Agricultural Transformation in Africa? Assessing the Evidence in Ethiopia

World Development 105 (2018) 286–298

Contents lists available at ScienceDirect

World Development journal homepage: www.elsevier.com/locate/worlddev

Agricultural Transformation in Africa? Assessing the Evidence in Ethiopia Fantu N. Bachewe, Guush Berhane, Bart Minten, Alemayehu S. Taffesse International Food Policy Research Institute, Ethiopia

a r t i c l e

i n f o

Article history: Available online 3 July 2017 Key words: agriculture growth productivity sources of growth adoption Solow decomposition Ethiopia

s u m m a r y Despite significant efforts, Africa has struggled to imitate the rapid agricultural growth that took place in Asia in the 1960s and 1970s. As a rare but important exception, Ethiopia’s agriculture sector recorded remarkable rapid growth during 2004–14. This paper explores this rapid change in the agriculture sector of this important country – the second most populous in Africa. We review the evidence on agricultural growth and decompose the contributions of modern inputs to growth using an adjusted Solow decomposition model. We also highlight the key pathways Ethiopia followed to achieve its agricultural growth. We find that land and labor use expanded significantly and total factor productivity grew by about 2.3% per year over the study period. Moreover, modern input use more than doubled, explaining some of this growth. The expansion in modern input use appears to have been driven by high government expenditures on the agriculture sector, including agricultural extension, but also by an improved road network, higher rural education levels, and favorable international and local price incentives. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Despite the significant recovery and continued output growth recorded in the last two decades, African agriculture still scores low in terms of productivity, measured in yield levels, relative to other parts of the world (NEPAD, 2013, Block, 2014; Nin-Pratt, Johnson, Magalhaes, You, Diao, & Chamberlin, 2011). Further improving agricultural performance in the continent is clearly vital to improve food and nutrition security, accelerate poverty reduction, and boost overall growth (Badiane, Makombe, & Bahiigwa, 2014; Bahiigwa, Samuel Benin, & Samson Jemaneh, 2015). This improvement needs to happen concurrently with still rapid population growth, climate change, and a potentially more constrained international economic environment. Relatedly, a broad set of recommendations have been proposed for African countries to follow to meet this challenge (NEPAD, 2013; Otsuka & Kijima, 2010; Timmer, 1988). Two solutions are deemed rather critical: agricultural intensification and commercialization through market development. The paper explores these and related issues in the context of Ethiopia. Ethiopia was one of the fastest growing economies in the world in the last decade, an impressive feat for a low-income African country that exports relatively little natural resources. National official data show that real agricultural output grew on average by 7.6% per year over the 2004–2014 period, and this growth in particular was a

http://dx.doi.org/10.1016/j.worlddev.2017.05.041 0305-750X/Ó 2017 Elsevier Ltd. All rights reserved.

major contributor to important rural poverty reductions observed during that period; i.e., rural poverty fell from 45% in 1999/00 to 30% in 2010/11 (World Bank, 2014a).1 Agricultural growth in Ethiopia as a major contributor to overall economic growth was a remarkable occurrence for Africa, which lags in agricultural performance globally and is increasingly dependent on imported staple foods to feed its population (Diao, Headey, & Johnson, 2008; Jayne, Anriquez, & Collier, 2013).2 Some potentially important lessons can thus be learnt from Ethiopia’s experience. The purpose of this paper is to document the rapid change in Ethiopia’s agriculture observed since 2004. The analysis of Ethiopia’s agricultural growth experience is conducted using several large agricultural household datasets, including a nationally representative dataset collected by the Central Statistical Agency of Ethiopia (hereafter CSA) through an annual survey of over 40,000 farm households. This paper contributes to the literature in three main ways. First, it reviews the evidence on the rates of agricultural growth in an important country – the second most populous in Africa – and identifies the sources of this growth using an adjusted Solow decomposition model, which allows measurement 1 Such high linkages between agricultural growth and poverty reduction have been noted before in these settings (Christiaensen, Demery, & Kuhl, 2011; Datt & Ravallion, 1998). 2 However, it is to be noted that overall agricultural growth improved significantly in Africa in the last decades compared to earlier periods (Badiane & Makombe, 2014).

287

F.N. Bachewe et al. / World Development 105 (2018) 286–298

of the extent to which modern inputs contributed to this growth. To our knowledge, such a recent and comprehensive study is nonexistent in Africa. Second, using a set of comprehensive datasets, this paper provides new (and updates existing) evidence on changes in modern input adoption over the 2004–2014 period. It also assesses the main sources of Ethiopia’s agricultural modernization process.3 While researchers have looked at determinants of specific improved technology adoption in Africa such as subsidies (e.g., Jayne & Rashid, 2013), roads (e.g., Minten, Koru, & Stifel, 2013), land constraints (e.g., Jayne, Chamberlin, & Headey, 2014), extension (e.g., Krishnan & Patnam, 2014), and cereal intensification (e.g., Spielman, Byerlee, Alemu, & Kelemework, 2010), we discuss these important factors in a more comprehensive way. Third, on a more general level, the paper unpacks Ethiopia’s agricultural growth over the 2004–2014 period, corroborating the nationally representative dataset with a set of micro-datasets – thereby contributing to the debate on quality of data vis-à-vis the reliability of national growth rate estimates in Africa. In brief, this paper examines the incidence, sources, and proximate causes of agricultural growth in Ethiopia over the last decade. In so doing, it provides up-to-date and wide-ranging country-level evidence on agricultural growth and its correlates in Africa. In contrast, almost all other such explorations rely on cross-country datasets and analyses. The paper also combines decomposition of output growth with a model of farmers’ modern input adoption decisions to identify and assess relevant correlations. We find that increasing adoption of improved seeds and chemical fertilizer contributed to agricultural output growth. While starting from a low base, the adoption of improved seeds and chemical fertilizer more than doubled over the last decade (but use is still far from universal). This increasing adoption of modern inputs was facilitated by large investments in agriculture and beyond, leading to improved road and communication networks, a better educated rural population, and a large agricultural extension workforce. Incentives for agricultural intensification were better because of more favorable international prices for export crops and improvements in modern input–output price ratios for locally consumed crops. Agricultural growth further benefited from the absence of major droughts, unlike previous decades (von Braun, Teklu, & Webb, 1998), and, more broadly, from the cessation of widespread civil conflict. Our findings have important implications for the ongoing debate on agricultural transformation in Africa as they show that under certain conditions, significant agricultural growth can be achieved in Africa in a relatively short period. It has been argued that the preconditions for fast, intensification-driven output growth might not be present in Africa (e.g., Diao et al., 2008; Pingali, 2012). However, this situation might be rapidly changing – partly driven by rapid population growth, increasing land scarcity, urbanization, better transport and communication infrastructure, higher incomes, and an emerging middle class – at least in parts of Africa (Reardon et al., 2015). These changing incentives combined with an enabling environment might then lead to improved agricultural performance across the continent. The remainder of the paper is structured as follows. The second section presents the analytical methods used to decompose growth in agricultural output and to study the associates of modern input adoption; it also discusses the sources and contents of the data. The third section provides evidence on agricultural growth and presents results of analyses conducted to indicate the sources of growth. The fourth section describes trends in modern input 3

In doing so, this paper focuses mainly on crop production, which accounted for nearly 70% of the real value of agricultural output during the last decade and grew in importance relative to other subsectors in agriculture.

adoption and presents results of analyses on associates of modern input adoption. The fifth section discusses the evidence on potential pathways to Ethiopia’s rapid agricultural growth, particularly looking at the role of agricultural extension, improved marketing, rural education, and incentives. The final section concludes. 2. Methodology (a) Agricultural growth decomposition This study decomposes growth in crop output using a modified version of Solow’s (1957) growth accounting method. The version used in this study decomposes growth in output into changes in exogenous factors that contribute to changes in output, in addition to changes in input use and total factor productivity (TFP). The method begins by assuming a functional relationship between crop output and inputs used in production. Suppose, in a given year, t, the crop production function in Ethiopia is given as:

Q t ¼ AðtÞf ðLt ; K t ; T t ; F t ; Mt ; It ; P t ; St ; Et ; Z t Þ

ð1Þ

where Q is the real value of crop output and AðtÞ stands for the cumulative effect of technical change.4 Nine production inputs are included: farming labor (Lt ), capital (K t ), land (T t ), chemical fertilizers (F t ), improved seeds (Mt ), irrigation (It ), agro-chemicals (P t ), extension (Et ), and service sector outputs (e.g., transportation and banking services) used in agriculture (St ). The vector Z t in f (. @Q =@t) stands for exogenous factors that affect production but are not directly put into production, such as infrastructure and returns to changes in the scale of agricultural production. Differentiating both sides of (1) with respect to time and dividing the result by Q gives:

DQ DAðtÞ X @Q DJ @Q DZ ¼ þ þ Q A @J Q @Z Q J

where

J 2 ½L; K; T; F; M; I; P; S; E

ð2Þ

where DQ (or) stands for the time derivative of output, DAðtÞ for technical change, and DJ and DZ for time derivatives of the nine inputs and exogenous factors, respectively. Eqn. (2) can be rearranged as:

DTFP ¼

DAðtÞ DQ X DJ wJ  aDRTS  bDRR ¼  A Q J J

ð3Þ

We use wJ ¼ ð@Q =@JÞ  ðJ=Q Þ to get from Eqns. (2) and (3), where wJ is the relative share of input J in crop output. Eqn. (3) indicates that the contribution of a given input/factor to output growth that occurred between periods t and t + 1 is determined by how much its use changed, and its share between t and t + 1. The vector of exogenous factors, Z t , in Eqn. (1) is represented by the last two expressions in (3): DRTS stands for changes in returns to scale (RTS), and DRTS stands for changes in infrastructure, which we proxy by rural roads. a and b represent the rate of change in output per unit of change in RTS and rural roads, respectively. Eqn. (3) can be used to estimate DAðtÞ=AðtÞ using time series data on real crop output, shares and changes in factors used in crop production, length of rural roads, and estimates of b, a, and DRTS. We follow Carlaw and Lipsey (2003) to estimate the effect of returns to scale, as the excess of the sum of shares of factors put into production over 1, which occurs if returns to scale are conP stant, or a ¼ J wJ  1. Moreover, DRTS is given as the excess of the sum of payment to inputs used in production weighted by 4 Eqn. (1) assumes neutral technical change that shifts the production function without affecting the marginal rates of substitution (Solow, 1957). For a review of the growth accounting method see also Carlaw and Lipsey (2003), Romer (1990), and Jorgenson and Griliches (1967).

