Testing Claims about Large Land Deals in Africa: Findings from a Multi-Country Study

Abstract

Despite much research on large land deals for plantation agriculture in Africa, reliable data remain elusive, partly because of limited access to information and practical and methodological challenges. International debates are still shaped by misperceptions about how much land is being acquired, where, by whom, how and with what consequences. This article aims empirically to test some common perceptions through an analysis of findings from research conducted in three African countries: Ethiopia, Ghana, and Tanzania. The article presents new evidence on the scale, geography, drivers and features of land deals, relates findings to data from earlier research and international efforts to monitor land deals, and outlines possible ways forward for ongoing monitoring of the deals.

1. Introduction

Recent years have witnessed renewed business interest in African agriculture. Evolutions in local-to-global economies (demographic growth, shifting consumption, constraints in food supply, climatic uncertainties, financialisation of commodity markets,¹ to name a few factors) have resulted in higher and more volatile agricultural commodity prices, and have made agriculture a more appealing business proposition, particularly in developing countries. Since the mid-2000s, this renewed interest has been associated with widely reported large-scale land deals in many parts of Africa, Asia and Latin America. Dubbed ‘land grabs’ in the media and in the critical literature, land deals have sparked much polarised discussion, in which strong positions are taken on the impacts of such investments on livelihoods, rights, sovereignty, development and conflict at local, national and international levels. The polarisation in the debate reflects radically different views of the best options for poor countries in terms of their agricultural development and policy priorities. While some emphasise the positive role that large-scale investors can play in agrarian transformations and a transition towards a more developed agriculture capable of feeding a growing population with higher productivity and superior technology, others oppose large-scale land deals that are seen to threaten local livelihoods, to accelerate the demise of peasantries and to contribute little to the development of recipient countries.

Since media reports about ‘land grabbing’ began flowing in the late 2000s, much research has emerged that aims to shed light on the scale, geography, drivers, features and early outcomes of large land deals. But reliable data remain elusive, partly because of limited access to information and practical and methodological challenges. International debates are still shaped by misperceptions about how much land is being acquired, where, by whom, how and with what consequences. This article aims empirically to test some common perceptions through an analysis of findings from country studies conducted in three African countries: Ethiopia, Ghana, and Tanzania. These three countries have experienced much investor and public interest in the recent wave of land deals, and several studies have already been published on the significance and implications of the deals in the three countries. The article presents new evidence on the scale, geography, drivers and features of land deals for plantation agriculture, relates empirical evidence from these country studies to findings from earlier research and international efforts to monitor land deals, and outlines possible ways forward for ongoing monitoring of the deals.

The research involved systematic data collection from relevant government ministries (responsible for land, agriculture and/or investment²), cross-checked for accuracy through third-party sources. Where land administration responsibilities are decentralised, data were also collected from selected regions.³ These systematic national inventories cover land deals over 1,000 hectares concluded between 1 January 2005 and 31 August 2012. The parameters for inclusion are assessed in relation to the first contract relating to a given project. For example, a deal is included in the inventory if the memorandum of understanding (MOU) is for a land area over 1,000 hectares, even if the leases issued thus far to implement that MOU are for a land area below 1,000 hectares. Each national inventory includes data on project location, timeline, land area size, land acquirer (for example, foreign/national; public/private; country of origin), land provider, land tenure (for example, land ownership; land rights acquired; pre-existing land rights), economic activity, investment amount, target markets for produce, key data on employment and fiscal regime, and any information about social and environmental impacts.

Data collection from official sources was complemented through a critical appraisal of the available literature in the three countries, through semi-structured interviews with national and sub-national government officials, company representatives, development agencies and researchers, and, in Ethiopia and Ghana, through limited fieldwork in a small number of sites.⁴ Limited fieldwork in Ethiopia and Ghana involved qualitative data collection aimed at identifying issues through interviews with company management (when accessible), and interviews and focus group discussions with selected local stakeholders adversely or positively affected by the project (for example, villagers who lost land to the project, current or former employees). In Ethiopia, about 50 interviews were carried out; in Ghana, 37 interviews, and focus group discussions with a total of 48 people were conducted. The aim of this limited fieldwork was to complement findings from the systematic inventories, to clarify points in relation to processes of land acquisition and allocation and, only to a very limited extent, to capture some early impressions and perceptions in selected sites about the initial and potential positive and negative socio-economic outcomes of a few selected deals.

In each country, data collection through both systematic inventories and limited fieldwork was undertaken by a national research team, leading to three country studies.⁵ The country study for Ethiopia is publicly available (Keeley, Eid, Kidewa, & Seide, 2014). This article summarises findings from the three country studies. The first lead author of this article led overall research design and coordination, and provided technical input into the country studies. Both lead authors of this article developed the synthesis analysis reflected in the article.

It is important to recognise the limitations of the evidence base that this article analyses. Carrying out research on a highly sensitive topic within the context of increasingly polarised debates has proven a major challenge – even more so because the study relies heavily on government-held data that are not in the public domain and on a limited pool of interviewees who may have their own biases about large land deals. In addition, the ability of governments to record land deals and manage data varies considerably. So the problem is not only data access, but also data availability. As a result of these difficulties, the three national inventories present major gaps, although the Ethiopia inventory is significantly more complete than the other two. The qualitative component of the country studies also suffered from important limitations. The number of people reached through interviews and focus groups was limited. Selection bias could also be an issue, as project selection for the site visits was heavily influenced by company willingness to engage and by decisions made by the national research teams, which also had to take into account practical considerations. Therefore, discussion of data drawing on the qualitative component only serves as preliminary illustration of aspects of a complex phenomenon that require more in-depth research.

Beyond limitations linked to the design and implementation of the three country studies, more fundamental methodological challenges constrain a rigorous assessment of important aspects of the land deals. These challenges include, for example, conceptual difficulties in classifying ‘approved’ deals (discussed below), the lack of baseline data to measure the socio-economic changes associated with the deals, and the very recent nature of many deals, which prevents an analysis of long-term outcomes. These challenges and limitations are discussed in greater detail in the relevant sections of this article.

Despite these limitations, which are shared by most of the literature on ‘land grabs’ (Oya, 2013a), we managed to test some common perceptions emerging from the available literature. Our findings can be grouped in four main areas. First, we find that the scale of land acquisition accounts for a small share of land suitable for agriculture in each country, although there is significant variation across the three countries. This finding puts into perspective claims that the current wave of large land deals signals a rapid, transformational transition from small- to large-scale farming in Africa. Long-term demographic and socio-economic change in many parts of Africa is likely to have a greater impact on evolving land relations than the much publicised transnational land deals. However, the scale of the transnational deals appears more significant once the short period of time within which the deals have been signed is considered, and once the strong geographic concentration of the deals and the prospects for continued deal-making, particularly in Ethiopia, are factored in.

Second, we confirm the important role played by national investors, which belies the excessive focus in the literature on foreign players. Although the literature is catching up with this phenomenon, the methods deployed and the challenges in obtaining reliable information on this issue make it difficult to ascertain the real significance of national investors. In Ethiopia at least, much land acquisition is led by national parastatal agencies. In addition, this research challenges enduring perceptions that investors from China and the Middle East are the leading land acquirers. Our findings suggest that companies originating in Europe and the United States have played a leading role in Ghana and Tanzania, while Indian companies have been particularly active in Ethiopia.

Third, this research nuances the often-held idea that governments in countries where land deals abound are either passive recipients of the deals, or the main drivers of deal-making. Our research highlights significant diversity in the ways in which governments facilitate or manage the deals. This is a sensitive political issue, and complex national and local political economies are at play.

Fourth, the article provides some warnings about the difficulty of assessing impacts due to both temporal and methodological challenges. What our research does find, however, is that many deals have had limited implementation, so that many of the benefits promised have not (yet) materialised. While more in-depth research is needed to fully understand the factors underpinning this limited implementation, the finding offers a cautionary tale on the potential of large land deals to contribute to poverty reduction and inclusive development.

Finally, the article points to areas for further research. The full implications of the new wave of land deals can only be assessed if the deals are examined not in isolation, but within the wider political and economic projects they form part of (for example, agro-industrial transformation, state building). This calls for research that goes beyond the case study approach that has thus far prevailed in the ‘land grab’ literature, and that explores the cumulative outcomes, both positive and negative, that the deals are having within those wider processes of change.

The remainder of the article is structured as follows. Section 2 discusses the evidence concerning scale, geography and drivers, comparing findings from our national inventories with available international databases. Section 3 examines key aspects of deal making in the three countries, and touches on the socio-economic outcomes of the deals. While recognising the importance of environmental considerations in assessing outcomes, the datasets emerging from the country studies are particularly incomplete on environmental aspects, so this article touches on these issues only very briefly. Section 4 suggests next steps in the effort to track large-scale land deals.

