Lessons of the Global Structural Transformation Experience for the East African Community

Size: px
Start display at page:

Download "Lessons of the Global Structural Transformation Experience for the East African Community"

Transcription

1 Lessons of the Global Structural Transformation Experience for the East African Community Uma Lele, Manmohan Agarwal, and Sambuddha Goswami 1,2 1 Uma Lele is independent scholar and former Senior Advisor in the World Bank, Manmohan Agarwal is Senior Fellow at the Center for International Governance, Waterloo, Canada and his work was conducted as part of Ontario Research Fund project on China; and Sambuddha Goswami is Research Associate of the Uma Lele research team. 2 The views expressed in this paper are those of the authors and do not represent views of the sponsoring organizations. i

2 Introduction Since the food price rises starting 27 and the financial crisis that followed in 28, the development community has been wrestling with the dual questions of threats to food security and the growing inequality accompanying economic growth. Almost all of the poverty and hunger is now concentrated in the low and lower-middle income countries of Asia and Sub-Saharan Africa, most of it in rural areas. In this paper we examine the performance of a selected number of countries in Asia and Africa by analyzing agriculture s historic role in structural transformation. A paper drawing on the experience of 19 countries over a 3 year period with a focus on China, India, Indonesia and Brazil was prepared as an input into India s 12 th Five Year Plan in 211 (Lele et al 211). By extending that analysis to the East African countries, we demonstrate the extent to which the late developers of the East African Community face odds which are greater than those of densely populated Asian countries in achieving structural transformation. We further demonstrate the urgency of focusing attention on agricultural productivity growth as the critical ingredient for achieving food security, poverty reduction and overall economic development. This analysis is of considerable relevance to contemporary debates on the role of agriculture. Our own analysis and that of several others has demonstrated that India s structural transformation has slowed (Hazell et al 211; & Binswanger and D Souza 211). Others have noted that with the exception of South Africa, in Sub-Saharan Africa total factor productivity growth during the period may be just enough to bring them up to par with the productivity levels they had experienced in the 196s (ReSAKSS 211). By estimating growth in total factor productivity for a large number of countries Fuglie has demonstrated similar slow growth in TFP in African agriculture (Fuglie 212). Still others have argued that population pressure on the land is immense in some African countries, and whereas landlessness is seen as the key dimension of rural poverty in South Asia, the term is rarely used in the African context, even in those regions where many farms are inadequately small, where landlessness already exists, and will surely be an emergent phenomenon in decades to come (Jayne and Heady Forthcoming). On the other hand some influential analysts have criticized the simplistic narratives that are put forward for agriculture in sub-saharan Africa (Derkon 213). They ask: Is agricultural growth really necessary for economic growth? Is it really the most poverty reducing kind of growth? Is raising smallholder productivity really the best way of getting the economy jolting, poverty reducing growth? By undertaking systematic analysis of structural transformation we address the significance of these debates for policymaking if overall development of countries is the objective. 1

3 Structural Transformation Economic development is accompanied by structural transformation (ST) consisting of several quite distinct processes: (1) declining share of agriculture in Gross domestic product (GDP), (2) declining share of agriculture in employment, (3) rural-urban migration, (4) growth of the service and manufacturing sectors and (5) a demographic transition with reduction in the population growth rates (Kuznets 1955 & 1966; Chenery and Syrquin 1975; & Timmer 29). Past analysts of structural transformation have highlighted the importance of agricultural productivity growth for overall economic growth. 3 Differences in labor productivity between the agricultural and non-agricultural sectors disappear in the final outcome of ST. However before convergence in productivity among sectors occur a huge and often even a widening gap in labor productivities occurs between the agricultural and non-agricultural sectors (Lewis 1954; Johnston and Mellor 1961; & Timmer and Akkus 28). It widens income inequalities and leads to concentration of poverty in the agricultural sector. The later narrowing of income inequality is explained by many factors including public policy (progressive taxation), changing patterns of savings and investment, and the nature of technical change. The eventual income distribution within agriculture and within the urban sector depends on the initial income distribution, land person ratios in agriculture etc., and this latter provides useful insights into the performance, for example, of the Latin American region as compared to Asia and Sub-Saharan Africa presented in this paper. In an open economy the differences in labor productivities between developed and developing countries would lead to capital flows from developed to developing countries, according to Lewis. Today to this can be added foreign direct investment for land acquisition by land short countries in land surplus countries, labor migration, restrictive policies of labor importing countries regarding migration and the increasing South-South trade among others. The transformation literature has focused on changes in labor productivity and implications for inter-sectorial labor transfers. The turning point is reached when labor productivities in the two sectors begin to converge. The turning point for today s developing countries is taking longer time to occur than their industrial counterparts and occurs at a higher income than was the case for the industrial countries (Timmer and Akkus 28). Previous analysts have shown that investment in agricultural research and innovation, education, transport and other supportive policies, institutions and investments are necessary to achieve ST (Clark 194; Kuznets 1955 & 1966; Chenery and Syrquin 1975; Chenery et al 1974; & Chenery and Taylor 1968, Lewis 1954; Johnston and Mellor 1961; Ranis and Fei 1961; Mellor and 3 Increased labor productivity in agriculture provides the food and savings for the development of the non-agricultural sector (Kuznets 1955 & 1966; Lewis 1954; Johnston and Mellor 1961; Lele and Mellor 1981; &Mellor and Lele 1973). 2