288

F.N. Bachewe et al. / World Development 105 (2018) 286–298

P the rate of change in the inputs ( J wJ DJ=J) over the rate of growth or expansion in the crop production subsector (f Þ or P DRTS ¼ wJ DJ=J  f , where the minimum of growth in cultivated area andJ labor, inputs indispensable in crop production, is used as a proxy for f. We also conduct the analyses assuming that the rate of expansion of the crop production subsector is given by the minimum growth rate in all inputs. Those results are close to what we report in Section 3. Finally, we replace b by Zhang and Fan’s (2004) estimate of elasticity of TFP with respect to rural roads (0.042). Crop production in Ethiopia was likely affected by other factors. For instance, farmers’ use of organic fertilizer and improved land and water management (ILWM) practices – that the Government of Ethiopia (GoE) invested in expanding – likely contributed to growth in crop output. Furthermore, as discussed in Section 4, crop production in Ethiopia is largely rainfed, implying that weather conditions affect changes in crop outputs. Insofar as these and other factors are not included in the analyses, changes in TFP obtained from the analyses also include the effects of those factors (Sumner, 2014). For example, since the period under discussion was largely characterized by good weather, the TFP results may be biased upward.

where xi are explanatory variables referring to individual farmer i, b is a vector of parameters to be estimated, and ei is a stochastic term assumed to be jointly distributed multivariate normal. Let m ¼ 0; 1; . . . ; M define the different unordered outcomes of the adoption decision. Given the latent variable, the observable choice of the decision maker can be written as:

yi ¼ 1 if yi ¼ max½yim ; m ¼ 0; 1; . . . ; M;

yi ¼ 0;

otherwise

ð5Þ

We thus estimate a multinomial probit model of the general structure in (5), using recent developments in simulation methods that circumvent computational difficulties, namely the Maximum Simulated Likelihood Estimation method due to GewekeHajivassiliou-Keane (Cappellari & Jenkins, 2006). We use CSA’s annual Agricultural Sample Survey (AgSS) data, which extend from 2004/05 to 2013/14. The model takes adoption of improved seeds and chemical fertilizers as dependent variables, in which non-adopters are contrasted with those adopting chemical fertilizer (=1), improved seeds (=2), or both (=3), for the four main cereals in the country: teff, maize, wheat, and barley. Household, plot, and climatic characteristics are used as right-hand side variables.5 (c) Data sources and coverage

(b) Adoption model This section briefly outlines the econometric model estimated to analyze factors associated with adoption of modern inputs, specifically, chemical fertilizer and improved seeds. The decision to adopt these two inputs is often interrelated for at least three reasons. First, average yield increases could be higher when the inputs are adopted simultaneously than when adopted separately due to complementarity effects (Feder, 1982). Second, the combined use of two or more inputs can at times be risk-reducing, in the sense that the synergistic use leads to better outcomes, as in the case of a particular seed becoming more drought- or damage-resistant if used with specialized nutrients. Related to the two circumstances above, Just and Zilberman (1983) show, for example, that the ‘‘correlation of outputs under alternative technologies plays an important role in determining adoption rates.” An important consideration related to our analysis is that although the proportion of farmers adopting only chemical fertilizer is higher, farmers using improved seeds are also more likely to adopt chemical fertilizers, suggesting that adoption decisions of such inputs are rather interdependent. Third, in the context of Ethiopia, agricultural input supply strategies in the last decade encouraged farmers to adopt chemical fertilizers and improved seeds as a package, at times bundled with input credit, making adoption of these two inputs an inherently simultaneous decision or one between sets of possible technology bundles. Feder (1985) suggests modeling this technology adoption decision as a multiple choice facing the farmer, who maximizes expected utility of net benefits derived conditional on adoption of a specific technology or set of technologies. Since we are interested to understand, relative to non-adopters, factors associated with adoption of chemical fertilizers, improved seeds, or both chemical fertilizers and improved seeds, we follow Dorfman (1996) and Barham, Foltz, Jackson-Smith, and Moon (2004) to model this decision in a multivariate framework. More specifically, we use a multinomial probit model to assess factors associated with adoption of only chemical fertilizer, only improved seeds, or both chemical fertilizers and improved seeds, in contrast to the benchmark category of adoption of neither input. Consider the unobserved latent variable yit , which represents the net benefit from adopting the technology or set of technologies, described in the following linear function:

yi ¼ b0 xi þ ei

ð4Þ

CSA publishes annual statistical reports on agricultural output and input use that are representative at national, regional, and zonal levels. Most of the descriptive analyses in this study use data taken from these reports,6 which can be accessed at www.csa.gov. et. The data pertain to smallholder farmers, who dominate agricultural land use in Ethiopia, accounting for 94% of the nationwide area cultivated in 2013/14 (CSA, 2014a). CSA’s AgSS, an exceptionally massive survey in Africa, collects data from over 40,000 agricultural households. For instance, 44,993 households were surveyed in 2013/14 (CSA, 2014a). During the period covered in this study, the sampling frame of the AgSS included the entire rural parts of the country except the non-sedentary population of three zones of Afar and six zones of Somali regions (CSA, 2005a–2014a). Data on input use levels rely on farmers’ recall, while crop output/yield data combine farmers’ recall and crop-cut measures. Out of the households sampled from each enumeration area, the output of up to five farmers’ plots is measured for each crop using the crop-cut method.7 Area cultivated is collected through interviews. Input use data used in the growth accounting analysis are obtained from CSA annual reports (CSA, 2005a–2014a; 2005b– 2014b; 2005c–2014c). Rural road and real crop output (Q) data are obtained from National Bank of Ethiopia (NBE, 2014). We proxy changes in intermediate service sector inputs by changes in real agricultural output, which assumes that changes in intermediate service sector input use are proportional to growth in crop output. Unlike data on output and input use levels, which often are available annually, data on factor shares are infrequent. We derive factor shares for labor, capital, land, chemical fertilizer, and services from two Social Accounting Matrices (SAMs) of Ethiopia: the 2005/06 (EDRI, 2009) and 2009/10 (Engida, Tamru, Tsehaye, Debowicz, Dorosh, & Robinson, 2011) SAMs. We complement the latter by the relative output elasticities with respect to quantity of improved seeds, pesticides, irrigation, and area covered by extension, obtained by estimating a crop pro-

5 In the online appendix, we summarize the variables used in the analyses and provide results obtained using binary probit and ordered probit models. 6 These data pertain to the main cropping season locally known as meher. Only 10% of all land was cultivated in belg, the second cropping season. Data on belg-season are infrequent. However, the data available indicate that no major changes occurred in the relative importance of the two seasons over the last decade. 7 For a summary on how CSA conducts crop cuts, see the online appendix.

F.N. Bachewe et al. / World Development 105 (2018) 286–298

289

During the 2004/05–2013/14 period,9 Ethiopia’s real gross domestic product (GDP) and real per capital GDP grew by average annual rates of 10.7 and 7.9%, respectively (NBE, 2014). Real agricultural GDP grew at an average annual rate of 7.6%. Within agriculture, crop production is the most important subsector. On average, crop output represented 32% of real GDP and grew at an annual rate of 8.8% (NBE, 2014). CSA data indicate that total crop output grew by over 120%, from 142.4 million metric tons (mmt) in 2004/05 to 319.7 mmt in 2014/15 – an average annual growth rate of 9.4%. Total crop area grew at an average annual rate of 2.7% (CSA, 2005a–2014a). Growth in total crop output and area appears to have mainly been driven by growth in grain crops, which on average accounted for 96.5% of the total crop area and 84% of the volume of output. Annual growth in grain output and area averaged 2.6 and 8.7%, respectively (CSA, 2005a–2014a). Furthermore, grains on average accounted for over 24% of real GDP (NBE, 2014). Crop yields have demonstrated major improvements since their low levels at the beginning of the decade (CSA, 2005a–2014a). Annual growth in cereal yields, which averaged about 7%, was faster relative to other crop groups. Figure 1 illustrates the contributions of expansion in cultivated area and yield growth to grain output growth; the contribution of area expansion declined over time. Yields grew consistently faster than area expansion, but the contribution of yield growth to output growth in the second half

