4.1 Introduction

4.2 Data and Measurement Issues in Estimating SSA Agricultural Productivity

4.3 Existing Research on Sub-Saharan Agricultural Productivity

4.4 Bottlenecks to Growth

4.5 The Role of Policy

4.6 Conclusions

References

Agricultural Productivity for Sustainable Food Security
in Sub-Saharan Africa
Keith D. Wiebe, Meredith J. Soule and David E. Schimmelpfennig¹

This study examines trends in agricultural productivity in sub-Saharan Africa, identifying sources of growth as well as bottlenecks to growth. Existing research has consistently found that about three-quarters of the variation in agricultural productivity in sub-Saharan Africa can be explained by use of conventional inputs. For sub-Saharan Africa to meet its food security needs in the next ten years, will require one to two percent greater agricultural production per year than even the most optimistic projections. Policy reforms to improve physical infrastructure, political stability and the institutional environment are needed to facilitate access to and incentives to use conventional inputs as well as increase the application and returns to research.

4.1 Introduction

Agriculture is the principal source of food, livelihood and foreign exchange earnings in sub-Saharan Africa (SSA) (Badiane and Delgado, 1995). Production is a particularly important component of food security in SSA; commercial imports will contribute less than 10 percent of food supply over the next ten years (Rosen, 1997). As a result, agricultural productivity is critical to SSA's ability to meet food security and economic development objectives in the face of rapid population growth. SSA is projected to have the world's highest population growth rate over the next decade, at 2.5 percent per year, with nearly three-quarters of its workforce employed in agriculture (World Bank, 1998).

Yet evidence of agricultural performance in SSA is mixed at best. Total factor productivity in agriculture is estimated to have grown by an average of 1.3 percent annually between 1961 and 1991 for Africa as a whole (Lusigi and Thirtle, 1997). Land productivity in SSA agriculture rose by an average of 1.9 percent per year between 1980 and the mid-1990s, while increasing by 3.4 percent and 2.0 percent annually in South Asia and Latin America and the Caribbean, respectively (World Bank, 1998). Over the same period crop production in SSA grew by 2.7 percent per year, and food production grew by 2.4 percent per year. By contrast, labour productivity fell by an average of one percent per year in SSA agriculture. It increased by 1.9 percent and 2.5 percent per year, respectively, in South Asia and Latin America and the Caribbean. Complicating the differences in these indicators of agricultural productivity at the regional level are differences in the level and rate of change in each indicator across subregions and countries within SSA.

TABLE 4.1
Agricultural Productivity Levels and Trends

	Land productivity		Labour productivity		Labour/Land	Total factor productivity
	Level ($/ha, 1993)	Growth rate (%/year, 1980-1993)	Level ($/worker,1995)	Growth rate (%/year, 1980-1995)	Ratio (workers/ha, 1991)	Level (index, 1991)	Growth rate (%/year, 1961-1991)
Sub-Saharan Africa	68	1.9	392	-1.0	0.30a	0.8a	1.3a
Central Africa
Cameroon	313	1.7	827	-0.3	0.17	0.9	1.8
Central African Rep.	119	1.7	516	0.8	0.18	0.6	2.7
Congo	28	2.2	629	1.0	0.04	0.9	1.2
Gabon	74	0.8	1 516	0.5	0.05	0.3	-2.3
Zaire	113	2.4	219	0.0	0.34	0.6	8.1
East Africa
Burundi	270	1.9	177	-1.4	1.01	2.9	3.4
Ethiopia	116	na	181	na	0.22	0.2	-1.7
Kenya	90	2.2	240	-0.7	0.13	0.6	1.9
Madagascar	34	2.1	178	-0.4	0.09	0.7	-0.1
Rwanda	378	-1.2	206	-2.6	1.42	0.7	6.1
Uganda	515	na	592	na	0.65	2.9	7.8
Sahel
Burkina Faso	93	2.9	182	1.1	0.25	0.1	0.8
Chad	10	4.0	198	2.0	0.16	0.6	0.2
Gambia	199	1.6	167	-1.7	1.00	0.4	-1.5
Mali	33	2.5	259	0.2	0.06	0.8	0.8
Mauritania	7	2.6	na	na	0.01	0.1	-0.3
Niger	63	0.8	256	-0.9	0.20	0.7	1.5
Senegal	118	1.9	375	0.9	0.34	0.6	1.5
Somalia	na	na	na	na	0.05	0.8	1.2
Sudan	na	na	na	na	0.04	0.8	0.1
West Africa
Benin	321	4.2	563	2.8	0.57	1.1	1.2
Côte d'Ivoire	212	0.6	1 354	-0.8	0.13	1.1	0.9
Ghana	227	0.4	684	-1.1	0.29	0.7	-0.5
Guinea	54	na	225	na	0.24	0.5	1.2
Guinea-Bissau	78	2.9	292	3.1	0.21	0.1	-2.1
Liberia		na	na	na	0.08	1.1	0.0
Nigeria	150	2.3	684	2.4	0.28	0.6	-0.3
Sierra Leone	123	0.4	344	-0.4	0.30	0.5	0.5
Togo	189	3.6	461	0.9	0.32	0.4	-1.3
Southern Africa
Angola	9	na	149	na	0.07	0.4	-0.8
Botswana	5	1.7	483	1.4	0.01	0.2	1.3
Lesotho	24	-2.9	194	-2.7	0.23	0.1	-1.7
Malawi	153	0.4	156	-0.3	0.69	0.7	0.3
Mozambique	12	na	92	na	0.11	0.3	0.3
Namibia	9	0.9	1 458	0.8	0.004	1.4	1.0
South Africa	49	0.7	2 870	1.3	0.02	1.4	1.3
Swaziland	na	na	na	na	0.12	1.2	3.3
Tanzania	na	na	na	na	0.20	0.4	0.2
Zambia	7	1.2	100	-1.0	0.04	0.5	1.5
Zimbabwe	41	1.5	266	-0.7	0.27	0.7	2.0

Sources: Lusigi and Thirtle (1997), World Bank (1998). ^aAverages for all Africa. "na" no data available.

Trends in agricultural productivity in sub-Saharan Africa

Table 4.1 presents three measures of agricultural productivity and their rates of change, by region and country, based on data from the World Bank (1998) and Lusigi and Thirtle (1997). It is important to note that the value of agricultural output used in generating these statistics is itself difficult to estimate because a large share of agricultural output is either consumed within the household that produces it, or is exchanged for commodities other than money (World Bank, 1998). The value of output is often estimated indirectly using a combination of methods, including reliance on estimates of yields and area under cultivation. To the extent that such attempts underestimate the true magnitude of agricultural output, estimates of agricultural productivity will themselves be biased downwards.

Output per unit of land, or crop yield, is commonly used by agricultural scientists to assess the success of new production practices. Land productivity is also used by national policy-makers to assess agricultural production for meeting national food security needs. Output per agricultural worker, on the other hand, may be a more important indicator of rural standards of living and welfare (Block, 1995). Recognizing that food may be acquired through exchange as well as production, income becomes an important determinant of access to food and thus of food security. As a result, labour productivity may be particularly important as an indication of the ability of agricultural workers to acquire sufficient food, whether or not they produce food themselves.