4 Lele 1973; Timmer 29; Hazell et al 211; Binswanger and D Souza 211; & Badiane 211). The nature of technical change in agriculture has important implications. 4 Labor intensive technical change in agriculture increases employment, generates rural demand and growth linkages leading to multiplier effects of agricultural growth, but contributes less to the growth of the marketed surplus of food due to the higher income elasticities of demand for food among laboring classes (Lele and Mellor 1981). Hence the extent to which labor intensive technical change in agriculture will affect non-agricultural wages and prices depends on the employment created and the food demand elasticities. The reverse would be true when the distributive bias is against labor, leading to greater growth of marketed food supply helping to keep wages and prices lower than they would otherwise be in the non-agricultural sector. 5 These findings are pertinent to the transformation processes of the agricultural and rural sectors as demonstrated in this paper. Chenery & Syrquin (1975) made a major contribution in characterizing the transformation process and the factors explaining differing patterns. In the same tradition Timmer and Akkus analyzed ST using regression analysis. Our analysis builds on this body of literature. Scope of This Paper We first analyze the structural transformation processes using an extended data set, involving 19 developed and developing countries compared to the 86 contained in the Timmer and Akkus work. We then focus more sharply on the performance of China, Indonesia, India and Brazil and four of the five members of the East African Community, Kenya, Uganda, Rwanda and Burundi for which data covering a period starting 198 are available, when the Food and Agriculture Organization of the United Nations (FAO) began to publish labor data, to 29. Including the post 2 period of dynamic economic growth and accelerated pace of globalization allows us to see if faster growth helps countries to transform faster. We also include South Africa since in several 4 In a formal model with separate but interacting labor and food markets how the nature of distributive bias of technical change in the agricultural sector affects labor supply and inter-sectorial terms of trade and thus on the pace and pattern of growth of employment in the non-agricultural sector is analyzed (Lele and Mellor 1981). 5 In cross country analysis Christiansen et al (211) recently empirically confirmed the arguments of Lewis. Johnston and Mellor, that agriculture is significantly more effective in reducing poverty-- up to 3.2 times better at reducing $1-day headcount poverty in low-income and resource rich countries but non-agriculture has the edge in dealing with the better off poor (reflected in the $2-day measure). The larger participation of poorer households in growth from agriculture more than compensates for the slower growth of the sector. 3

5 ways South Africa s structure of agriculture is similar to that of Brazil, and it helps us to explore the issues of agrarian structure for generating employment in agriculture. Our specification includes population as a variable as included in the Chenery-Syrquin analysis but not in the Timmer-Akkus work, and we include the terms of trade used in Timmer-Akkus but not in Chenery-Syrquin. The analysis offers new insights into whether the pattern of ST has changed and, if so, the causes of this change. Our analysis includes: Regressions for the entire sample of 19 developed and developing countries, Regressions for only the 88 developing countries because of the large difference between developed and developing countries even at the start of the period of analysis in 198, Regressions for developing countries within each region to understand neighborhood patterns. We also analyze the performance of four large countries, Brazil, India, Indonesia, and China (BIIC) and five or six small countries in East and Southern Africa and Egypt in North Africa in each of the above three contexts (1) developed and developing countries, (2) developing countries only and (3) the performance of their regions. The BIIC are of interest because of their scale, significant roles in the world food and agricultural markets and contribution to global economic growth. 6 East African Community countries are of interest because after a prolonged period, they have begun to grow and agricultural development is receiving attention as part of NEPAD (The New Partnership for Africa s Development) and CADAAP (The Comprehensive Africa Agriculture Development Programme). Yet they are different in their resource endowments and agro-ecological systems, their political systems, the size of their internal markets, their institutional choices and capacities. Their performance is also symptomatic of the behavior in the regions where they are located and of initial structural conditions in those regions. These include acute land and income inequalities in Latin America and South Africa, labor surplus in Asia, and despite the general view about land extensive agriculture in Sub-Saharan Africa, the considerable and growing population pressure on the land, among countries of the East African Community, most notably Rwanda, Burundi and Kenya. Indeed their population pressure is already greater than that in India, and similar to China s (Figure 1.1 and 1.2). 6 Together they contain 43 percent, provide a quarter of the global GDP in purchasing power parity terms, with China alone providing16 percent, account for 32 percent of the global area harvested for cereals and produce 36.7 percent of the global cereal production. In recent years they have been some of the fastest growing economies and all are members of the G 2. 4

