Renewable and Sustainable Energy Reviews - Gasparatos Lab

Renewable and Sustainable Energy Reviews 154 (2022) 111875

1364-0321/© 2021 Elsevier Ltd. All rights reserved.

Local food security impacts of biofuel crop production in southern Africa

A. Gasparatos a,*, S. Mudombi b, B.S. Balde a, G.P. von Maltitz c, F.X. Johnson d, C. Romeu-Dalmau e, C. Jumbe f, C. Ochieng d, D. Luhanga f, A. Nyambane d, C. Rossignoli g, M. P. Jarzebski a, R. Dam Lam a, E.B. Dompreh a, K.J. Willis e

a Institute for Future Initiatives (IFI), University of Tokyo, 7-3-1 Hongo, 113-8654, Tokyo, Japan b Trade & Industrial Policy Strategies (TIPS), 234 Lange St, Pretoria, 0108, South Africa c Council for Scientific and Industrial Research (CSIR), Meiring Naude Rd., 0184, Pretoria, South Africa d Stockholm Environment Institute (SEI), Linnegatan 87D, 104 51, Stockholm, Sweden e Department of Zoology, University of Oxford, 11a Mansfield Rd, OX1 3SZ, Oxford, UK f Lilongwe University of Agriculture and Natural Resources, P.O Box 219, Lilongwe, Malawi g WorldFish, Jalan Batu Maung, 11960, Penang, Malaysia

A R T I C L E I N F O

Keywords: Bioenergy Jatropha Sugarcane Smallholders Plantations Sub-Saharan Africa Food consumption score (FCS) Household food insecurity access scale (HFIAS)

A B S T R A C T

Biofuels have been promoted as a renewable energy option in many countries, but have also faced extensive scrutiny over their sustainability. Food security is perhaps the most debated sustainability impact of biofuels, especially in regions such as Sub-Saharan Africa that experience high rates of malnutrition and have been a major destination for biofuel-related investments. This study assesses the local food security impacts of engagement in biofuel crop production using a consistent protocol between multiple crops and sites. We use standardized metrics of food security related to dietary diversity and perceptions of hunger, and focus on feedstock small-holders and plantation workers in four operational projects: a large-scale jatropha plantation (Mozambique), a smallholder-based jatropha project (Malawi) and two hybrid sugarcane projects (Malawi, Eswatini). Collectively these reflect the main feedstocks, modes of production and land use transitions related to biofuel projects in Sub- Sahara Africa. Inverse Probability Weighting analysis indicates that involvement in sugarcane production improved household food security for plantation workers and feedstock smallholders. Conversely, involvement in jatropha production does not have a statistically significant positive effect on household food security for both workers and smallholders. Regression models indicate that the factors driving food security indicator levels vary between study sites. Wealth indicators influence food security indicators in several sites, but the absolute level of income plays a smaller role, while income stability/regularity, access to credit and stable markets for selling sugarcane be important drivers as indicated by the strong effect of proxy variables on indicators.

1. Introduction

Liquid biofuels have emerged as an important renewable energy option for substituting conventional fossil fuels in the transport sector [1]. Globally, the production of liquid biofuels from food crops, ligno-cellulosic material, and waste has increased 12-fold since 1990, reaching 131,224 kt in 2019 [2]. Many countries have enacted biofuel blending mandates to accelerate the penetration of biofuels in the transport sector, with a large constellation of policies implemented to facilitate the expansion of the sector [3].

However, biofuel production, use and trade have a series of sus-tainability impacts. Some biofuel pathways have been associated with

positive sustainability outcomes related to energy security [4] and economic growth and rural development [5]. In some cases some biofuel options can have emissions saving compared to conventional fuels [6,7] and store carbon on the biomass of the feedstock and the soil [8,9], having the potential to offer larger climate-related benefits compared to other land-based mitigation options [10]. However, many types of first-generation biofuels have been linked to some negative sustain-ability impacts such as, among others, land use change and deforestation [11,12], biodiversity loss [13,14], increased GHG emissions (especially if direct and indirect land use change is factored) [15,16], carbon loss from soil [17], water overexploitation [7,12], water pollution [18,19], loss of land tenure and land-grabbing [20,21], social conflicts [22,23], gender inequality [24,25]. However, food security is perhaps the most

* Corresponding author. E-mail address: [email protected] (A. Gasparatos).