Land productivity averaged US$68 per hectare of agricultural land (measured as the sum of arable land, permanent cropland and permanent pasture) for SSA as a whole in 1993, compared with $519 in South Asia and $116 in Latin America and the Caribbean. Values ranged from $5-10 per hectare in the drier countries of Southern Africa and the Sahel to $200 per hectare and more in the East African highland countries and tropical West Africa. For SSA as a whole, land productivity grew at an average rate of 1.9 percent per year between 1980 and 1993, with slow to moderate growth in most countries. Land productivity grew most rapidly in the Sahelian countries and West Africa, and more slowly in Eastern and Southern Africa (Table 4.1). These patterns in land productivity levels and growth are also depicted in the second map in Figure 4.1.

Labour productivity averaged $392 per agricultural worker for SSA as a whole in 1995, this compares with $383 in South Asia and $2 292 in Latin America and the Caribbean. Values ranged from $100-$200 per worker in many countries in Eastern and Southern Africa and the Sahel to more than $500 per worker in parts of West and Central Africa. Labour productivity declined at an average rate of 1.0 percent per year for SSA as a whole between 1980 and 1995, with modest growth in West Africa and declines in Eastern and Southern Africa (Table 4.1). The first map in Figure 4.1 illustrates patterns of labour productivity. In contrast to the pattern evident for land productivity, the most rapid growth in labour productivity appears to be taking place in those countries that already have the highest levels of labour productivity. Perhaps this reflects the pull of off-farm employment opportunities, suggesting that disparities in labour productivity across countries may increase over time.

Low (or declining) labour productivity is consistent with high (or growing) land productivity in the context of a large (or expanding) agricultural labour force. Such patterns are evident in the agricultural labour/land ratios reported by Lusigi and Thirtle (1997) and presented in Table 4.1. The labour/land ratio is generally high in East Africa and low in Central and Southern Africa and the Sahel.

Land and labour productivity measures are both incomplete indicators of agricultural productivity, since they measure the productivity of only a single factor of production, and may well move in opposite directions. In an effort to address this problem, economists estimate total factor productivity (TFP), which measures changes in agricultural output relative to changes in an aggregated index of multiple inputs. TFP growth reflects factors such as technical change or improvements in infrastructure or research. It can also reflect failure to include or measure inputs such as depletion of soil or other natural resources.

FIGURE 4.1
Agricultural Productivity Levels and Trends in Sub-Saharan Africa

The the last two columns of Table 4.1 and third map in Figure 4.1 report estimates by Lusigi and Thirtle (1997) of TFP levels for 1991 and rates of change for 1961-91.² For Africa as a whole the TFP index estimated by Lusigi and Thirtle averaged 0.8. This indicates that on average, the productivity of the set of inputs measured was 0.8 times as high in 1991 as it was in the most efficient countries in 1961, up from an average of 0.7 in 1961. Levels are generally mixed, with the highest estimates in Uganda and Burundi, suggesting the importance of land quality, availability of water, and labour supply in driving agricultural TFP (Lusigi and Thirtle, 1997). Uganda also has one of the highest rates of growth in TFP, averaging 7.8 percent per year since 1961. For Africa as a whole, TFP grew at an average rate of 1.3 percent per year over the period.

4.2 Data and Measurement Issues in Estimating
SSA Agricultural Productivity

Data on agricultural outputs and inputs are costly to collect. Sub-Saharan African countries have limited budgets devoted to data collection, with the result that data on both conventional and non-conventional inputs are often unavailable or incomplete. For example, in the United States, the definition of conventional inputs has expanded to include pesticides, energy, feed, seeds and intermediate livestock inputs. When an input such as pesticides is left out, increased output that might be attributed to increased pesticide use may be incorrectly attributed to TFP growth instead.

Inadequacies in the international data set for productivity analysis have been pointed out by a number of researchers (Wiggens, 1998; Craig, Pardey and Roseboom, 1997; Trueblood, 1991). To date, most research has concentrated on measuring productivity in new ways with the same existing and insufficient data. Efforts have been made in recent years to improve the data, such as constructing a data set of public agricultural research expenditures by country (Pardey, Roseboom and Anderson, 1989), but much work still needs to be done.

Aggregating Agricultural Output

In order to aggregate agricultural output for international consistency, output must be measured in a common unit. Typically, output has either been reported in terms of dollars or in terms of "wheat units" (Hayami and Ruttan, 1985; Block, 1995). The wheat units approach was developed by Hayami and Ruttan (1985) and is based on the ratio of each individual commodity price to the price of wheat in India, the United States and Japan. Official exchange rates are generally considered to be a poor choice for converting output in local currency units to dollars due to the biases introduced by fixed exchange rates or sudden devaluations. Most researchers use the purchasing power parity exchange rates inherent in the Food and Agriculture Organization of the United Nations (FAO) international dollar concept. However, Block (1995) argues that wheat units are preferable due to the impact of annual price movements that can affect the agricultural value added to which the international dollar conversions are applied. However, most recent studies have used FAO's international dollars (Trueblood and Coggins, 1997; Craig, Pardey and Roseboom, 1997).

Conventional Factors of Production

Conventional inputs to agricultural production are land, labour, physical capital, livestock and fertilizer. For international comparison studies, the source of most data on these inputs is FAO. Conventional inputs are typically measured in relatively simple physical terms that mask potentially important qualitative variations.

Land is typically measured as hectares of agricultural land, i.e. arable and permanent cropland and permanent pasture (Fulginiti and Perrin, 1997). FAO statistics indicate that Africa had just over one thousand million hectares of agricultural land in 1990, up 0.1 percent annually in the previous two decades (Table 4.2). This measure does not account for land quality. Failure to account for land quality may lead researchers to incorrectly attribute to other inputs differences in production that are actually due to differences or changes in land quality. Some attempts have been made to control for differences in land quality by including a land quality index as a non-conventional input. Such attempts are noteworthy (Craig, Pardey and Roseboom, 1997), but they have been able to apply only one land quality indicator per country. This is problematic for large countries that span several ecozones.

TABLE 4.2
Levels and Trends in African Agricultural Output and Inputs, 1970-1990

	Level		Growth rate
	1970	1990	(percent/year, 1970-1990)
Agricultural output (index, base 1989-1991)	66	98	2.0
Agricultural land (millions of hectares)	1 059	1 090	0.1
Agricultural labour (millions)	120	167	1.7
Tractors in use (thousands)	334	521	2.2
Cattle (millions)	149	188	1.2
Fertilizer consumption (thousands of tonnes)	1 615	3 686	4.2

Source: FAO (1999).

The proxy for agricultural labour is often the economically active population in agriculture. Early FAO statistics only included males in the agricultural labour force. More recent data have included both males and females. Table 4.2 shows that Africa's agricultural labour force grew 1.7 percent annually between 1970 and 1990, to 167 million. However, this agricultural labour force variable still does not control for differences across countries in the composition (and thus potentially the quality) of the agricultural labour force by age and education. An additional problem with the FAO data is that the economically active population in agriculture is defined to include workers in agriculture, forestry and fisheries. This implies that the number of workers is overstated for every country, and is more heavily overstated for countries with large forestry and/or fishery sectors relative to their basic agriculture sectors. A few researchers have made attempts to correct for the quality of the agricultural labour force by including national-level measures of education or literacy as non-conventional inputs. No researchers of SSA have adjusted the quality of labour directly by sex, age and education.

Another problem is that many of these economically active agricultural workers are not employed full-time in agriculture. Evidence from Africa suggests that many farmers are heavily involved in off-farm work to supplement their farm incomes (Reardon, Delgado and Matlon, 1992). In such cases, the agricultural labour force may look unduly large and thus bias estimates of labour productivity downwards.

