WPS7342 Policy Research Working Paper 7342 Using National Statistics to Increase Transparency of Large Land Acquisition Evidence from Ethiopia Daniel Ali Klaus Deininger Anthony Harris Development Research Group Agriculture and Rural Development Team June 2015 Policy Research Working Paper 7342 Abstract The 2007/08 commodity price boom triggered a ‘rush’ for address this by (i) allowing assessment of the productivity land in developing countries. Yet, many affected countries of land use and taking measures to increase it; (ii) compar- lacked the regulatory infrastructure to cope with such ing productivity between large and small farms to identify demand and reliable data on investors’ performance. This spillovers and ways to improve these; and (iii) setting in study uses the example of Ethiopia to show how simple motion a process of continuing improvement. Implica- improvements in administrative data collection can help to tions for global investment in this area are drawn out. This paper is a product of the Agriculture and Rural Development Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at dali1@worldbank.org, kdeininger@worldbank.org, and aharris3@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Using National Statistics to Increase Transparency of Large Land Acquisition: Evidence from Ethiopia¶ Daniel Ali, Klaus Deininger, Anthony Harris World Bank, Washington DC dali1@worldbank.org, kdeininger@worldbank.org, aharris3@worldbank.org JEL Classification: O13, Q12, Q15 Keywords: large scale land acquisition, commercial farms, productivity, farm size distribution, Africa, Ethiopia ¶ The research leading to this paper would not have been possible without the initiative by Hashim Ahmed, Office of the Prime Minister and Head Economic and Policy Analysis Unit, Government of Ethiopia, the active collaboration by the Ethiopian Central Statistical Agency, especially Biratu Yigezu, Habekristos Beyene, and Ahmed Ebrahim, and support by Alemayehu Ambel, Andrew Goodland, Firew Bekele, Asmelash Haile, Lars Moeller, and Lou Scura. Insightful comments from Gero Carletto, Luc Christiaensen, Issa Faye, Thea Hilhorst, Bart Minten, James Opio-Omoding and seminar participants at the African Development Bank and the 2015 Annual Bank Conference on Land and Poverty greatly helped to improve the quality of the paper. Funding from DFID and the German Government is gratefully acknowledged. 1. Introduction Nearly a decade after the 2007/08 commodity-price boom triggered a global ‘rush’ for agricultural land by investors, a number of stylized facts regarding this phenomenon seem agreed. First, weak or fragmented institutions compromised countries’ ability to channel such demand towards areas where it would yield the highest returns or to reject non-viable proposals from inexperienced investors. As a result, the benefits have been less than expected and a sizeable share of investors either went out of business or failed to fully utilize all the land allocated to them. Second, although land demand retreated from 2008 levels, it is expected to continue, albeit at levels much lower than what had been observed at the height of the land rush. In fact, if guided by a coherent and enforceable policy, responsible agricultural investment could provide countries whose economy depends on agriculture with opportunities to add value and generate local benefits. To realize these benefits, policy decisions are required on a number of issues, including (i) where in the value chain—up-stream in agro-processing, in mixed nucleus estate models with outgrowers, or fully own production—investment would be most desirable and what complementary public inputs may encourage such investment; (ii) how well land that has been transferred to investors is utilized and if not what remedial action (e.g., canceling licenses) may be needed or appropriate; and (iii) how performance compares between local producers and outside investors and what this implies for regulating the sector. Policy decisions on these as well as private investment will require regular access to up-to-date information. Yet, few countries developed the systems to regularly provide the data needed and even much of the empirical literature is still based on case studies the representativeness of which is difficult to establish, often focusing on process rather than quantification of outcomes. Lack of reliable and regular information will make it difficult to manage investment- or country- risks, enforce laws, and document compliance with global standards to attract capable and responsible (institutional) investors. To explore ways of satisfying such information needs in a sustainable and effective way, we draw on a nationally representative large farm survey in Ethiopia conducted between 2010/11 to 2013/14. Ethiopia is an interesting case as it has a long tradition of collecting systematic data on large (state) farm performance which, due to poor data quality, were often not reported or even stopped intermittently. Data collected with an improved survey instrument allow to draw interesting conclusions on commercial farming in the country. First, even at the peak of the ‘land rush’, the amount of land transferred to investors in commercial farms that are at least partly operational, the vast majority of which was to Ethiopians rather than foreigners, was much less than claimed in some widely quoted reports (Land Matrix 2015; Oakland Institute 2011; Rulli et al. 2013). Since 2007/08, the level of annual land transfers reverted more or less to what was observed before 2007. Second, about 55% of the land transferred remains unutilized with main constraints to expansion related to technology and labor. Third, with one permanent job per 20 ha plus some temporary 2 jobs, commercial farms fail to generate much employment. Finally, for most crops, commercial farms’ yields are roughly double those by smallholders though the highest yields are normally obtained by those in the 10-20 ha category which often also managed to expand area cultivated quite rapidly. While lack of data on prices and inputs makes it impossible to assess how much of this difference is due to higher efficiency rather than just more intensive use of inputs, in particular improved seeds, or locations, commercial farms could create spillovers for small farms through a number of channels that would merit further study. Beyond providing analytical insights that are of interest in their own right, having reliable data on outcomes by commercial farms can generate feedback loops for policy formulation. In our case, these include a continued impetus to improve the survey instruments used, efforts to better link to existing data, such as investment licenses and survey plans, that are issued and maintained by other agencies, and exploration of ways, such as yield monitoring based on remotely-sensed data, that could help to complement, cross-check, and eventually even partially replace survey information. The paper is structured as follows. Section two provides background information on the ‘global land rush’ and discusses the Ethiopian context with efforts to improve the quality of data collected in the sector. Section three and four use the case of Ethiopia to illustrate how a relatively modest investment can significantly improve availability of information on key parameters in the debate, including (i) the share of national vs. foreign investors; (ii) the amount of large farm land that is not utilized; (iii) key constraints used by large farmers; (iv) input use and yields for main crops in different farm size groups; and (v) comparing yields between large and small farmers. Finally, in addition to describing outcomes, we highlight potential alternatives for situations where no regular farm surveys are conducted and draw out how these data can inform the research agenda. 