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

https://doi.org/10.1016/j.rser.2021.111875 Received 23 June 2020; Received in revised form 27 October 2021; Accepted 6 November 2021

mailto:[email protected]

www.sciencedirect.com/science/journal/13640321

https://www.elsevier.com/locate/rser

https://doi.org/10.1016/j.rser.2021.111875



http://crossmark.crossref.org/dialog/?doi=10.1016/j.rser.2021.111875&domain=pdf


2

polarizing sustainability impact of biofuels [26]. Many publications have sought to assess the local, national and international impacts of biofuels on food security [27,28], exploring different aspects of this interface such as the effects of land use change on food availability [29], the effects of income/employment generation on access to food [30], the effects of diversion of female labour diversion on nutrition [31] and the effects on food prices [32], among others. However, the actual sus-tainability impacts of biofuels (including on food security) have varied due to a large number of factors such as the feedstock, the fuel, region, the socioeconomic and environmental conditions in areas of biofuel production and use, and the underlying policies and trade regimes [28].

Sub-Saharan Africa (SSA) has been one of the parts of the globe that was particularly affected by the expansion of the biofuel sector, mainly due to the production of biofuel feedstock. Essentially, SSA was an attractive destination for domestic and foreign agricultural investments for non-food commodity crops that can be used for the production of biofuels [33]. After the mid-2000s, a large proportion of these in-vestments reflected wider market and policy expectations for biofuel crops [33]. Jatropha and sugarcane were the two most popular biofuel crops in SSA during that period, accounting for the largest fraction of large-scale land acquisitions [33,34]. Sugarcane has been steadily pro-moted and increased prominence in many SSA countries such as Malawi, Ethiopia, Mozambique, and Zambia for ethanol production [35–37], while jatropha investments materialized in many SSA countries, but the sector eventually collapsed [38,39].

Such investments sought to address a series of interrelated policy priorities in SSA related to energy security, economic growth and rural development [34]. However the actual drivers varied between crops and countries, depending on national priorities and contexts. For example, sugarcane for ethanol production was promoted in some countries such as Malawi and Ethiopia a means for enhancing national energy security and rural development [34,40]. Conversely jatropha was mostly pro-moted for exports to the EU biofuel market, but due to low cost-effectiveness it was re-oriented to domestic markets in many SSA countries before collapsing [34,38]. However, in some countries such as Malawi, jatropha was promoted from the onset as a fuel for domestic use to meet energy security and rural development objectives [39]. This suggests that the drivers, history and trajectories of biofuel investments varied substantially between SSA contexts.

Similarly, biofuel crops are grown in large-scale plantations, smallholder-based systems, or hybrid systems, adopting very diverse technical and organizational configurations [40,41]. Equally diverse between SSA contexts are the organisations promoting biofuel produc-tion, which have included government agencies, civil society, and the private sector (both convectional companies and social enterprises) [34–38,40,41]. Similarly, various factors dictate the type and magnitude of feedstock production impacts, including the type of feedstock, mode of production, local environmental and socioeconomic context, and the institutions regulating biofuel production, use and trade [34,42,43].