The use of physical capital is typically measured by the stock of tractor horsepower (Hayami and Ruttan, 1985). The number of tractors in use in African agriculture increased 2.2 percent annually between 1970 and 1990, to 521 thousand (Table 4.2). Such a measure is problematic in SSA where many farmers continue to use hand implements, especially in hilly regions where tractors are ill-suited. Farmers' investments in hand hoes, carts, ploughs, fencing, buildings and other locally produced capital inputs have not been accounted for in national and international productivity studies. Incomplete measurement of physical capital, in terms of quality and quantity, will bias productivity estimates. In an effort to improve measurement of physical capital, Craig, Pardey and Roseboom (1997) have updated the physical capital variable to include two- and four-wheel tractors that are converted to horsepower using regional averages.

Livestock is a difficult input to measure since livestock may serve as both an input and an output in agricultural production. As an input, livestock has been measured as the number of livestock on farms at a given point in time (Hayami and Ruttan, 1985). Kawagoe, Hayami and Ruttan (1985) included all livestock as an input, arguing that they represent long-term capital formation in the agricultural sector. Arnade (1997) did not include livestock as an input to production, arguing that in developing countries that do not have meat-processing sectors, livestock is usually sold directly as an output. Craig, Pardey and Roseboom (1997) included livestock as an input, but only included those animals that are primarily used for traction or breeding services. Clearly, differences in how the livestock variable is treated will affect estimates of the levels, sources, and changes in agricultural productivity. Table 4.2 shows a total of 188 million cattle in Africa in 1990, up 1.2 percent annually since 1970.

Commercial fertilizer inputs are measured as tons of nutrient units of nitrogen, phosphorus and potash. FAO data indicate that fertilizer consumed in African agriculture in 1990 totalled 3.7 million tonnes, up 4.2 percent annually since 1970 (Table 4.2). Fertilizer consumption subsequently declined by 1.1 percent per year between 1990 and 1995. However, the fertility benefits of organic sources of nutrients, such as manure and legumes, are not accounted for in this measure. This omission is potentially significant given widespread reliance on organic sources of nutrients in SSA.

Non-Conventional Factors

Non-conventional factors include private and public agricultural research, education, infrastructure, government programmes and policies, and environmental degradation. Sometimes in an attempt to adjust conventional inputs for quality, researchers have included these variables in the set of non-conventional inputs. Examples include land quality indicators (Frisvold and Ingram, 1995), or proxies for agricultural labour quality, such as literacy and life expectancy (Craig, Pardey and Roseboom, 1997).

At the national level, public agricultural research expenditures are generally used as a proxy for research and development. Public agricultural research expenditures are typically lagged for a number of years to compensate for the time required for research to reach fruition. However, this measure does not account for the spillover of research that is easily transferred from other countries. Private research expenditures have not been included in studies of developing countries since that information has not been collected.

Education is related to the quality of the agricultural labour force. For example, literacy would be expected to improve a farmer's ability to make use of information provided by extension services, or to keep better track of the costs and returns to alternative inputs or marketing opportunities. More generally, a more educated populace may also provide better services to agriculture, improving agriculture's productivity even without changing the quality of the agricultural labour force directly. Since no data are available specifically on the educational level of the agricultural labour force in most countries, national-level proxies are used. Education has been measured by the school enrolment ratio or the adult literacy rate (Hayami and Ruttan, 1985). More generally, the overall quality of the labour force has been measured by national life expectancy (Craig, Pardey and Roseboom, 1997) and by historic calorie availability (Frisvold and Ingram, 1995). In an effort to focus more specifically on the education achievements of the agricultural labour force, Hayami and Ruttan (1985) also looked at the number of agricultural college graduates as a proxy for the level of advanced technical education in agriculture.

Public investments in infrastructure such as roads, utilities, and communications can increase agricultural productivity as well, by lowering the cost of inputs at the farm level and increasing farmers' access to marketing opportunities. Proxy variables include paved road density (Craig, Pardey and Roseboom, 1997) or by gross domestic product of each country's transportation and communication sectors (Hu and Antle, 1993).

Government programmes and policies also affect agricultural productivity. For example, Fulginiti and Perrin (1993) argue that historic agricultural output and input prices affect the technology chosen by farmers, and thus drive observed productivity patterns. Prices may be affected by government policies that tax or subsidize agriculture, and a "net protection coefficient" is used to capture the effect of these policies on agricultural productivity by Hu and Antle (1993) and Fulginiti and Perrin (1997). Block (1995) used depreciation of the real exchange rate as a proxy for government policy reform. The past export growth rate and export instability (Frisvold and Ingram 1995) have also been used as proxies for government policies that might affect productivity. They argue that export growth tends to stimulate overall economic development and productivity growth. They also note that export instability might slow productivity growth.

Researchers have used several variables in an attempt to adjust for the impact of land quality differences on productivity. Several studies have used a land quality index created by Peterson (1987) that indexes land quality at the national level as a function of historic precipitation and the share of a country's land area devoted to pasture and crops. Researchers have also used the percentage of a country's land that is arable, the percentage of land that is irrigated and mean rainfall to adjust for variations in land quality across countries.

Environmental degradation and actions that farmers take to reduce or reverse degradation have been recognized as potentially significant inputs to the production process (Thrupp, 1997), but they have not yet been measured and included as explanatory variables in productivity studies due to the scarcity of nationally or internationally comparable data.

4.3 Existing Research on Sub-Saharan Agricultural Productivity

Research on cross-country agricultural productivity comparisons has concentrated on two areas. The first type of studies has estimated econometrically multi-country aggregate or meta-production functions to explain cross-country variation in agricultural productivity and to estimate production elasticities. The second, more recent type of studies has used data envelopment analysis (DEA) to construct Malmquist TFP indices to show which countries are experiencing the highest or lowest rates of growth.

Aggregate Production Functions

Table 4.3 compares estimated Cobb-Douglas coefficients from studies that estimated separate equations for developed countries (DCs), developing countries (LDCs) and African countries. Kawagoe, Hayami and Ruttan (1985) split their sample of 43 countries into 21 DCs and 22 LDCs. They found all the conventional variables as well as technical education to be important in explaining output levels for the DCs. For the LDCs, land and fertilizer were not found to be significant explanatory variables, but livestock was more important when compared to the DCs. In addition, both the school enrolment ratio and technical education were found to be significant. Summing over the coefficients of conventional inputs, Kawagoe, Hayami and Ruttan found increasing returns to scale in the DCs and constant returns to scale in the LDCs. They explain that an increasing substitution of large machines for labour has led to the pattern of increasing returns to scale observed in DC agriculture. On the other hand, they argue that many of the productivity increases realized in LDC agriculture have come from the increased use of high-yielding seed varieties and fertilizer, which are scale-neutral. They conclude that there is still much scope in LDC agriculture to increase productivity, and in particular labour productivity, by increasing investment in education, research, and modern inputs.