2. Scale, Geography and Drivers

2.1 Scale – How Much Land has been Transacted?

The polarisation in debates about ‘land grabbing’ is partly fuelled by competing evidence on the scale of the deals, coming from different institutions, researchers, governments and investors themselves. There is a wide range of different figures regarding the scale of the deals, their trends and the factors that drive them (Cotula, 2012, 2013; Oya, 2013b). Yet the evidence base remains patchy.

Online databases run by the non-governmental organisation GRAIN and by the Land Matrix consortium led by the International Land Coalition contain extensive information on publicly reported land deals.⁶ These databases have played an important role in increasing public awareness of the phenomenon, and in catalysing citizen action to hold governments and companies to account, during what Edelman, Oya, and Borras (2013, p. 1520) call the ‘making sense period’. The Land Matrix is mainly based on publicly available data, but also involves cross-checking of information. Some commentators have raised concerns about the comprehensiveness and reliability of the information contained in the Matrix dataset.⁷ A major constraint is that publicly available information remains limited, and reliability is often a challenge. Negotiations typically occur behind closed doors, and only a few contracts are in the public domain. Even access to land registries is often constrained. In addition, important features of the deals, including their scale and level of implementation, can evolve very rapidly, as projects may be announced, terminated or transferred. This makes it difficult for any system for ongoing monitoring to keep data up-to-date. However, several revisions to the Matrix dataset and less reliance on media reports have increased its accuracy over time.

In addition to these global datasets, some studies have carried out systematic inventories of land deals based on official government records in the countries where the land has been acquired. Most commonly, these official lists were cross-checked with third-party sources within the country (for example Cotula, Vermeulen, Leonard, & Keeley, 2009; Deininger & Byerlee, 2011; GTZ, 2009). The methodology used for this present research is a further refinement of these earlier efforts. But even these systematic national inventories can be challenged. Access to government records is typically constrained. In all cases, the quality of estimates from country-level inventories ultimately depends on how well national systems record land allocations and on how easy it is to access those data. It is significant that the Ethiopia inventory was by far the most complete out of several country studies both in Cotula et al. (2009) and in this study’s new inventory exercise – reflecting, perhaps, a higher level of government involvement in deal-making. In other countries, governments themselves do not have systematic information about the deals that their various ministries or customary authorities are handing out. In addition, inventories that restrict their research to deals recorded with central government may not capture the smaller deals – even though their cumulative impact may be greater than that of a few large deals.

Even defining what constitutes a ‘deal’ to be entered into a database is not as straightforward as it might seem. A ‘memorandum of understanding’, ‘letter of intent’ or ‘convention of establishment’ may commit the government to allocate land, but the actual lease may concern a much smaller area. In fact, conventions of establishment or letters of intent may never translate into land leases. In Ghana, a memorandum of understanding with traditional authorities for a jatropha plantation of over 300,000 hectares resulted in a lease for about 13,000 hectares (Wisborg [2012] and contracts reviewed by the authors). On the other hand, some contracts entitle the company to acquire additional land and expand the plantation beyond the size of the original deal. In sum, the degree of precision for data on land deals is extremely problematic and usually subject to correction.

Both global online databases and national systematic inventories point to a phenomenon of considerable scale in absolute terms, although with considerable variation in figures – as will be discussed below. We now turn to discussing findings from the national inventories prepared for this article. Key findings are presented in Table 1. Overall, the national inventories confirm the large scale of cumulative land acquisition in absolute terms, but also highlight considerable variation across the three countries: in Ethiopia, land leases account for almost 10 times the aggregate land area leased in Ghana over the same period (see Table 1). As discussed, the national inventories also differ substantially in terms of data gaps, with the Ethiopian inventory being far more complete than the ones prepared for Ghana and Tanzania, so comparisons between these three inventories should be treated with caution. It is possible that the inventories for Ghana and Tanzania underestimate scale more than the Ethiopia inventory – though underestimation of scale is a risk in all three inventories for the reasons outlined above, and also because in Ethiopia and Ghana the research teams could only visit selected regions.

Table 1. Summary of scale and size of land deals in national inventories

Country	Total number of deals recorded	Aggregate land area under MOU or establishment convention^a	Aggregate land area leased^b	Total land area under transaction^c	Average size^d	Median size	Total land under transaction as a percentage of land suitable for agriculture^e
Ethiopia	174	NA	1,055,975	1,055,975	4,516	2,000	1.9%
Ghana	28	402,941	112,485	402,941	9,374	8,825	1.9%
Tanzania	64	707,175	203,347	896,742	7,262	5,236	1.1%

Source: Inventory data – authors’ estimates.Notes: ^aFigure for Ghana refers to 13 deals for which data is available. ^bLand area leased for Ethiopia includes land allocations for state-run sugar plantations amounting to 324,436 hectares, which are not counted in the list of 174 deals; land area leased for Ghana is based on the figure for MOU, adjusting figure for one large deal to ‘land under lease’ instead of ‘land under MOU’. ^cTotal land under transaction includes MOUs not (fully) followed through as well as leases provided the latter are for land areas not already covered by the MOUs. ^dThe average and median sizes are calculated only on the deals with information on land amounts for land lease (162 in Ethiopia, excluding state-run sugar plantations, 12 in Ghana and 28 in Tanzania). ^ePercentages calculated on the basis of Food and Agriculture Organisation of the United Nations/Global Agro-Ecological Zones data for land suitable for rain-fed cereal farming.

A few additional clarifications are needed to interpret the findings presented in Table 1 correctly. The figure for Ethiopia refers to the land area actually leased over the study period. The aggregate land area covered by MOUs is not reported in Table 1, due to the many gaps in this part of the inventory. For the few deals for which we have figures for both MOUs and actual leases (70 deals out of 174), land area sizes are identical for virtually all deals – with the exception of one deal for 50,000 hectares under MOU, and 4,000 hectares under lease. This suggests that the land area under lease may not depart considerably from the land area under MOU. Entries are also missing for a few land leases, including a couple of MOUs involving large land areas (31,300 and 15,000 hectares).

On the other hand, the Ghana inventory contains only two entries for land under lease, so Table 1 contains an aggregate figure based on land under MOUs or Establishment Conventions. Even for MOUs, there are numerous data gaps (land area data are missing for 15 deals out of 28), so we expect our inventory to significantly underestimate scale, at least in terms of MOUs. Table 1 includes an estimate for land actually leased. This estimate is calculated by taking the total area under MOU, replacing figures for the two deals presenting leased area data, and assuming that area sizes for all other MOUs and leases are the same. One ‘Heads of Agreement’ for 303,514 hectares in Ashanti Region accounts for the vast majority of the land area under MOU, and also for the difference between aggregate land area under MOU and under lease (the actual land lease for this deal is for 13,058 hectares; Wisborg [2012] and contracts reviewed by the authors). Assuming the same land area size for all other MOUs and leases may result in significant overestimation of aggregate scale, which might balance significant underestimation due to the fact that data are missing for over 50 per cent of the deals.

In Tanzania, the inventory contains at least one land area size figure for all land deals. However, in the vast majority of cases data refers to either MOUs or land leases – there is very little overlap between land area size under MOU and lease. As a result, the total land area under transaction needs to be calculated by adding up land areas under MOU and under land lease. One exception is one deal that involves an MOU for 324,800 hectares but leases for a total of 13,780 hectares.

Apart from the large differences between the three countries in terms of number of deals and total land area transacted, there are also remarkable differences in the average size of the deals. Since the cut-off point was set at 1,000 hectares, all these deals are on average quite large, as expected, but also with much heterogeneity within each country. This is particularly the case in Ethiopia and Tanzania, where there are significant differences between average values (pushed upwards by a few ‘mega’ deals) and median values (see Table 1). Ethiopia displays the smallest median values, which reflects the presence of a greater number of smaller deals identified between 1,000 and 2,000 hectares, almost all by Ethiopian investors. In Tanzania, apart from one very large MOU over 300,000 hectares, most deals are below 10,000 hectares. In Ghana, the median size is larger, but also there are no documented examples of ‘megafarms’ over 100,000 hectares.⁸ Overall, ‘megafarms’, which tend to receive much publicity in the literature, are actually a small minority in the total count of deals above 1,000 hectares.