Similar to the global patterns discussed above, the food security impacts of biofuel feedstock production have been perhaps the most hotly debated aspect of biofuel production in SSA, considering that re-gion still experiences high malnutrition and endemic [34,44]. However,

very different mechanisms mediate the actual impact of biofuel crop production on local and household food security [26,30,45–53]. A recent meta-analysis identified no less than 25 different mechanisms mediating the food security impacts of industrial crops in SSA, with studies having reported at least 11 mechanisms for sugarcane and 20 mechanisms for jatropha [46]. Some of the most common mechanisms identified in the studies outlined above include (a) waged plantation employment or feedstock cultivation on own land that can reduce the land and labour availability for food crop production, (b) the diversifi-cation of farm output and income streams (both in terms of magnitude and stability), and (c) the better access of feedstock smallholders to technical training, fertilisers and other agricultural inputs.

As outlined above, several studies in SSA have explored different food security outcomes of biofuel crop production at the household level, but most have focused on single sites, crops, or modes of pro-duction. There is a lack of comparative studies using a consistent pro-tocol, largely because the underlying mechanisms have a combined effect on food security, complicating the elicitation of the actual impacts [45,46]. The practically complete collapse of the jatropha sector [34,38, 39] has prevented the emergence of a comprehensive evidence base about the impacts of operational jatropha projects, as most such studies focused on the early project phases.

This paper provides the first-ever comparative and consistent anal-ysis of the household-level food security outcomes of involvement in operational biofuel feedstock projects in SSA. We focus on four opera-tional projects in southern Africa that reflect the main configurations of feedstock production encountered in the region (see Methodology). We use robust direct measures of household food security, including the Food Consumption Score (FCS), the Household Food Insecurity Access Scale (HFIAS) and the Months of Inadequate Household Food Provi-sioning (MIHFP) [54]. We compare the levels of these indicators for groups with different engagement in biofuel crop production (e.g. smallholders, plantation workers) and households not engaged (control groups). Section 2 outlines the study sites and the methods for data collection and analysis. Section 3 outlines the results, and Section 4 discusses the main findings and implications.

2. Methodology

2.1. Study sites

Through an extensive literature review and multiple site visits across the region, we identified the main biofuel feedstocks (i.e. jatropha, sugarcane) and modes of production (i.e. plantation-based, smallholder- based, hybrid) [34,41]. During this preparatory phase we also identified the main factors that can affect food security such as the type/magnitude of land use change, land allocation decisions, income, and payment structure among others [34,55]. Based on this information we identified four study sites representing the main biofuel crops, production models and types of land use change encountered in southern Africa, see Ref. [56].

Fig. 1 identifies the location of the study sites and biofuel projects, and Table 1 the main characteristics of the study sites and projects. More details about the land use change patterns in each site are included in a study on carbon stock change following land use and cover change (LUCC) for biofuel crop production [50]. Below we describe in greater detail some of the main characteristics of each site and biofuel project.

The Dwangwa (Malawi) study site consists a large sugarcane plan-tation and mill operated by the multinational company Illovo. The plantation and mill are surrounded by several hundred rainfed and irrigated sugarcane smallholders that sell sugarcane directly to Illovo. The rainfed sugarcane farmers self-organise in small associations, while the irrigated farmers are part of the Dwangwa Cane Growers Limited (DCGL) [40]. Ethanol is distilled from molasses purchased through the Illovo mill by an independent company (EthCo).

The Tshaneni (Eswatini, formerly known as Swaziland) study site

Abbreviations

ATE Average Treatment Effect ATT Average Treatment Effect in the Treated FCS Food Consumption Score HFIAS Household Food Insecurity Access Scale MIHFP Months of Inadequate Household Food Provisioning IPW Inverse Probability Weighting SSA Sub-Saharan Africa

A. Gasparatos et al.


3

consists a large sugarcane plantation, two mills and an ethanol distillery operated by the Royal Swazi Sugar Company (RSSC). The Swazi gov-ernment incentivised in 1990s the local communities adjacent to the Komati River to consolidate their land and undertake sugarcane pro-duction through an extensive irrigation and rural development pro-gramme. The smallholders that consolidated land formed 28 associations that sell sugarcane to the RSSC mills, with each member becoming an equal partner and receiving annual dividends.