2. Background and literature Although the volume of new land deals has declined, large-scale land acquisition will not disappear quickly. To regulate it in a way that allows to attract capable investors and to address the legacy of failed land acquisition, good information on the overall level of land transfers and the efficiency of land use is indispensable. Yet, while a review of the literature points towards a number of stylized facts regarding such investment, it is often based on one-off case studies the representativeness of which is not clear. The case of Ethiopia provides some lessons in the area. 2.1 The global land rush (nearly) a decade on After decades of stagnant or declining commodity prices when agriculture was considered a “sunset industry”, increases in the level and volatility of commodity prices and a concomitant rise in global demand 3 for land have led to what many observers view as a “land grab” (Pearce 2012) that risks depriving local farmers and communities of resources rather than providing them with opportunities for development. The size of land demand took many host country governments by surprise. Without the requisite institutional infrastructure to check investors’ business plans, provide information or assistance with negotiation to communities where land demand had surfaced, and document and verify compliance with contractual terms and the impact of large land acquisitions, many of the relevant cases did indeed fail to increase productivity of land use or generate local benefits (Deininger and Byerlee 2011).1 There is indeed evidence for investment having failed to not only live up to high and possibly exaggerated expectations by its proponents, something that should not have come as a surprise given the historically high failure rate of such investments (Tyler and Dixie 2013). At the same time, it is argued that in many developing countries, there is a need to go beyond the dichotomy of large vs. small and look instead at new ways of combining the two (Collier and Dercon 2014). There is now indeed a wealth of guidelines to support responsible investment throughout the agricultural supply chain (OECD 2014) and it is often argued that the most effective arrangements may be those aiming at exploiting synergies between upstream investments in agro-processing industries or through nucleus estate or outgrower arrangements. This leads to two challenges. First, there is a need to distinguish between ‘pioneers’ increasing productivity and generating local benefits and ‘speculators’, including possibly urban elites (Sitko and Jayne 2014), who acquire large tracts of land in the hope of benefiting from future land price appreciation (Collier and Venables 2012). Pro-active steps to protect existing rights, strengthen negotiating and monitoring capacity, and establish the regulatory framework and transparency in terms of access to information and accountability are needed to attract investors who are financially and technically capable as these are not just going to show up at a country’s doorstep. Second, a strategy on discovering and dealing with failed enterprises, including ways for failed ventures to exit the sector in ways that do not imperil local people, ia important. This is particularly relevant in light of the fact that in many of the cases of large-scale investment, land markets do not exist or may be legally prohibited. History demonstrates that large land owners who are unable to make effective use of land may use political channels to affect factor prices, e.g., by trying to keep down labor cost or constrain access to capital, with unfavorable long-term consequences.2 1 The land available for expansion in Africa, most of it is concentrated in few countries (Deininger and Byerlee 2012), with poor access to infrastructure and low levels of profitability (Chamberlin et al. 2014), and often also very weak governance (Arezki et al. 2013). 2 The importance of this issue is demonstrated by the many historical examples where accumulation of large tracts of land by large but relatively inefficient farms led to rent-seeking behavior and, using their locally dominant position, to monopolize input or output markets (Binswanger et al. 1995), subvert provision of public goods such as education (Nugent and Robinson 2010; Vollrath 2009), undermine financial sector development (Rajan and Ramcharan 2011), or restrict political participation (Baland and Robinson 2008). 4 Although the uptick in land-related investment prompted efforts at systematic data collection (Anseeuw et al. 2012), available global databases suffer from considerable weaknesses (Arezki et al. 2015) and fail to reflect the complexities of land acquisitions (Hall 2011) or representativeness (Cotula 2014). Still, a large amount of the current knowledge comes from case studies, which can provide qualitatively valuable insights but, without information on the universe from which cases were selected, will still have to be accompanied by ways to test their representativeness and replicability (Schoneveld 2014). Moreover, many of the case studies focus on process rather than robust evidence on outcomes. From an analytical point of view, systematic data at country level are needed to explore the extent to which these investments target easily accessibly areas or compete with smallholder agriculture (Messerli et al. 2014), investors live up to their commitments (Burnod et al. 2013), and to identify indirect effects of agricultural investment (Andrade de Sa et al. 2013). In terms of policy, lack of good and regular data makes it more difficult to attract capable investors by (i) increasing the transaction costs of identifying good opportunities; and (ii) making it more difficult to access institutional sources of funding which require minimum levels of reporting and are increasingly bound by a trend towards traceability and ‘responsible’ investment. It may also reduce security of tenure for investors as, without a basis to document claims, they may be subject to frequent and disruptive renegotiation. 2.2 Ethiopia’s context Large farm investment is not new to Ethiopia; in fact in the pre-revolutionary period, the state used large subsidies to attract commercial investment in agriculture. This led to establishment of highly mechanized cash crop production in so-called ‘model farms’ the establishment of which was often associated with tenant evictions.3 After the 1974 revolution, the Derg converted most of these into state farms for food production. Though their yields were above those by peasants, their efficiency and contribution to national agricultural output (2%) remained low (Abebe 1990). After 1990, a strategy of market liberalization and agriculture-led industrialization focusing on small-scale producers was adopted.4 To assess its success, Ethiopia’s Central Statistical Agency (CSA) started collecting of annual data on smallholders’ productive performance using a survey administered by resident enumerators to a representative sample of some 40,000 farmers nationally. It complements information on inputs application with a crop cutting exercise for a minimum number of plots in each enumeration area. The transition to a free market economy was followed by higher private investment in commercial farms whose share of the country’s gross agricultural output increased to about 7%. In fact, reference to potential 3 Main areas included Awassa, Arbaminch, Zeway and Shewarobit. 4 Small producers are defined as those with a size below 10 hectares and all farms above this size fall into the ‘commercial’ category. 5 benefits in terms of food security, job creation, technology transfer, capital accumulation, export promotion, and environmental sustainability led the government’s latest 5-year Growth and Transformation Program (GTP) to put considerable focus on investment in commercial farms. To assess if such benefits materialize, reliable information on large farms’ productivity is essential. A ‘State Farms Survey’ completely enumerate the universe of commercial farms had been part of the country’s annual agricultural sample survey from the very beginning. Yet, the underlying data were based on farms’ administrative records rather than actual enumeration with often little relationship to field realities.5 To reduce such reliance on secondary sources, CSA included a complete enumeration of commercial farms, using a sample frame assembled from information provided by zonal investment offices,6 in its 2001/02 agricultural sample enumeration. However, high non-response and weak supervision reduced data quality. Although large and medium commercial farm surveys (LMSCFS) were conducted on a regular basis in subsequent years, results were not published or underlying data disseminated. Eventually, CSA’s program of large- and medium-scale commercial farm surveys was suspended in 2005/06, just before investor interest in large-scale land-based investment peaked. In 2007/08, such surveys were thus resumed, with a number of modifications, namely (i) use of a smaller number of more highly qualified field