6 Furthermore their population growth rates are far higher than those in Asia (Figure 2). Due to slowing growth rate, China s economically active population in agriculture had already peaked in 23 whereas India s economically active population in agriculture is still growing (Figure 3.1 and 3.2). China s total population will peak at 1.45 billion in 23, where India will peak at 1.65 billion in 263 on a surface area nearly 4 percent of China s (Figure 4.1 and 4.2). This means increasing population pressure on the land requiring substantial employment generation in agriculture as well as non-agriculture for a growing population. By comparison to Asia, land pressure is immensely greater in the countries of the East African Community. First, their economically active population in agriculture is already a higher share of the total population, their population is growing more rapidly than either that of India, China or Indonesia, and will continue to grow way beyond when populations of India and Indonesia peak, increasing pressure on land and requiring even greater employment generation, if poverty, food insecurity and social conflict are not to increase. Figure 1.1 and 1.2: Arable Land and Permanent Crops/Agricultural Population (ha/capita) ( ) (Five Member Countries of the East African Community+ CII+ Egypt) and (Brazil and South Africa) Figure Figure Source: FAOSTAT Source: FAOSTAT Burundi Egypt Indonesia Rwanda United Republic of Tanzania China India Kenya Uganda Brazil South Africa 5

7 Base Year 195=1 Figure 2: Total Population (Estimated & Projected) Growths (Base Year 195=1) (195-21) (5 member countries of the East African Community + BIIC + Egypt and South Africa) Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda United Republic of Tanzania Source: Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat, World Population Prospects: The 212 Revision, 6

8 Figure 3.1 and 3.2: Total Economically Active Population in Agriculture (1) ( ) (Five Member Countries of the East African Community+ Egypt and South Africa) and (BIIC) 2 Figure Figure Source: FAOSTAT Source: FAOSTAT Burundi Egypt Kenya Rwanda Brazil China South Africa United Republic of Tanzania Uganda India Indonesia Whether considered in terms of total economically active populations in agriculture or total populations Egypt and South Africa had already seen a decline by 212whereas those in the East African countries continue to grow. 7

9 Figure 4.1 and 4.2: Total Population (Estimated & Projected) (in millions) (195-21) (5 member countries of the East African Community + Egypt and South Africa) and (BIIC) Figure 4.1 Figure Burundi Kenya South Africa Egypt Rwanda Uganda Brazil India China Indonesia United Republic of Tanzania Source: Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat, World Population Prospects: The 212 Revision, Data Used Availability of the time series and cross country data in the public domain from international organizations, including the FAO, the World Bank, and the Asian Development Bank etc. make analysis of structural transformation easier today than in the days of Kuznets and Chenery. Yet, there are differences in the concepts used by different international organizations, related to seemingly similar issues, such as labor employed in agriculture. International organizations report data provided by governments. Improving data quality requires working with governments and other stakeholders to ensure data accuracy. The World Bank, FAO and other UN organizations, supported by donors are giving high priority to improving data quality, a message reinforced by the UN Secretary General s High Level Taskforce on Post 215 8

10 agenda (UN System Task Team Report 213). Thus notwithstanding data weaknesses analysis such as the one contained in this paper can help to identify data weaknesses and create demand for improved data, which this paper does as well. Analysis of Structural Transformation Our quadratic log form specification allows for many different types of behavior, including initial stagnation a low level trap followed by accelerating and then decelerating increase, stagnation and other non-linear patterns. The form also allows the share of agriculture to settle somewhere above zero. We used two alternative dummies, annual dummies and decadal dummies instead of Chenery Syrquin s five year dummies and Timmer-Akkus annual dummies. Annual dummies capture short term changes in policies, institutions etc., decadal dummies capture long term changes in institutions, technology, infrastructure etc. which are unlikely to be captured by annual dummies. We also introduced dummies for the three regions, Asia, Latin America and Sub-Saharan Africa for characteristics of the specific regions that may affect outcomes, e.g., the structural inequality in Latin America, land pressure and intensive agriculture in Asia and land surplus, extensive agriculture in Africa. The specification is X = a + b. Ln Y + c. (Ln Y) ^2 + d. Ln Pop + e. (Ln Pop) ^2 + f. TOT Where TOT, the terms of trade, is the deflator for value-added in agriculture divided by the deflator for non-agriculture, Y is per capita income and Pop is population. X, the dependent variable, represents in different equations, share of value added in agriculture in GDP, share of employment in agriculture to total employment, value added in agriculture in 2 US dollars, value added per worker in agriculture, and the difference between agriculture s share in value-added and agriculture s share in employment. The results of the regressions for the 19 countries using annual dummies are in Table 1; those for 88 developing countries are available on demand. There is relatively little difference between estimates based on the decadal and annual dummies, except in a few key areas, discussed later. 9