The Mangochi (Malawi) study site contains a few hundred jatropha smallholders cultivating jatropha on their own farms. In particular, they grow jatropha at the boundaries of their plots, as these areas are

considered to be underutilized and thus able to accommodate jatropha without compromising food crop production [57]. This model was promoted through a social enterprise (BERL) to approximately 100,000 farmers across Malawi [39], with the study farmers being some of the earliest targeted by BERL. By the time of the fieldwork, the selected farmers had already sold jatropha seeds for at least 2–3 times, receiving some level of income depending on harvest.

The Buzi (Mozambique) study site consists of a large jatropha plan-tation, which is operated by a private company (Niqel) funded from Dutch investors [39]. The plantation was established in 2007–2008, and while it was initially expected to convert 10,000 ha of miombo wood-land to a jatropha monoculture, only 1700 ha had been established by the time of the fieldwork [50]. This site does not contain any small-holder- or outgrower-jatropha production [57]. Most of the initial plans for the on-site processing of jatropha oil into fuel, fertiliser, animal feed and organic pellets was abandoned following the restructuring the company in 2016–2017.

We selected study sites where biofuel crop production was mature during field data collection (August 2014–April 2015, see below) in that the actual projects were operational for at least five years. This is because project maturity is an essential factor for the manifestation of food security effects. Both sugarcane projects were operational for more than two decades [58,59]. Both of the selected jatropha projects were operational (i.e. had not collapsed), had started their operations in 2008, and by the time of data collection they had reached maturity, having provided multiple payments to the involved smallholders and plantation workers [57]. Even though both jatropha projects collapsed in 2016–2017, this study provides a very rare insight of the impacts of fully mature jatropha projects, which is a major gap in the SSA biofuel literature. Land use change effects in each site are explained in more detail elsewhere [50].

Fig. 1. Location of study sites.

Table 1 Study groups in each study site, before and after data cleaning.

Project Respondent type Original number Final number

Dwangwa Plantation workers (Illovo) 104 102 Irrigated sugarcane farmers 104 98 Rainfed sugarcane farmers 107 96 Control group, close 104 97 Control group, far away 99 92

Tshaneni Plantation workers (RSSC) 103 97 Plantation workers (Community) 113 109 Community plantation farmers 93 85 Control group, close 101 92 Control group, far away 104 98

Buzi Plantation workers (Niqel) 98 90 Control group, close 104 94 Control group, far away 108 99

Mangochi Jatropha growers 101 96 Control group, close 101 98

Note: Final sample numbers for each study group were achieved after removing cases with missing values.



4

2.2. Data collection

We assess household-level food security impacts from engagement in biofuel feedstock through a structured household survey in the vicinity of the biofuel crop production sites described above. In each site we sampled groups with different engagement in biofuel crop production (i. e. plantation workers, irrigated/rainfed feedstock smallholders) and members of the local community not engaged in feedstock production (i. e. control group). We sampled a total of 1544 households across all sites, across groups that represent the unique characteristic of each site and project (Table 1). Below we outline sampling approach in each site, but more detail can be found elsewhere [56].

In the study sites containing large plantations we sampled two con-trol groups, (a) a control group in close proximity to the plantation and (b) a control group further away in areas that sharing similar charac-teristics. Below we provide basic sampling details, with the full infor-mation about the protocol and underlying datasets found elsewhere [56].

In Dwangwa (Malawi) three intervention groups were sampled namely (a) formal plantation workers (Illovo), (b) rainfed sugarcane smallholders (from relevant associations), and (c) irrigated sugarcane smallholders (for DCGL). One control group containing food crop farmers was also sampled within the sugarcane-growing areas, and one control group containing food crop farmers about 50 Km from the plantation. This distance is considered to be a good cut-off point after which it becomes uneconomical to supply sugarcane to the Illovo mill (sugarcane is a highly perishable crop) [56].