Our datasets present differences compared to internationally available databases like the Land Matrix, discussed above. Our inventories suggest that the aggregate scale of land acquisition is smaller than was previously suggested, especially if actual land leases (as opposed to MOUs) are considered. This may be due to overestimation of earlier figures, but also possibly to a slowdown in actual deal-making and termination of deals. The latter hypothesis is supported by other research that has observed a decrease in the volume of activity (Sulle and Nelson [2009] and Locher and Sulle [2013], writing on Tanzania at two different points in time). Discrepancies between datasets affect different types of deals unevenly, so differences in aggregate scale are likely to also reflect differences in key features of land acquisition (for example, with internationally available datasets downplaying the significance of land acquisitions by national investors).

There are some caveats to note before comparing figures from our inventories to Land Matrix data. First, Land Matrix figures have changed considerably over time, reflecting a trend towards improved accuracy of the database. These revisions have resulted in a considerable downscaling of aggregate data. As of late 2011, the database included reports for over 200 million hectares worldwide, and cross-checked deals for 71 million hectares worldwide (Anseeuw et al., 2012). As of December 2012, the aggregate figure was down to 48 million hectares, and a new revision in June 2013 brought the figure to some 33 million hectares.⁹ The June 2013 version also more clearly differentiates between deals under MOU and deals under actual lease. We compare our figures to data from both the December 2012 and June 2013 versions of the Land Matrix.

Second, our inventories are limited to agricultural investments initiated between 2005 and 2012 and involving areas over 1,000 hectares; the Land Matrix aims to cover investments concluded since 2000 and with areas over 200 hectares. Also, the Land Matrix includes land acquisitions related to extractive industry, conservation and tourism developments. So the data are not strictly comparable because of the different criteria for inclusion (land size, timeframe, land use). In comparing data from our inventories and the Land Matrix, we restricted the analysis to Land Matrix deals over 1,000 hectares and concerning agricultural projects. For the June 2013 Land Matrix dataset, we limited our analysis to deals concluded during the period covered by our study (January 2005–August 2012),¹⁰ thereby achieving consistency also for the time criterion for inclusion. The vast majority of the deals in the Matrix for the three countries did refer to transactions since 2005, so this adjustment involved only minor changes to the figures. This adjustment was not possible for figures from the December 2012 Land Matrix dataset, however, which is no longer available on the Internet.

We now compare our figures to data from the December 2012 and June 2013 versions of the Land Matrix. We focus on lease (not MOU) figures for both Land Matrix and inventory datasets. However, December 2012 Land Matrix figures did not clearly differentiate between MOUs and leases, so we use the aggregate figures. In Ethiopia, the December 2012 Land Matrix figure of 2,412,362 hectares was roughly 2.4 times higher than our inventory figure of 1,055,965 hectares (a discrepancy of almost 1.4 million hectares). The June 2013 Land Matrix figure for land leases is 905,342 hectares, which seems lower than our inventory figure. However, our Ethiopia inventory includes some 325,000 hectares of land acquired for state-run sugarcane plantations, which are not included in the Matrix. So the Land Matrix figure is higher than ours once our inventory figure is adjusted accordingly (731,528 hectares).

In Ghana, the December 2012 Land Matrix figure (258,950 hectares) was 2.2 times higher than that from our dataset (113,337 hectares). The June 2013 Land Matrix figure is substantially higher – 715,776. A single deal for 400,000 hectares, included in the Matrix, accounts for a large share of this difference. It must be remembered that our Ghana inventory only contains land area data for 13 out of 28 MOUs, and that the lease area figure is extrapolated by adjusting the MOU figure for one large deal. So it is possible that our inventory particularly underestimates scale. In Tanzania, the December 2012 Land Matrix figure was nearly five times higher than our inventory figure (988,453 versus 203,347 hectares). The June 2013 Land Matrix data is substantially closer to ours: 256,105 hectares.

In Ethiopia and Tanzania the trend following the June 2013 Land Matrix revision is towards greater convergence between datasets, which suggests improving accuracy on both sides (systematic inventory exercises and Land Matrix). However, the composition of the aggregate figures can also differ significantly. In Ethiopia, aggregate figures from our inventory and from June 2013 Land Matrix figures are broadly comparable. But our inventory includes 174 deals, while the Land Matrix lists 53. This suggests that average land area sizes are smaller than what would appear from the Matrix. Also, June 2013 Land Matrix data appears to only include land acquisitions by foreign investors, while our datasets would in principle cover both foreign and domestic investors (although data for the latter is particularly difficult to come by). So discrepancies in aggregate scale continue to exist if only foreign investment is considered.

It may be useful to also compare data from our inventories to figures from other published research. Again, our figures of scale appear lower than some earlier estimates. For example, the figure for Ethiopia is broadly comparable to the 1.2 million hectares-worth of deals documented by Deininger and Byerlee (2011), who covered a different timeframe (2004–2009) and visited all regions. But it is substantially lower than the figure of 3,619,509 hectares suggested by Oakland Institute (2011), which drew on communications with government officials, published research and media reports. Similarly, our figure for Ghana is considerably lower than the estimate of over 1 million hectares for selected biofuel projects alone, suggested by Schoneveld, German, and Nutakor (2011), though this particular inventory also includes deals that had not been recorded at the Lands Commission. On the other hand, in Tanzania our figure is higher than the 50,000 hectares worth of deals documented by Deininger and Byerlee (2011) for the period 2004–2009. But Sulle and Nelson (2009), who covered biofuel developments alone, listed land applications for over 4 million hectares, MOUs for 640,000 hectares and formal rights of occupancy for 100,000 hectares up to 2009.

Other more recent studies based on in-country research and reliance on government data have also pointed to a smaller aggregate scale than had been assumed. Locke and Henley (2013) carried out inventories of land deals for biofuel projects in five countries, including Ethiopia and Tanzania. For Ethiopia, this inventory documented 575,902 hectares for the period 2007–2012, including all deals over 200 hectares. For Tanzania, it includes deals for a total of 66,211 hectares authorised between 2006 and 2012, with the smallest deal being for 163 hectares.

A different matter is what land area size figures represent in relative terms – that is, relative to existing land resources. For this calculation we use the Global Agro-Ecological Zones (GAEZ) database available on the website of the Food and Agriculture Organisation of the United Nations (FAO), which is deemed to provide a more accurate picture of ‘suitable’ land than figures for ‘arable’ land.¹¹ For each country, we consider the area of land presenting ‘moderate’ to ‘very high’ crop suitability index for low-input level rain-fed cereal farming. In Ethiopia, the area under land lease accounts for just under 2 per cent of national suitable land; in Ghana, this percentage is 0.53 per cent for land under lease and 1.9 per cent for total land transacted; and in Tanzania, land leases account for 0.25 per cent of suitable land and total land transacted accounts for 1.12 per cent of suitable land.

These figures must be taken with considerable caution, given limitations both in our inventory figures and in available national land use data.¹² But they do suggest that, despite the significant increase in the interest of private capital to invest in agriculture in these countries, national agrarian structures are unlikely to change dramatically as a result of these deals, unless deal making continues to be sustained or even accelerated in future. Population growth, rural-to-urban migration and commercial intensification in agriculture are likely to have a more substantial effect on agrarian transformations and on the prospects for small-scale agriculture.

However, the deals can still exacerbate competition for land, especially if one considers that the deals are often geographically concentrated in some regions or provinces within a country, and that pressures from the deals may be compounded by other sources of increased competition for land, including demographic change and extractive industry developments (see below).

Land size is not the only indicator of scale. The planned and actual investments are also important, given that land is allocated in order to attract private investment into the agricultural sector. These data are even more difficult to obtain on a consistent basis than acreage of land leases. We could not obtain systematic, reliable data on actual investment flows. On the other hand, we have some figures for investment plans, based on public announcements, business plans and investment proposals. Of course, promised investments do not necessarily materialise and they may be inflated to provide more justification to proposed deals.

We have a more complete dataset for Ethiopia. So, for purposes of illustration, we discuss data for Ethiopia in more detail. The most striking result is that there is huge variation in the range of investment plans. One of the largest deals in Ethiopia plans a total investment of around US$2 billion dollars over several years, while many smaller investors report plans under US$1 million for deals worth more than 1,000 hectares. These are only investment plans and subject to reporting errors. Assuming the figures are correct, the planned investment per hectare can range from over US$ 200,000, a figure closer to capital-intensive projects with large infrastructure outlays (as in floriculture, for example), to a number of cases where investors report less than US$ 1,000 per hectare – an investment that could barely cover working capital for wages and basic tools, something that is plausible in large-scale but low-input forms of agricultural production. In Ghana and Tanzania, datasets are considerably patchier. But here too, there is huge variation in the few available figures. In Tanzania, two deals reported widely different investment plans, namely US$ 500 million and US$ 50 million, for farms of about 10,000–15,000 hectares – that is, ranging between US$ 3,300 and US$ 45,000 per hectare leased.¹³

In sum, inventory results confirm that there is growing demand for land for large-scale commercial agriculture, and that the individual scale of most of these deals is significant for the standards of recipient countries. But the relative aggregate scale in relation to suitable land in each country is still small and unlikely to induce substantial transformation in national agrarian structures and land use patterns, although the deals can still increase pressures on land if they cluster in specific regions and provinces. And it is to the geographical distribution of the deals within each country that we now turn.