11 Table 1: Regression Result for the 19 Developed and Developing Countries using Regional Dummies and Annual Dummies Agriculture share in VA Employment VA in Agriculture VA/L Ag share in VA minus Ag share in employment Constant 1.95* 1.82* 3.15* 5.6*.16* Ln Y -.39* -.27*.416* * (Ln Y)^2.2*.1* *.1* Ln P -.9* -.1*.97*.3*.4* (Ln P)^2 -.1*.3*.9* -.2* -.4* TOT.5*.8* -.32* -.35*.5* d1 Asian Countries.2.84*.93** -.338* -.8* d2 LAC Countries ** -.26*.1 d3 SSA Countries -.51*.162* -.414* -.992* -.21* R * ** Significant at 5%; * significant at 1%. Key Conclusions of Our Analysis The share of value added and share of employment in agriculture declines with per capita income, confirming the results of previous studies; these shares fall at a declining rate (The squared income term is positive whereas the linear term is negative). Total value added in agriculture increases with per capita income but at a decreasing rate. However, the behavior of labor productivity measured as value added per worker, is different when only the 88 developing countries are considered, rather than 19 countries including developed countries. The r 2 s are slightly smaller because of greater variability among developing countries. But more importantly, per worker value added in agriculture declines at a diminishing rate if all 19 countries are considered, where increases at a declining rate for the 88 developing countries. 7 Further elasticity with respect to per capita income is higher than elasticities for population in all structural changes with the notable exception of value added in agriculture. Elasticity of value added in agriculture with respect to per capita is lower than for elasticity of value added with respect to population. This means labor intensity makes a greater difference to value added in agriculture than the general increase in income. 7 The linear term is negative but is not significant at the 1 percent level and the squared income term is positive in the regression with 88 developing countries. 1

12 Turning Points The gap between the value added in agriculture and the share of employment in agriculture reflects the differences between per worker productivity in the agricultural and non-agricultural sectors and is important for the process of convergence in incomes between the two sectors. In the regression with 19 countries, the gap narrows over time with an r 2 of.52, with all the variables in the equation being significant at the 1 percent level. But the gap first increases with income, thereby enhancing the duality between the agriculture and non-agricultural sectors and later tapers off. There is a very sharp difference when the regression is fitted only to 88 developing countries. Then there is no convergence as income increases. The r 2 is very low and only the population square and TOT variables have significant coefficients. The turning point is not reached, until per capita income reaches a very high level. We return to these issues below. We next estimated the turning point incomes for the four countries using the equation with country dummies which assumes that these 4 countries follow the same average pattern demonstrated by the 19 developed and developing countries, or the 88 developing countries. Furthermore we performed this analysis for regional groups of countries, e.g. 19 Asian and the remaining Non-Asian Countries, (as did Timmer and Akkus). But we also performed the analysis by further dividing the countries into sub groups of 38 developing countries in Sub-Saharan Africa, and 24 developing countries in Latin America. In each case we performed this analysis using both decadal dummies and annual dummies (the latter to compare our results with those of Timmer and Akkus). The per capita income at which the turning point is reached is much higher if developed and developing countries are included than for only the developing countries (Table 2). This happens because there are basically two clusters of data. One is for the developing countries which have much lower per capita income and larger share of agriculture in value added and in employment with the agricultural employment share being greater than the value added in agriculture. The second cluster is of data for the developed countries, which have much higher per capita income and a lower share of agriculture in value added and in employment and so is to the South east of the data cluster for the developing countries. Therefore when the developed countries are included in the sample the regression equation is more to the right in terms of the X-axis, namely, the income axis, and so the turning point occurs at a higher level of income. The time 11

13 dummies are negative so that the so that the turning point is occurring at a higher level of income. 8 The turning point for only the Asian countries is considerably lower than if all developing countries, or developing and developed countries are considered, a result consistent with that obtained by Timmer Akkus. But as Binswanger and D Souza (211) note the turning points are quite unstable. Table 2: Estimates of Average per Capita Income at Which Turning Points are Reached Our Estimates Using Decade Dummies [ydum1( ) and ydum2 ( )] Region Ln Y (Ln Y)^2 R^2 Turning Point of LnGDPpc Turning Point of GDPpc (constant 2 US$) 19 Countries (88 Developing Countries+21 Developed Countries) $ Developing Countries $3824 Asia (19 Developing Countries) $1681 Sub-Saharan Africa (SSA) (38 Developing Countries) Latin America and Caribbean (LAC ) (24 Developing Countries) $ $4272 Non-Asian Countries (88 Countries 69 Developing+19 Developed) $146 4 Countries (Brazil+ China+ India+ Indonesia) $1488 However, when we used annual dummies, like Timmer and Akkus, turning points for non-asian countries, particularly Sub-Saharan Africa, turned out to be at extraordinarily high levels of income perhaps because of collinearity among income increases and annual dummies. Number of Years to Reach a Turning Point Turning points are quite sensitive to the choice of countries, choice of period (i.e., depends on per capita income and the growth rate of per capita income) and the specification of the model. E.g., considering developed and developing countries results 8 These results are consistent with those obtained by Timmer and Akkus. 12