of the decade, which averaged 76%, was considerably larger than in the first half (59%) (CSA, 2005a–2014a). We use complementary data sources to triangulate the robustness of these estimates, an important exercise as CSA data have been challenged by a number of authors (e.g., Dercon & Hill, 2009; Gollin, 2011; Mandefro & Jerven, 2015). We present evidence from four groups of complementary methods and data sources. First, CSA data on growth in yields of five main cereals, which on average accounted for 73% of the total cultivated area during the period studied, are compared with those obtained from other large-scale rural household survey datasets collected over the last decade. The surveys and the objectives for which they were conducted are discussed in the online appendix. While the surveys cover relatively large parts of the country, caution is required in comparing yields over time since the surveys were fielded in different areas and periods, and for different purposes. However, growth rates in yields of the five major cereals calculated using these surveys are generally similar to the growth rates obtained from CSA data, except for sorghum (Table 1, third column). Second, we use data from the Ethiopian Rural Household Survey (ERHS), a unique longitudinal household survey that contains information on agriculture, consumption, assets, and income of almost 1,500 households in 15 villages across the country surveyed for over a decade (1994 to 2009). The ERHS dataset indicates considerable improvements in yields, as the real value of output grew annually by 5.2% during 1994–2009. In particular, growth was rapid during 2004–2009, when it averaged 6.7%. Growth rates in the real value of agricultural output as well as crop yields implied by ERHS data were lower than the corresponding values from CSA data (fourth column), however. Third, as the last decade was characterized by relatively good weather, we use CSA data to estimate growth rates covering longer time periods, including the recovery from a major drought at the beginning of the 2000 s. The last two columns of Table 1 show that growth in yields over these longer periods was considerably lower than for those estimated for the period 2004–2014 for all cereals (see also Mellor, 2014; Mellor & Kuma, 2014). Cereal yields declined during 2014/15–2015/16 (CSA, 2016a) because agricultural production in 2015/16 was disrupted by irregular rain patterns (EHCT, 2015), which implies growth rates would be lower if the yield growth analysis was extended beyond 2014. Fourth, focus groups in the 304 kebeles (villages) in the AGP baseline survey were asked to compare average yields in their community in 2011 relative to 10 years earlier. Focus groups included at least five people knowledgeable about the community, such as community leaders, kebele chairpersons, elders, priests, and teachers. All cereals were perceived to have shown considerable growth in yield. The highest yield growth was noted in maize (44%) and wheat (35%). Growth in white teff, barley, black teff, and sorghum yields averaged 20, 18, 17, and 16%, respectively. These growth rates were similar to those implied by CSA data for the 2001–2012 period for maize and wheat but were considerably lower for teff and sorghum. Moreover, increases in cultivated land amounted to approximately 28% in AGP communities over the 2001–2011 period. This is in line with CSA estimates at national level. Therefore, these complementary data sources illustrate significant yield and production growth in the cereal subsector in the last decade, as shown earlier using CSA data. However, documented growth rates from these complementary methods are generally also slightly lower.10 Differences in survey methods may have

8 In the online appendix, we provide two tables and a figure that show the trend in cultivated area, yields, and the contribution of yields and area to growth in grain output. 9 Unless stated otherwise the description in this section pertains to the 2004/05– 2013/14 period.

10 However, they are not far off the 6% target set by CAADP (see also Mellor, 2014). CAADP stands for the Comprehensive Africa Agriculture Development Programme signed by African Heads of State Summit as a New Partnership for Africa’s Development Program. Ethiopia signed this agreement in 2003.

Figure 1. Growth in area cultivated and yield of grains. Source: Authors’ computation using CSA data (CSA, 2005a–2014a). Note: Grains include 8 cereals, 9 pulses, and 6 oilseeds or a total of 23 crop types.

duction function using data from the Agricultural Growth Program (AGP) of Ethiopia baseline survey. The AGP dataset used in the analysis pertains to the 2010/11 main agricultural season and includes about 7,000 households, sampled to represent 8 million households residing in Tigray, Amhara, Oromiya, and SNNP regions. Furthermore, since data on factor shares are available for only 2005/06 and 2009/10, we considered different scenarios to compute factor shares for the remaining years. Results obtained from different scenarios, which can be obtained upon request, are similar in most factors except land and labor, mainly because factor shares change little in the remaining inputs. The results presented in Section 4 assume that differences in factor shares in the 2005/06 and 2009/10 SAMs occurred uniformly before 2005/06, during 2005/06–2009/10, and after 2009/10.

3. Agricultural growth in Ethiopia (a) Evidence of agricultural growth, 2004–20148

290

F.N. Bachewe et al. / World Development 105 (2018) 286–298

Table 1 Estimates from alternative datasets of annual cereal yield growth in Ethiopia, % Survey Period Number of surveys

CSA 2005–2014 Annual

Ad hoc surveys 2008–2013 8 surveys

ERHS 2004 – 2009 2 surveys

CSA 1997–2012 Annual

CSA 2001–2012 Annual

5.8 6.2 4.8 5.4 5.4

4.7 6.2 6.8 6.3 1.8

1.7 0.4 10.1 3.6 –

4.2 3.4 – 4.0 4.5

4.6 4.2 – 3.6 4.8

Teff Maize Barley Wheat Sorghum Source: Authors’ computation.

Table 2 Contribution of inputs, other factors, and TFP to crop output growth, % Year (1) 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 2004/05–08/09 2009/10–13/14 2004/05–13/14

Crop output (2)

Labor

Capital

Land

Fertilizer

Impr. seeds

Pesticides

(3)

(4)

(5)

(6)

(7)

15.0 11.0 8.0 6.5 8.7 10.3 5.0 8.2 6.6

3.5 2.1 2.7 2.6 3.6 3.4 1.6 1.5 1.4

0.1 0.0 0.1 0.1 0.0 0.0 0.0 0.0 0.0

0.9 1.1 1.1 0.9 1.0 1.2 1.2 1.0 0.6

2.2 0.2 0.1 0.1 0.0 0.7 0.6 1.3 1.0

0.4 0.4 0.3 1.1 2.4 2.2 0.4 0.4 0.4

10.11 7.52 8.81

2.72 1.99 2.50

0.05 0.02 0.03

1.01 0.97 1.00

0.65 0.90 0.69

Average during 0.56 0.08 0.83 0.09 0.88 0.09

Rural roads (13)

D TFP

(11)

Ret. to scale (12)

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

0.3 0.0 0.0 0.1 0.3 0.4 0.1 0.1 0.1

0.4 0.5 0.3 0.3 0.2 0.6 0.1 0.1 0.1

5.7 6.4 3.1 1.0 1.0 0.4 0.6 3.4 1.5

0.02 0.02 0.02

0.11 0.17 0.16

0.36 0.24 0.29

4.07 0.96 2.34

Irrigation

Extension

Services

(8)

(9)

(10)

0.1 0.1 0.1 0.0 0.1 0.1 0.0 0.1 0.1

0.6 0.2 0.1 0.2 0.1 0.1 0.1 0.0 0.0

0.7 0.4 0.0 0.0 0.0 1.9 1.5 0.3 1.3

0.19 0.06 0.12

0.28 1.28 0.70

(14)

Source: Authors’ computation.

contributed to these deviations. CSA statistics on yields are based on data collected through crop-cut and recall methods while the others rely on the recall method. It seems that the higher yield levels found in crop-cuts would benefit from further analysis. For example, Detriere (2016) shows how yield levels from crop-cuts are significantly higher for larger plots compared to self-reported production. On the other hand, it is also to be noted that some datasets, for example the ERHS, oversample low agricultural potential or drought-prone areas (Stifel & Woldehanna, 2016). (b) Sources of agricultural growth11

made since the late 1990 s, and the relatively lower levels of productivity at the beginning of the period. Results of the growth accounting analyses not only show the importance of labor and land in the output growth recorded in the last decade, but also indicate the growing importance of modern inputs and agricultural extension. Relative to the contribution of labor and land, modern inputs contribute less in absolute terms, but their contribution increased over time. This is also the case for the area covered by extension packages. The contribution of the latter set of inputs was higher in the second half of the period relative to the first half while the reverse held for labor and land. The performance of Ethiopia’s agriculture is consistent with the recent recovery and growth of agriculture in many African countries (e.g., Nin-Pratt, 2015; Yu & Nin-Pratt, 2011). Nin-Pratt (2015) reports that agricultural output per worker grew by 2% during 2001–2012. This compares to 0.6% growth during the 1990s and no growth in the 1970s and 1980s. He also estimates an annual average TFP growth rate of 2.2% for the best performers during the period 1995–2012. Ethiopia’s TFP growth rate for the same period stands at 2.6%.

Table 2 shows results of the growth accounting analysis. The second column of the table provides changes in real crop output between consecutive years, starting from 2004/05. The remaining columns provide the contribution of inputs and exogenous factors to growth in crop output. The table indicates that annual growth in the real value of crop output averaged 8.8% during 2004/05– 2013/14 (last row). Out of the 8.8% growth in output, 2.5% was accounted for by labor (column 3), while expansion in cultivated land (column 5) accounted for 1.0%. A further 0.9, 0.7, and 0.7% of the 8.8% output growth originated from increases in improved seeds, area covered by extension packages, and chemical fertilizer use, respectively. Table 2 also provides a basis for exploring growth in crop output resulting from changes in TFP. Annual increase in TFP on average accounted for about 2.3% of the growth in crop output. It is interesting to note that the contributions of TFP were higher during the first half of the period (4%) relative to the second half (1%), possibly reflecting the considerable investment in extension provision

Technological change in agriculture – such as the replacement of traditional seed varieties with improved cultivars and increased adoption of chemical fertilizers, often aided by improved irrigation

11 In the online appendix, we provide more details of the data and results of the growth accounting analyses.

12 We provide a detailed summary of fertilizer and improved seeds application for all crop groups in the online appendix.

4. Intensification and the adoption of improved agricultural technologies (a) Adoption of chemical fertilizer, improved seeds, and other inputs12

291

F.N. Bachewe et al. / World Development 105 (2018) 286–298 Table 3 Proportion of farm holders applying improved seeds,% Crop Barley Maize Sorghum Teff Wheat Cereals

2004/05

2009/10

2013/14

0.8 11.6 0.9 1.0 4.5 10.1

1.2 15.7 1.8 2.4 4.1 11.3

0.8 27.6 0.4 4.6 7.7 21.5

Source: Authors’ calculations using CSA data (CSA, 2005b–2014b).

Figure 2. Proportions of cereal farmers applying fertilizer and applied area. Source: Authors calculations using CSA data (CSA, 2005b–2014b).