2.2 Geographical Distribution within Country – Some Patterns

The three countries, and especially Ethiopia and Ghana, show that the distribution of land deals is not uniform across regions. There is a significant degree of concentration of deals, especially when we consider the amount of land allocated. In Ethiopia, this follows federal policy to grant land concessions in particular regions where private investment is scarce or non-existent and where there is a perception by the central and local governments of abundant unused land (Lavers, 2012). Regions like Gambella and Southern Nations, Nationalities, and Peoples’ Region (SNNPR) in Ethiopia, broadly associated with the so-called ‘lowlands’ where demographic pressure is considered less significant, figure prominently in the ranking of target regions.

In Ethiopia, political considerations also appear to be at play in government prioritisation of certain regions as targets for land deals, because allocating a long-term lease to a large-scale operator appears to be a mechanism to consolidate government control over areas where central authority has so far been limited. Indeed, the integration of lowland regions into the national economy and the reach of the central government have been weak for a long time. Today, the Ethiopian government is taking steps to more fully integrate these peripheral regions into the national economy, and the land deals appear to be part of this strategy (Mosley, 2012).

In Ghana, two regions feature more prominently, namely the Brong Ahafo Region and the Northern region. This reflects investors’ preferences, particularly in light of agro-ecological conditions, land availability and infrastructure. Indeed, several projects are located in the Transitional and Guinea savannah zones, which together constitute the major agricultural production zone in Ghana. However, it is not possible to reach a firm conclusion about the dominance of one particular region, given reports of deals in other regions and the fact that many parts of the country are also suitable for the crops favoured by investors.

In Tanzania, the pattern shows – perhaps not surprisingly – two types of concentration: large number of deals located in regions (Morogoro and Coast) close to the capital Dar es Salaam and its main transport infrastructure (for example, harbour, airport), and some large-scale allocations located in areas where land is perceived by governments and companies to be underutilised, namely Lindi, Kigoma and Rukwa, which concentrate the biggest deals, particularly one MOU that represents almost 40 per cent of all the land reported to be under transaction in our inventory. Most of the deals in Tanzania also appear to be concentrated along the infrastructure route that is planned for upgrading through the government-led initiative on Southern Agricultural Growth Corridor of Tanzania (SAGCOT). In regions like Arusha and Manyara, where land is scarce, land appears to be only acquired through subleases of estates run by parastatals.

In sum, regional concentration of land deals is likely to be determined by different factors, depending on the crop and the nature of the investor: agro-ecological features, proximity to main transport infrastructure and hubs, and relative land availability. In Ethiopia the latter prevails, while in Ghana and Tanzania a combination of the above three factors seems to be at work.

An important insight is that, while land deals have an important national dimension, not least given the important role of national governments in enabling or promoting the deals (see below), they tend to only or mainly affect certain parts of the country. Conversely, the smaller aggregate scale of land acquisition that our inventories indicate, discussed above, needs to be placed within this context of this high spatial concentration. For equal aggregated land areas, greater spatial concentration would increase the potential of the deals to exacerbate competition for land in particular places, thereby potentially creating conflict that will require careful management. In other words, while the deals account for small percentages of suitable land in any given country, the impact of the deals on land competition is increased by their geographical concentration. This is even more so where the deals intervene in contexts where pressures on land are already increasing, as a result of diverse factors like demographic growth or mining, for example (as suggested by evidence from Ghana).

The need to take seriously the implications of the deals for pressures on land is compounded by trends in deal-making over time, and future prospects, to which we now turn.

2.3 Trends and Prospects

Generally, some acceleration is observed in the three countries, in terms of both the frequency of land deals and their aggregate land area, since the food crisis of 2007–2008. This is consistent with the picture presented in much of the literature. However, land deals are not a new phenomenon. In Ethiopia, for example, land acquisition has been ongoing since the early 2000s, partly associated with the drive to privatise state farms, partly linked to a government strategy to attract private investment (whether domestic or foreign) into agriculture in order to boost foreign exchange revenues from agricultural exports. In Ethiopia, and based on information available for 153 deals, there was a peak in land allocations in the period 2008–2010, in terms of total land area, but since then deal making appears to have significantly slowed down (see Figure 1).

Figure 1. Land deals in Ethiopia.

Source: Inventory data.

Note: timeline data available for 153 out of 174 deals.

Nonetheless, given the explicit objectives of the Growth and Transformation Plan (GTP) to devote millions of hectares to large-scale commercial agriculture, we would expect significantly more land to be allocated in future. In fact, the ambitious GTP targets for crop production in cotton, rubber, sugar and other crops, for export or to supply domestic industries, can only be achieved with a very substantial increase in the amount of land allocated to large-scale commercial agriculture. So far, and given the data of our inventory, progress has actually been relatively limited, more so if we consider how much allocated land has come under full implementation, as will be discussed below.

In Ghana, there seems to be a clearer acceleration after 2008, although the number of deals remains limited. But reports on foreign direct investment in agriculture in Ghana also show dynamic activity prior to 2007, and a phase of particularly high levels of activity in the mid-1990s, following economic and political liberalisation. Thus, actual land deals may not be a very substantial part of the stock of agriculture-related foreign direct investment in the country (which includes agroprocessing, out of the scope of this research). In Tanzania, we lack sufficient information on the year of MOU or land lease, so it is hard to establish a clear trend. Most of the land transacted appears to concentrate in the year 2006 and then tends to decrease thereafter, but data gaps do not allow a more precise assessment. Moreover, the biggest MOU, which explains the peak in 2006, is still under partial implementation and negotiation for one of the plots initially agreed.

In Tanzania, prospects for large-scale land investments look more uncertain than in Ethiopia. On the one hand, land deals in the country present a high association with biofuel development. But in recent years, there has been a drastic reduction in applications for land to develop biofuels in the country. On the other hand, ambitious government plans to modernise agriculture, including through the development of infrastructure corridors aimed at unleashing agricultural potential in currently remote areas, may well create renewed interest among private investors. In Ghana, a significant association between land deals and biofuel development also makes future prospects more uncertain than in Ethiopia.

To sum up, the period 2006–2010 witnessed a substantial increase in the number and size of land deals in the three countries, although in Ethiopia and Ghana, and especially in the former, large-scale deals preceded the 2007–2008 food crisis following an increase in land allocations since the late 1990s. The trends would generally suggest that the 2007–2008 food crisis may have given impetus to land requests but cannot be considered as the chief driving factor. In the three countries, there is an apparent slowing down of land allocations since 2010. These findings are in line with some earlier analyses, particularly an analysis of global trends based on Land Matrix data (Anseeuw et al., 2012). While prospects would suggest a continuation of large-scale land allocations in Ethiopia, future trends appear more uncertain in Ghana and Tanzania.

2.4 Who are the Land Acquirers?

A key finding of the national inventories is the prominent role of domestic investors in land acquisition. This is particularly clear in Ethiopia, which displays the largest number of deals and acreage reported. Here, more than 80 per cent of deals were by Ethiopian nationals and more than 5 per cent by a combination of Ethiopian diaspora and Ethiopian–foreign joint ventures. In terms of acreage, over 50 per cent of total land leased was for Ethiopians and diaspora (the former accounting for more than 40 per cent of total land leased). These figures exclude data concerning deals aimed at establishing state-run sugarcane plantations, which account for over 300,000 hectares. These plantations are part of a strategic drive to expand sugar production in Ethiopia, aimed at covering national consumption needs. If we include data for these deals, the percentage of domestic ownership is even higher. This finding also suggests that a significant share of land acquisition in Ethiopia is driven by state agencies, rather than private investors.

Inventory data on origin of investors in Ghana and Tanzania is very patchy, so the following results should be seen as indications only, considering the large number of deals without information on this aspect. In relation to the few deals for which we hold confirmed information, foreign investors are more dominant in Ghana, where as discussed the number of deals and estimated acreage is much smaller than in Ethiopia. And Ghanaian investors, too, play an important role, leading 4 out of the 15 deals with information on investor origin. However, in Ghana one single investor from Norway accounts for more than 80 per cent of total land transacted if we consider the MOU; this MOU has thus far resulted in a land lease for a very small fraction of the land area under MOU. In Tanzania, where we have information on country of origin for only 32 of the 64 recorded deals, foreign investors dominate. But a single US investor accounts for a large proportion of total land transacted, which may skew the picture, given missing data on many other projects. Charts based on available data are presented in Figure 2.