14 in a higher per capita income at which turning point are reached than if only developing countries-- or only Asian countries-- are considered. In our analysis of the number of years a country takes to reach the turning point depends on per capita income in 21 and the growth rate of per capita income. 9 We estimate that Brazil and South Africa have already reached the turning point; China will take 4 to 5 years, India over 2 years, Indonesia almost thirty years, Rwanda and Uganda almost 55 years and 65 years respectively; where Kenya will take additional years due to its lower growth rate of per capita though Kenya s initial per capita income was the highest among the Five East African Community Members countries and Burundi is far behind due to its lowest per capita income and low growth rate of per capita income among the Five East African Community Members countries (Table 3). Table 3: Our Estimates of Number of Years to Reach Turning Point Using Decade Dummies Country 24-8 Growth Rate 29-1 Growth Rate Brazil Burundi Already There Not in the Trajectory China 4 5 Egypt India Indonesia Kenya Rwanda 4 67 South Africa Already There Uganda Differences in Behavior among Regions and Countries The lack of definite behavior among all 88 developing countries with respect to the gap equation as regards income is explained by the different behavior among regions. The gap has been decreasing in LAC and SSA and increasing in East and South Asia. The decline in duality between agriculture and non-agriculture in LAC is not only because productivity (i.e., value added per worker) increased in agriculture by 87 percent but because it fell in non-agriculture by over a quarter. The outcome is similar in SSA as in Latin America but is more worrisome because there was only a small rise in value added per worker in agriculture of only 2 percent and a sharper fall in non- agricultural value 9 We use both the average annual growth for the period 24-8 as well as the lower growth rate for the period

15 added per worker, i.e., by 3 percent. Growths in land productivity and agricultural Total Factor Productivity (TFP) have also been slowest in SSA. We performed this same exercise for the East African community. Agricultural value added per worker increased in Rwanda and Tanzania, it declined only slowly in Uganda but decidedly declined in Kenya and Burundi. Figure 5.1 and 5.2: Agricultural Value Added Per Worker (Constant 25 US$) ( ) (Five Member Countries of the East African Community) and (BIIC + Egypt and South Africa) Figure 5.1 Figure Source: WDI, World Bank Note: For United Rep of Tanzania data is available since 1991 Source: WDI, World Bank Burundi Rwanda United Republic of Tanzania Kenya Uganda Brazil China Egypt India Indonesia South Africa 14

16 Figure 6.1 and 6.2: Non-Agricultural Value Added Per Worker (Constant 25 US$) ( ) (Five Member Countries of the East African Community) and (BIIC + Egypt and South Africa) 6 Figure 6.1 Figure Source: WDI, World Bank and FAOSTAT Note: For United Rep of Tanzania data is available since 1991 Source: WDI, World Bank and FAOSTAT Burundi Rwanda United Republic of Tanzania Kenya Uganda Brazil China Egypt India Indonesia South Africa Non agricultural value added per worker increased the fastest in Rwanda, followed by Uganda and Tanzania but it declined in Kenya and Burundi. Per worker value added increased in South Africa, and increased the fastest in China followed by India. The increase in the gap in Asia in the value added per worker in the two sectors is because value added per worker increased more rapidly in non-agriculture than agriculture though it increased considerably in both. The productivity increase was greater in both sectors in East Asia than in South Asia. Value added per worker in non-agriculture relative to agriculture more than doubled in East Asia (Figure 7). The relative value added per worker in the non-agricultural sector in South Asia increased only by about 6 percent while it increased in agriculture also coincidentally by about 6 percent. 1 Thus the fall in value added per worker in agriculture relative to that in non-agriculture is very sharp in East Asia relative to South Asia (Figure 7). The ratio of value added per worker in non-agriculture to agriculture also increased in Rwanda, Uganda and Burundi but not in Tanzania and Kenya (Figure 8). 1 These results are consistent with the evidence on agricultural land productivity and total factor productivity presented in the larger monograph. 15

17 Figure 7: Ratio of Value Added per Worker (Non-Agriculture/ Agriculture) by Region (198-29) Source: WDI & Global Development Finance, World Bank and FAOSTAT East Asia & Pacific (developing only) Middle East & North Africa (developing only) Sub-Saharan Africa (developing only) Latin America & Caribbean (developing only) South Asia High income However, the duality decreases in all other regions, developed and developing alike. Yet this similarity hides important differences among regions. Value added per worker rises in agriculture in all regions, albeit very slowly in SSA. But while value added per worker in non-agriculture also rises in East and South Asia as well as in the industrial countries, it falls in LAC and SSA. So the decreasing duality in SSA and LAC is accompanied by decreasing labor productivity in non-agriculture. But there is great variation among countries in regions with Rwanda, Burundi and Uganda showing increased intersectorial duality. The difference between LAC and Asia is mirrored in the difference between Brazil on the one hand and China, India and Indonesia on the other. Inter-sectorial duality has increased sharply in China and gradually in India, and increased in Indonesia too until the Asian financial crisis in 1997 slowed Indonesia s growth. But it has decreased in Brazil (Figure 8). 16