– drove the dramatic agricultural growth in Asian countries in the 1960s and 1970s (Evenson & Gollin, 2003; Feder, 1980; Hiebert, 1974; Kislev & Shchori-Bachrach, 1973). Significant effort has since been made to replicate such rapid growth in Africa, particularly in Ethiopia. Since the early 1990s, Ethiopia has implemented several cereal intensification programs promoting the adoption of modern technologies. At the center of these strategies was the push for adoption of chemical fertilizer and improved seed packages by smallholders (Spielman et al., 2010). This section assesses the extent to which changes in the adoption of improved technologies occurred in the last decade. (i) Chemical fertilizer Since chemical fertilizer is not produced in Ethiopia, trends in chemical fertilizer imports are a good indicator of total chemical fertilizer use in the country. Rashid, Tefera, Minot, and Ayele (2013) indicate that Ethiopia’s chemical fertilizer imports remained below 100,000 metric tons (mt) per year up to the mid-1990s. Chemical fertilizer imports increased by over 124%, from 346,000 mt in 2004/05 to 778,000 mt (the average of 2012 and 2013 imports). Chemical fertilizer use by smallholders increased by 144% (CSA, 2005b–2014b) over the same decade (i.e., excluding commercial farms). Owing partly to the attention given to cereal production to achieve food security, most of the chemical fertilizer was used on cereals. According to CSA, 4.7 million smallholders (46%) growing cereals used chemical fertilizer in 2004/05, and this number increased to 10.1 million (76%) in 2013/14 (Figure 2). Cereal area applied with chemical fertilizer – which nearly doubled from 2.7 million hectares (36%) in 2004/05 to 5.2 million hectares (53%) in 2013/14 – accounted for at least 91% of total fertilized area in all years except 2009/10. The intensity of chemical fertilizer use on areas covered with chemical fertilizer was 28% higher in 2013/14 (122 kg/ha) than in 2004/05 (95 kg/ha) (CSA, 2005b– 2014b). Although chemical fertilizer use on other crops was less prevalent, in cereals it increased considerably over the period studied.13

from 4% of the area in 2004/05 to 10% in 2013/14 (CSA, 2005b– 2014b). This seems to have been driven especially by the rapid increase of improved seed adoption in maize. Large increases are also noted in the case of teff and wheat. While these data show substantial improvement in adoption over time, significant measurement errors in CSA’s improved seed adoption data might be present, as discussed below. The two main types of improved seeds are hybrid and openpollinated. Hybrid seeds often show considerable yield gains, but these gains often decrease rapidly – the rate depends on the type of hybrid and on the environment – inducing farmers to purchase new seeds regularly to maintain hybrid vigor (heterosis). With hybrid maize seeds accounting for 62% of the improved maize varieties released during 2000–2009, they are especially important for that crop in Ethiopia. Benson, Spielman, and Kasa (2014), Zeng, Alwang, Norton, Shiferaw, and Jaleta (2013), and Abate et al. (2015) estimate that 35–40% of maize area is applied with hybrid and improved seeds. Therefore, CSA adoption rate estimates are lower than those mentioned in these studies. In contrast to hybrid seeds, open-pollinated varieties do not exhibit such sharp yield decreases after the first year (but yields do deteriorate as well and it is recommended that seeds be replaced every 3–4 years, at least in the case of maize). Improved varieties can therefore be multiplied by farmers themselves, and can often be easily obtained through informal distribution channels. This makes it difficult to estimate the extent to which openpollinated varieties have spread in the country. For example, while only 8% of farmers report using improved wheat varieties, some key informants estimate that use of improved cultivars might be considerably higher. CSA survey questions regarding improved seed use are framed in such a way that the responses cannot easily capture whether farmers adopted improved seeds, particularly open-pollinated ones. While only 62% of wheat farmers in the sample indicated using improved seeds, recent DNA fingerprinting exercises in three zones in Oromiya region put this number at 96%.14 The difference is lower for maize (56% indicate using improved seeds relative to the 64% implied by DNA testing), seemingly because annually purchased hybrid maize seeds are dominant.

(ii) Improved seeds The share of farmers that reported using improved seeds and area applied with improved seeds in the cereal subsector increased substantially over the last decade (Table 3). While overall adoption rates were low, the share more than doubled over the last decade, from 10% of cereal producers in 2004/05 to 21% in 2013/14, and

(iii) Irrigation and pesticides CSA data show that irrigation use was low and did not change in the last decade for any of the crop categories (Table 4). However, other data sources find larger irrigated areas than reported by CSA. For example, Hagos, Makombe, Namara, and Awulachew (2009) and Mulat (2011) place the proportion of irrigated area at about 5% of all crop area. Moreover, the Growth and Transformation Plan (GTP) progress report (MoA, 2013) indicates significantly larger irrigated areas. These differences are possibly explained by the fact that irrigation is practiced more by larger farms and that irrigation is used more during the belg period for off-season crops. CSA data further indicate considerable increases in pesticide use

13 These figures are consistent with other datasets. Using the AGP datasets, Berhane, Woldu, and Ragasa (2017) report that more than 67% of households used fertilizer at least once in the last seven years prior to 2011.

14 http://www.slideshare.net/futureagricultures/results-dna-finger-printing-pilotpresentation-nairobi-July-14-2014

292

F.N. Bachewe et al. / World Development 105 (2018) 286–298

Table 4 Area under irrigation and pesticide use Variable and data source Irrigation 1. Irrigated area (‘000 ha, only meher) – CSA annual reports Proportion of area irrigated (meher, %) All crops Cereals Pulses Oilseeds Vegetables Root crops 2. Irrigated area (‘000 ha) – MoA Annual GTP progress reports Proportion of total crop area applied with pesticides (percent) All crops Cereals Pulses Oilseeds Vegetables Root crops

2004/05

2009/10

2012/13

154

152

0.8 0.8 0.5 0.2 5.9 6.8

1.0 1.0 0.5 0.1 3.7 5.4 853

0.8 0.6 0.2 0.3 5.2 9.1 1830

13.0 16.7 0.8 1.1 2.3 3.0

12.4 13.4 9.6 8.7 4.1 3.9

21.5 26.1 6.5 2.8 4.1 12.6

Source: Authors’ computations using CSA annual reports (2005b–2014b).

over time (Table 4). While 13% of the crop area was exposed to pesticides in 2004/05, this increased to 21% in 2013/14. Particularly, the proportion of total area on which pesticides were applied grew rapidly during 2009/10–2013/14. This might have particularly been driven by increasing herbicide use (Tamru, Minten, Bachewe, & Alemu, 2017). Finally, it is to be noted that land and labor use expanded considerably over the last decade and are among the important contributors to Ethiopia’s agricultural growth. However, growth in the number of farmers (39%) was faster than area expansion (27%), suggesting smaller farm sizes over time; therefore, more intensive labor use per unit of land is likely, given the relatively few off-farm opportunities in rural areas (Bachewe, Berhane, Minten, & Taffesse, 2016; World Bank, 2014a). Headey, Dereje, and Taffesse (2014) confirm these stylized facts for Ethiopia and document that farm sizes have declined rapidly, young farmers cultivate substantially less land than previous generations, and family labor use per hectare increases substantially with increasing land pressure, leading to higher gross incomes per hectare.

(iv) Factors associated with sources of growth The evidence presented in the last section illustrates that important changes took place in the adoption of improved agricultural technologies over the last decade. The uptake in improved technologies was especially faster in the second half of the period, 2009/2010–2013/14, when agricultural growth was linked more with increased use of modern inputs (Tables 3 and 4; Figure 2). This evidence leads to the question of which factors are associated with the increased modernization and intensification in agriculture. We use two criteria to identify the drivers of the increasing adoption of improved technologies in the last decade. First, drivers need to be linked with significantly greater adoption of improved practices. Second, they need to have shown major positive changes over the last decade. Table 5 presents the average marginal effects from the multinomial probit model estimated (as described above). The results show large and significant associations of improved technology adoption, with, among others, extension, remoteness, and education. Farmers who received visits from extension workers are associated with a higher likelihood of adopting improved technologies. Less remote and more educated farmers are also more likely to adopt the inputs. These results are consistent with the correlates of improved technology adoption observed in the literature. Moreover, these factors changed considerably during the period

studied (as posited below) and are therefore arguably among the main drivers for improved technology adoption. Other factors also show significant association with improved technology adoption. First, larger plots are associated with a higher likelihood of improved technology adoption and, except for maize and wheat, households with larger farm size are more likely to adopt these technologies. Second, owner-managed plots mostly have a lower likelihood of adoption of improved technologies. Third, households’ access to credit is associated with a higher likelihood of adoption of improved agricultural technologies, likely by reducing households’ financial constraints. Finally, it is important to note that while these results illustrate significant associations between improved technology adoption and potential sources of growth in Ethiopia, the results do not unambiguously show that these changes can be attributed to these factors because of potential methodological issues (most importantly, endogeneity). Nevertheless, given the depth of the nationally representative plot-level data used covering several years and hence the consistency of the results across models estimated, we believe the results can be plausibly taken as strong suggestive evidence in the direction of causality we cautiously interpret. 15 A number of studies that use other methods but smaller datasets corroborate our results.16 Another important factor contributing to technology adoption is the gains in output-input price ratios in recent years (see later), with substantial positive implications for technology adoption (Morris, 2007). Along this line, Spielman, Kelemwork, and Alemu

15 Furthermore, we conduct Granger causality tests on the variables. The results, which can be obtained upon request, imply, among others, that the null hypotheses that travel time, education, and access to credit do not cause adoption of only improved seeds, and improved seeds and fertilizer are rejected. Moreover, extension visit and adoption of only fertilizer and only improved seeds appear to have causation that runs in both directions. 16 A number of studies assess the impact of increased coverage by extension agents. Ragasa, Berhane, Tadesse, and Taffesse (2013) and Minten et al. (2013) show that a strong association exists between increased use of technologies, mainly use of improved seeds, fertilizer, and pesticides, and extension services received. Using large-scale panel data in high-potential agricultural areas, Berhane et al. (2017) find that agricultural extension increases productivity indirectly through its effects on adoption of improved technologies. Using data collected in a quasi-experimental setting from a remote area in Ethiopia, Minten et al. (2013) also show that a 20 km increase in the distance from the farm to a modern input distribution center and output market leads to a 47 kg/ha and 6 kg/ha reduction in chemical fertilizer and improved seed use, respectively. Stifel and Minten (2017) show a strong association between remoteness with modern input use and agricultural production, consistent with our finding that recent investments in Ethiopia’s transportation network might have contributed to increases in modern input use.