Figure 2. Data on country of origin of land acquirers.

Source: Inventory data – authors’ estimates.

Note: the percentages only refer to the deals for which relevant data on land transactions (MOU and/or lease) and investor origin was available (that is, subsamples of the total number of deals recorded in each country).

Among the foreign investors reported in the inventories for Ghana and Tanzania, a large proportion of deals and acreage has been allocated to investors from Europe and the United States. In both Ghana and Tanzania, investors from the United States, Norway and the United Kingdom clearly dominate the reported acreage allocated to foreign investors. In Ethiopia, on the other hand, there is relatively more variety of origins and more presence of Indian investors, especially in very large deals. Chinese investors are virtually absent from the deals included in our inventories. The role of Middle Eastern investors also appears to have been overplayed. This is even more so when account is taken of a large land deal in Ethiopia that is led by a company with a Saudi name but owned by a dual Ethiopian/Saudi national. Because of features of the investor’s profile, this deal is best seen as part of land acquisition and investment strategies by national elites, rather than as a deal led by Saudi investors. This evidence challenges widespread perceptions about the key players in land acquisition, and supports research on this issue carried out by Brautigam and Zhang (2013), specifically on China, and more generally by Cotula (2012, 2013).

2.5 Drivers: Crops and Target Markets

The relative importance of different crops varies significantly across the three countries. Evidence suggests that biofuels were an important driver in Ghana and Tanzania (where a significant number of deals are associated with biofuels, including some of the largest land allocations) but not in Ethiopia, and that the relative importance of biofuels as a driver of land deals is now decreasing. Despite global narratives that link large land deals to the policy imperative of ‘feeding the world’, evidence also suggests that overall demand for energy (biofuels) and consumer goods (reflected in land deals for agroindustrial crops like cotton) has been a leading driver of land acquisition in the three countries.

In Tanzania, extensive gaps in this part of the inventory make it impossible to draw firm conclusions on prevailing crops. But available data suggest that an early predominance of biofuels is now giving way to a wider range of crops, including food crops. This transition is linked to the failure of some large biofuel projects, and to a shift in national policy towards biofuels (namely, a moratorium on new land concessions for biofuels adopted in 2009). This finding confirms other research that pointed both to biofuels being a main driver of land acquisitions in Tanzania, at least in the period up to 2009 (Sulle & Nelson, 2009), and to the relative importance of biofuels having more recently decreased (Locher & Sulle, 2013; Locke & Henley, 2013). Similarly, in Ghana, biofuels appear to have constituted a main driver, but here too the relative importance of biofuels appears to be changing. The country study documented cases of biofuel ventures (particularly jatropha) that switched to food crops following disappointing results.

In Ethiopia, the importance of biofuels seems to be much smaller, with only one main land deal specifically dedicated to biofuel production, although some state-led deals for sugar cane cultivation may also be devoted to the production of ethanol. We could not document the significant level of biodiesel projects (nearly 576,000 hectares) identified by Locke and Henley (2013), though that study also found the relative importance of biofuels in Ethiopia to be decreasing. In Ethiopia, the main crops include ‘industrial’ crops like cotton, which, according to the patchy data in our inventory, may account for over 40 per cent of land transacted, and also edible oil crops. Food crops and cereals appear to represent a relatively lower share, around 30 per cent of allocated land.

It is also worth pointing out that the boundaries between food crops and biofuels are not neat. This is because the same crop may be used for multiple purposes (for example sugarcane, palm oil), but also because some biofuel ventures adversely affected by disappointing agronomic results and the global economic crisis have reinvented themselves with a new focus on food crops. This phenomenon has been documented, for example, by the Ghana study, as one of the projects visited experienced a transition from jatropha for biofuel to maize for food. This finding also shows that intentions on crop production stated in the original business plan may not be matched by actual production – with changes depending on profitability, the situation in domestic and international markets and supply constraints.

There is significant variation in terms of whether crops are for national distribution or for export, although export crops seem to dominate. This is not surprising, since export markets often tend to be more profitable. In Ethiopia, export orientation is an explicit government policy to generate foreign exchange through agricultural exports from commercial farming, since foreign exchange is one of the critical macroeconomic constraints. The Ethiopian government is open and proactive in terms of its drive to attract private (domestic and foreign) investment into Ethiopian agriculture and to expand large-scale commercial agriculture as a complement to its support to smallholder farming, as noted in the current Growth and Transformation Plan 2011–2015. Again, gaps in datasets, particularly in Tanzania and Ghana, make it impossible to generalise. Even in Ethiopia, the most complete of the three datasets on this point, target markets are undetermined for the vast majority of deals, though ‘100 per cent export’ entries dominate in terms of deals for which information is available.

Equally, target markets can shift over project duration. Interviews with a biofuel operator in Ghana indicated that export markets become a real option only when minimum volumes are achieved through expansion on domestic markets, as scale is necessary to offset the high transport and transaction costs. Other factors are likely to be at play: foreign companies with experience in international markets and aiming to expand supply for these markets from new production sites may have a preference for exports; some of the crops involved are essentially export crops, for example if the domestic market is limited.

Overall, it is hard to generalise from the data collected in the three countries, but evidence seems to paint a more mixed picture than the available literature usually presents. Some points are worth noting. First, intentions may not translate into actual production realities. Companies and governments may switch targets in accordance to what happens in international markets as well as domestic demand. Second, some targeted crops may turn out to be unprofitable or unviable once investors ascertain the actual production and marketing conditions. Third, many deals appear to target export markets as a result of a variety of considerations – from government policy to company profile and preferences, to shifts in the conditions that prevail in relevant markets.

3. Features and Socio-economic Outcomes

3.1 Features

Features of land deals in the three focus countries present similarities, but also important differences, reflecting diversity in national policies and political economies. Land leases, rather than purchases, predominate, accounting for virtually all documented deals in Ethiopia, Ghana and Tanzania. This finding is in line with the wider trends emerging from the literature (for example, Anseeuw et al., 2012; Cotula et al., 2009; Deininger and Byerlee, 2011; GTZ, 2009). Also in line with trends documented in the literature, the government plays a critical role in deal-making. But the extent, modalities and motivations of this role vary considerably across the three countries.

In Ethiopia, government control over deal-making appears to be particularly high, in line with a wider state-driven development model. All deals involve leases on state-owned land. This reflects provisions of the national Constitution, whereby land is nationalised. Leases have been issued by central and regional government bodies. However, reforms in 2009 have curbed the power of regional governments to issue leases above 5,000 hectares. A ‘land bank’ administered by a directorate within the Ministry of Agriculture has been established to help investors acquire rights to land that the government considers to be ‘available’. In addition, as already mentioned, about a quarter of land transacted in Ethiopia has been acquired by state entities to establish sugar plantations.

Government policy appears to be shaped by both political and economic considerations. The latter relate to a policy of promoting economic development through realising unfulfilled potential in agriculture and supporting the emergence of a large-scale, commercial farming sector alongside (rather than replacing, as often claimed) the existing small-scale sector (Lavers, 2012).¹⁴ Improving much-needed foreign exchange reserves through export promotion is a critical part of this strategy. But as discussed, geopolitical considerations are also at stake, because allocating a long-term lease to a large-scale operator appears to be a mechanism to consolidate government authority over areas where it has so far been limited. In this sense, a proactive government role in deal-making may be seen as an element of a wider strategy of state-building.

Tanzania presents a different context. Here, too, the government plays an important role. The national inventory presents gaps on key features of the leases, but the limited data available confirms the central role of government authorities in making land available to investors. However, in contrast to many other African countries, an estimated 70 per cent of the land in Tanzania is under the control of village authorities (the lowest level of local government), rather than the central government. And only an estimated 2 per cent of the country is classified as ‘general land’ – that is, the land area that is under direct management by the central government. This situation would be expected to trigger greater involvement of village authorities in deal-making.

However, findings from the Tanzania country study suggest that district authorities (a level of local government above villages) are playing an important role in deal-making, well beyond the advisory role in land administration envisaged by legislation. This is to the detriment of village authorities, who are responsible for managing village land under national law. The central government also wields considerable power in the allocation of village lands, as all land deals above 250 hectares must be approved by the Ministry of Lands. The central government has made efforts to help investors acquire land rights. For example, the Tanzania Investment Centre, which is the government agency mandated with promoting investment, sought to establish a ‘land bank’ to help investors gain access to village lands. The land bank is effectively a database of land deemed to be ‘available’ for allocation to investors, but the initiative has since lost steam. In addition, several agricultural investments involved transferring land from ‘village land’, which is managed by village authorities, to ‘general land’. The law explicitly states that investments of national interest constitute public purpose that justifies effecting this transfer on a compulsory basis.¹⁵ Some investors have also acquired land through subleases from parastatal agencies.