staff to gather information; (ii) a reduction of the sample size by only covering the universe of farms above a certain cut-off point (50 ha) and using a sample for commercial farms below this size; (iii) streamlining the survey instrument to focus on essential information first with the option of expanding as experience accumulates; (iv) collaboration with other agencies, in particular the Investment Agency, to ensure a sample frame; and (iv) area measurement by GPS to get a point of reference that makes it more difficult for respondents to intentionally provide mis-information. Still, when policy makers required information on the contribution of different types of commercial farm investment to the national economy and regulatory measure to maximize such impact, survey data proved insufficient to answer any questions beyond simple output estimates for national production statistics. To remedy this, a technical cooperation with the World Bank’s research department was initiated starting with the 2013/14 survey focusing on two aspects. First, the questionnaire was expanded to include time and amount of original land receipt, basic data on labor use, investment, and markets, and perceived obstacles to further expansion. Second, efforts were made to link data over time and use the frame to obtain an 5 Information obtained included area and production of temporary and permanent crops; land use, i.e., total cropland area, fallow land, grazing land; agricultural practices, such as use of various inputs, quantity and cost of inputs used; number of livestock and poultry by type, purpose, age and sex; number of farm machineries used by type and market value of each; number of permanent and temporary employees by type of profession. 6 In principle, the most immediate candidate for a sample frame is the list of potential investors who either applied for or obtained an investment license. This would include those who proceeded to invest and those who did not, thus allowing to obtain information on obstacles encountered throughout the process. Unfortunately, although a list of more than 10,000 investors was available, it was not a viable strategy due to limited documentation gathered in the process and the fact that most of the phone numbers provided could not be reached. 6 estimate of the amount of land transferred to commercial farms that is entirely unused and to link different survey rounds over time to measure changes in performance over time by any given farm (Beyene 2015). The positive impact of these changes, as described below, led CSA to further modify the instrument for the 2014/15 season to collect data on inputs and outputs at the plot rather than the farm level. For 2015/16, it is planned to (i) have plot boundaries recorded to allow yields to be cross-checked against independent estimates obtained from satellite imagery; and (ii) link the survey frame more clearly to data from other sources, especially the Agricultural Investment Office’s list of investment licenses to determine what types of projects were abandoned and possibly probe into reasons why this happened. 3. Descriptive evidence Survey data on operational farms allow us to characterize overall land transfers as well as performance of commercial farms in a way that can dispose of some common myths. First, with some 160 ha per farm, mean area transferred to large farms is below what has been reported in other databases (Land Matrix 2015) and studies (Oakland Institute 2011; Rulli et al. 2013; Rahmato 2011). With land disputes, technology, and manpower given as main reasons for 43% of commercial farms’ land not being fully utilized, tighter screening of applications in the pre-investment phase or monitoring once ventures started may improve outcomes. Second, positive spillovers may be limited by the fact that most investment is focused on on- farm infrastructure, less than 5% of large farms have outstanding credit, and the amount of lease fees paid seems modest. Third, with one permanent employee per 20 ha, job creation is below the amount of labor that could be productively absorbed by smallholder agriculture, possibly through investment upstream in the value chain. 3.1 Scale and use of lands transferred Starting from 1991, some 1.33 million ha had been allocated to 6,600 commercial farms with the rate of transfer having peaked in 2008 but since then fallen to pre-2007 levels.7 As large farms’ plots were geo- referenced and the start date for each farm in the sample is known, we can map the evolution of commercial farms over time.8 Results show an initial cluster of sesame and sorghum farms comprising transformed state farms in the North-West, followed by expansion of commercial farming for maize and wheat in the center, and eventually a movement towards the border regions of Benishangul-Gumuz and Gambella (fig. 1-4).9 7 Note that the sampling strategy implies that, subject to non-response, the survey provides a reliable measure of the area transferred to farms that were fully or partly operational (CSA, 2011). A simple check of land transferred by the Investment agency against those that have been identified by CSA as operational on the ground could provide information on land area under non-operational farms. In terms of non-response, information on 300 farms with size 50 and above, who did not respond to the detailed survey, is not included. The total farm size of these farms (excluding Karuturi Global Limited with 100,000 ha), based on administrative information, is about 117,000 ha. 8 Although GPS was used to determine plot area, the waypoints generated in this way were not saved and instead only one point was recorded per plot. While this allows us to locate farms, it makes it impossible to obtain precise plot boundaries. 9 In interpreting these figures, it is important to note that there will be survivor bias, i.e., farms that have gone out of business are no longer included. 7 Table 1 provides information on total land area allocated to commercial farms in different regions by time of establishment, owner’s nationality, main crop, and the main negotiating partner (direct negotiation with locals, woreda, region, or federal government). While commercial farms acquired sizeable amounts of land over time, neither total nor average amounts of land transferred are close to figures reported in some of the literature (Land Matrix 2015; Oakland Institute 2011; Rulli et al. 2013).10 A total of 1.33 million hectares was transferred to 6,612 commercial farms operational in 2013/14, i.e., some 200 hectares per farm, compared to 9.6 million hectares cultivated by smallholder farms. Average farm size varies across regions: it is highest in Gambella (462 ha per farm)11 followed by Benishangul and Oromia (293 and 290 ha per farm, respectively), and that of SNNP and Afar regions (163 and 140 ha, respectively). Average farm size in Tigray and Amhara regions is roughly a fourth and a fifth of what is observed in Gambella, and the lowest in Somali region with just 16 ha per farm. Of the 6,612 commercial farms, most (28%) are in of Amhara, followed by SNNP (25%), Tigray (15%), Benishangul (12%), Oromia (11%), and Afar, Gambella, and Somali (together 9%). The highest share of land area initially transferred to investors was in SNNP (25%), followed by Benishangul (21%), Oromia (18%), Amhara (15%) and Tigray (10%). Meanwhile, the contribution of Gambella (excluding land transferred to Karuturi Global Limited) and Afar was, respectively, about 7% and 4% of the total land transferred, and that of Somali region was a mere 0.01%. About 87% of the 6,344 farms that provided information on the size of the initial land allocation on average received area of land between 20 and 500 ha and only 6.4% received above 500 ha. Similarly, of the farms providing information on the owner’s nationality, 97% are purely Ethiopian owned and only 134 or 36 are, respectively, held purely by or jointly with foreigners. In line with the policy change since 1991, most of Ethiopia’s commercial farms are owned and operated by private owners (93% of the farms holding about 78% of initially land transfers) while some 2% of farms with 15% of initially awarded land are owned by government and the remainder are owned and operated by producer cooperatives. Government owned farms are primarily located in Oromia where they account for 93% of initial land transfers, with 83% of farms having been established before 1991. A total of 302 farms were established before 1991, mainly in Oromia and SNNP, with a mean initial land allocation of 684 ha per farm. Subsequently, the number of farms established per year reached 315 in 2002- 06, peaked at 793 during the 2007/08 commodity price boom, and fell to 546 and 371, respectively, in the 2009/10 and 2011/13 period. In other words new commercial farm establishment in 2013 had dropped to the level attained before the 2007 food price boom. Mean size of initial land transfer was, with some 80 ha, smallest in 1991/2, gradually increasing to 223 ha during the 2011-13 period. 