Permanent plantation workers below mid-level management were sampled randomly using employee lists provided by the Illovo’s HR office. Similarly rainfed and irrigated smallholders were randomly selected from lists of smallholders retrieved from the respective farmers’ associations. We used weights proportionate to the membership of each association to establish sample for each association, as a means of avoiding oversampling from any individual association. Full details about the sampling can be found elsewhere [56].

In Tshaneni (Eswatini) three intervention groups were sampled (a) formal workers in the RSSC plantation, (b) formal workers in community plantations, and (c) irrigated sugarcane smallholders (Table 1). Two control groups were also sampled following the criteria outlined for the sugarcane area in Dwangwa, Malawi (see above). Plantation workers below mid-level management were randomly selected from employee lists obtained from RSSC’s HR department. A 2-stage process was used to sample irrigated sugarcane smallholders (a) fourteen associations were randomly from the total 27 smallholder associations; (b) each associa-tion was visited to obtain lists of their constituent smallholders to select randomly respondents. We used weights proportionate to the overall membership of each association to avoid oversampling. This 2-stage sampling approach allowed for a good spread of respondent house-holds around the site, without diluting sample sizes between associa-tions. Full details about the sampling can be found elsewhere [56].

In Buzi, Mozambique we sampled formal Niqel workers as the intervention grops, and two control groups following the criteria out-lined above (Table 1). Due to its smaller size compared to the sugarcane plantations (both in terms of area and personnel) [50], workers were randomly selected from congregation areas such as the garage, picking points, and nurseries during breaks or after shifts. We sampled the close control group through transect walks from starting points that were randomly selected in the periphery of the plantation. The faraway control group was also identified through transect walks in two pre-identified sites that share similar characteristics with the area around the plantation. Due to the high forest density and very low population density in the study area [50,57], we selected every second household identified across the transect walks to achieve some degree of randomization. Full details about the sampling can be found elsewhere [56].

In Mangochi (Malawi) jatropha smallholders were the intervention

group and food crop farmers the control group. Due to the absence of a large plantation to cause positive or negative spill-over effects, a control group was not sampled in far away from the area of jatropha production as in other sites. Jatropha farmers were selected through information gathered from the lead farmers designated by BERL. This information was used to identify farmers that still sell jatropha as several jatropha farmers had stopped producing jatropha seeds. The selected households had already sold jatropha to BERL at least 2–3 times prior to the survey. Full details about the sampling can be found elsewhere [56].

2.3. Data analysis

2.3.1. Food security indicators There are multiple mechanisms that can affect household food se-

curity in sites of industrial crop and biofuel feedstock production [26,45, 46]. The study projects have been developed during different periods in time, from different organisations, and for different end goals (Section 2.1). Similarly, in each site feedstock smallholders and plantation workers are targeted and integrated in feedstock production value chains through very different approaches, ranging from strong targeting and support through government agencies for rural development (e.g. irrigated sugarcane farmers in Eswatini and Malawi), to targeting and basic support from social enterprises (e.g. jatropha smallholders in Malawi), to individual engagement through own motivation (e.g. rain-fed sugarcane farmers in Malawi, plantation workers).

The above suggest the very different motivations and approaches related to engagement in feedstock production. All of these suggest the existence of very different mechanisms through which food security outcomes emerge for those engaged in feedstock production and the surrounding communities. To avoid these complications we rely on different standardized metrics of household food security [54,60]. Despite some shortcomings [61], such standardized metrics can have large explanatory power and relevance for policy and practice [62]. In this study we use the Food Consumption Score (FCS), the Household Food Insecurity Access Scale (HFIAS) and the Months of Inadequate Household Food Provisioning (MIHFP) as indicators of food security. These measures have been used to capture complementary aspects of food security in agricultural contexts of SSA [63,64].