The context and features of deal-making in Ghana differ in important ways from both Tanzania and Ethiopia. Part of the land is owned by the state, but most belongs to customary chiefdoms, extended families and individuals. A few years ago, two Ghanaian experts estimated that 80 to 90 per cent of all undeveloped land in Ghana is held by these private entities (Kasanga & Kotey, 2001). Ghana’s Constitution recognises customary law as a source of law and the role of customary chiefs in land administration.¹⁶ Customary authorities constitute a powerful and well-organised constituency, relative to experience in other African countries. Hierarchical structures headed by paramount chiefs, assisted by traditional councils and chains of sub-chiefs, control much of the land and hold considerable power. So while inventory data are sketchy on this point and trends are impossible to quantify, many land deals in Ghana involve leases signed with customary chiefs, rather than the central government.¹⁷ The extent to which these leases are formalised through formal registration with the Lands Commission varies, with evidence suggesting that many leases are not formally recorded (see, for example, Schoneveld et al., 2011). Among the three focus countries, central government control over deal-making is least pronounced in Ghana. In fact, there is no central repository of information about land deals at government level, meaning that data are scattered across multiple agencies, and that government authorities are themselves lacking comprehensive, systematic data on land deals. This is in sharp contrast with the situation prevailing in Ethiopia.

For customary chiefs, land deals offer an opportunity to receive higher revenues than they would be able to extract from small-scale farmers. This can be a source of personal enrichment or, if the chiefs use monies for development projects, a way of consolidating the chiefly power base through patron–client networks. Also, compared to numerous small-scale tenancies, fewer, larger and more formalised transactions make it easier for the chiefs to collect revenues (Boamah, 2012). In addition, allocating large areas of land to companies can provide a means for chiefs to assert their ownership of the land: being able to lease out the land without contestation, and perhaps even with the official sanctioning of the transaction from government agencies, implies a recognition that the chief has ultimate authority over the land in question. This can be an important benefit for the chiefs, especially where customary landholdings, and the boundaries between them, are unclear or contested. In effect, allocating land to a company means staking a claim to radical title over that land (Boamah, 2012). These considerations present some parallels with the above discussion of the political economy of government-led deal-making in Ethiopia – because both government and customary authorities appear to be using, albeit in different ways, land allocation as a tool to consolidate their control over land or the national territory.¹⁸

While the three country experiences present different contexts and modalities of land deals, there is much diversity within each country too, and contexts and modalities in each country have experienced significant evolution in recent years. In Ghana, the nature of customary authorities varies considerably in different parts of the country, and in places extended families, rather than chiefdoms, control the land.¹⁹ The Ghanaian government recently adopted a new set of guidelines on agricultural investments to, among other things, discourage speculative acquisitions, protect local land rights and reassure genuine investors. In Ethiopia, federalism and devolution of land management responsibilities to regional governments have resulted in much diversity and evolutions in the ways in which regional authorities have fulfilled their responsibilities – though a trend toward convergence appears to be emerging. For example, in Afar, agreements were until recently made between clan elders and investors, with no role for regional government. However, a new regional land administration policy was adopted in 2011, and a regional land administration bureau has been set up. In Benishangul-Gumuz, the investment bureau used to allocate land to investors; now the bureau only grants a license and investors apply for land from the land bank through a new land administration bureau. As discussed, the very division of labour between federal and regional government bodies in Ethiopia has shifted, following a policy change in 2009 that recentralised authority over deal-making for leases above 5,000 hectares.

A recurring feature of deal-making relates to the multiple sites of authority involved in any given deal. Where the central government leads the action, several agencies are usually engaged. Even where the investment promotion agency acts as a central point of contact (‘one-stop shop’) for prospective investors, for example in Tanzania, this agency alone does not deal with all aspects of the land deal. Other government entities are usually involved, for instance, to approve impact assessments and issue environmental permits, and to issue land leases. In Ethiopia, a land application involves the Ethiopian Investment Agency, which issues a foreign investment licence based on pre-defined criteria; a directorate within the Ministry of Agriculture, which concludes a land lease, and government authorities responsible for approving the subsequent environmental impact assessment. In Ghana, where land leases are mainly signed by the chiefs, diverse government authorities deal with environmental permits or tax incentives, for example, and a government agency is responsible for collecting and distributing revenues paid under the leases. Lack of clear guidance or uncertainty in the procedures for acquiring land have resulted in different deals in the same country following diverse routes – as documented, for example, in Tanzania, where the process followed for acquiring land in two major deals presented irregularities vis-à-vis national law requirements.

The content of the deals presents much diversity within and across the three countries, and there is only limited scope for discerning general trends. For example, the duration of the leases varies widely, ranging from 10 to 99 years. In Ethiopia, some leases are for 10 years, many for 45 years, and data are missing for many deals. The vast majority of these leases are renewable. In Ghana, all land leases for which data are available are for a duration of 50 years, consistently with constitutional provisions setting a 50-year ceiling on leases to foreign nationals. In Tanzania, maximum durations are longer, with several deals having a documented 99-year term. The recurring use of similar durations for presumably different projects suggests that standard duration clauses are applied, reflecting limited or no alignment between contract duration and the specific economics of individual deals.

Concerns have been raised in the literature about the capacity of government agencies to scrutinise investment proposals, including business plans and impact assessments, and to monitor implementation and outcomes after the contract is concluded. Only limited data could be collected on this point as part of this study. Inventory data on social and environmental impact assessments in Ghana and Tanzania are lacking or incomplete, though the limited available data point to impact assessments of land deals having been weak or non-existent – in line with a wider literature highlighting shortcomings in the implementation of these safeguards (on Ghana, Schoneveld et al. [2011]; on Tanzania, Deininger and Byerlee [2011]). In Tanzania, the environmental impact assessment carried out for one project formed the object of controversy, including allegations that the company had doctored the impact study. In Ethiopia, none of the deals for which data are available formed the object of a duly approved social impact assessment, and only some 30 per cent of the deals in the inventory have an approved environmental impact assessment.²⁰

Anecdotal evidence suggests that federal government agencies in Ethiopia have built up capacity and experience in dealing with large-scale investors. The recent change in responsibility for allocating large land concessions in Ethiopia, meant that, in one very large land deal, the contract previously agreed by a regional government was revised by the federal government. Also, in Ethiopia new investors seeking land at the federal level need to deliver an environmental impact assessment within three months of signing a lease, and investors at regional level are also increasingly requested to do so. However, the fact that the lease is signed before the impact assessment study is conducted raises questions as to how impact assessment findings are used.

Government land allocations are usually subject to the investor’s compliance with investment plans for the first few years of the project, after which the allocation is confirmed. But in the past, African governments have rarely used this lever to hold investors to account. The wording of contracts may not be specific enough to be enforceable. Furthermore, one-off assessments at an early stage of implementation do not amount to continued monitoring and sanctioning of investment performance over a project’s lifespan. And capacity and resource limitations continue to constrain ongoing monitoring of approved projects. In addition, monitoring tends to focus on compliance with basic project indicators (for example, in Ethiopia, whether the investor has started to farm or not, and whether the land is being used for the agreed purpose), to the exclusion of wider socio-economic outcomes.

The central role of government and customary authorities in allocating land raises important issues, particularly about the extent to which these authorities take account of different local interests in land, and to the mechanisms that exist for constituents to hold authorities to account. It should be noted that, even where national law places control over land squarely in the hands of the government, investors may seek some engagement with local groups, especially local elites, to secure support for their activities. Very often, this means talking to customary chiefs and influential elders. Inventory data indicate that some deals in Ethiopia involved transactions with local chiefs. But host governments may contractually commit themselves to providing land before any consultation with local land users has taken place. Shortcomings in the quality and inclusiveness of local consultation processes have also been documented, and some land conflicts associated with land deals are ultimately rooted in failures affecting consultation processes. An important question for further research that emerges from contrasting the experiences of Ghana and Tanzania is whether, at the village level, for example, democratic frameworks (village authorities in Tanzania being the lowest level of elected local government) provide more robust pathways for citizen representation and public accountability than customary institutions or vice versa, and why (see Polack, Cotula, & Côte, 2013).

Importantly, investment decision-making and contract negotiation typically occur behind closed doors. As discussed, only a few contracts are in the public domain – though the government of Ethiopia has posted recent contracts online.²¹ This widespread lack of transparency creates a breeding ground for deals that do not maximise the public interest. More generally, opportunities for citizens to influence decision-making processes and hold decision makers to account appear limited, although a growing body of evidence points to the diverse types of citizen action through which local-to-global alliances of affected people, social movements, NGOs, diaspora associations and others are seeking accountability and justice (Wisborg [2012] on Ghana; Polack et al. [2013] covering examples from Ethiopia, Ghana and Tanzania).

