3): 116-127.
van Zyl, J., Fenyes, T.I. & Vink, N. 1992. Effects of the farmer support programme and changes in marketing policies on maize production in South Africa. Journal of Agricultural Economics, 43(3): 466-476.
van Zyl, J. & Groenewald, J.A. 1988. Effects of protection on South African commercial agriculture. Journal of Agricultural Economics, 39(3): 387-401.
van Zyl, J., Vink, N. & Fenyes, T.I. 1987. Labour-related structural trends in South African maize production. Agricultural Economics, 1(3): 241-258.
World Bank. 1998. 1998 world development indicators. Washington, DC: The World Bank.
World Resources Institute, United Nations Environment Programme, United Nations Development Programme & the World Bank. 1998. World resources 1998-99. New York: Oxford University Press.
1 The authors are grateful to Colin Thirtle, Rob Townsend and their colleagues for making their spreadsheets with original data and calculations available for further analysis and for frank discussions of the limitations of both the data and the calculations. The comments of Shahla Shapouri and Lydia Zepeda are gratefully acknowledged. The views expressed are those of the authors and may not be attributed to the Economic Research Service.
2 A crop-year is the period during which a crop acutally grows even if that period happens to include more than one calendar year, as it often does in the Southern Hemisphere.
3 An interesting element is included to reflect the opportunity cost of hloding farm capital, and this has led observers such as Paul and Abey (1984) to recommend incorporation of the real after-tax return on a safe asset. Such a series would include negative values and the choice of real or nominal rates affects the capital input series considerably during inflationary periods like the 1970s. These authors followed USDA (1980) and used an arbitrary fixed-rate. This appears to be the safest course, until more work has been done to support a more elaborate treatment of this difficult issue.