The FCS is a measure of dietary diversity, which estimates the fre-quency of household consumption across eight food groups (i.e., staples, pulses, vegetables, fruits, meat/fish/egg, milk, sugar, and oil) in the seven days before the survey [65]. These frequencies are aggregated and multiplied by standardized weights for each food group [65]. The cutoffs that were used for FCS are as follows: 0 to 28.4 (poor); 28.5–42 (borderline); >42 (acceptable), with high FCS indicating a higher diet diversity, as a proxy to higher food security [61].

The HFIAS elicits perceptions about hunger and the lack of food, using progressively more severe questions about their prevalence in the four weeks before the survey [66]. The HFIAS uses nine “occurrence” (i. e. yes, no) and then nine “frequency-of-occurrence” (i.e. rarely, some-times, often). The HFIAS scores range between 0 and 27, with low HFIAS scores indicating a lower prevalence of hunger, as a proxy to better food security.

Finally, we use the MIHFP as a simple proxy indicator of access to food [67]. The MIHFP indicates the number of months that the surveyed household lacked enough food to meet its needs. However, due to its simplicity and limited explanatory power [54], we report the statistical analysis, but we do not discuss it in depth.

Apart from the indicator scores we use t-tests to assess the signifi-cance of differences between groups using the svyttest function from the Survey package [68,69] in R. Boxplots are also plotted on the unadjusted sample (i.e. before IPW). The graphs are done using the ggplot2 package [70].

2.3.2. Food security determinants We examine the factors affecting food security using the generalized



5

linear regression models for the pooled samples in each study area. In particular, we regress the FCS, HFIAS, and MIFHP as the dependent variables, on a set of variables that include the different types of engagement in biofuel production (as dummy variables), and various demographic and socioeconomic characteristics that have been shown to affect household food security in rural SSA contexts (Table 2). The regressions follow Eqs. (1)–(3):

FCS= β0 + β1X1 + β2X2 + …βnXn + ε (1)

HFIAS= β0 + β1X1 + β2X2 + …βnXn + ε (2)

MIFHP= β0 + β1X1 + β2X2 + …βnXn + ε (3)

where MIFHP, FCS and HFIAS are dependent variables denoting the three food security indicators, β0 is the constant, ε is the error term, and X1 to Xn are the different covariates. Section S1 (Supplementary Elec-tronic Material) contains detailed explanation for these variables and scores for each group.

2.3.3. Inverse probability weighting We use, inverse probability weighting (IPW) as a causal inference

method [71,72]. We calculate two specific indicators, namely the average treatment effect (ATE) and the average treatment effect in the

treated (ATT). “The former denotes the average effect of treatment in an entire population/sample when that entire population/sample is moved from control to treated, while the latter denotes the average effect of treatment in those subjects who were ultimately treated” [73] (p. 1655). In our analysis, the main treatment denotes involvement in feedstock production as a plantation worker, irrigated feedstock smallholder or rainfed feedstock smallholder. An additional analysis involved assessing the causal effect of irrigation on food security, which involved the comparison of irri-gated sugarcane farmers as treatment and rainfed sugarcane farmers as the control (only in Dwangwa, Malawi). The robustness of the results was assessed using the robust standard errors and bootstrapping tech-niques [74] (Section S2; Table S29-31, Supplementary Electronic Material).

Two possible treatments are assumed for households (active treat-ment vs. control treatment), with each subject having two potential ef-fects: Yi (0) and Yi (1), which are the effects under the control and active treatment, respectively under the same conditions [75]. However, each household receives either the active (Z = 1) or no treatment (control; Z = 0) [75]. In this study we use three outcomes (FSCi = Yi, HFIASk = Yk, MIFHPj = Yj) for each household (see Eqs. (4)–(6)) that are equal to:

Yi =ZiYi(1) + (1 − Zi)Yi(0) (4)

Yk =ZkYk(1) + (1 − Zk)Yk(0) (5)