Table 5 Variables associated with adoption of improved seeds and chemical fertilizer in cereal production (multinomial probit model) Variablesa

Barley

Maize

Teff

Wheat

Fertilizer

Both

Improved seeds

Fertilizer

Both

Improved seeds

Fertilizer

Both

Improved seeds

Fertilizer

Both

Key variables of interest Received extension visit Avg. travel time to nearest city (hrs) Education (highest grade) Plot area in hectares Total crop area Did the farmer own the land? Had the farmer access to credit?

0.0004** 0.0001 0.0001*** 0.001** 0.0000 0.0004 0.0001

0.024*** 0.003*** 0.003*** 0.171*** 0.001 0.022*** 0.053***

0.002*** 0.0005** 0.0002*** 0.003*** 0.0003*** 0.0001 0.001***

0.013*** 0.0004 0.0005*** 0.020*** 0.0003 0.011*** 0.003***

0.022*** 0.007*** 0.001*** 0.098*** 0.005*** 0.034*** 0.030***

0.065*** 0.005*** 0.003*** 0.143*** 0.001 0.051*** 0.024******

0.0005 0.0003 0.0002*** 0.002*** 0.0002** 0.0001 0.001***

0.056*** 0.006*** 0.007*** 0.112*** 0.003*** 0.020*** 0.081***

0.006*** 0.001*** 0.001*** 0.003*** 0.0004* 0.001 0.004***

0.007*** 0.001 0.001*** 0.009*** 0.001 0.0001 0.004***

0.018*** 0.002 0.002*** 0.201*** 0.001 0.024*** 0.068***

0.020*** 0.006*** 0.002*** 0.020*** 0.0001 0.005*** 0.013***

Household characteristics Age of farm holder Farm holder is female Household size

0.0000 0.0001 0.0001*

0.001*** 0.002 0.001***

0.0000 0.0001 0.0001**

0.0000 0.003*** 0.001***

0.0004*** 0.004** 0.001***

0.0002*** 0.001 0.002***

0.0000 0.001 0.0001*

0.0000 0.004 0.0003

0.0000 0.001 0.0002**

0.0000 0.002* 0.001***

0.001*** 0.004 0.002***

0.0001* 0.002 0.001***

Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes

Other controls Fertilizer users in woreda (%, lagged) Imp. seeds users in woreda (%, lagged) Zonal average crop output sold (%) Rainfall and other climate variables Year dummies Zone dummies Chi2 Number of observations

Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes 38,403.3 149,356

Yes Yes Yes Yes Yes Yes 91,636.4 297,708

Source: Authors’ analyses. Table shows (average) marginal effects. Coefficients with ***, **, and * are significant at 1, 5, and 10% levels, respectively. Note: a) Full set of estimates and standard errors provided in Table F.3 of the online appendix accompanying this paper.

Yes Yes Yes Yes Yes Yes 66,085.0 213,343

Yes Yes Yes Yes Yes Yes 36,222.5 108,785

Yes Yes Yes Yes Yes Yes

F.N. Bachewe et al. / World Development 105 (2018) 286–298

Improved seeds

293

294

F.N. Bachewe et al. / World Development 105 (2018) 286–298

inefficiencies that hampered past technology adoption (such as in markets of input, output, labor, and credit and in information and risk) (Jack, 2011). (a) Changes in informational efficiency and the role of agricultural extension

Figure 3. Number and share of holders and area covered through the public extension system, 2004/05–2013/14. Source: Authors calculations using CSA data (CSA, 2005b–2014b). Note: Ext-pkg = Extension-package.

(2012) document the extent to which incentives matter in the adoption of chemical fertilizer in Ethiopia. Minten et al. (2013) illustrate for northern Ethiopia how changes in value-costs ratios, driven by transportation and transaction costs, led to significantly lower adoption rates of chemical fertilizer and improved seeds. In sum, our results are consistent with the available evidence that suggests these factors are associated with increasing adoption of improved technologies in Ethiopia. The evidence on changes in the last decade of these important sources of growth – agricultural extension, road infrastructure, education, and incentives – as well as other factors is discussed next.

5. Evidence on potential pathways to rapid agricultural growth in Ethiopia The GoE has for a long time put agriculture at the center of its national policy priorities. The Agriculture Development Led Industrialization (ADLI) strategy was formulated in the mid-1990s to serve as a roadmap to transform smallholder agriculture in the country. Rural education and health, infrastructure, extension services, and strengthening of public agricultural research were among its top priorities.17 Ethiopia is one of only four African countries to have implemented the CAADP agreement of a 10% target of annual government expenditures going to agriculture over the 2003–2013 period (Benin, 2014). The sections below illustrate the extent to which some of these expenditures contributed to improved agricultural performance and adoption of improved agricultural technologies that likely contributed to a reduction in the 17 Government expenditures in Ethiopia over the years were guided by several plans that consistently advanced agriculture as an important sector in which to invest. The 2005–2010 period was guided by the Plan for Accelerated and Sustained Development to End Poverty (PASDEP). The 2010–2015 period was the first phase of the Growth and Transformation Plan (GTP).

Since 1992, the GoE has made large investments on an agricultural extension system focused on the provision of advisory and training services led by frontline development agents (DAs). To redress the challenges faced and to scale up best practices learned in an earlier period, the GoE started a more comprehensive largescale extension system in 2002. Ethiopia thus achieved one of the highest DA-to-farmer ratios in the world, estimated at roughly 1 DA for every 476 farmers or 21 DAs per 10,000 farmers, significantly higher than in other countries such as China, Indonesia, and Tanzania, where this ratio is 16, 6, and 4, respectively (Davis et al., 2010).18 CSA data allow us to track changes in the number of farmers with access to extension advice. Figure 3 clearly confirms the increasing presence of extension agents over the last decade. The number of smallholders who reported using extension advisory service tripled from 3.6 million in 2004/05 to 10.9 million in 2013/14, an increase from 33% of all smallholders to 71% (bottom graph). Those that used different crop extension packages more than doubled from 2.6 million in 2004/05 to 6.6 million in 2013/14. In the same period, cultivated area covered by the extension package program increased from 1.5 million hectares in 2004/05 to 3.9 million hectares in 2013/14. Moreover, most studies indicate that farmers seem to have been satisfied by the services delivered by extension agents (Berhane et al., 2017; Davis et al., 2010), although some are more skeptical (Berhanu & Poulton, 2014). (b) Changes in input and output market efficiency Well-functioning agricultural marketing systems are important anywhere in the world but more so in Ethiopia given that poorly functioning food markets resulted in disastrous food insecurity outcomes in the past, with food stocks available in some parts of the country alongside widespread acute food insecurity in others (von Braun et al., 1998). Major reasons for poorly functioning food markets in Ethiopia have been linked to lack of market information, bad road infrastructure, and high transaction costs (von Braun & Olofinbiyi, 2007). Ethiopia’s progress in the last decade in filling these gaps had enormous implications for growth, particularly in the agriculture sector, and for overall market performance. Most importantly, the GoE embarked on a large road investment program in the last two decades. The total length of allweather surfaced roads tripled in less than 15 years, from an estimated 32,900 km in 2000 to 99,500 km in 2013 (NBE, 2014). This substantially improved connectivity of agricultural markets in the country. In 1997/98 only 15% of the population was within three hours of a city with a population of at least 50,000, a proportion that increased to 47% by 2010/11 (Kedir, Schmidt, & Tilahun, 2015). The improved road network led, among others, to a reduction of travel times between wholesale markets by an estimated 20%. However, travel costs might have fallen even more than 20% with more competition and a shift to bigger and cheaper trucks (Minten, Tamiru, and Stifel, 2014). In this regard, the growing accessibility to and expansion of transport services have been shown to positively impact agricultural productivity (Li, Gorham, & Nuru, 2011).

18 This is with the caveat that these figures do not tell much about the quality of the extension system.

F.N. Bachewe et al. / World Development 105 (2018) 286–298

295

10 years was thus likely related to these improvements. Moreover, access to urban centers leads to increasing agricultural intensification, and urbanization can then act as an engine of agricultural transformation (Schultz, 1951; Vandercasteelen, Tamru, Minten, & Swinnen., 2016). For example, Minten, Tamru, Engida, and Tadesse (2016) show that adoption of chemical fertilizer for teff production was significantly higher in villages close to Addis Ababa than in more remote ones. Other factors might have contributed to the reduction of input and output market inefficiencies. In the last decade, the use of mobile phones significantly expanded in urban and rural areas of Ethiopia, despite the still limited penetration in many rural parts. This might possibly have impacted agricultural trade through better access to information, making markets more efficient (Minten et al., 2014). (c) Changes in human capital accumulation Accumulation of human capital can affect adoption of improved technologies in several important ways. Ethiopia made significant strides to achieve universal coverage of primary education, particularly in rural areas. The number of educated farmers therefore increased as agriculture retained some of these students afterward. CSA data show that the share of illiterate farmers declined at 1.8% per year over the 2004/05 to 2013/14 period. Moreover, efforts were made to make adult education more accessible. Figure 4. Output – fertilizer price ratio (top) and export price indices (2003/ 04 = 1.0) (bottom). Source: Authors’ computations using CSA data (CSA, 2005b– 2014b).