18 Figure 8: Ratio of Value Added per Worker (Non-Agriculture/Agriculture) ( ) (Five Member Countries of the East African Community + BIIC + Egypt and South Africa) Source: WDI, World Bank and FAOSTAT Note: For United Rep of Tanzania data is available since 1991 Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda United Republic of Tanzania Value added per worker in non-agriculture increased less steeply in India than in China, but it has declined throughout the period in Brazil. The decline in value added per worker in non-agriculture and the increase in agriculture is consistent with other studies which suggest that Brazil s recent growth has been driven by agricultural growth (Contini et al 21). 11 The ratio of value added per worker in the nonagricultural sector to agricultural sector increased in Rwanda, Burundi and Uganda but declined in Kenya, Tanzania, Egypt and South Africa in the case of Africa, much like in Brazil. 11 The performance in exports of goods in China and India has been led by exports of manufactures. 17

19 Some Hypotheses of Causes of Behavioral Differences among Regions and Countries Distributive Bias of Agricultural Technology, Factor Efficiency and Factor Productivity Factor bias in agriculture determines labor s share and income distribution within agriculture and labor transfers to the non-agricultural sector. Even though we did not deal with within sector distribution, regression results shed important light on the issues highlighted by others which provide evidence on economies of scale in agriculture using new technologies of precision agriculture and factor market failures for small farmers (Deininger and Byrelee 211; and Helfand and Moreira 212). China and Brazil each behave quite differently than predicted by the regression analysis of 19 countries. China s agriculture is losing labor more slowly than the regression equation predicts, 12 while Brazil s agriculture is shedding labor much more rapidly than predicted (Figure 9). China s share of agriculture employment in total employment started out at the highest level (73.8 percent) among the four countries in 198, and declined to only 6 percent still the highest among the four countries using FAO data (FAOSTAT). FAO data includes both those employed in agriculture and those seeking employment; Asian Development Bank (ADB) data on the other hand only includes those employed and the difference would reflect unemployment and this seems to be increasing. However, ADB data which show a lower initial share of labor in agriculture and its faster decline are not available for the entire period covered in our analysis and do not change the conclusions but do change the extent of inter-sectorial disparities. 13 During the course of presentations in China, some economists attributed the high level of labor reported in agriculture due to under-reporting of migration to urban areas, in part due to the hukou system (Christiansen et al 211). Brazil s share of labor in agriculture was the lowest in 198 and declined the most rapidly, reflecting the acute dualism in its agriculture. Analysis of residuals of the regressions helps to understand deviation of countries from the predicted values based on the behavior of 19 countries. South Africa behaves much like Brazil. The residuals of both run almost parallel to each other (Figure 9). This means both have lost more labor from agriculture than the trend for the 19 countries predicted. In the case of the East African countries, what is surprising is that the agriculture of all four East African countries started with more labor in agriculture than the predicted 12 This led to a number of hypotheses and questions when the results were presented in various seminars in China and the World Bank. 13 Data are available on request. 18

20 Residuals value based on the trend for 19 countries. Over time Rwanda has maintained excess labor relative to the trend; excess labor has fallen slightly in Kenya and more in Burundi, but increased in Uganda. At the end of the period Burundi had less labor than predicted while the other three had more than predicted share of employment in agriculture (Figure 9). Indonesia s and India s decline in labor share was consistent with that predicted by the model using FAO data. Figure 9: Agricultural Employment Share Residuals ((198-29) [4 member countries of the East African Community (Burundi, Kenya, Rwanda and Uganda) + BIIC + Egypt and South Africa] Note: Tanzania has been excluded due to data unavailibility Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda Relative terms of trade moved sharply against agriculture in Brazil over time, as indeed they have in most other regions including LAC, whereas terms of trade moved in favor of agriculture in Asia, the most for China, particularly since 2. In Brazil this seems to be a result of a sharp rise in agricultural productivity, relatively low income elasticity of demand for food perhaps due to higher initial levels of income, and not the least technological change which created relatively little employment in agriculture and generated substantial agricultural surplus. 19

21 The behavior of agriculture s terms of trade reflects a combination of public policy, particularly in the form of price supports and market interventions in the agricultural sector, and movement in relative outputs prompted by technical change and relative price changes, exposure of agricultural to global competition, and in SSA due to a combination of structural adjustment leading to alignment of overvalued exchange rates and liberalization of economic policies and markets. Whereas the trend of terms of trade moving in favor of agriculture in Asia particularly since 2 are clear, the direction of causality between terms of trade and labor shares in agriculture is less clear. Timmer argued in earlier analysis that Asian countries used the terms of trade as a policy instrument for keeping labor employed in agriculture. The pattern of movements in the agricultural terms of trade saw these decline until 2 (the end of Timmer s period of analysis) in Asian countries at half the pace of the non-asian countries. Our analysis shows that in the 2 29 period the terms of trade moved very sharply in favor of agriculture in Asia, and more so in East than in South Asia despite it s relatively more open food trade policy which resulted in a greater share of net imports in total availability. 14 But the subsidization (and protection) has dramatically increased in China resulting in a nominal protection rate of 17 percent of gross farm receipts (Christiansen 211). The improved terms of trade for agriculture in China and East Asia suggest that worker productivity and incomes and so demand increased faster than agricultural supply. How much of this was due to increased employment in agriculture remains unclear. The movement of inter-sectorial terms of trade has tended to leave intersectorial nominal incomes less unequal than they would otherwise be. Terms of trade have also moved in favor of agriculture in two of the four countries-kenya and Uganda much like in Asia since 25 (Figure 1.1 and 1.2). 14 Brazilians attribute their agricultural success at least in part to the growth of markets in East Asia (Contini et al 21). A large share of growth in Soybean in Brazil is due to growth of imports by China, including use of grain for domestic ethanol production (211 Global Food Policy Report, IFPRI 212). 2