10 “Our research shows that approximately 3,619,509 ha of land have been awarded, as of January 2011” (Oakland Institute 2011: p. 18). 11 Note that Karuturi Global Limited, a company that acquired 100,000 ha of farmland in Gambella in 2008 and that is now deemed to be bankrupt, is not included in the computation of total area and average farm size. 8 Regionally, the location of newly established farms shifted from the central highlands (Tigray, Oromia, Amhara, and to some extent SNNP) to the periphery (Benishangul and, to a lesser extent, Afar, Gambella, and Somali regions). Across regions, the share of farms and area allocated are highest for sesame (38% of farms and 27% of area, with an average farm size of 142 ha primarily located in Amhara, Tigray, and Benishangul regions), followed by coffee (15% of farms and 12% of area, with a farm size of 167 ha mainly in SNNP and Oromia), maize (13% of farms and 12% of area, with a farm size of 183 ha mainly in SNNP and Oromia), sorghum (13% of farms and 7% of area with a mean farm size of 116 ha in Tigray, Amhara, and Benishangul), cotton (6% of farms and 9% of area, with a mean farm size of 324 ha focused on Afar region), wheat (4% of farms and 9% of area, with a mean farm size of 498 ha mainly in Oromia and SNNP regions), and all other crops (12% of farms and 23% of the area, with a mean farm size of 389 ha). Ethiopian law, similar to that in many other countries, allows land transfers below a certain size to be processed locally, with cascading thresholds, which were adjusted over time, requiring approval at woreda, regional, or national level. A total of 380 farms with an average size of 113 ha, mainly in Afar and SNNP, had been acquired through direct negotiation with farmers with little effective government. With 32% of area transferred, the majority of land acquisitions were negotiated at woreda level (3,499 or 56% of farms with a mean size of 115 ha) or the region (34% of farms with a mean size of 241 ha accounting for 41% of area). The federal government accounted for 252 land transfers (4% of the total) although larger land sizes (average farm size of 1,155 ha) imply that these accounted for 23% of total area transferred. 3.2 Structural characteristics Farms in our sample expanded from 1.33 million ha to 1.77 million ha, though only some 57% of this area (1.01 million ha or 153 ha per farm) is actually utilized (table 2). Levels of land utilization are highest for farms established before 1991 (72%) and those cultivating wheat (72%) and cotton (62%). Contrary to anecdotal evidence of larger farms utilizing less of the land transferred to them, we find that farms with a size above 1,000 ha use 64% of their land, compared to below 55% for the farm size group below 100 ha. Discrepancies between the size of land originally allocated and what is currently cultivated point towards dynamics that would be worth exploring in their own right. Farms with initial allocations of less than 500 ha expanded significantly so that they now hold areas that are, on average, much larger than what had been originally transferred to them. By contrast farms in the above 500 ha group cultivate much less than what they received initially. As Ethiopia does not allow land sales, it would be of great interest to find out how expanding farms received their land from and what happened to land transferred but no longer used. The average size of utilized land ranges from 7.4 ha in Somali region to 312 ha in Gambella and the time profile of land transfers illustrates that, in most regions, average utilized area per farm peaked in 2007/8 9 and decreased thereafter. Data on average used land by size of initial land transfer point towards some interesting dynamics by farm size. For example, in all regions except Benishangul (Somali and SNNP), commercial farms that had initially been allocated less than 20 ha (20-50 ha) expanded, in some cases quite significantly. At the same time, farms initially allocated between 500 and 1000 ha or above 1000 ha cultivated less than this amount in 4 of the 8 regions. While detailed analysis of farm dynamics, sources of additional land acquired, and underlying factors (e.g., productivity) will be of interest, opening up options for a gradual increase in farm size for investors who prove to be successful may thus be a better strategy than giving large initial amounts of land to relatively unproven investors. Average land size transferred and utilized is larger for foreign than for national investors throughout, being largest in Gambella (2,212 ha) and Tigray (1,441 ha). It is larger for farms approved at higher levels, although, with the exception of Tigray, Amhara, and SNNP, area currently used by farms acquired by direct negotiation exceeds that of farms established through negotiations at woreda level. Farm size is higher for government owned farms—predominantly located in Oromia (1,509 ha/farm) and SNNP (220 ha/farm) — and for cotton in Benishangul (1,126 ha) and Gambella (1,070 ha), followed by wheat in Oromia (671 ha). To allow a first assessment of the extent to which commercial farms generate benefits for locals, table 3 tabulates key parameters in this respect (land acquisition, investment, credit market participation, use of improved inputs, and job creation) overall and by main crop. It illustrates that, overall, less than half the land area acquired (95 of 197 ha) was utilized during the first year of operations, increasing to 153 ha in the 2013/14 growing season. Only some 37% (close to 60% for maize, wheat, coffee and ‘others’, a category that is likely to include horticultural crops) report either lease length or annual lease fee. For those who do so, average lease length is 32 years at a fee of B 474 per ha (about USD 24 at the current exchange rate), although it is more than double this amount for wheat and ‘other’ crops. With the exception of wheat farms, more than 90% of the farms reported to have made investments amounting to an average of B 13,000 per ha (USD 650 at today’s exchange rate), from B 8,000/ha for coffee to B 34,000/ha for others. While there are differences across crops, most investment was for tractors and other machinery (39%), followed by land clearing (37%), buildings (18%), and roads/infrastructure (6%). Contrary to the perception of investors having near unlimited access to working capital, most of the initial investments seem to have been made out of own equity rather than by tapping into credit markets: in fact, less than 20% of commercial farmers indicated having taken a loan during the last 5 years with a mean loan size close to B 14,000. Even less, only 5% of commercial farms, had outstanding credit (of some B 11,500 or less than USD 600) with the incidence of outstanding loans lowest for Sorghum (3.6%) and highest for maize (8.1%) with mean outstanding loan amounts of less than B 7,000 for sesame and cotton and more 10 than B 25,000 for ‘others’. Establishing commercial farms is unlikely to increase credit market access for neighboring small-scale farmers. While some 50%, 60%, and 25% of commercial farms overall applied fertilizer, crop protection chemicals and improved seeds, incidence and magnitude of the use of these inputs varied by crop and farm size. Fertilizer use averaged 200 kg/ha, with an incidence between 13% for cotton and coffee (using 17 and 60 kg/ha, respectively), 52% for sesame (33 kg/ha), 65% for ‘others’ (785 kg/ha), and 80% or 82% for maize and wheat (with 215 and 187 kg/ha, respectively). Incidence of crop protection chemical use ranged from 20% for coffee to 91% for cotton while, for crops where improved seeds are important, incidence of their use ranged between 57% for ‘others’ and 81% for maize. The notion that large farms do not create large amounts of permanent employment is supported by the fact that, with an average of 0.05 permanent jobs per ha or one permanent worker per 20 ha farmed (plus 6.4 temporary workers, most of them male), commercial agriculture generates much less gainful employment than smallholder agriculture. Interpreting these figures is made difficult by the fact that the survey does not provide information on number of days worked and failed to indicate whether such employment was during the startup phase or during regular operations. Also, as smallholders may grow crops similar to those cultivated by commercial farms, seasonality will be an issue. 