TABLE 4.3
Cobb-Douglas Aggregate Production Function Estimates (OLS), Grouped by
Developed Countries (DCs), Developing Countries (LDCs) and African Countries

	Kawagoe, Hayami & Ruttan (1985)	Kawagoe, Hayami & Ruttan (1985)	Fulginiti & Perrin (1993)	Craig, Pardey & Roseboom (1997)	Frisvold & Ingram (1995)	Lusigi & Thirtle (1997)a
years covered	1957-1980	1957-1980	1961-1985	1961-1990	1973-1985	1961-1991
no. of countries	21 DCs	22 LDCs	18 LDCs	67 LDCs	28 African	47 African
no. of African countries	0	1	3	25	28	47
Conventional inputs
Labour	0.71*	0.56*	0.25*	0.41*	0.59*	0.21*
Land	0.10*	-0.07	0.25*	0.33*		0.19*
Fertilizer	0.19*	0.09	0.18*	-0.01	0.022*	0.16*
Livestock	0.15*	0.32*	0.17*	0.21*	0.18*	0.23*
Machinery	0.18*	0.14*	0.21*	0.06*	0.04*	0.04*
Animal traction hp				-0.02
Non-conventional inputs
school enrollment ratio	-0.17	0.41*	0.3*
technical eduction	0.14*	0.17*
life expectancy				1.64*
adult literacy				-0.05
land quality index			0.51*		0.89*	0.28*
mean rainfall				0.27*
% land arable				0.35*
% land (not) irrigated				-0.37*	0.45*
research expenditures			-0.02	0.09*	0.09	0.03*
infrastructure				0.48
past output price			0.13*
past wages			-0.09*
past fertilizer price			0.03
export growth					0.03*
export instability index					0.0004
calorie availability					0.35*
Dummy variables
time (t+1)	0.09	-0.22*				0.0004*
time (t+2)	0.17*	-0.43*
sum of conventional Coefficients	1.32	1.04	1.06			0.83

a This study used corrected ordinary least squares.
* indicates significance at the 0.05 level.

Fulginiti and Perrin (1993) conducted a study of 21 countries that included three African countries. They were particularly interested in examining how past pricing policies affected technology choice and thus current agricultural productivity. In addition to conventional inputs, they included past output prices, past wages and past fertilizer prices as well as the school enrolment ratio, a land quality index and research expenditures as non-conventional inputs. They found all of the conventional inputs to be significant as well as fairly close to the range of estimates found in other studies, except for land, which had a larger coefficient (0.25). They argued that previous studies, which did not include country-specific effects such as the land quality index used in their study, tended to bias the land elasticity downward. Higher school enrolment ratios are associated with higher productivity. The coefficient on the past output price variable was 0.13, indicating that a one percent change in past output price expectations would lead to a 0.13 percent shift in the production function. Echoing Hu and Antle's (1993) study of non-African countries, they argue that eliminating policy interventions that tax agriculture would increase productivity, while elimination of subsidies would decrease productivity.

Fulginiti and Perrin (1993) also estimated productivity increases that would be possible from eliminating output price policies that tax agriculture. The estimated productivity increases associated with such policy reforms ranged from 1.4 percent in Colombia to 129 percent in Zambia. Côte d'Ivoire and Ghana had estimated productivity increases of 88 percent and 16 percent, respectively. Fulginiti and Perrin tried several different formulations of the research expenditure variable, and found it to be significant only when measured as the work-years of government agricultural research, cumulated into a stock using a five-year lag. The sensitivity of these results with respect to different variable specifications reinforces the importance of measurement issues in evaluating agricultural productivity.

Frisvold and Ingram (1995) estimated land productivity for 28 countries in sub-Saharan Africa between 1973-1975 and 1983-1985. They estimated land productivity grew at an annual rate of 1.5 to 1.8 percent in most regions over the period. All conventional inputs were found to be significant, and the coefficient on labour was particularly large at 0.59. Frisvold and Ingram found that increased application of agricultural labour was the single most important factor in explaining growth in land productivity, and concluded that substantial increases in land productivity should not be expected until land becomes relatively scarce, echoing Binswanger and Pingali (1988) and Boserup (1965). They also found land quality, as measured by the Peterson (1987) index, to be an important explanatory variable, as was the percentage of area irrigated. Similar to Fulginiti and Perrin (1993), they did not find research expenditures to be significant. However, they did show that export growth and historic calorie availability have contributed positively to productivity growth in SSA. Growth in the stock of conventional inputs as a whole accounted for more than two thirds of growth in land productivity, which in turn accounted for the majority of growth in agricultural output.

Lusigi and Thirtle (1997) used a sample of 47 countries in Africa over the period 1961-1991. They estimated an average rate of TFP growth of 1.3 percent per year over that time period. Livestock, labour and land are the most important inputs explaining variation in productivity across the countries, with population pressure (and thus expansion of the agricultural labour force) an important factor driving productivity growth. Fertilizer and machinery are significant but of less importance. Like Frisvold and Ingram (1995), Lusigi and Thirtle stressed the contribution of population pressure to faster growth, arguing that land abundance depresses farmer incentives to increase land productivity by adopting yield-increasing technologies. They also found land quality and research expenditures to be important explanatory variables.

While the significance of specific inputs varies across studies using the production function approach, these studies have found in general that increased use of conventional inputs, especially those other than land, are the most important factors in explaining growth in agricultural productivity and output. Studies that have included measures of infrastructure and of land and labour quality have shown that variation in those factors is also important. The estimated effect of research on agricultural productivity varies by region studied, how the research variable is specified, and on the lag length selected.

The Malmquist Index

A traditional Tornqvist TFP index, which requires data on input prices and is used for calculating agricultural productivity in the United States, has never been constructed for SSA or used in multi-country production studies because the required data are not available. Instead, studies of agricultural productivity in SSA have used either a production function approach or a Malmquist TFP index, neither of which requires data on input prices. The Malmquist TFP studies are relatively new and have provided some insight into the relative ranking of countries in terms of productivity. However, researchers have not yet attempted to explain the differences in productivity across countries.

Data envelopment analysis (DEA) for creating Malmquist indexes of total factor productivity is a non-parametric programming approach that uses data on physical inputs and outputs to model efficiency levels. DEA identifies the best-practice countries out of a set of countries under study. The best-practice countries are those which minimize inputs per unit of output. Those countries then define the efficient production frontier. The efficiency of all other countries under study is measured relative to that efficient production frontier. Technical change is determined by the position of a country relative to the changing position of the efficient frontier over time.

Since the efficient production frontier depends on the set of countries included in the study, the productivity growth rates calculated by each study will depend on the countries included in the analysis. For example, Table 4.4 compares the TFP growth rates calculated from Malmquist indexes for selected SSA countries included in three different studies. Lusigi and Thirtle (1997) calculated the Malmquist TFP growth rates based on a set of 47 African countries for the period 1961-1991. Trueblood and Coggins (1997) constructed their Malmquist TFP growth rates by analysing 117 countries from all world regions for 1962-1990. Thirtle, Hadley and Townsend (1995) estimated growth rates for 22 SSA countries over the period 1971-1986.

The differences in results due to the selection of countries and time periods are striking. Relative to the efficient frontier for Africa alone, Lusigi and Thirtle (1997) show that the majority of SSA countries had positive TFP growth rates. Zaire, Uganda, Rwanda and Burundi (3.4 to 8.1 percent per year) achieved the highest rates of growth. In comparison, Trueblood and Coggins (1997) showed negative TFP growth rates for most SSA countries over the period 1962-1990 when they were compared to a global set of countries encompassing a much wider range of technologies, suggesting the technology gap between Africa and the rest of the world is widening. The ranking of SSA countries with positive growth rates were Congo (Brazzaville), South Africa, Uganda, Benin, Sierra Leone and the Central African Republic. Two of the highest-ranking countries in the study by Lusigi and Thirtle (1997), Rwanda and Burundi, had nearly the lowest growth rates of TFP in the worldwide study (-13.9 and -16.4 percent, respectively). Thirtle, Hadley and Townsend (1995) find low but positive TFP growth rates for most of the 22 SSA countries they studied for 1971-1986, with the highest rates in Rwanda and Burundi.