Over the last decade, prices in input and output markets changed considerably, leading to improved incentives for agricultural intensification. This was the case for both tradable and nontradable agriculture subsectors. First, the ratios of output (of the five main cereals) to chemical fertilizer prices were twice as high in 2012 than in 2004, indicating improved incentives favoring application of chemical fertilizer in these crops (Figure 4).19 This change in output-input ratio over time seems to be linked to a number of factors, including fixing the margins for chemical fertilizer distributing cooperatives to keep prices low; overvaluation of the Ethiopian currency (birr), making imports cheaper (World Bank, 2014b); a decline in international chemical fertilizer prices since 2008 (FAO, 2012); and high output prices, especially during 2008–2010. Second, international prices were significantly higher for most export crops at the end of the decade than at the beginning. We examine the price evolution of Ethiopia’s four most important export commodity groups. Figure 4 shows that the price of coffee was 2.5 times higher in 2012/13 than in 2003/04, while the prices of oilseeds and pulses were twice as high. The increase in the price of these commodities led to significantly higher export revenues, and further incentivized investments in these commodities, as shown, for example, by the rapid expansion of sesame cultivation during the period (CSA, 2005a–2014a). Third, rising market demand due to urbanization and economic growth was also an important contributor to changes in market performance. Compared to the beginning of the decade, an estimated 3.7 million more people were living in urban settings. Combined with an expanding transport network, this may have made supplying to the market more rewarding for farmers. The increase in commercial surplus from agricultural production over the last 19 To calculate these ratios, we rely on average cooperatives union prices of chemical fertilizer reported at the regional level and CSA producer prices collected in all rural zones of the country.

(d) Changes in other factors Since Ethiopia’s agriculture is predominantly rainfed, production is heavily dependent on timely and sufficient rainfall. During the period covered in this study, no major incidences of the large-scale droughts that plagued Ethiopia in the past occurred, and rainfall levels were relatively stable (Bachewe, Berhane, Minten, & Taffesse. A.S., 2015). Moreover, Ethiopia is equipped with a good early warning system. When droughts arise, GoE and donors address production shortfalls sufficiently, as manifested by interventions during the Horn of Africa drought in 2012 (Maxwell et al., 2014) and the major drought in 2015/16. Moreover, other factors contributed to improved agricultural productivity, albeit to a limited extent compared to the drivers mentioned above. We discuss these other factors below. First, lack of access to credit is often seen as one of the major constraints to increasing agricultural productivity and rural transformation. Following the international microfinance revolution in the 1980s and 1990s, Ethiopia witnessed a remarkable progress in this sector, starting in the early 2000s. The number of microfinance institutions (MFIs) in the country grew from 1 in 1994 to 32 in 2004 and Ethiopia became home to 2 of the 50 largest MFIs in the world, measured in terms of gross loan portfolio size (Forbes., 2007). The MFI industry expanded enormously in the last decade, both in terms of number of active borrowers and in the outstanding loan portfolio, which rose from a little less than 2 billion birr in 2003 to around 11 billion birr in 2014 (in 2011 prices). In addition, significant amounts of government program loans were disbursed through Rural Saving and Credit Cooperatives in the last decade.20 Second, the risk situation might have changed in favor of higher agricultural productivity over the last decade. Dercon and

20 However, while there has been significant growth in this area, it is unclear how much of the micro-finance has been used toward the agricultural sector. For example, CSA data show that the share of farmers that used credit for agricultural purposes has not changed significantly over time – it varied between 22% and 28% over the 2005– 13 period. It therefore seems that credit might have been more readily available in rural areas, but it might not have directly impacted agricultural activities.

296

F.N. Bachewe et al. / World Development 105 (2018) 286–298

Christiaensen (2011) show that uninsured risk is a significant constraint to technology adoption in Ethiopia. Although insurance markets may be only slightly different in 2014 than in 2004, it is plausible that the value of risk (relative to income) of technology adoption, or its perception, changed due to a number of factors: (i) the doubling of output-input price ratios is more likely to reduce losses on returns to chemical fertilizer use in the event of bad weather (Dercon & Hill, 2009); (ii) a well-functioning social protection program put in place in the last decade in most drought-prone areas means that poor households are likely to be incentivized to take more risks; (iii) a relatively long period of good weather may make the occurrence of bad weather shocks more muted (perhaps changing farmers’ expectations of the risk); and (iv) farmers are seemingly richer, have more assets, and are better able to selfinsure. Third, a large-scale land certification program was set up in the country in the 1990s to ensure more secure land property rights. This land certification program was one of the largest, cheapest, and fastest in Africa (Deininger, Ayalew, Holden, & Zevenbergen, 2008). While land still remains the property of the state, the certificates have been found to have increased security of property rights and encouraged investments, increased land rentals, and increased productivity and food security (Deininger, Ayalew, & Alemu, 2011; Ghebru & Holden, 2013). These more secure property rights may therefore have been enablers for the increase in agricultural productivity seen in the last decade. Fourth, in 2005 Ethiopia put in place a large safety net program – the Productive Safety Net Programme (PSNP) – that covers over 7 million vulnerable people in large parts of the country. An important feature of this program is the Public Works component, which generates employment on improving communal infrastructure, including rural roads, irrigation systems, and terraces. Taffesse, Yimer, and Dereje (2016) show to what extent investments in these communal assets contributed to higher agricultural productivity, echoing similar findings of increased adoption of modern inputs by households participating in the PSNP (Hoddinott, Berhane, Gilligan, Kumar, & Taffesse, 2012).

6. Conclusion and implications Agricultural productivity in Sub-Saharan Africa is significantly lower than in most other regions of the world. This has been linked with a number of structural issues such as land and water constraints, low human capital accumulation, deficient institutional and physical infrastructure, uninsured risk, and weak governance (Barrett, Christiaensen, Sheahan, & Shimeles, 2017). However, the beginning of the 21st century saw significant improvements in agricultural productivity in some parts of Sub-Saharan Africa. Despite this progress, it is still not well understood how some countries were able to overcome these structural constraints. Ethiopia is a case in point. This paper examines the incidence, sources, and proximate causes of agricultural growth in Ethiopia over the last decade using a number of datasets that are unavailable for most other SubSaharan African countries. The paper’s focus is to better understand the country’s agricultural growth process over the 2004–2014 period. In so doing, it provides up-to-date, wide-ranging, country-level evidence on recent agricultural growth and its correlates in Africa. In contrast, almost all other such explorations rely on crosscountry datasets and analyses. The paper also combines decomposition of output growth with a model of farmers’ modern input adoption decisions to identify and assess relevant correlations. We find that significant changes occurred in Ethiopia’s agriculture sector in the last decade. Agricultural output more than doubled, driven in part by area expansion, but more importantly by

significant yield increases. The real value of agricultural GDP increased by 7.6% per year. This agricultural growth is shown to be associated with significant poverty reduction (Hill & Tsehaye, 2014). The increased productivity is partly explained by rapid uptake of several improved agricultural technologies. However, part of this agricultural growth cannot be explained by increased adoption of modern inputs and other production factors. Significant growth in TFP, 2.3% per year on average, therefore also contributed to this growth. We further note that the adoption of modern technologies was rapid and its contribution to agricultural growth was especially higher in the second half of the last decade. In the first half, agricultural growth was relatively more driven by area expansion and TFP growth. Major drivers for the increasing adoption of modern inputs seem to be multiple, and linked with significantly higher expenditures in the agriculture sector. First, Ethiopia built up a large agricultural extension system in the last decade, with one of the highest extension agent-to-farmer ratios in the world. Second, access to markets significantly improved. While 67% of the population lived more than five hours from a city in 1997/98, this declined to 26% in 2010/11. Third, improved access to education led to a significant decrease in illiteracy in rural areas. Fourth, high international prices of export products as well as improved modern input–output ratios for local crops provided better incentives for the agriculture sector over the last decade. These factors all show a strong association with increasing adoption of improved technologies, and consequently agricultural productivity. However, other factors played a role as well, including good weather, better access to MFIs in rural areas, improved tenure security and a well-functioning safety net. The analyses from different data sources show a highly dynamic agricultural environment in Ethiopia in the last decade but several data-related caveats remain. While we infer major plausible pathways that might have led to productivity increases in the country, we lack the data to quantitatively measure the specific contributions of drivers to agricultural growth. Ethiopia is also an extremely diversified country with large variation in agro-ecologies. While we analyze the issue at the national level, more in-depth research is needed to understand specific issues at the disaggregated level. Moreover, we have no data on the cost (private as well as public) of achieving this growth. Finally, our analysis is mostly based on annual CSA data, and while we triangulate our findings with other large-scale data sources, alternative large-scale, nationally representative data sources with similar frequency are unavailable. The findings also point to several issues that require more investigation. First, while growth rates are found to be rather similar, notable differences in yield levels arise between the different data sources. Experiments to understand and explain the differences in yield data collection methods in the Ethiopian context are therefore called for (see Beegle, Carletto, & Himelein, 2012; Carletto, Gourlay, & Winters, 2013; Deininger et al., 2011; Deininger, Carletto, et al., 2011). Second, growth in cereal output over most of the decade studied is estimated to be larger than growth in cereal consumption (Worku, Dereje, Berhane, Minten, & Taffesse, 2015). Research is needed to understand changes in the food balance, wastage and food losses, feed demand (Mellor, 2014; Mellor & Kuma, 2014), seed demand, formal and informal export and imports, possible issues with comparability of statistics over time,21 and changes in food demand by an increasing population (including international refugees). Third, more detailed and updated analysis on land use changes in the country would be useful to better understand from where the additional agricultural land was derived and the implications of this change for future land use. 21 For a discussion on consumption numbers and possible issues with these, see Stifel and Woldehanna (2016) and Mandefro and Jerven (2015); for issues with production numbers, see Dercon and Hill (2009) and Gollin (2011).