3.3 Constraints to expanding operations and interactions with smallholders As large amounts of land are transferred but not fully utilized, assessing perceived constraints to better land utilization may help identify areas to be strengthened in the pre-transfer review process. Indeed, as illustrated in table 4, some 50% of commercial farms, from 17% in Tigray to of 88% in Benighangul and some 70% in Gambela and SNNP indicate that they do not effectively utilize all of the land received. Key constraints to full use of available land relate, in descending order, to technology, manpower, land disputes, infrastructure access, and lack of resources for land clearing or credit. Regional differences are, not surprisingly, considerable: delayed availability of machinery, possibly due to import regulations, comes top in Gambella, followed by manpower which is a top issue in Benishangul and the second most important one in SNNP. Capital needs for land clearing need are important in Amhara and marketing in Afar. While earlier figures suggest limited credit market access, almost two thirds of farms, from 80% in Tigray, 78%, 68%, and 67% in SNNP, Oromia, and Gambella, respectively, to 33% in Afar and Somali regions, reported having given advice or other support to neighboring smallholders. If true and if no countervailing negative effects (e.g., from monopolizing access to resources) exist, one would expect commercial farm establishment to generate positive spillover for smallholders. Testing existence or magnitude of such effects empirically, though beyond the scope of this paper, would be of interest. 11 4. Resource use and productive performance across the farm size distribution To assess relative productive performance, we pool data on commercial farm with those from smallholders. Overall yields by commercial farms’ are double those by smallholders most often achieved by farms in the 10-20 ha category. More intensive use of inputs by large farms may account for part of this advantage, but lack of data on these, especially on labor, makes it difficult to assess productivity. The inverse relationship between farm size and productivity tends to disappear for commercial farms. 4.1 Descriptive evidence While there has been a large literature on the relationship between farm size and productivity (Ali and Deininger 2015), most analyses use data from smallholders or commercial farms separately, a separation that can yield biased estimates (Muyanga and Jayne 2014). To overcome this, we pool smallholder and commercial farms to explore the farm size-productivity relationship using data for the entire farm size distribution. Table 5 reports summary statistics of cultivated area and yield for selected major crops, as well as incidence and intensity of chemical fertilizer and improved seeds use by crop.12 The top panel highlights the marked differences in area cultivated between smallholders who all use less than one hectare on average for each specific crop and commercial farms, the largest group of which use 1,650 ha for wheat and about 1,200 and 1,000 ha for cotton and sesame. Comparing yields between smallholders and commercial farms suggest that, except for teff and coffee,13 the latter obtain about double the yield of the former, though better information on specific inputs, especially labor, and their cost would be required to translate this into productivity. Among commercial farms, those between 10 and 20 ha obtain the highest yields for most crops (maize with 4.2 t/ha, sorghum with 3.1 t/ha, teff with 0.9 t/ha, wheat with 4.2 t/ha, sesame with 1.4 t/ha, and soya beans with 2.4 t/ha). Exceptions are beans and coffee where yields are highest in the 20-50 ha category, and cotton, where they peak in the 50-100 ha bracket. Beyond yield differences, incidence of fertilizer use varies across farm size groups. In maize, only 35% of smallholders but 53% of commercial farms below 20 ha and 84% of those in the 100-500 ha group use fertilizer although, with 160 to 190 kg/ha, amounts applied by users are very similar. Differences are similarly pronounced for sesame, where 10% of smallholders vs. 65% of the largest farms use fertilizer, coffee (5% vs. 60%), and sorghum (17% vs. 45%) but less strong for wheat (65% vs. 75%) and teff, a typical smallholder crop where fertilizer use is highest for commercial farmers below 20 ha. The gap in improved input use between large and small farms is even more pronounced for improved seeds which was used by 20%, 1%, 4%, 12%, 1% of smallholders in maize, sorghum, teff, wheat, and sesame, respectively, 12 The number of observations in each of the cells is reported in the bottom panel of table 5. 13 Estimating coffee output could be tricky as it can be reported in different conditions and states (e.g., cherry coffee harvest vs. washed and dried coffee beans). Apparently, without the conditions and states reported in the two surveys, we need to be cautious while comparing coffee yield between smallholder and commercial farms. 12 compared to 80%, 42%, 85%, 74% and 60% in these crops for the largest commercial farms. As fertilizer will be most effective if applied together with seed, this could pose a serious constraint; in fact smallholders may even apply more fertilizer than can productively be absorbed by their traditional seed material. 4.2 Regression results While lack of prices and labor input in our data makes it difficult to estimate a production function, we use pooled smallholder and commercial farm data to run naïve regressions to estimate the relationship between farm size and yields as well as the likelihood of applying fertilizer or improved seeds. Assuming a quadratic relationship and allowing for structural differences in technology between smallholders and commercial farms, our estimating equation takes the form: ‫ݕ‬௜ ൌ ߙ଴ ൅ ߙଵ ‫ܣ‬௜ ൅ ߙଶ ‫ܣ‬ଶ ଶ ௜ ൅ ߚ଴ ‫ܦ‬௜ ൅ ߚଵ ‫ܣ‬௜ ൈ ‫ܦ‬௜ ൅ ߚଶ ‫ܣ‬௜ ൈ ‫ܦ‬௜ ൅ ߠ௪ ൅ ߳௜ (1) where ‫ݕ‬௜ is the log of physical yield of farm i from a given crop (maize, sorghum, teff, wheat, haricot beans, sesame or coffee) in quintals per ha; ‫ܣ‬௜ is the log of cropped area allocated to this crop; ‫ܦ‬௜ is a commercial farm dummy; ߠ௪ denotes district (woreda) fixed effects to capture differences in agro-ecological conditions or infrastructure; ࢻ and ࢼ are vectors of parameters to be estimated, and ߳௜ is a random error term. We use a similar specification to estimate a linear probability model for the incidence of chemical fertilizer and improved seed use. As output is in physical units and no information on prices was collected, computing monetary output values is impossible. Results from regressions for yield and the incidence of fertilizer and improved seed use are given in table 6, 7, and 8, respectively. Figures 2-6 graphically depict the relationship between farm size and yield as well as incidence of fertilizer use by plotting the marginal effects together with the associated 95% confidence interval for different farm sizes (in logs), separately for smallholders and commercial farms in maize, sorghum, teff, wheat, and sesame. We focus on three parameters, namely (i) if marginal effects are different from zero at 5% statistical level (as indicated by the 95% confidence band in the graph); (ii) if the marginal effect, at a given farm size, is greater or smaller than zero; and (iii) the slope of the marginal effect. In line with what was obtained earlier, our analysis illustrates that underlying relationships may be crop specific. For maize, we obtain a significant negative and slightly downward-sloping relationship between farm size and yield for smallholders, implying that the marginal effect is negative and decreases with farm size. For commercial farms, the marginal effect of farm size is marginally positive at the smallest farm size (1.4 ha) but turns negative above a farm size of some 30 ha. By contrast, the impact of farm size on fertilizer application is positive but downward-sloping, i.e., larger farm size makes fertilizer application more likely but at a decreasing rate. For wheat, marginal effects are also negative throughout but increase and become indistinguishable from zero for the largest smallholder and commercial farms. The likelihood of fertilizer 13 use in wheat fields is unaffected by size for commercial farms and is positive, but with a decreasing slope, for smallholders. By contrast, for sesame and sorghum, we find a negative relationship between size and productivity for small farms, approximate neutrality of yields with respect to size for large farms, and a positive relationship between fertilizer application and farm size for commercial farms. 