TABLE 4.4
Comparisons of Malmquist TFP Growth Rates for African Countries

	Lusigi & Thirtle	Trueblood & Coggins	Thirtle, Hadley & Townsend
	(percent/year, 1961-91)	(percent/year, 1962-90)	(percent/year, 1971-86)
Central Africa
Cameroon	1.8	-2.8	1.0
Central African Republic	2.7	0.0	1.8
Congo	1.2	1.6	-1.4
Gabon	-2.3	-25.8	na
Zaire	8.1	-2.1	1.5
East Africa
Burundi	3.4	-16.4	2.7
Ethiopia	-1.7	-2.5	0.7
Kenya	1.9	-0.2	0.6
Madagascar	-0.1	-5	na
Rwanda	6.1	-13.9	3.2
Uganda	7.8	1.2	na
Sahel
Burkina Faso	0.8	-4.6	1.6
Chad	0.2	-2.5	na
Gambia	-1.5	na	na
Mali	0.8	-4.1	2.2
Mauritania	-0.3	-12.3	na
Niger	1.5	-5.7	na
Senegal	1.5	-3.4	-0.0
Somalia	1.2	-0.3	-0.3
Sudan	0.1	-2.5	-0.3
West Africa
Benin	1.2	0.8	na
Côte d'Ivoire	0.9	-1.0	1.3
Ghana	-0.5	-2.5	0.6
Guinea	1.2	-1.7	na
Guinea-Bissau	-2.1	na	na
Liberia	0.0	-4	na
Nigeria	-0.3	-4.7	1.0
Sierra Leone	0.5	0.4	0.3
Togo	-1.3	-6.4	0.1
Southern Africa
Angola	-0.8	-3.2	na
Botswana	1.3	-0.1	na
Lesotho	-1.7	-2.7	na
Malawi	0.3	-3.1	0.6
Mozambique	0.3	-1.0	na
Namibia	1.0	na	na
South Africa	1.3	1.5	na
Swaziland	3.3	na	na
Tanzania	0.2	-3.0	2.0
Zambia	1.5	-1.9	-1.0
Zimbabwe	2.0	-0.4	0.3

Sources: Lusigi & Thirtle (1997); Trueblood & Coggins (1997); Thirtle, Hadley & Townsend (1995).

Thirtle, Hadley and Townsend (1995) decompose the Malmquist TFP growth rates they find for 22 SSA countries into technical progress (from the time series for this panel of countries) and efficiency changes (from the cross-section). Investments in infrastructure, extension and the level of real protection on international agricultural markets are shown to be significant in explaining efficiency change, while tractors, the labour-land ratio, R&D and secondary education are found to explain the variation in technical progress. They find the labour-land ratio, or population density, to be the single most important explanatory variable, again suggesting that productivity growth will accelerate in land abundant countries as population density increases.

Growth Accounting

Paralleling the two types of productivity studies, aggregate production function and TFP indexes, are two types of growth accounting exercises. In the first case, the coefficients from the aggregate Cobb-Douglas production function are used to calculate the percentage growth in output due to changes in the level of each input. Changes in the residual or TFP are attributed to non-conventional factors.

Arnade (1997) conducted this exercise for a set of 77 countries, including five in Africa (Kenya, South Africa, Sudan, Zaire and Zimbabwe). He included livestock as a conventional factor only in Asia, where bullocks are commonly used for ploughing. He did not include livestock as a factor of production in other regions, arguing that there is little meat processing in the developing countries so livestock is sold as an output. For the Cobb-Douglas coefficients, Arnade used 0.20 for labour, 0.45 for land, 0.15 for fertilizer, and 0.20 for tractors. Constant coefficients such as these are consistent with marginal factor productivities that may vary with input levels across countries and across time. Arnade's coefficients are within the range of estimates found in the empirical studies reported in Table 4.3, except for the coefficient on land, which is larger than those found in other studies. This is likely due to the exclusion of livestock from the aggregate production function.

Results from Arnade's study for the five African countries and a representative sample of other countries are reported in Table 4.5. As expected, during the period 1961-1987, the growth rate of productivity was much higher in the "advanced technology" countries than in Africa or Asia. In the advanced technology countries, the level of land and labour inputs dropped while output increased, due mainly to increased fertilizer use and improvements in productivity. In the Asian countries reported, the contribution of land expansion to output growth was low, while tractor use showed a high rate of increase. The growth rate of labour and fertilizer was moderate, while productivity growth was mixed. For the African countries, the increases in agricultural land were relatively small, while increases in labour, tractor, and fertilizer use contributed significantly to output growth.

The differences between Asia and Africa and the advanced technology countries are stark. The latter have been able to maintain increases in agricultural output while decreasing labour and land inputs. Asia and Africa have relied much more heavily on increases in the conventional inputs. This is hardly surprising, given their relatively low levels of conventional input use, other than labour and land. For Africa, excluding South Africa (because of its advanced agricultural technology relative to the rest of SSA) and Zambia (due to its negative growth rate of productivity), growth in the use of conventional inputs explains approximately 60 to 80 percent of the growth in agricultural output over the period under study.

Most of the studies reviewed above that calculated aggregate Cobb-Douglas production functions did not go on to perform a growth accounting exercise. However, Craig, Pardey and Roseboom (1997) noted that the conventional inputs in their equation accounted for 72 percent of the variation in productivity, both for LDCs alone and also when DCs were included. Frisvold and Ingram (1995) did perform an in-depth growth accounting study for their sample of 27 SSA countries. For three of the four SSA regions under study (the Semi-Arid Tropics, the Sub-Humid Tropics and the Humid Tropics), they found that conventional inputs explained 84 to 93 percent of the increase in agricultural output. For the Lowland Humid Tropics (which consisted only of Mauritius and Madagascar) the figure was only 22 percent.

Projections of Future Growth in African Agricultural Production

A number of research efforts have generated projections of future growth in agricultural production in SSA. For example, FAO projects that gross agricultural production in SSA will increase at an average annual rate of three percent between 1988-1990 and 2010 (FAO, 1993). Of this increase, 53 percent is projected to come from increases in yield, 30 percent from increases in arable land, and 17 percent from increases in cropping intensity. The International Food Policy Research Institute projects similarly strong growth in grain production in SSA, averaging three percent annually between 1990 and 2020 (Rosegrant, Agcaoili-Sombilla and Perez, 1995). Area in cereals is projected to increase 1.3 percent annually over the period, and yields are projected to increase 1.7 percent per year. World Bank simulations for selected developing countries include projections of 2.5 percent annual growth in grain production in Central Africa between 2000 and 2010, and 1.5 in Nigeria and South Africa, from a combination of yield and area increases (Mitchell and Ingco, 1995). The USDA's Economic Research Service (Shapouri and Rosen, 1998) projects that food production in SSA will grow at an average rate of 2.3 percent per year between 1995-1997 and 2008 through a combination of area expansion (1.3 percent per year) and yield increases (1.0 percent per year).