F.N. Bachewe et al. / World Development 105 (2018) 286–298

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.worlddev.2017. 05.041. References Abate, T., Bekele, S., Abebe, M., Dagne, W., Yilma, K., Kindie, T., ... Tolera, K. (2015). Factors that transformed maize productivity in Ethiopia. Food Security, 7(5), 965–981. Bachewe, F., Berhane, G., Minten, B., & Taffesse, A. S. (2015). Agricultural growth in Ethiopia (2004-2014): Evidence and drivers (ESSP working paper 81. Addis Ababa). Bachewe, F., Berhane, G., Minten, B., & Taffesse, A. S. (2016). Non-farm income and labor markets in rural Ethiopia (ESSP working paper 90, Addis Ababa). Badiane, O., & Makombe, T. (2014). Beyond a Middle Income Africa: Transforming African Economies for Sustained Growth with Rising Employment and Incomes. ReSAKSS Annual Trends and Outlook Report. Badiane, O., Makombe, T., & Bahiigwa, G. (2014). Promoting agricultural trade to enhance resilience in Africa: ReSAKSS annual trends and outlook report 2013. Promoting agricultural trade to enhance resilience in Africa: ReSAKSS annual trends and outlook report 2013. Bahiigwa, Godfrey, Samuel Benin, Wondwosen Tefera, & Samson Jemaneh (2015). ‘‘Tracking Key CAADP Indicators and Implementation”, In Badiane, O. and Makombe, T. (Eds). 2015. Beyond a Middle Income Africa: Transforming African Economies for Sustained Growth with Rising Employment and Incomes. ReSAKSS Annual Trends and Outlook Report 2014. International Food Policy Research Institute (IFPRI). Barham, B. L., Foltz, J. D., Jackson-Smith, D., & Moon, S. (2004). The dynamics of agricultural biotechnology adoption: Lessons from series rBST use in Wisconsin, 1994–2001. American Journal of Agricultural Economics, 86(1), 61–72. Barrett, C. B., Christiaensen, L., Sheahan, M. B., & Shimeles, A. (2017). On the structural transformation of rural Africa. Policy Research working paper; no. WPS 7938. Washington, D.C.: World Bank Group. Beegle, K., Carletto, C., & Himelein, K. (2012). Reliability of recall in agricultural data. Journal of Development Economics, 98, 34–41. Benin, S. (2014). Implementation performance and progress towards core CAADP targets, presentation at the 2014 ReSAKSS annual conference, October 8-10. Addis Ababa. Benson, T., Spielman, D., & Kasa, L. (2014). Direct seed marketing program in Ethiopia in 2013: An operational evaluation to guide seed sector reform, IFPRI Discussion Paper 01350. Washington, DC. Berhane, G., Woldu, T. A., & Ragasa, C. (2017). Understanding Ethiopia’s Extension System. ESSP Discussion Paper (in press). Berhanu, K., & Poulton, C. (2014). The political economy of agricultural extension policy in Ethiopia: Economic growth and political control. Development Policy Review, 32(S2), S197–S213. Block, S. A. (2014). The post-independence decline and rise of crop productivity in sub-Saharan Africa: Measurement and explanations. Oxford Economic Papers, 66 (2), 373–396. Cappellari, L., & Jenkins, S. P. (2006). Calculation of multivariate normal probabilities by simulation, with applications to maximum simulated likelihood estimation (IZA DP No. 2112). Carlaw, K., & Lipsey, R. G. (2003). Productivity, technology, and economic growth: What is the relationship? Journal of Economic Surveys, 17(3), 457–495. Carletto, C., Gourlay, S., & Winters, P. (2013). From Guesstimates to GPStimates: Land area measurement and implications for agricultural analysis. Washington, DC: World Bank Policy Research Working Paper 6550. Central Statistical Agency of the Federal Democratic Republic of Ethiopia (CSA). 2005a–2016a. Agricultural Sample Survey: Volume I–Report on Area and Production of Crops (Private Peasant Holdings, Meher Season), Addis Ababa. Christiaensen, L., Demery, L., & Kuhl, J. (2011). The (evolving) role of agriculture in poverty reduction—An empirical perspective. Journal of Development Economics, 96(2), 239–254. CSA. 2005b–2014b. Agricultural Sample Survey: Volume III–Report on Farm Management Practices (Private Peasant Holdings, Meher Season), Addis Ababa. CSA. 2005c–2014c. Agricultural Sample Survey: Volume II–Report on Livestock and Livestock Characteristics (Private Peasant Holdings, Meher Season), Addis Ababa. Datt, G., & Ravallion, M. (1998). Why have some Indian states done better than others at reducing rural poverty? Economica, 65(257), 17–38. Davis, K., Swanson, B., Amudavi, D., Mekonnen, D. A., Flohrs, A., Riese, J., Lamb, C., & Zerfu, E. (2010). In-depth Assessment of the Public Agricultural Extension System of Ethiopia and Recommendations for Improvement. Washington, DC: IFPRI Discussion Paper 01041. Deininger, K., Ayalew, D., & Alemu, T. (2011). Impacts of Land Certification on Tenure Security, Investment, and Land Market Participation: Evidence from Ethiopia. Land Economics, 87(2), 312–334. Deininger, K., Ayalew, D., Holden, S., & Zevenbergen, J. (2008). Rural Land Certification in Ethiopia: Process, Initial Impact, and Implications for Other African Countries. World Development, 36(10), 1786–1812. Deininger, K., Carletto, C., Savastano, S., & Muwonge, J. (2011). Can diaries help improve agricultural production statistics? Evidence from Uganda. Washington, DC: The World Bank.

297

Dercon, S., & Hill, R. V. (2009). Growth from agriculture in Ethiopia: Identifying Key Constraints. Paper prepared as part of a study on Agriculture and growth in Ethiopia by the Department of International Development (DFID). Dercon, S., & Christiaensen, L. (2011). Consumption risk, technology adoption and poverty traps: Evidence from Ethiopia. Journal of Development Economics, 96(2), 159–173. Detriere, S. (2016). The inverse productivity size relationship: Can it be explained by systematic measurement error in self-reported production. Invited paper presented at the 5th International Conference of the African Association of Agricultural Economists, September 23–26, 2016, Addis Ababa, Ethiopia. Diao, X., Headey, D., & Johnson, M. (2008). Toward a green revolution in Africa: What would it achieve, and what would it require? Agricultural Economics, 39, 539–550. Dorfman, J. H. (1996). Modeling multiple adoption decisions in a joint framework. American Journal of Agricultural Economics, 78(3), 547–557. Engida, E., Tamru, S., Tsehaye, E., Debowicz, D., Dorosh, P., & Robinson, S. (2011). Ethiopia’s Growth and Transformation Plan: A Computable General Equilibrium Analysis of Alternative Financing Options (ESSP II Working Paper 30, Addis Ababa). Ethiopian Development Research Institute (EDRI) (2009). Ethiopia: Input Output table and Social Accounting Matrix (Addis Ababa). Ethiopian Humanitarian Country Team (EHCT). (2015). Ethiopia: Government and humanitarian partners scale up to meet additional immediate relief needs of El Niño-driven crisis, Press release, United Nations, October 13th, 2015. Evenson, R. E., & Gollin, D. (2003). Assessing the impact of the Green Revolution. Science, 300, 758–762. Feder, G. (1980). Farm size, risk aversion and the adoption of new technology under uncertainty. Oxford Economic Papers, 32(1980), 263–284. Feder, G. (1982). Adoption of interrelated agricultural innovations: Complementarity and the impacts of risk, scale, and credit. American Journal of Agricultural. Economics, 64(1), 94–101. Feder, G. (1985). The relation between farm size and farm productivity: The role of family labor, supervision and credit constraints. Journal of Development Economics, 18(2), 297–313. Food and Agricultural Organization of the United Nations (FAO) (2012). Current world fertilizer trends and outlook to 2016. Rome: FAO. Forbes. (2007). The 50 Top Microfinance Institutions. Accessed in June 2015, available at http://www.forbes.com/2007/12/20/microfinance-philanthropycredit-biz-cz_ms_1220microfinance_table.html. Ghebru, H., & Holden, S. (2013). Links between tenure security and food security: Evidence from Ethiopia. Washington, DC: IFPRI discussion paper 1288. Gollin, D. (2011). Crop production in Ethiopia: Assessing the evidence for green revolution. Technical report. International Growth Centre. Hagos, F., Makombe, G., Namara, R. E., & Awulachew, S. B. (2009). Importance of irrigated agriculture to the Ethiopian economy: Capturing the direct net benefits of irrigation. Colombo: IWMI Research Report 128. Headey, D., Dereje, M., & Taffesse, A. S. (2014). Land constraints and agricultural intensification in Ethiopia: A village-level analysis of high-potential areas. Food Policy, 48, 129–141. Hiebert, D. (1974). Risk, Learning, and the Adoption of Fertilizer Responsive Seed Varieties. American Journal of Agricultural Economics, 56, 764–768. Hill, R. V., & Tsehaye, E. (2014). Growth, Safety Nets and Poverty: Assessing Progress in Ethiopia from 1996 to 2011. Background paper for the Ethiopia Poverty Assessment. Washington, DC: World Bank. Hoddinott, J., Berhane, G., Gilligan, D. O., Kumar, N., & Taffesse, A. S. (2012). The impact of Ethiopia’s productive safety net programme and related transfers on agricultural productivity. Journal of African Economies, 21(5), 761–786. Jack, B. K. (2011). Constraints on the adoption of agricultural technologies in developing countries. UC Berkeley: White paper, Agricultural Technology Adoption Initiative, J-PAL (MIT) and CEGA. Jayne, T., Chamberlin, J., & Headey, D. (2014). Land pressures, the evolution of farming systems, and development strategies in Africa: A synthesis. Food Policy, 48, 1–17. Jayne, T., Anriquez, G., & Collier, E. (2013). African agriculture toward 2030: Changes in urbanization and agricultural land dynamics and their implications for CGIAR research. In Independent Science and Partnership Council (ISPC) Foresight meeting. Jayne, T., & Rashid, S. (2013). Input subsidy programs in sub-Saharan Africa: A synthesis of recent evidence. Agricultural Economics, 44(6), 547–562. Jorgenson, D. W., & Griliches, Z. (1967). The explanation of productivity change. The Review of Economic Studies, 34(3), 249–283. Just, R. E., & Zilberman, D. (1983). Stochastic structure, farm size and technology adoption in developing agriculture. Oxford Economic Papers, 35(2), 307–328. Kedir, M., Schmidt, E., & Tilahun, H. (2015). Production Patterns and Fertilizer Use. Mimeo. Kislev, Y., & Shchori-Bachrach, N. (1973). The process of an innovation cycle. American Journal of Agricultural Economics, 55, 28–37. Krishnan, P., & Patnam, M. (2014). Neighbors and extension agents in Ethiopia: Who matters more for technology adoption? American Journal of Agricultural Economics, 96(1), 308–327. Li, J., Gorham, R., & Nuru, M. (2011). Road Development, Agglomeration and Agricultural Growth in Ethiopia: An Empirical Analysis. Manuscript. Mandefro, H., & Jerven, M. (2015). GDP growth and poverty decline in Ethiopia: Who owns the numbers? Addis Standard, February 2015: 12–14. Maxwell, D., Majid, N., Stobaugh, H., Kim, J. J., Lauer, J., & Paul, E. (2014). Lessons learned from the Somalia famine and the greater Horn of Africa 2011–2012: Desk Review of Literature. Medford: Feinstein International Center, Tufts University.