Yj =ZjYj(1) +(1 − Zj

)Yj(0) (6)

where Yi or Yk is equivalent to Yi(0) (if Zi = 0), and equivalent to Yi (1) (if Zi = 1). The effect of treatment is the difference between the two po-tential outcomes: Yi (1) - Yi (0) [75]. In this case the average treatment effect (ATE) is E [Yi (1) - Yi (0)] [71], and defined as the average effect, at the group level, of moving an entire group from the control to the treated. If Z denotes treatment assignment (Z = 1 treatment; Z =0 absence of treatment), and X is vector of observed baseline covariates, then the propensity score is defined as the probability of a subject receiving the treatment of interest conditional on their observed base-line covariates: e = P (Z = 1/X) [75]. The inverse probability of treat-ment weight (Eq. (7)) is then defined as:

W =Ze+

1 − Z1 − e

(7)

The weight of each household is equal to the inverse of the proba-bility of receiving the treatment that the subject received [75]. If the outcome variable is denoted with Y variable (Eq. (8)), the ATE is esti-mated as:

1n

∑n

i=1

ZiYi

ei−

1n

∑n

i=1

(1 − Zi)Yi

1 − ei(8)

… where n is the number of households. Another estimator of the ATE is (Eq. (9)): (∑n

i=1

Zi

ei

)∑n

i=1

ZiYi

ei−

(∑n

i=1

1 − Zi

1 − ei

)− 1∑n

i=1

(1 − Zi)Yi

1 − ei(9)

The weights described earlier are estimated as:

WATE =Ze+

1 − Z1 − e

(10)

Similarly, to the weights described earlier, this formula allows the estimation of the ATE. By using different weights it is possible to esti-mate the ATT [75] (Eq. (11)):

WATET =Z +e(1 − Z)

1 − e(11)

In observational studies it is important to define explicitly the causal effect and the focal population subset [76]. In Dwangwa (Malawi), we consider engagement in irrigated and rainfed sugarcane production and

Table 2 Definition of variables used in the analysis.

Category Variable Metric

Demographic Gender Binary (1 = male-headed; 0 = female- headed)

Age of household head Continuous (years) Household size Continuous (years) Education of household head

Ordinal (1 = no formal schooling; 2 = some primary; 3 = completed primary; 4 = some secondary; 5 =completed secondary or equivalent; 6 = completed college/pre- university/university; 7 = completed post-graduate

Education (highest level in the household)

Ordinal (1 = no formal schooling; 2 = some primary; 3 = completed primary; 4 = some secondary; 5 =completed secondary or equivalent; 6 = completed college/pre- university/university; 7 = completed post-graduate

Farming Land ownership (agricultural)

Hectares

Crop diversity - food Total number of food crops grown Fertilizer usage Kilograms Irrigation usage Binary (1 = uses irrigation, 0 =

otherwise) Asset ownership Household domestic asset index

(HDAI) Livestock ownership Tropical livestock holding (TLH)

Biofuel crop engagement

Biofuel plantation worker

Binary (1 = plantation worker, 0 =otherwise)

Biofuel farmer Binary (1 = biofuel crop farmers (sugarcane or Jatropha); 0 =otherwise)

Biofuel feedstock experience

Years of involvement in biofuel feedstock farmers or working as plantation workers

Livelihood and income

Market orientation Ratio of products sold to those produced, in energy units

Ecosystem services (food)

Summation of the frequencies of getting various food products from nature

Total borrowed Total borrowed per year (converted to US$ equivalen

Total borrowed for food

Total borrowed for food per year (converted to US$ equivalent)

Total income Total income per year (converted to US$ equivalent)



6

waged employment in the sugarcane estate (Illovo) as the involvement in the biofuel value chain. The first considers sugarcane estate worker as the treated group, and the non-sugarcane growers within the sugarcane area as the control (close control). The second considers sugarcane es-tate worker as the treated group, and the non-sugarcane growers outside the sugarcane area as the control (far away). The third compares irri-gated sugarcane farmers as the treated group, and the non-sugarcane growers within the sugarcane area as the control (close control). The fourth compares irrigated sugarcane farmers as the treated group, and the non-sugarcane growers outside the sugarcane area as the control (far away). The fifth compares irrigated sugarcane farmers as the treated group, and the non-sugarcane growers within the sugarcane area as the control (close control), sixth is with the non-sugarcane growers outside the sugarcane area as the control (far away).