TABLE 4.5(a).
Contributions to Average Rates of Output Growth, Based on Coefficients
from an Aggregate Cobb-Douglas Production Function, 1961-1987
(Growth Rates in Percent per Year)

Country	Output	Labour	Land	Tractors	Fertilizer	Productivity	Animals
C-D coefficients		0.20	0.45	0.20	0.15
Africa
Kenya	2.81	0.67	0.13	0.32	1.09	0.60
South Africa	2.14	-0.19	-0.10	0.31	0.78	1.34
Sudan	3.05	0.39	0.05	0.92	0.53	1.15
Zaire	2.03	0.18	0.07	0.91	0.17	0.71
Zambia	2.40	0.44	0.02	0.99	1.45	-0.49
Zimbabwe	2.67	0.46	0.18	0.29	0.68	1.05
Asia
China	4.10	0.35	-0.03	2.40	0.05	1.27	0.06
India	2.50	0.28	0.05	2.34	1.31	-1.44	-0.04
Indonesia	3.80	0.14	0.18	2.60	0.37	0.46	0.05
Philippines	3.44	0.32	0.36	1.26	0.50	1.05	-0.06
Advanced technology
Australia	2.16	-0.15	-0.05	0.12	0.23	2.02
France	1.66	-0.83	-0.15	0.49	0.47	1.65
United Kingdom	1.60	-0.34	-0.11	0.11	0.36	1.58
United States	1.85	-0.41	-0.03	-0.05	0.45	1.89

Source: Arnade (1997).

TABLE 4.5(b).
Contributions to Average Rates of Output Growth, Based on Coefficients
from an Aggregate Cobb-Douglas Production Function, 1961-1987
(Percentage of Output Growth Due to Each Factor)

Country	Output	Labour	Land	Tractors	Fertilizer	Productivity	Animals
C-D coefficients		0.20	0.45	0.20	0.15
Africa
Kenya	2.81	23.8	4.6	11.4	38.8	21.4
South Africa	2.14	-8.9	-4.7	14.5	36.4	62.6
Sudan	3.05	12.8	1.6	30.2	17.4	37.8
Zaire	2.03	8.9	3.4	44.8	8.4	35.0
Zambia	2.40	18.3	0.8	41.3	60.4	-20.4
Zimbabwe	2.67	17.2	6.7	10.9	25.5	39.3
Asia
China	4.10	8.5	-0.7	58.5	1.2	31.0	1.5
India	2.50	11.2	2.0	93.6	52.4	-57.6	-1.6
Indonesia	3.80	3.7	4.7	68.4	9.7	12.1	1.3
Philippines	3.44	9.3	10.5	36.7	14.5	30.5	-1.7
Advanced technology
Australia	2.16	-6.9	-2.3	5.6	10.6	93.5
France	1.66	-50.0	-9.0	29.5	28.3	99.4
United Kingdom	1.60	-21.3	-6.9	6.9	22.5	98.8
United States	1.85	-22.1	-1.6	-2.7	24.3	102.2

Source: Arnade (1997).

4.4 Bottlenecks to Growth

The studies reviewed provide a guide to the factors that have historically affected agricultural productivity in SSA. Since conventional inputs explain most of the variation in productivity between countries in SSA, it is apparent that many of these countries still have considerable potential to raise productivity through increased use of fertilizer, machinery and livestock inputs. It has been argued that barriers to increased use of these inputs include lack of appropriate infrastructure, poor policy environments, and lack of cash to increase input purchases (Byerlee and Heisey, 1996; Heisey and Mwangi, 1996; Larson and Frisvold, 1996).

The importance of conventional inputs in SSA suggests that factors limiting their use are the most critical constraints on continued growth in agricultural productivity. Foremost among these are inadequacies in the provision of basic infrastructure, both physical and institutional. For example, limited surface transportation and communication networks in SSA increase the cost of inputs, inhibit the timely acquisition and application of inputs, and decrease access to output markets. Examples of institutional bottlenecks with similar effects include elements as diverse as political instability and constraints on access to credit and extension services. Credit market constraints are in turn driven, at least in part, by the complexities of land tenure that characterize much agricultural land in SSA. In particular, lack of individual private tenure and associated land titles as collateral may inhibit access to formal credit sources, even though customary tenure systems may offer no less security than individual private property systems (Bruce and Migot-Adholla, 1994).

Even more basic than concerns about physical and institutional infrastructure, however, are questions about the potential for continued increases in application of the conventional inputs that have contributed to growth in SSA agricultural productivity in the past. For example, Crosson and Anderson (1995) report that just over one billion hectares of land are considered by FAO to be at least marginally suitable for crop production in SSA. About 213 million hectares, or just over a fifth of that, are currently in crops. Crosson and Anderson note that if all the remaining suitable land were to be brought under crop production in the coming decades, output would increase more than enough to meet a tripling in demand by 2025, even without any increase in crop yields. Expansion on such a scale is of course unlikely, as the authors argue, because of the economic and environmental costs involved. In fact, FAO (1993) estimates that cropland area in SSA will expand by 0.9 percent per year over the next decade, which, if continued, would result in a 37 percent expansion by 2025. By contrast, the World Bank suggests that 0.5 percent annual expansion may be the maximum rate consistent with long-term sustainability (Cleaver and Schreiber, 1994).

Given the importance accorded to physical infrastructure and education as non-conventional inputs in other multi-country studies of agricultural productivity as well (Craig, Pardey and Roseboom, 1997; Antle, 1983), it is surprising that these variables have not been included in the studies exclusive to Africa. It may be that data on infrastructure are not available for a sufficiently large set of African countries. In addition, the sequence of non-conventional inputs may be important. A study of agricultural productivity in the US has shown that infrastructure investments made important contributions to agricultural productivity through the 1960s (Shane, Roe and Gopinath, 1998). Since that time, however, public and private R&D have become more important in spurring productivity growth in the United States. If a similar trend holds for countries where infrastructure is not yet well developed (as in much of Africa), large increases in agricultural productivity may be possible from investments in rural roads and utilities.

Other variables that deserve closer attention in studies of agricultural productivity include changes in resource quality over time and measures of political and institutional instability. Messer, Cohen and D'Costa (1998) estimated that cessation of armed conflict would have added two to five percent annually to Africa's per caput food production since 1980. Peterson's (1987) useful land quality index, which controls for irrigation, precipitation and soil nitrogen, has been used frequently in international agricultural empirical work, but provides only one (constant) number per country that fails to reflect possible changes in land quality over time. If a portion of growth in agricultural output is actually due soil fertility depletion, but soil depletion is left as an unmeasured explanatory variable, then growth in output may be incorrectly attributed to productivity growth.

Based on the limited data currently available on land degradation and its productivity consequences in SSA, Crosson and Anderson (1995) estimate the average loss in agricultural productivity due to historic land degradation for Africa as a whole is about 12 percent. The authors conclude that land expansion and restoration will together contribute only about a third of the increased production necessary to meet anticipated demand in 2025. They caution that the potential for increased water supply is too limited to make a major contribution to increased production.

Crosson and Anderson argue that the remainder of the necessary production increases will have to come from adoption of a variety of more productive technologies, including improved crop varieties, increased use of fertilizer and pesticide and mechanization. Policies that will help widespread adoption of such technologies include reform of foreign exchange and tax policies that discriminate against agriculture, improvement of transportation and communications infrastructure, improved education and extension services, support for research and increased recognition of the security of property rights in land afforded by evolving local tenure systems.