5. Conclusion and policy implications A key premise of this paper is that, without sustained improvements in the quality and publicity of data on large-scale land-based investment, target countries may struggle to put in place a regulatory framework to increase the likelihood of such demand being translated into sustained local benefits or to attract responsible investors and institutional capital. The case of Ethiopia shows that a modest investment in data collection can have high returns and can incrementally build local data collection and analytical capacity. Although the descriptive evidence reported here only scratches the surface, it provides an interesting basis for future research. From a research perspective, it will be interesting to link this more directly to ongoing data collection efforts in the smallholder sector, among others, to look at the interaction between large farms and neighboring smallholders; in particular, the nature and size of potential spillover effects and the channels through which they materialize. Among commercial farms, determinants of entry and exit, and more generally the dynamics of firm performance over time would also be of interest. The usefulness of such data for government decision-making can be improved by linking more closely to data that are—or should be—held by other parts of the administration in at least two respects. First, business plans or licenses held by investment promotion agencies can serve as a basis for enforcement of contractual terms (e.g., fee collection) and meaningful monitoring of compliance with contractual obligations and, if necessary, adoption of remedies to bring about improvements. Second, linking geographic data on land cultivated with parcel boundaries kept by land administration institutions will be useful, as it allows comparison between the area adjudicated and what is actually used, thus possibly providing an opportunity to either cleanse or establish records through a process of negotiation to avoid border disputes. The fact that, in Ethiopia, the statistical agency is now initiating experiments to use the parcel boundaries as a basis for yield monitoring illustrates that, beyond its immediate usefulness in terms of providing vital information to assess and take regulatory action to manage an important part of the agricultural sector, such investment in data also create opportunities for realizing synergies with other institutions that can help to increase transparency to better assess and manage risk in production or upstream in the value chain. This can in turn be an important input into attracting higher levels of investment to the sector, something that by itself can generate the returns needed to justify such an investment. 14 Table 1: No of large farms in Ethiopia and land area initially transferred to them Farms Area Avg. Tigray Afar Amhara Oromia Somali Benish. SNNP Gambella No. ha ha Number of farms Total1 6,612 1,328,883 200.99 1,028 295 1,857 701 60 811 1,686 169 By Year of establishment Before 1991 302 211,074 698.41 17 1 36 103 46 1 86 10 1991-1992 394 31,431 79.72 372 14 1 6 1992-2002 1,234 206,587 167.46 372 30 342 175 1 18 265 30 2002-2006 1,261 168,211 133.40 93 39 694 124 4 54 242 10 2007-2008 1,586 314,775 198.47 105 93 280 173 6 289 587 53 2009-2010 1,093 231,651 211.99 39 66 287 72 324 254 51 2011-2013 742 165,154 222.62 30 66 204 53 3 125 246 15 By Size < 20 ha 392 4,940 12.62 155 58 57 34 15 2 63 4 20 - 50ha 1,031 32,533 31.54 466 59 239 110 6 20 120 11 50 - 100 ha 2,224 140,260 63.06 112 93 1020 131 4 150 690 23 100 - 500 ha 2,292 495,104 215.99 228 75 494 278 1 487 656 74 500 - 1000 ha 243 147,966 609.47 26 1 16 30 1 109 32 28 > 1000 ha 162 508,079 3,143.28 12 6 4 30 29 61 19 By owner’s nationality Ethiopian 6,287 1,085,593 172.67 1018 296 1824 566 60 786 1570 160 Foreign2 134 112,363 840.60 3 10 46 18 56 3 Joint 36 18,368 506.28 27 5 5 By ownership type Government3 121 203,304 1,677.70 1 1 8 67 1 33 10 Private 6,170 1,038,373 168.31 1001 290 1797 549 42 738 1606 143 Cooperatives 306 81,679 267.17 24 5 52 75 18 69 47 13 Others 15 5,526 361.67 1 10 1 2 By main crop cultivated Maize 885 162,063 183.09 23 89 178 7 94 452 43 Sorghum 843 97,466 115.59 429 1 238 30 5 103 31 6 Wheat 242 120,497 498.21 1 10 120 25 79 5 Sesame 2,494 354,688 142.19 581 18 1390 46 418 27 15 Coffee 977 162,944 166.76 1 202 749 26 Cotton 373 120,959 324.29 236 41 6 17 32 42 Other 797 310,267 389.32 16 18 88 120 23 178 317 32 By mode of original land acquisition Direct negot. 380 43,143 113.64 24 223 18 23 4 4 64 17 Woreda 3,499 400,883 114.58 933 4 1429 297 9 218 580 29 Region 2,162 520,603 240.75 53 20 372 230 571 823 94 Fed’l Gov't 252 291,533 1,155.32 1 11 5 66 3 17 130 19 Initial land acquisition by region Area in ha 114991 41201 172451 384968 1610 237540 293117 830052 Avg size in ha 112 140 93 549 27 293 174 491 Source: Own computation from 2013 CSA large farm survey 1About 299 (excluding Karuturi Global Limited) farms with size 50 ha and above did not respond to the detailed survey, and hence are not included in the analysis. The total area of these farms is 118,000 ha, and of which the total area for those with 1000 ha (24 farms) and above is 57,000 ha. 2A notable omission from the data is Karuturi Global Limited that acquired 100,000 ha of farm land in 2008 in Gambella region, but it is now deemed to be bankrupt. 3Most of the government owned farms are located in Oromia region accounting about 93% the initially transferred land, and note also that of which 83% of the land was acquired before 1991. 15 Table 2: Area transferred to and cultivated by commercial farms in different regions Area Tigray Afar Amhara Oromia Somali Benish’gl SNNP Gambella Transferred Cultivated Total (ha) 1,768,540 1,008,809 200,310 70,651 186,360 284,141 1,487 70,343 142,019 53,498 …ha/farm 267.5 152.6 195 239.1 100.4 405.9 24.8 86.9 84.2 317.9 By Year Before 91 216,278 155,873 107.5 74 133.1 1333.8 22.7 22 114.8 86.3 1991-92 71,703 36,369 91.1 114.9 51 138.7 92-2002 492,181 254,134 339.6 191.2 123.8 244.2 3 146.3 97.9 280.8 02-2006 293,027 158,101 105.7 501.6 89.4 231.3 14.5 131 116.8 260.4 07-2008 394,800 227,945 142.7 275 132.9 276.8 3.3 124.5 65.4 526.4 09-2010 180,043 104,569 275.4 132.3 85 165.4 59.3 67.2 245.7 11-2013 120,508 71,816 90.9 166.3 68.5 292.5 120.3 43.5 88.1 78.5 By Size (ha) < 20 ha 45,409 23,921 29.5 166 40.4 126.7 66.9 9 24.1 141 20-50 252,779 129,940 188.1 133.4 51.9 131 16.3 85.9 43.4 51 50-100 329,036 176,438 90.9 131.7 76.4 306.2 51 39.2 37.3 182.3 100-500 682,137 396,692 264.9 489.4 161.5 319.6 179 52.1 115.8 397.3 500-1000 155,970 88,106 628.1 1115 498.1 916 2 157.1 320.9 279 >1000 ha 303,211 193,712 1762 514.7 1497.3 3633.1 697 382.3 577 By owner's nationality Ethiopian 1,570,323 859,211 192 238.7 98.3 304.6 24.8 87 79.4 292.7 Foreign 80,445 47,677 1441 476.5 497.7 57.2 143.4 2,212.3 Joint 11,989 7,087 198 148.4 200 By ownership type Government 175,760 120,658 31.0 15,638.1 320.8 1,509.3 21.5 220.9 101.4 Private 1,495,524 833,108 194.0 202.7 97.4 298.5 31.7 89.8 78.2 337.8 Cooperatives 87,440 49,276 232.1 519.0 170.8 243.8 8.9 56.7 191.4 60.0 Others 9,815 5,767 483.8 177.1 102.0 1,593.7 By main crop cultivated Maize 165,995 93,493 41.6 182.5 209.8 9.1 73.3 70.1 6.9 Sorghum 117,612 62,316 71.9 58 74.2 163.7 8.6 76 29 13.3 Wheat 118,816 85,419 41 262.2 670.7 15.1 23.9 0.8 Sesame 607,417 314,268 290.1 18.9 83.1 160.8 51.4 28.4 18.4 Coffee 209,152 124,579 204 245.1 92.3 219.3 Cotton 263,526 163,442 253.3 369 326.3 1125.9 703.2 1,069.8 Other 286,021 165,293 57.1 528.5 216.2 854.5 43.6 84.2 47.6 69.1 By mode of original land acquisition Direct neg. 125,118 64,669 141.7 208.2 106.7 227 39.5 64 51.2 235.6 Woreda 759,636 396,857 185 568.8 89.1 147.9 6.8 54.4 64.3 53 Region 585,337 342,313 348.1 89.9 136.4 468.7 93 83.9 440.3 