Irrespective of whether the deals produce positive and/or negative socio-economic outcomes, this lack of control by potentially affected people over decisions and processes that can so profoundly impact on their lives is in itself an important source of concern. This is all the more important given the divides that may exist between the views and aspirations of the urban and rural elites in charge of government and those of more vulnerable rural people. In Ethiopia, much land allocation affects lowland areas where pastoralism is an important livelihood strategy, particularly in Afar and Gambela. Pastoralism is a way of life, not just as an economic activity, even though it may be subject to change like all ways of life. The constitution protects the rights of pastoralists, but perceptions of pastoralism as a backward and unproductive system remain widespread – despite much evidence that has proved the economic and ecological rationale of pastoralism in climatically unstable environments.²² In addition, many farmers value having control over their land and crops, rather than becoming wage labourers on somebody else’s plantation, assuming they actually have a choice and can survive on their own-account farming.

Agricultural modernisation based on large land deals may well create jobs for farm labourers – but an important question is who among rural people want to or can trade their current ways of life for those jobs. Responses are likely to be differentiated depending on socio-economic status, age, gender, ethnicity and political and ideological preferences.²³ While government authorities are the legitimate driver of national development strategies, this question highlights the need to strengthen mechanisms for public participation and accountability in decision-making processes.

3.2 Socio-economic Outcomes

Assessing the socio-economic outcomes of large land deals is riddled with difficulties. In many localities affected by the deals, no reliable baseline is available that would enable rigorous monitoring of changes in key socio-economic indicators over project duration. Many land deals are recent and are only just getting established. Many have fallen behind schedule, others are on track but their investment targets are spread over long timeframes. Implementing large investments inevitably requires time – to source the financing, build processing facilities and scale up cultivation. Also, the full impacts of an investment may only become apparent a long time after full-scale implementation begins. It is therefore necessary to distinguish between different types of impacts at different times in the lifetime of the project. Initial impressions may be too skewed and misleading. Indeed, the time distribution of costs and benefits in large-scale investments is often uneven, so looking at short-term outcomes alone may result in a skewed picture: negative impacts – loss of land, for instance – are often felt first, while jobs, opportunities for local businesses and government revenues may only fully materialise at a later stage. For many recent agricultural investments in Africa, it is just too early to tell. This is an important caveat, even more so given that evidence collected for this study suggests that implementation is still very limited, especially in Ghana and Tanzania. So any attempt to study outcomes will be severely constrained by the fact that very few outcomes can yet be assessed.

As can be seen from Figure 3, information on implementation is still patchy and limited, especially in the case of Ghana and Tanzania, where we have information on implementation for only a subset of deals, and only on the degree of implementation (unlike the Ethiopia inventory, the inventories for Ghana and Tanzania provide no information on projects that have had no implementation at all). In all countries, there are very few deals that are being fully implemented. The vast majority of deals display a degree of implementation usually limited to less than 30 per cent of the allocated land. In each of Ghana and Tanzania, only one of the deals for which we have data is being fully implemented. In Ethiopia, there is information on just above 50 per cent of all the deals counted. Of these, only 36 per cent are being fully implemented, and almost 50 per cent have gone through very limited implementation.

Figure 3. Implementation by country.

Source: Inventory data.

Note: Number of deals with information on implementation: Ethiopia N = 91 out of 174; Ghana N = 11 out of 28; Tanzania N = 7 out of 64.

This evidence confirms the findings of earlier studies, which have pointed to high failure rates and low implementation rates in the recent wave of land deals (Anseeuw et al. [2012]; Deininger and Byerlee [2011]; and specifically on Ethiopia and Tanzania, Locke and Henley [2013]). Understanding the reasons for the limited implementation would require more in-depth case studies, but anecdotal evidence from our research suggests that relevant factors can vary substantially. Some deals have been adversely affected by financing problems and underestimation of set-up costs. Other deals may be deemed insufficiently profitable and therefore suspended for some time. Some deals may be the result of pure speculation, that is no production was initially intended and companies aim to speculate on the possible value of land or may use these deals to boost their share value. Each of these scenarios has very different consequences for the future of the project and for the people living in the areas allocated to it. This situation underscores the urgent need for more rigorous monitoring systems, especially through government-led institutions, so that failed or speculative deals are cancelled and other forms of noncompliance with investment plans are sanctioned. This is not just a technical question but also a political one: weak monitoring of implementation may be function of a complex political economy of deals, with various vested interests involved.

The country studies for this project were not designed to undertake primary research on the socio-economic outcomes of land deals. But the research does provide pointers for future research on these issues. First, while the country studies generated evidence of some localised conflict and dispossession, evidence on employment dynamics is more limited. However, data suggest that some of the employment promises and expectations are not realistic. Indeed, the promise, or expectation, that plantation agriculture could bring substantial numbers of permanent and skilled jobs in the production of basic crops is inconsistent with the realities of agriculture and farm labour in developing countries. Most agricultural work, whether in self- or wage-employment, is seasonal and temporary in nature. Therefore, a more realistic approach to understanding the employment benefits of land deals is needed, in terms of job numbers (person-days of employment effectively created) and quality, and also of the long-term sustainability and security of the new jobs. It may well be that non-permanent employment of better quality compared to existing wage-labour opportunities is created, but this requires detailed evidence on the issues mentioned above. It is also important to consider possible employment losses associated with the deals, for example whether the deals displace small-scale farms and whether displaced farmers have not been allocated land elsewhere which they can put to use. Contrary to prevailing generalisations, livelihood outcomes are likely to be differentiated by groups of people, so research needs to disaggregate analysis of socio-economic outcomes to a greater extent than has so far been the case.

Second, most available studies that make any claim on the impact of land deals have taken a case study approach – assessing outcomes of a given investment project, and often based on limited field research (Oya 2013b). But to assess the socio-economic outcomes of the deals, it is critical to understand the role that their cumulative footprint is playing in transforming economies and societies at the national level – or in particular regions or districts within a country. Very little is known about the cumulative impacts, both direct and indirect, of the land deals on macroeconomic aspects like economic growth, balance of payments, tax revenue and opportunities for local businesses, and whether and to what extent these translate into benefits for the poor, including those displaced by the deals. Yet, not considering these ‘macro’ issues would result in a skewed picture of socio-economic outcomes.

Overall, the limited implementation of most land deals offers a cautionary tale on the real development potential of these deals. While the diverse reasons behind this limited implementation are not yet fully understood, it is clear that many hopes were misplaced, and that many deals have so far not lived up to their promises. In addition, reports of conflict and dispossession, albeit thus far not properly quantified, point to the fact that, even where the deals do prove viable, they will involve losers as well as winners, and that the social transformations at stake can have far-reaching and painful consequences.

4. Conclusion

This contribution aimed to test some common perceptions on land deals in Africa through research in three countries that have experienced much investor and public interest: Ethiopia, Ghana, and Tanzania. The research involved systematic data collection from relevant government ministries, cross-checked for accuracy through third-party sources, as well as limited field research to complement information on land acquisition processes and outcomes. Despite the limitations of this study, acknowledged in detail in the contribution, a number of findings provide insights for efforts to develop a system to systematically track and assess large-scale land transactions.

The study found that attempts to ascertain the precise number and size of land acquisitions are fraught with methodological difficulties, even when careful work of triangulation and verification is undertaken. These difficulties arise from various factors, including the lack of accessible, centralised information on land deals, their implementation and their outcomes; conceptual difficulties in classifying ‘approved’ deals; a fast evolving context that makes it difficult to keep track of the deals; and deliberate opacity about the deals among some public authorities and private companies.

Considering the above, it is perhaps not surprising that data from our three national inventories of land deals present discrepancies compared to widely used international datasets like the Land Matrix, both in terms of scale and composition of deals. The deals reported in our inventories generally resulted in significantly less land allocated since 2005, compared to earlier estimates based on data from alternative sources. Also, data from our national inventories highlight a much greater role of national players in land acquisition, especially in Ethiopia, while existing databases tend to focus on deals involving international investors. Our data point to a significant role of European and North American companies in land acquisition, especially in Ghana and Tanzania, and cast doubt on earlier perceptions about a key role of China and Gulf countries in land acquisition. Research from Ethiopia highlights the prominent role of Indian investors, and also of national parastatals.