Following Pingali and Heisey (1996), the technological transformation of crop production systems can be characterized in various stages, as different factors of production become scarce in succession. Pingali and Heisey describe three stages in particular with regard to cereal production, as land, labour and factors such as knowledge and management intensity become increasingly valuable. Thus, cropland expansion alone will no longer satisfy needed output growth, and further increases will need to come from intensification of production on existing cropland. Such intensification will require investment not just in basic transportation infrastructure but in the physical and institutional infrastructure necessary to improve delivery of irrigation, commercial fertilizer, extension services and other conventional and non-conventional inputs (Pingali and Heisey, 1996). Pingali and Heisey argue that for maize, an important food crop in much of SSA, there remains an economically exploitable gap between farmer performance and the technology frontier as represented by the yields achieved on experiment stations (in contrast to rice and wheat yields). They argue further that the technology frontier itself could be shifted more readily for maize than for rice or wheat, through transfer of technology from the more advanced countries. They caution, however, that such transfers are much less likely in SSA than in parts of Asia, where rising feed demand coincides with institutional environments that are more attractive to large private-sector seed companies.

4.5 The Role of Policy

Agriculture in SSA is characterized by multiple constraints on accelerated productivity growth. In the absence of broad improvements in physical infrastructure, political stability and the institutional environment, the returns to any given intervention in isolation are likely to be limited as other constraints quickly become binding. In such an environment, the role of policy is twofold.

First, governments and international agencies need to invest in underlying physical and institutional infrastructure to improve the basic performance of markets by reducing the costs of transportation and transactions and by facilitating the transmission of goods, services and market signals. Improved access to fertilizer, credit and roads are among the most promising steps that could be taken along these lines. Such improvements can be expected to reduce input costs and increase access to output markets, providing both demand- and supply-side incentives for increased use of conventional inputs and output growth.

Even when markets can be structured to perform more efficiently, a second role of policy remains critical. This is the mitigation of externalities. For example, reduced fallow periods by one farmer might pose erosion problems that result in sedimentation or increased flood risk to producers downstream, or in eventual on-site resource degradation that threatens farm yields in the future. Externalities highlight the importance of well-defined institutions governing property and the distribution of costs and benefits associated with various technical and institutional innovations. They also highlight the importance of policy in influencing how these costs and benefits are distributed spatially and temporally. Hazell and Fan (1998) note the importance of investing in measures to improve productivity not only in prime agricultural areas but in less-favoured lands as well. Their results are based on analysis of Indian data; additional research is needed to determine whether similar patterns may characterize sub-Saharan Africa.

Several of the studies reviewed above looked explicitly at policy reform as an explanation for productivity growth, focusing particularly on institutional reforms that affect the performance of markets. Block (1995) found that countries that depreciated the real exchange rate tended to have higher growth rates of total factor productivity. Fulginiti and Perrin (1997) used nominal price protection as a proxy for policy reform and concluded that the countries that tax agriculture the most tend to have the most negative rates of productivity change. Fulginiti and Perrin (1993) and Hu and Antle (1993) found that an indicator of the degree of subsidization or taxation of agriculture is significant in some ranges; reducing protection would increase (decrease) productivity in countries that have been taxing (subsidizing) agriculture.

4.6 Conclusions

Agricultural production has been increasing in SSA at over two percent per year in recent years (FAO, 1999). Land used in agricultural production has increased less rapidly, resulting in an average annual increase in land productivity of 1.9 percent in SSA between 1980 and 1993 (World Bank, 1998). Labour used in agriculture has increased more rapidly than land and labour productivity declined at an average annual rate of 1.0 percent between 1980 and 1995 (World Bank, 1998). Levels of physical capital, livestock, fertilizer and non-conventional inputs have also changed, contributing to an estimated 1.3 percent annual increase in total factor productivity between 1961 and 1991 (Lusigi and Thirtle, 1997).

Existing research has consistently found that about three-quarters of the variation in agricultural productivity in SSA is explained by the use of conventional inputs. This is not surprising, given the low levels of use of some conventional inputs in SSA, such as physical capital and fertilizer, relative to other developing regions and the developed countries. As a result, research suggests that there remains significant scope to improve productivity in many SSA countries through increased use of conventional inputs, particularly fertilizer, physical capital and livestock.

Based in part on such improvements in productivity, several recent research efforts have generated projections of future growth in agricultural production in SSA. FAO projects agricultural production will increase at an average annual rate of 3.0 percent between 1988-1990 and 2010, driven primarily by increases in yield (FAO, 1993; Alexandratos, 1995). The International Food Policy Research Institute projects 3.0 percent growth in grain production annually between 1990 and 2020, also due primarily to yield increases (Rosegrant, Agcaoili-Sombilla and Perez, 1995). The World Bank projects annual rates of growth in grain production of 1.5 to 2.5 percent for selected SSA countries between 2000 and 2010 (Mitchell and Ingco, 1995). The USDA's Economic Research Service (Shapouri and Rosen, 1998) projects that food production in SSA will grow at an average rate of 2.3 percent per year over the next decade through a combination of area expansion (1.3 percent per year) and yield increases (1.0 percent per year).

The ERS analysis further projects that food production in SSA would have to grow at a rate of 3.3 to 4.5 percent annually to maintain per caput consumption levels or meet nutritional requirements over the next decade. Given projected production increases on the order of 2.3 to 3.0 percent per year, these rates suggest that additional increases on the order of one to two percent per year will be required to make progress towards food security. If the World Bank's recommendation (Cleaver and Schreiber, 1994) that agricultural area expansion in SSA be limited to 0.5 percent per year on sustainability grounds is further incorporated, this gap increases to two to three percent per year.

How might such gains be realized? The studies reviewed above on the sources of productivity growth in SSA agriculture indicate that continued growth of the agricultural labour force on the order of two percent per year (FAO, 1999) can be expected to increase agricultural output by about one percent per year. As land becomes increasingly scarce relative to labour, farmers will increasingly seek ways to augment land through increased application of other inputs as well. Fertilizer application rates have been declining in Africa by an average of 1.1 percent per year since 1990 (FAO, 1999); reversing this trend and increasing fertilizer use by five percent per year could increase agricultural output by an additional 0.5 percent per year. Proportionate increases in the use of machinery and in research expenditures could be expected to add similar increases to output. Expected increases in output from improved infrastructure and price policies are difficult to quantify, but such improvements are probably prerequisites to make possible the increases in productivity from the use of conventional inputs and research.

Achieving these goals poses significant challenges for policymakers. Agriculture in SSA is characterized by multiple constraints, principal among which are poverty, poor infrastructure and political instability, which combine to limit increased use of conventional inputs in SSA agriculture. Other important constraints to agricultural productivity are the quality and availability of education, research and extension services, as well as institutional uncertainties that weaken incentives to invest in the maintenance or improvement of land quality.

In the absence of broad improvements in physical infrastructure, political stability and the institutional environment, the returns to any given intervention in isolation are likely to be limited as other constraints quickly become binding. Policy reforms directed at improving physical and institutional infrastructure may not only increase use of inputs by lowering prices, but may also improve farm-gate prices of agricultural output and thus more directly stimulate output.

In addition to facilitating increased use of conventional inputs, education of the rural labour force as well as agricultural research will improve the future prospects for productivity growth in SSA. The full benefits of research are unlikely to be realized before more basic constraints are surmounted. Nevertheless continued investment in research (alongside attention to more basic sources of productivity growth) remains important due to potentially long lags in application.

References

Alexandratos, N. 1995.. The outlook for world food and agriculture to year 2010. In Nurul Islam,ed. Population and food in the early twenty-first century. Washington, DC, International Food Policy Research Institute.

Antle, J. M. 1983. Infrastructure and aggregate agricultural productivity: international evidence. Economic Development and Cultural Change, 31(3):609-619.