Fed’l Gov’t 175,203 100,952 4158 1257.4 549.2 670.8 120.3 302.7 188.8 310.3 No. of obs. 6,612 6,612 1,027 296 1,857 700 60 810 1,686 168 Source: Own computation from 2013 CSA large farm survey 16 Table 3: Key production decisions by commercial farms overall and by main crop cultivated Total Maize Sorghum Wheat Sesame Coffee Cotton Other Land Area acquired (ha) 197.29 179.93 115.59 410.66 142.19 166.69 322.86 390.14 .. developed first year (ha) 94.55 78.80 62.88 214.93 76.62 60.91 106.32 219.08 Cultivated now (ha; GPS) 152.59 105.62 73.90 353.39 125.99 127.50 438.19 207.42 Has irrigation facility (%) 13.78 19.65 6.88 9.17 5.39 3.43 66.87 30.25 Attempted to expand operation 38.75 27.96 39.01 27.28 45.73 42.08 35.35 28.18 Length of lease reported (%) 37.65 56.36 20.97 61.67 20.72 59.50 22.56 60.52 .. if yes, years 31.77 31.77 24.31 26.43 23.01 39.54 31.26 36.25 Annual lease fee reported (%) 36.43 54.89 20.12 52.81 20.22 59.38 20.83 58.06 . if yes, lease fee (Birr/ha) 474.07 190.44 115.56 1539.29 167.13 210.67 904.57 1202.01 Other annual payments reported (%) 10.09 18.65 6.59 10.54 6.91 8.16 10.47 16.27 . if yes, amount (Birr/ha) 375.31 724.47 86.38 438.17 212.84 99.44 99.48 510.91 Incidence and type of investment Made any investment (%) 94.10 90.44 99.01 61.30 99.47 92.97 91.48 88.70 If yes, size (Birr/ha) 13067 12,352 12,809 17,229 8,887 7,920 14,680 34,006 … of which on roads (%) 6.67 8.02 1.89 28.24 2.95 9.24 23.38 7.91 … of which on land clearing 36.51 20.49 43.86 14.16 42.82 41.30 24.20 28.28 … of which on buildings 18.21 24.93 10.80 14.54 13.00 30.59 12.44 25.26 … of which on tractors & machines 30.16 40.95 33.82 30.80 33.79 10.54 24.00 28.93 Credit market participation Took any loans last 5 years (%) 19.08 11.36 31.99 10.73 22.19 10.31 26.72 12.83 .. if yes, amount (Birr/ha) 13783 17,260 10,310 54,250 10,973 19,833 15,838 28,169 Has outstanding loans (%) 5.05 8.06 3.63 4.20 4.63 3.84 4.89 6.36 Outstanding loans (Birr/ha) 11560 10,085 15,944 14,249 6,584 9,351 6,991 25,634 Input use Used chemical fertilizer (%) 49.18 80.15 43.78 82.06 51.72 12.51 12.87 64.54 .. if yes, amount (kg/ha) 205.23 215.10 63.70 186.65 33.31 59.91 17.46 784.75 Used chemicals (pest/herb/fungic.) 60.22 48.49 81.16 79.65 68.65 20.62 90.85 53.01 Used improved seed 25.16 81.05 8.34 62.44 6.47 4.94 15.64 57.30 .. if yes amount (kg/ha) 156.85 60.66 4.41 210.64 26.49 11.45 1.92 393.14 Job creation Permanent farm workers/ha 0.05 0.02 0.03 0.03 0.04 0.03 0.05 0.15 Male temporary workers/ha 5.42 3.38 4.02 38.84 4.03 1.22 10.74 8.37 Female temporary workers/ha 1.39 2.76 0.22 0.16 0.28 0.76 5.31 4.04 Temporary workers per/ha 6.37 6.13 4.25 39.00 4.32 2.14 7.35 12.43 Total workers on the farm/ha 7.41 10.68 4.41 40.72 4.54 2.74 7.66 13.84 Crop shares (% of total land) Maize 11.64 74.32 1.56 1.79 1.26 0.56 0.32 7.03 Sorghum 16.96 3.56 73.74 2.47 16.79 0.21 2.74 3.91 Whet 3.58 0.87 0.15 87.33 0.08 0.00 0.03 1.84 Sesame 33.48 3.32 20.25 0.05 79.37 0.11 3.23 2.53 Coffee 14.49 0.13 0.06 0.02 0.02 97.69 0.00 0.14 Cotton 5.76 0.06 0.49 0.02 0.90 0.00 92.75 0.99 Others 14.09 17.74 3.74 8.32 1.59 1.43 0.93 83.56 No. of obs. (farms) 6612 885 843 242 2494 977 373 797 Source: Own computation from 2013 CSA large farm surveys 17 Table 4: Reasons for commercial farms’ failure to cultivate all of the area allocated to them Total Tigray Afar Amhara Orom Som. B'gl SNNP Gmbla Incidence & reasons for cultivating less than the total allocated area Cult. <100% of area 48.22 16.62 13.51 26.95 59.71 40.52 88.22 70.88 69.54 Technology 30.91 43.73 96.55 29.48 39.74 37.76 14.57 33.55 41.01 Manpower 29.77 4.23 0.00 11.21 24.93 0.00 50.60 30.54 37.26 Land dispute 21.09 26.33 1.90 18.76 19.80 10.32 31.57 15.83 22.95 Land clearing need 18.07 24.12 0.00 40.25 11.82 6.87 24.05 6.97 15.15 Infrastructure 17.79 10.20 0.00 10.23 24.22 3.46 23.26 17.50 17.74 Credit 17.71 7.46 1.90 17.85 11.25 54.09 22.24 17.38 23.92 Delayed machinery 9.87 8.67 2.95 2.25 3.13 23.74 9.37 12.24 44.07 Lack of irrigation 6.26 3.75 1.56 2.15 6.55 31.47 7.07 7.44 4.31 Lack of skilled workers 3.88 3.88 0.00 5.67 4.84 0.00 6.04 1.11 7.61 Input shortage 3.67 0.92 0.00 2.19 5.56 7.45 7.07 0.83 13.34 Marketing 3.47 6.53 37.20 3.16 3.85 0.00 3.39 1.26 10.08 Foreign exchange 1.06 0.00 0.00 0.69 3.70 0.00 0.00 0.84 3.73 Scope for expansion Tried to expand 38.75 23.72 37.63 60.52 38.33 3.04 38.62 26.98 18.05 If yes, faced problems 64.50 39.49 47.30 68.82 68.61 0.00 70.01 64.40 72.01 Gave advice to farmers 63.79 79.54 32.99 50.20 68.13 34.16 55.40 77.84 66.72 Source: Own computation from 2013 CSA large farm survey 18 Table 5: Productive performance of smallholders vs. commercial farms in different farm size classes Maize Sorghum Teff Wheat H. beans Soya Sesame Coffee Cotton Area cultivated Smallholder 0.25 0.43 0.46 0.34 0.11 0.69 0.14 < 20 ha 4.96 7.08 3.90 6.00 5.81 2.87 8.57 12.66 3.28 20-50 14.95 13.59 6.20 15.15 4.88 13.56 24.07 32.07 23.26 50-100 29.4 21.9 7.1 46.6 16.4 26.8 45.5 65.4 33.5 100-500 96.2 55.5 10.0 118.7 31.3 45.3 114.4 201.7 214.5 > 500 ha 339.0 217.1 32.3 1654.6 147.6 496.7 1008.7 672.0 1194.3 Output (quintal/ha) Smallholder 18.73 14.90 8.44 14.59 13.74 6.00 16.79 < 20 ha 42.04 30.88 9.18 41.69 23.86 24.39 13.94 4.24 5.18 20-50 37.42 24.52 8.79 33.68 38.31 13.91 10.74 8.43 17.75 50-100 36.89 25.64 8.61 26.11 20.95 18.82 8.46 6.38 30.81 100-500 39.30 28.21 7.75 24.64 30.52 23.71 9.77 6.37 29.31 > 500 ha 33.81 29.51 9.90 28.32 34.96 22.50 10.84 7.02 18.92 Used chemical fertilizer Smallholder 34.52 17.06 61.56 65.01 18.10 10.40 4.82 < 20 ha 52.85 19.48 85.44 65.79 30.52 23.83 12.99 0.00 0.00 20-50 58.39 20.18 77.00 59.62 20.76 2.05 33.29 0.91 0.00 50-100 73.79 18.28 61.34 74.22 18.26 2.88 48.19 3.13 10.46 100-500 84.31 33.18 62.44 64.74 25.77 27.83 51.25 14.99 12.32 > 500 ha 82.70 44.79 48.01 74.45 31.69 40.82 64.93 59.41 15.09 Quantity of chemical fertilizer used (kg/ha) Smallholder 179.32 112.01 133.07 187.57 105.32 77.86 69.98 < 20 ha 165.43 84.99 122.63 363.65 32.79 340.37 171.62 20-50 187.22 80.60 155.68 270.65 115.41 48.33 38.53 4.68 50-100 151.59 51.86 101.79 97.74 78.08 73.85 32.33 17.99 12.04 100-500 187.09 63.42 89.83 118.91 127.90 129.88 40.73 96.37 51.04 > 500 ha 163.68 82.05 64.34 72.46 163.64 32.19 59.02 88.05 64.57 Used improved seed Smallholder 20.30 1.77 4.23 12.51 1.59 0.45 1.24 0.38 < 20 ha 51.99 17.58 85.44 65.79 30.52 23.83 11.18 0.00 0.00 20-50 57.87 18.66 76.23 58.25 17.03 0.82 29.67 0.91 0.00 50-100 69.57 15.88 58.66 74.22 18.26 1.44 40.22 2.71 10.46 100-500 86.21 4.21 45.86 40.32 26.84 21.19 3.72 0.95 10.14 > 500 ha 80.30 42.08 40.93 74.45 21.23 40.82 59.81 56.85 12.64 Quantity of improved seed used (kg/ha) Smallholder 40.93 33.09 53.17 183.24 121.14 73.31 < 20 ha 53.90 9.92 36.12 581.88 36.79 125.20 17.24 20-50 46.49 25.71 79.62 262.04 115.11 21.22 7.11 6.38 50-100 28.93 17.30 40.67 138.65 59.42 91.68 3.26 8.21 100-500 24.47 10.01 52.12 156.68 69.57 148.37 5.35 116.92 8.39 > 500 ha 47.44 2.86 33.91 67.52 171.41 60.91 3.27 1.37 No. of observations Smallholder 1703 1216 1189 794 521 180 838 < 20 ha 358 295 291 162 46 79 349 87 73 20-50 479 1122 212 109 40 122 1207 220 27 50-100 351 833 124 28 100 104 1000 358 90 100-500 382 724 165 128 142 178 792 315 234 > 500 ha 89 103 34 37 29 33 109 39 96 Source: Own computation from 2013 CSA large farm and smallholder farm surveys 19 Table 6: Naïve estimates of the farm size productivity relationship for all Ethiopian producers Maize Sorghum Teff Wheat H. Beans Sesame Coffee Panel A: All observations Cropped area (α1) -0.337*** -0.294*** -0.270*** -0.134** -0.602*** -0.434*** -0.279*** (-8.213) (-8.129) (-5.553) (-2.032) (-9.064) (-6.512) (-3.357) Cropped area squared(α2) -0.007 0.020** 0.015 0.023 -0.019** -0.007 0.012 (-1.016) (2.282) (1.029) (1.626) (-2.020) (-0.330) (1.050) Commercial farm dummy 1.236*** 1.345*** 0.333*** 1.584*** 3.319*** 0.883*** 1.167* (9.116) (10.084) (2.620) (4.938) (10.259) (5.349) (1.842) Cropped area * commercial (β1) 0.454*** 0.146** 0.203* -0.250* 0.507*** 0.561*** -0.270 (5.654) (2.061) (1.883) (-1.834) (3.270) (6.402) (-0.998) C. area squared * commercial (β2) -0.021 -0.002 -0.016 0.002 -0.016 -0.014 0.062** (-1.592) (-0.169) (-0.446) (0.124) (-0.673) (-0.669) (2.076) Constant 1.892*** 1.877*** 1.407*** 2.063*** 0.246* 1.061*** 1.047*** (32.994) (35.378) (34.047) (28.930) (1.845) (9.348) (7.203) Number +of observations 2,124 2,559 1,403 970 625 1,699 1,019 R2 0.148 0.135 0.114 0.083 0.379 0.082 0.163 Note: Woreda fixed effects included throughout. Cropped area is in logs, denominated in hectares. 