22 Figure 1.1 and 1.2: Terms of Trade (Deflator for Agriculture/Deflator for Non-Agriculture [Industry + Service]) ( ) (in US$) (5 member countries of the East African Community) and (BIIC+ Egypt and South Africa) Figure Figure Source: WDI, World Bank Note: For Tanzania data is available since 199 Source: WDI, World Bank Burundi Kenya Rwanda Brazil China Egypt Tanzania Uganda India Indonesia South Africa The Changes in Inequality Each in Rural and Urban Areas Per capita GDP in four of the five East Africa countries was comparable to that in China and India in the early 198s but Kenya s was more than twice that of China and India and comparable to Indonesia s. Kenya s growth rate was the highest in periods but then declined till 2. Burundi s growth rate was negative in the periods and all these five East African Community Members have improved from that phase of poor performance. The story is almost similar in the GDP per capita annual average growth rates for Rwanda, Uganda and Tanzania after 2. Overall annual average GDP growth rate in these countries were comparable to those in Asian countries, Kenya s growth rate was above 4 and Burundi s was near 3.5, whereas South Africa s growth rate was much lower than of East African countries Brazil s overall growth rate had decelerated too but was worse than South Africa post 2, but after 28 was better than South Africa s. 21

23 Share of Africa s total agricultural export in World agricultural exports had declined from 11.7 in 1961 and to 3.4 in 211 and share of Africa s agricultural import in World s imports had increased from 4.7 in 1961 to 6 in 211. But the five East African Community Members performed better. Their share in the Africa s total exports had increased from 9.3 in 1961 to 11 in 211 and share in Africa s total import declined from 5.5 in 1961 to 5.3 in 211. Kenya s total agricultural import and export value (in current US$) was highest among the five East African Community Members and the gap between Kenya s level and that of the other four countries had widened substantially by the end of the period. Despite good agricultural export performance Kenya has not done well on growth. The economic growth performance and changes in labor productivity are reflected in changes in inequality each in rural and urban areas. Brazil had the most unequal income distribution with a Gini coefficient of.57 in 1981, which further increased to.62 by the end of the 198s, i.e. income distribution worsened (Ferreira et al 26). But then income distribution improved as the Gini declined to.53 in 29 mainly because per worker output in non-agriculture which was almost 8 times the level in agriculture declined to 3. The greater use of distributive safety nets such as Bolsa Familia also contributed to a lower Gini (Higgins 211). Gini coefficients for China, India or Indonesia were much lower than in Brazil and closer together at the end of the 197s. 15 But it increased in these three countries as the gap in rural urban incomes grew; The Gini increased the most in China to.46, which had the greatest disparity in productivity growth between agriculture and non-agriculture has. The inequality between urban and rural areas contributes about 6 percent of the overall inequality more than increasing intra-sector inequality (Chen et al 21). However, the Terms of Trade behave differently in South Asia than in East Asia. South Asia experienced slower growth in both agriculture and non-agriculture sectors and therefore did not see as big a terms of trade effect as East Asia. The moderate movement in TOTs in South Asia may be explained by slower growth in effective demand for food as reflected in the large incidence of hunger and malnutrition. 15 See Pal and Ghosh (27) for India, Chen et al (21) for China, Asra (2) for Indonesia. The figure for India for 25 and for Indonesia for 29 is from World Development Indicators,

24 Table 4: Gini Co-efficient (Brazil, China, India and Indonesia) Year * Brazil Year China Year / / /2 25* India Year 1969/ * Indonesia Source: Ferreira, Leite & Litchfield 26; Pal & Ghosh 27; Chen, Dai, Pu, Hou & Feng; Asra 2; and GDI and Global Development Finance, World Bank. * is from World Development Indicators 211. Table 5: Gini Co-efficient (5 member countries of the East African Community + Egypt and South Africa) Year Burundi Year Egypt Year Kenya Year Rwanda Year South Africa Year Uganda Year United Republic of Tanzania Source: WDI, World Bank. South Africa had the most unequal income distribution with a Gini coefficient of.59 in 1993 which further increased to.63 in 29. Kenya s Gini coefficient was also high in 1992, i.e.,.57 but decreased to.48 in 25. Rwanda s income distribution worsened most strikingly (from.29 in 1985 to.51 in 211) among the Five East African Community Members followed by Tanzania (from.34 in 1992 to.38 in 27), whereas Gini coefficient has not changed for Burundi, Uganda and Egypt. 23