Arnade, C.A. 1997. Agriculture growth sources: a look at 77 countries, Staff Paper No. 9709. Washington, DC, US Department of Agriculture, Economic Research Service

Badiane, O. & Delgado, C. 1995. A 2020 vision for food, agriculture, and the environment in Sub-Saharan Africa. Food, Agriculture, and the Environment Discussion Paper 4. Washington, DC, International Food Policy Research Institute.

Binswanger, H. & Pingali, P. 1988. Technological priorities for farming in Sub-Saharan Africa. World Bank Research Observer, 3: 81-98.

Block, S.A. 1995. The recovery of agricultural productivity in Sub-Saharan Africa. Food Policy, 20(5):385-405.

Boserup, E. 1965. The conditions of agricultural growth: the economics of agrarian change under population pressure. London, George Allen & Unwin.

Bruce, J.W. & Migot-Adholla, S.E. 1994. Searching for land tenure security in Africa. Dubuque, Iowa, USA, Kendall-Hunt, for the World Bank.

Byerlee, D. & Heisey, P.W. 1996. Past and potential impacts of maize research in sub-Saharan Africa: a critical assessment. Food Policy, 21(3):255-277.

Cleaver, K.M. & Schreiber, G.A. 1994. Reversing the spiral: the population, agriculture, and environment nexus in Sub-Saharan Africa. Washington, DC, The World Bank,

Craig, B.J., Pardey, P.G. & Roseboom, J. 1997. International productivity patterns: accounting for input quality, infrastructure and research. American Journal of Agricultural Economics, 79(November):1064-1076.

Crosson, P. & Anderson, J.R. 1995. Achieving a sustainable agricultural system in Sub-Saharan Africa. World Bank, Building Blocks for Africa 2025 Paper No. 2. March.

FAO. 1999. FAOSTAT Database. Http://apps.fao.org. Accessed 23 March 1999.

FAO. 1993. Agriculture: towards 2010. Rome.

Frisvold, G. & Ingram, K. 1995. Sources of agricultural productivity growth and stagnation in Sub-Saharan Africa. Agricultural Economics, 13:51-61.

Fulginiti, L.E. & Perrin, R.K. 1993. Prices and productivity in agriculture. The Review of Economics and Statistics, 75(August, no. 3):471-482.

Fulginiti, L.E. & Perrin, R.K. 1997. LDC agriculture: nonparametric Malmquist productivity indexes. Journal of Development Economics, 53:373-390.

Hayami, Y. & Ruttan, V.W. 1985. Agricultural development: an international perspective. Baltimore, USA. The Johns Hopkins University Press.

Hazell, P. & Fan, S. 1998. Balancing regional development priorities to achieve sustainable and equitable agricultural growth. Washington, DC, International Food Policy Research Institute. Prepared for the AAEA International Conference on Agricultural Intensification, Economic Development, and the Environment, July 31-August 1, Salt Lake City.

Heisey, P.W. & Mwangi, W. 1996. Fertilizer use and maize production in Sub-Saharan Africa. CIMMYT Economics Working Paper 96-01. Mexico, D.F., International Maize and Wheat Improvement Center (CIMMYT).

Hu, F. & Antle, J.M. 1993. Agricultural policy and productivity: international evidence. Review of Agricultural Economics, 15(September, no. 3):495-505.

Kawagoe, T., Hayami, Y. & Ruttan, V.W. 1985. The intercountry agricultural production function and productivity differences among countries. Journal of Development Economics, 19(Sept-Oct):113-132.

Larson, B.A. & Frisvold, G.B. 1996. Fertilizers to support agricultural development in Sub-Saharan Africa: what is needed and why. Food Policy, 21:509-525.

Lusigi, A. & Thirtle, C. 1997. Total factor productivity and the effects of R&D in African agriculture. Journal of International Development, 9(4): 529-538.

Messer, E., Cohen, M.J. & D'Costa, J. 1998. Food from peace: breaking the links between conflict and hunger. 2020 Brief 50. Washington, DC, The International Food Policy Research Institute.

Mitchell, D.O. & Ingco, M.D. 1995. Global and regional food demand and supply prospects. In N. Islam, ed. Population and Food in the Early Twenty-First Century. Washington, DC, International Food Policy Research Institute.

Pardey, P., Roseboom, J. & Anderson, J. 1989. Agricultural research policy: international quantitative perspectives. Cambridge, UK: Cambridge University Press.

Peterson, W. 1987. International land quality indexes. Department of Agricultural and Applied Economics Staff Paper P87-10, University of Minnesota.

Pingali, P.L. & Heisey, P.W. 1996. Cereal crop productivity in developing countries: past trends and future prospects. Conference Proceedings, Global Agricultural Science Policy for the Twenty-First Century, Melbourne, Australia, 26-28 August.

Reardon, T., Delgado, C. & Matlon, P. 1992. Determinants and effects of income diversification amongst farm households in Burkina Faso. Journal of Development Studies, (January).

Rosegrant, M.W., Agcaoili-Sombilla, M. & Perez, N.D. 1995. Global Food Projections to 2020: Implications for Investment. Food, Agriculture, and the Environment Discussion Paper No. 5. Washington, DC, International Food Policy Research Institute.

Rosen, S. 1997. Sub-Saharan Africa. In S. Shapouri & S. Rosen, eds. Food Security Assessment. International Agriculture and Trade Report No. GFA-9, Economic Research Service, US Department of Agriculture (November), Washington, DC.

Shane, M., Roe, T. & Gopinath, M. 1998. US agricultural growth and productivity: an economywide perspective. Market and Trade Economics Division, Economic Research Service, US Department of Agriculture, Washington, DC. Agricultural Economic Report No. 758.

Shapouri, S. & Rosen, S., eds. 1998. Food security assessment. International Agriculture and Trade Report No. GFA-10. Economic Research Service, US Department of Agriculture, Washington, DC.

Thirtle, C., Hadley, D. & Townsend, R. 1995. Policy induced innovation in Sub-Saharan African agriculture: a multilateral Malmquist productivity index approach. Development Policy Review, 13(4): 323-342.

Thrupp, L.A. 1997. Linking biodiversity and agriculture: challenges and opportunities for sustainable food security. Washington, DC, World Resources Institute.

Trueblood, M.A. 1991. Agricultural production functions estimated from aggregate intercountry observations: a selected survey. Agriculture and Trade Analysis Division, Economic Research Service, US Department of Agriculture, Washington, DC. Staff Report No. AGES 9132.

Trueblood, M.A. & Coggins, J. 1997. Nonparametric estimates of intercountry agricultural efficiency and productivity. Economic Research Service, US Department of Agriculture, Washington, DC. Mimeo.

Wiggens, S. 1998. African farming seen from village studies: changes from the 1970s to the 1990s. Paper for the ASA-Supported Conference on Africa and globalisation: towards the millennium, University of Central Lancashire at Preston, 24-26 April 1998.

World Bank. 1998. World development indicators 1998. Washington, DC, The World Bank.

1 Comments from Shahla Shapouri and Lydia Zepeda and GIS assistance from Vince Breneman are gratefully acknowledged. The views expressed here are those of the authors, and may not be attributed to the Economic Research Service.

2 Several recent studies have presented Malmquist TFP indexes for various sets of countries (Fulginiti and Perrin, 1997; Lusigi and Thirtle, 1997; Trueblood and Coggins, 1997). The results of Lusigi and Thirtle are presented here, since it is the only study which focuses exclusively and exhaustively on Africa.