20 Table 7: Farm Size and Incidence of fertilizer use Maize Sorghum Teff Wheat H. Beans Sesame Coffee Panel A: All observations Cropped area (α1) 0.027* -0.006 0.054** 0.022 -0.037* 0.012 -0.000 (1.798) (-0.370) (2.378) (0.911) (-1.716) (0.328) (-0.011) Cropped area squared(α2) -0.004 -0.000 0.001 -0.008* -0.008*** -0.001 -0.001 (-1.582) (-0.151) (0.094) (-1.706) (-2.867) (-0.214) (-0.406) Commercial farm dummy 0.152*** 0.082 0.053 0.042 -0.076 -0.166* 0.159** (3.404) (1.379) (0.914) (0.439) (-0.787) (-1.790) (2.116) Cropped area * commercial (β1) 0.044* -0.045 -0.025 -0.021 -0.030 0.083* -0.196*** (1.818) (-1.480) (-0.612) (-0.530) (-0.659) (1.667) (-5.637) Cropped area sq. * commercial (β2) -0.000 0.021*** -0.003 0.010 0.027*** 0.004 0.039*** (-0.014) (4.144) (-0.203) (1.548) (3.541) (0.522) (9.355) Constant 0.416*** 0.155*** 0.646*** 0.690*** 0.201*** 0.231*** 0.048* (18.726) (6.088) (31.806) (24.819) (4.366) (3.682) (1.802) Number +of observations 2,276 2,620 1,425 998 674 1,724 1,147 R2 0.106 0.048 0.024 0.037 0.041 0.109 0.184 Note: Woreda fixed effects included throughout. Cropped area is in logs, denominated in hectares. 21 Table 8: Farm Size and incidence of improved seed use Maize Teff Wheat Panel A: All observations Cropped area (α1) 0.037** 0.001 0.037 (2.510) (0.091) (1.397) Cropped area squared(α2) -0.002 -0.000 0.008 (-1.028) (-0.012) (1.641) Commercial farm dummy 0.350*** 0.408*** 0.306** (7.155) (10.142) (2.295) Cropped area * commercial (β1) 0.077*** 0.059** 0.149** (2.991) (2.111) (2.389) Cropped area sq. * commercial (β2) -0.009** 0.012 -0.033*** (-2.176) (1.261) (-3.844) Constant 0.315*** 0.014 0.091*** (14.547) (0.988) (3.068) Number +of observations 1,760 1,228 547 R2 0.249 0.236 0.224 Note: Woreda fixed effects included throughout. Cropped area is in logs, denominated in hectares. 22 Figure 1: Changes in coverage with commercial farms over time 23 Marginal Effects of Land Size on Maize Yield Smallholders Commercial Farms .2 .2 0 0 -.2 -.2 -.4 -.4 -3.95 -2.90 -2.06 -1.38 -0.50 0.47 2.10 3.18 4.21 5.31 Incidence of Fertilizer Use Smallholders Commercial Farms .15 .15 .1 .1 .05 .05 0 0 -.05 -.05 -.1 -.1 -3.95 -2.90 -2.06 -1.38 -0.50 0.47 2.10 3.18 4.21 5.31 Figure 2: Marginal effects of land size on maize Marginal Effects of Land Size on Sorghum Yield Smallholders Commercial Farms .2 .2 0 0 -.2 -.2 -.4 -.4 -.6 -.6 -3.10 -2.07 -1.40 -0.79 0.01 1.65 2.56 3.11 3.74 4.67 Incidence of Fertilizer Use Smallholders Commercial Farms .3 .3 .2 .2 .1 .1 0 0 -.1 -.1 -3.10 -2.07 -1.40 -0.79 0.01 1.65 2.56 3.11 3.74 4.67 Figure 3: Marginal effects of land size on sorghum 24 Marginal Effects of Land Size on Teff Yield Smallholders Commercial Farms .4 .4 .2 .2 0 0 -.2 -.2 -.4 -.4 -2.63 -1.78 -1.19 -0.64 0.02 -0.22 1.09 1.67 2.19 2.89 Incidence of Fertilizer Use Smallholders Commercial Farms .2 .2 .1 .1 0 0 -.1 -.1 -.2 -.2 -2.63 -1.78 -1.19 -0.64 0.02 -0.22 1.09 1.67 2.19 2.89 Figure 4: Marginal effects of land size on teff Marginal Effects of Land Size on Wheat Yield Smallholders Commercial Farms .4 .4 .2 .2 0 0 -.2 -.2 -.4 -.4 -.6 -.6 -3.47 -2.27 -1.64 -1.06 -0.35 -0.13 1.87 3.40 4.91 6.25 Incidence of Fertilizer Use Smallholders Commercial Farms .15 .15 .1 .1 .05 .05 0 0 -.05 -.05 -.1 -.1 -3.47 -2.27 -1.64 -1.06 -0.35 -0.13 1.87 3.40 4.91 6.25 Figure 5: Marginal effects of land size on wheat 25 Marginal Effects of Land Size on Sesame Yield Smallholders Commercial Farms .2 .2 0 0 -.2 -.2 -.4 -.4 -.6 -.6 -.8 -.8 -3.53 -2.13 -1.25 -0.55 0.70 2.12 3.19 3.74 4.38 5.61 Incidence of Fertilizer Use Smallholders Commercial Farms .2 .2 .1 .1 0 0 -.1 -.1 -.2 -.2 -3.53 -2.13 -1.25 -0.55 0.70 2.12 3.19 3.74 4.38 5.61 Figure 6: Marginal effects of land size on sesame 26 References: Abebe, H. G. 1990. "Generating Marketed Surplus of Food through State Farms: A Critical Evaluation of the Ethiopian Experience." Insitute for Social Studies, The Haque. Ali, D. A. and K. Deininger. 2015. "Is There a Farm Size–Productivity Relationship in African Agriculture? Evidence from Rwanda." Land Economics, 91(2), 317-343. Andrade de Sa, S., C. Palmer and S. di Falco. 2013. "Dynamics of Indirect Land-Use Change: Empirical Evidence from Brazil." Journal of Environmental Economics and Management, 65(3), 377-393. Anseeuw, W., et al. 2012. "Transnational Land Deals for Agriculture in the Global South: Analytical Report Based on the Land Matrix Database." Bern, Montpellier, Hamburg. Arezki, R., K. Deininger and H. Selod. 2015. "What Drives the Global 'Land Rush'?", World Bank Economic Review,. Baland, J. M. and J. A. Robinson. 2008. "Land and Power: Theory and Evidence from Chile." American Economic Review, 98(5), 1737-1765. Beyene, H. 2015. "Large and Medium Scale Commercial Farms Annual Surveys: The Experience of Ethiopia." Paper presented at the 16th Annual World Bank Conference on Land and Poverty, World Bank, Washington DC. Binswanger, H. P., K. Deininger and G. Feder. 1995. "Power, Distortions, Revolt and Reform in Agricultural Land Relations." Handbook of development economics, 3B, 2659-2772. Burnod, P., M. Gingembre and R. Andrianirina Ratsialonana. 2013. "Competition over Authority and Access: International Land Deals in Madagascar." Development and Change, 44(2), 357-379. Central Statistical Agency, Federal Democratic Republic of Ethiopia 2011. Large and medium scale commercial farms sample survey 2010/11 (2003 E.C): Resutls at country and regional levels, Addis Ababa Chamberlin, J., T. S. Jayne and D. Headey. 2014. "Scarcity Amidst Abundance? Reassessing the Potential for Cropland Expansion in Africa." Food Policy, 48, 51-65. Collier, P. and A. J. Venables. 2012. "Land Deals in Africa: Pioneers and Speculators." Journal of Globalization and Development, 3(1). Collier, P. and S. Dercon. 2014. "African Agriculture in 50 Years: Smallholders in a Rapidly Changing World?" World development, 63, 92-101. Cotula, L. 2014. "Testing Claims About Large Land Deals in Africa: Findings from a Multi-Country Study." Journal of Development Studies, 50(7), 903-925. Deininger, K. and D. Byerlee. 2011. Rising Global Interest in Farmland: Can It Yield Sustainable and Equitable Benefits? Washington, D.C.: World Bank. Deininger, K. and D. Byerlee. 2012. "The Rise of Large Farms in Land Abundant Countries: Do They Have a Future?" World Development, 40(4), 701-714. Hall, R. 2011. "Land Grabbing in Southern Africa: The Many Faces of the Investor Rush." Review of African Political Economy, 38(128), 193-214. Land Matrix 2015. Country Inventory Ethiopia http://www.landmatrix.org/en/get-the-detail/by-investor-country/ethiopia/ Messerli, P., et al. 2014. "The Geography of Large-Scale Land Acquisitions: Analysing Socio-Ecological Patterns of Target Contexts in the Global South." Applied Geography, 53(0), 449-459. Muyanga, M. and T. S. Jayne. 2014. "Effects of Rising Rural Population Density on Smallholder Agriculture in Kenya." Food Policy, 48, 98-113. Nugent, J. B. and J. A. Robinson. 2010. "Are Factor Endowments Fate?" Revista de Historia Economica, 28(1), 45-82. Oakland Institute (2011), Understanding land investment deals in Africa: Country Report Ethiopia, The Oakland Institute, Oakland, CA. OECD. 2014. "Responsible Agricultural Supply Chains." Paris. Pearce, F. 2012. The Landgrabbers: The New Fight over Who Owns the Earth. Transworld Publishers, London. Rahmato, D 2011, Land to investors: Large scale land transfers in Ethiopia, Addis Ababa, Forum for Social Studies. Rajan, R. and R. Ramcharan. 2011. "Land and Credit: A Study of the Political Economy of Banking in the United States in the Early 20th Century." The journal of finance, 66(6), 1895-1931. 27 Rulli, M. C., A. Saviori, and P. D'Odorico. 2013. "Global land and water grabbing." Proceedings of the National Academy of Sciences of the United States of America 110 (3): 892-7. Schoneveld, G. C. 2014. "The Geographic and Sectoral Patterns of Large-Scale Farmland Investments in Sub-Saharan Africa." Food Policy, 48(0), 34-50. Sitko, N. J. and T. S. Jayne. 2014. "Structural Transformation or Elite Land Capture? The Growth of “Emergent” Farmers in Zambia." Food Policy, 48(0), 194-202. Tyler, G. and G. Dixie. 2013. "Investments in Agribusiness: A Retrospective View of a Development Bank's Investments in Agribusiness in Africa and East Asia." Washington, DC. Vollrath, D. 2009. "How Important Are Dual Economy Effects for Aggregate Productivity?" Journal of Development Economics, 88(2), 325-334. 28