25 Residuals Figure 11: Agricultural Value Added Share Residuals (198-29) [4 member countries of the East African Community (Burundi, Kenya, Rwanda and Uganda) + BIIC + Egypt and South Africa] Note: Tanzania has been excluded due to data unavailibility Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda 24

26 Residuals Figure 12: Ln Agricultural Value Added in Million (constant 2 US$) (198-29) Residuals [4 member countries of the East African Community (Burundi, Kenya, Rwanda and Uganda) + BIIC + Egypt and South Africa] Note: Tanzania has been excluded due to data unavailibility Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda The residuals of actual share of value added in agriculture as compared to the predicted values shown in Figure 11 increased most rapidly in China and Brazil which means the share of value added increased relative to the predicted in these two countries. In Brazil value added in agriculture increased even in the face of the decline in Terms of Trades (TOTs), reflecting the rapid increase in efficiency of agriculture. In India, Indonesia, Egypt and South Africa actual values do not differ from the predicted values. But Burundi, Kenya, Rwanda and Uganda s actual values have decreased relative to predicted values which means share of value added in agriculture declined more rapidly than predicted. These trends in share of agriculture in value added and in employment imply that the gap between agriculture s share in value added and employment has been increasing the 25

27 Residuals most in China followed by followed by East African Countries, Indonesia and India, but decreasing in Brazil and South Africa. The residuals from the gap equation are positive for Brazil and South Africa and have been increasing over time (Figure 13). On the other hand the residuals from the gap equation have fallen sharply for China and East African countries with Burundi and Uganda showing maximum decreasing trend as share of agriculture in value added has fallen much more sharply than its share in employment. That means these countries have retained more labor in agriculture relative to predicted values in contrast to Brazil and South Africa that have more from agriculture. Figure 13: Agriculture Value Added Share minus Agricultural Employment Share Residuals (198-29) [4 member countries of the East African Community (Burundi, Kenya, Rwanda and Uganda) + BIIC + Egypt and South Africa] Note: Tanzania has been excluded due to data unavailibility Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda 26

28 Taking into account the five processes of structural transformation listed at the outset in this paper, India is clearly behind China and Indonesia in the process of transformation. It has the highest share of agricultural value added in GDP among the three countries, higher share of labor force in agriculture than Indonesia but lower than China, the lowest valued added per worker, and lower total value added in agriculture than China though higher than Indonesia. It also has the highest birth and death rates among the three Asian countries. China s share of employment in agriculture would be even lower if agricultural employment is not overstated in the data China reports to FAO. Similarly the East African countries are substantially behind India because they have high shares of population in agriculture, as yet no demographic transition and increased population pressure on land. Land Productivity and Total Factor Productivity in Agriculture Much of the previous discussion was on labor productivity. But land productivity measured as yields per ha of major crops is also far lower in Eastern Africa than in other countries and has not increased much in contrast to Egypt, China, Indonesia, South Africa and Brazil followed by India. In the latter set of countries irrigation and technological change has increased productivity. Once again, South Africa behaves almost similarly to Brazil on growth in cereal productivity per hectare. 27

29 Figure 14: Total Cereals Yield (hg/ha) ( ) (5 member countries of the East African Community + BIIC+ Egypt and South Africa) Source: FAOSTAT Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda United Republic of Tanzania Total Factor Productivity Total Factor Productivity measures growth in all output relative to all inputs. Once again, Egypt has had very rapid productivity growth, followed by China and Brazil, followed by South Africa and Indonesia. India has lagged behind But East African countries are lagging further behind in total agricultural productivity growth. 28

30 Base Year 1961=1 Figure 15: Agricultural Total Factor Productivity (TFP) Index Growth (Base Year 1961=1) ( ) (5 member countries of the East African Community + BIIC+ Egypt and South Africa) Source: Fuglie, Wang and Ball 212 Brazil Burundi China Egypt India Indonesia Kenya Rwanda South Africa Uganda Tanzania Role of Savings, Investment, Foreign Direct Investment and Aid in Productivity Growth Saving rates have been and increasing more rapidly than in Sub-Saharan Africa. Gross domestic savings as share of GDP was highest and growing most rapidly in China reaching well over half of GDP in 212, followed by India and Indonesia at above 3 percent in the same year, while saving rates have been much lower at about 2 percent in Brazil throughout the period, South Africa s rate declined from 28 percent in 1973 to near 2 percent by 211. In Eastern Africa, three year moving averages suggest Tanzania s saving rate increased since mid 199s to 2 percent in 212 from a very low level in the mid 9s and Uganda s which had been near zero around 199 and increased to 12 percent. Kenya s saving rate, declined from 2 percent in 1973, remaining steady until the mid 199s and then declined to 8 percent. Burundi s rate has been negative since 199s and now at -1 percent. For Rwanda savings have recovered to a positive rate. 29