298

F.N. Bachewe et al. / World Development 105 (2018) 286–298

Mellor, J. (2014). Rapid cereals production growth in Ethiopia: How rapid, how achieved, can it continue, what impact?. In D. Chanyalew, J. Mellor, B. Minten, & T. Kuma (Eds.), Cereals price policy and analysis for Ethiopia in the context of rapid production growth, proceedings of conference held in Addis Ababa, June 24; (pp. 31–48). Mellor, J., & Kuma, T. (2014). Rapid growth in cereals production temporarily generates much of its own demand-modelling for Ethiopia, In D. Chanyalew, Mellor, B. Minten, & T. Kuma (Eds), Cereals price policy and analysis for Ethiopia in the context of rapid production growth, proceedings of conference held in Addis Ababa, June 24, (pp. 49–67). Ministry of Agriculture of the Federal Democratic Republic of Ethiopia (MoA) (2013). GTP Annual Performance Report, EFY 2005 (Addis Ababa). Minten, B., Koru, B., & Stifel, D. (2013). The last mile(s) in modern input distribution: Pricing, profitability, and adoption. Agricultural Economics, 44, 1–18. Minten, B., Tamru, S., Engida, E., & Tadesse, K. (2016). Transforming staple food value chains in Africa: The case of teff in Ethiopia. Journal of Development Studies, 52(5), 627–645. Minten, B., Tamru, S., & Stifel, D. (2014). Structural transformation in cereal markets in Ethiopia. Journal of Development Studies, 50(5), 611–629. Morris, M. L. (2007). Fertilizer use in African agriculture: Lessons learned and good practice guidelines. World Bank Publications. Mulat, Y. S. (2011). The status of irrigated agriculture in Ethiopia. Research report submitted to the Ethiopian Development Research Institute. National Bank of Ethiopia (NBE) (2014). Annual report 2013-2014 (Addis Ababa). New Partnership for Africa’s Development (NEPAD) (2013). Transforming Africa. Annual Report 2013. Accessed November 2016, available at http://www.nepad. org/resource/nepad-agency-2013-annual-report-0. Nin-Pratt, A. (2015). Inputs, productivity, and agricultural growth in Africa South of the Sahara. Washington, DC: IFPRI Discussion Paper 01432. Nin-Pratt, A., Johnson, M., Magalhaes, E., You, L., Diao, X., & Chamberlin, J. (2011). Yield gaps and potential agricultural growth in West and Central Africa, (Vol. 170). IFPRI, Washington, DC. Otsuka, K., & Kijima, Y. (2010). Technology policies for a green revolution and agricultural transformation in Africa. Journal of African Economies, 19(Suppl. 2), ii60–ii76. Pingali, P. L. (2012). Green Revolution: Impacts, limits, and the path ahead. Proceedings of the National Academy of Sciences, 109(31), 12302–12308. Ragasa, C., Berhane, G., Tadesse, F., & Taffesse, A. S. (2013). Gender differences in access to extension services and agricultural productivity. The Journal of Agricultural Education and Extension, 19(5), 437–468. Rashid, S., Tefera, N., Minot, N., & Ayele, G. (2013). Fertilizer in Ethiopia: An assessment of policies, value chain, and profitability. IFPRI Discussion Paper No. 01304, Washington, DC. Reardon, T., Tschirley, D., Minten, B., Haggblade, S., Liverpool-Tasie, S., Dolislager, M., Snyder, J., & Ijumba, C. (2015). Transformation of African Agrifood Systems in the New Era of Rapid Urbanization and the Emergence of a Middle Class. In Badiane, O. and T. Makombe. 2015. Beyond a Middle Income Africa: Transforming African Economies for Sustained Growth with Rising Employment and Incomes. ReSAKSS Annual Trends and Outlook Report 2014. International Food Policy Research Institute, Washington, DC. Romer, P. (1990). Are nonconvexities important for understanding growth? (No. w3271). National Bureau of Economic Research.

Schultz, T. W. (1951). A framework for land economics. the long view. Journal of Farm Economics, 33(2), 204–215. Solow, R. M. (1957). Technical change and the aggregate production function. Review of Economics and Statistics, 39, 312–320. Spielman, D. J., Byerlee, D., Alemu, D., & Kelemework, D. (2010). Policies to promote cereal intensification in Ethiopia: The search for appropriate public and private roles. Food Policy, 35(3), 185–194. Spielman, D., Kelemwork, D., & Alemu, D. (2012). Seed, fertilizer, and agricultural extension in Ethiopia. In P. Dorosh & S. Rashid (Eds.), Food and agriculture in Ethiopia: Progress and policy challenges (pp. 132–183). Philadelphia: IFPRI/ Pennsylvania University Press. Stifel, D., & Minten, B. (2017). Market Access, Welfare, and Nutrition: Evidence from Ethiopia. World Development, 90, 229–241. Stifel, D., & Woldehanna, T. (2016). Poverty in Ethiopia, 2000–11: Welfare Improvements in a Changing Economic Landscape. In C. Arndt, A. McKay, & F. Tarp (Eds.), Growth and Poverty in Sub-Saharan Africa. United Kingdom: Oxford University Press. Sumner, D. A. (2014). American farms keep growing: Size, productivity and policy. Journal of Economic Perspectives, 28(1), 147–166. Taffesse, A. S., Yimer, F., & Dereje, M. (2016). The impact of the PSNP-created community assets on crop yield growth (ESSP working paper, Addis Ababa, (in press)). Tamru, S., Minten, B., Bachewe, F., & Alemu, D. (2017). The rapid expansion of herbicide use in smallholder agriculture in Ethiopia: Patterns, drivers, and implications. European Journal of Development Research (in press). Timmer, C. P. (1988). The agricultural transformation. Handbook of development economics, 1. –331. Vandercasteelen, J., Tamru, S., Minten, B., & Swinnen, J. (2016). ities and agricultural transformation in Africa: The case of Ethiopia (ESSP Working Paper 91). von Braun, J., & Olofinbiyi, T. (2007). Famine and food security in Ethiopia. Case study #7–4 of the program: ‘‘Food policy for developing countries: The role of government in the global food system” 2007, edited by Pinstrup-Andersen, P., and F. Cheng. Ithaca, NY: Cornell University. von Braun, J., Teklu, T., & Webb, P. (1998). Famine in Africa. Causes, Responses and Prevention. Washington, DC: International Food Policy Research Institute. Worku, I., Dereje, M., Berhane, G., Minten, B., & Taffesse, A. S. (2015). Teff and its role in the agriculture and food economy. Washington, DC: International Food Policy Research Institute, mimeo. World Bank. (2014a). Ethiopia: Poverty Assessment. Report No. AUS6744. Poverty Global Practice. Africa Region. World Bank Group. Washington, DC. World Bank (2014b). 3rd Ethiopian economic update: Strengthening export performance through improved competitiveness. Washington, DC: World Bank. Yu, B., & Nin-Pratt, A. (2011). Agricultural productivity and policies in Sub-Saharan Africa. IFPRI Discussion Paper No. 01150. Zeng, D., Alwang, J., Norton, G. W., Shiferaw, B., & Jaleta, M. (2013). Ex-post impacts of improved maize varieties on poverty in rural Ethiopia: Diffusion and impact of improved varieties in Africa (DIIVA). Rome: CGIAR Standing Panel on Impact Assessment (SPIA). Zhang, X., & Fan, S. (2004). How productive is infrastructure? A new approach and evidence from rural India. American Journal of Agricultural Economics, 86(2), 492–501.