Evidence collected in the three countries reveals substantial variation both between and within countries. There are considerable differences in terms of overall scale of land acquisition in the three countries, with Ethiopia containing a substantially larger number of deals and more land transacted, especially compared to Ghana. There are cross-country differences in terms of prospects as well, for example, in Ethiopia, because a sustained government drive to attract investors into the agricultural sector makes it particularly likely that deal-making will continue in the foreseeable future – despite a dip in transactions since 2009.

In terms of the drivers of land deals, evidence from the three countries paints a more complex picture than the available literature usually presents. While biofuels provided impetus in Ghana and Tanzania for a few years, other interests and markets are increasingly driving the interest of investors – not least because several biofuel ventures in the two countries have collapsed, or switched to other crops. In Ethiopia, the state is a key driver and the motivations are both political and economic, but all clearly framed within a long-term development strategy of agricultural transformation that promotes large-scale commercial farming to boost export earnings and agricultural production. Hence, in Ethiopia, a wider variety of crops is being promoted, often associated with export markets.

The country studies have also shed light on the complexity of land allocation processes, and the various layers of decision-making involved. While in Ethiopia the federal state plays a central, proactive role, in Ghana, customary authorities seem pivotal in land allocation, whereas in Tanzania, there is a mix of situations depending on the nature and scale of the deal. Therefore, the role of government in deal-making varies considerably in the three countries, as does the ability of government to control and regulate the process. In fact, even the ability of government agencies to keep track of land acquisition processes varies substantially in the three countries – with Ethiopia top of the three, and Ghana at the bottom.

The evidence about the outcomes of the deals for local livelihoods and – even more so – for national economies remains very patchy. The recent nature of many deals, coupled with constrained access to data and limited rigorous research, are important factors underpinning this situation. However, a common feature in the three countries is the limited progress with implementation of the deals. This situation creates significant opportunity costs for recipient countries, and makes it less likely that negative outcomes are offset by positive ones. While more in-depth research is needed to understand the factors underpinning this limited implementation, the finding offers a cautionary tale on the potential of large-scale land deals to contribute to poverty reduction and inclusive development.

The practical and methodological challenges constraining assessments of scale, coupled with cross-country diversity of contexts and investment processes, limit the usefulness of attempts to provide aggregate global estimates of scale for such a heterogeneous mix of land deals. In each country, the scale of land acquisition is perhaps best expressed as a range than a ‘definitive’ number, and initiatives to assess how the cumulative effects of the deals are increasing pressures in given subnational contexts would seem more promising than efforts to produce an aggregate national figure of scale. Furthermore, country- and even local-level variation calls for caution in assessing and interpreting global and regional scale on the basis of data for selected countries. It also provides a warning against rapid uptake of ‘big’, over-generalised narratives on the nature of land deals, their trends and outcomes.

The particular circumstances of the countries reviewed in this research, their different histories, polities, societies and economies also suggest that a careful political economy approach is necessary. It is crucial to understand both economic and political drivers, the role of the state (by commission or omission), the variety of interests and factions of capital (the investors), and the differential outcomes for different sections of society. Political economy analysis tools may contribute to identify similarities and differences in patterns of land acquisition, so that more context-specific narratives can be developed.

These findings provide methodological insights for informing a research agenda and for monitoring land deals more systematically and rigorously. ‘Quick and dirty’ research approaches that have so far abounded in the literature on land deals in Africa are not enough and can be counterproductive if they introduce analytical and empirical biases. More in-depth and longer-term research is necessary to fully understand the outcomes of land deals. Research could go beyond the current emphasis on case studies to examine the cumulative outcomes of the deals in given subnational contexts (for example, districts or provinces), even more so given that the deals may be geographically concentrated within a few districts or provinces in the same country. Research could also explore a fuller range of livelihood outcomes linked not only to individual deals but also to other sources of pressures on land and to the wider transformations in local and national economies of which the deals form part. Involving key local and national stakeholders, in government and beyond, throughout the research process might help to increase local ‘ownership’ of evidence and facilitate more informed policy debate on options for agricultural investment.

Baseline information is crucial to effectively assess differential outcomes in the short, medium and long term. In the absence of baseline evidence from existing sources (national-, regional- and local-level data from official statistics or previous research), new baseline information would need to be constructed, particularly in cases where land application processes are at an advanced stage (that is, before potential implementation). Rigorous, independent and publicly accessible social and environmental impact assessments can play a role in constructing solid baseline data.

Research on land deals, on the one hand, and monitoring systems aimed at promoting public awareness and mobilisation, on the other, would benefit from being more clearly separated and independent from each other. International efforts to track land deals need to be complemented by transparent national systems, not just in terms of civil society scrutiny, but also of government capacity to monitor and manage investments. From a host country perspective, national capacity for research and analysis, both within and outside government, is crucial to establish an effective and sustainable monitoring system and to properly regulate and manage the renewed interest in investing in agriculture.

Acknowledgements

This contribution draws on a multi-country research project funded by the Bill and Melinda Gates Foundation and coordinated by the International Institute for Environment and Development (IIED). The views expressed in the report do not necessarily reflect those of the institutions involved.

Notes

1. See Ghosh (2010) and Spratt (2013).

2. In Ethiopia, data were collected from the Agricultural Investment Support Directorate (AISD) in the Ministry of Agriculture and from regional offices in selected regions (see Note 3) based on a memorandum of understanding with the AISD. In Ghana, data were collected from the Ghana Investment Promotion Centre, the Ghana Free Zones Board, the Lands Commission, the Office of the Administrator of Stool Lands, the Environmental Protection Agency and the Land Title Registry; from the Ministries of Lands and Natural Resources, of Environment, Science and Technology, of Food and Agriculture, and of Local Government and Rural Development; and from the regional offices of the Lands Commission in selected regions (see Note 3). In Tanzania, data were collected from the Ministry of Lands, Housing and Human Settlement Development, the Tanzania Investment Centre and the Ministry of Livestock and Fisheries Development (in relation to the acquisition of land formerly held by a parastatal controlled by this ministry).

3. In Ethiopia, data were collected from regional government offices in Afar, Benishangul-Gumuz, Gambella, Oromia, Somali and SNNPR regions. In Ghana, data were collected from authorities in Ashanti, Brong Ahafo and Northern regions. In both countries, these regions are understood to have been the focus of much recent land acquisition.

4. In Ethiopia, the research covered one site in Gambella, two in SNNPR, two in Somali region and two in Afar; in Ghana, two sites in Brong Ahafo, one in Ashanti and one in Northern Region.

5. The country study for Ethiopia was authored by James Keeley, Abdurehman Eid, Admasu Lokaley Kidewa and Wondwosen Michago Seide; the one for Ghana by Mark Kakraba-Ampeh, Richard Owusu Asare and Emmanuel A. Codjoe; and the Tanzania study was authored by William Ole Nasha and Matteo Rizzo. Melissa Makwarimba carried out an international scoping on options for ongoing monitoring of land deals.

6. See http://www.grain.org/article/entries/4479-grain-releases-data-set-with-over-400-global-land-grabs and http://landportal.info/landmatrix.

7. See, for example, Verhoeven and Woertz (2012).

8. As discussed, an MOU originally involving 300,000 hectares has since been substantially scaled down.

9. http://landmatrix.org/.

10. We also included deals reported in the Land Matrix without date, or with a 2012 date (which could include deals concluded after our cut-off date of 31 August 2012).

11. This database is available at http://gaez.fao.org/Main.html#. Data used in this article relates to land suitability, not land availability.

12. On the latter, see Hannerz and Lotsch (2006).

13. The US$500 million project is for sugarcane and involves plans for a processing factory and power plant.

14. Indeed, one of Ethiopia’s flagship programmes for agriculture, written by the Agricultural Transformation Agency, states: ‘The Plan’s objectives focus on enhancing productivity and production of smallholder farmers and pastoralists, strengthening market systems, improving participation and engagement of the private sector, expanding the amount of land under irrigation, and reducing the number of chronically food insecure households.’ Extracted from the Agency’s website http://www.ata.gov.et/.

15. Article 4 of the Village Land Act of 1999.

16. Articles 11 and 267 of the Constitution of 1992.

17. Out of the 28 deals listed in the inventory, four leases have been granted by customary chiefs, one by government agencies, with the remainder presenting incomplete data.

18. These reflections are inspired by Lund (2011). See also work by Lavers (2012) and Mosley (2012) on Ethiopia.

19. For a comprehensive analysis of the relationships between customary and statutory tenure, as well as the institutional interactions between the state and traditional authorities, see the edited collection by Ubink and Amanor (2008).

20. Thirty-one deals, out of a total of 101 for which this data is available.

21. See http://www.moa.gov.et/node/150.

22. See Behnke and Scoones (1992); Hesse and MacGregor (2006); Thébaud (2002)

23. See Borras and Franco (2013) on differentiated political reactions ‘from below’.