In Tshaneni (Eswatini), we consider irrigated sugarcane production and waged employment in sugarcane estates as the involvement in the biofuel value chain. The first and second comparison considers waged employment in the RSSC sugarcane estate as the treated group, in relation to the close control and far away control. The third and fourth compares waged employment in the community sugarcane plantations as the treated group, to the close control and far away control. The fifth and sixth comparison is between irrigated sugarcane growers as the treated group, against the close control and far away control.

The jatropha sites do not contain coupled plantation and smallholder systems, and hence there is a small number of study groups in each area. In Buzi (Mozambique), we consider employment in the jatropha estate (Niqel) as the involvement in the biofuel value chain. Estate workers are considered as the treated group, while the close control group is the only control. This is because there is no smallholder-based biofuel production in the area (see Section 2.1). In Mangochi (Malawi), we consider engagement in smallholder jatropha production as the involvement in the biofuel value chain. The jatropha growers are the treated group, which is compared to the close control of non-jatropha farmers. This is because there is no plantation-based jatropha production in the area (Section 2.1).

The covariates used for labour group comparisons include: age of household head, gender of household head, household size, education of household head, highest education within the household, and asset ownership index (HDAI). The covariates used for farmer group com-parisons include: age of household head, gender of household head, household size, education of household head, highest education within the household, assets ownership index (HDAI), livestock ownership (TLH), and agricultural land size.

2.3.4. Data analysis steps in R programming The data handling approach entailed:

• Import data from Excel using readxl package [77]. • Organise the sub-datasets for various group comparisons in a lists of

dataframes and respective computation using the map () from purrr package [78].

For the descriptive analysis we estimate:

• Means, standard deviations, standard errors, confidence intervals through the summarySE () function from Rmisc package [79].

• Frequencies and proportions through xtabs () and prop. table () • Boxplots through geom_boxplot () in ggplot2 package [70]. • T-tests through svyttest () from survey package [69].

For the Inverse Probability Weighting we use two main R packages the WeightIt [81] and cobalt [80]. The IPW is conducted for both the ATT and the ATE for all metrics of food security. We follow the steps t outlined by Ref. [74], and summarized below:

• Examine the initial imbalance on the covariates using bal. tab () function through the cobalt package;

• Estimate weights using weightit () from the WeightIt package to establish balance on the covariates;

• Trim off extreme weights using a 0.95 level through the trim () function from the WeightIt package;

• Check if these weights manage to yield balance on the covariates using the bal. tab ();

• Apply the derived weights using svydesign () function from the survey package;

• Estimate the treatment effect using svyglm () function from the survey package;

• Design the graphs using the ggpredict () function from ggeffects package [82] and ggplot2 package [70].

The distribution of propensity scores before and after weighting adjustment was assessed using bal. plot () function from the cobalt package (Figure S3-S4, Supplementary Electronic Material). The bal-ance on covariates using absolute standardised mean differences was also assessed using the love. plot () function from cobalt package (Figure S5-6, Supplementary Electronic Material). For most of the groups the IPW improved that balance, however, some groups that balancing did not improve much, particularly in Tshaneni, for the RSSC plantation workers and the respective control groups.

An important aspect in the analysis was to assess the robustness of the results. Some authors recommend using robust standard errors, while others recommend using bootstrapping [74]. In this study both methods are applied. The output from svyglm () has robust standard errors and summ () function from jtools package [83] was used to summarise the results and get the robust confidence intervals. To esti-mate confidence intervals using the bootstrapping approach, the boot package [84] was used. A total of 2000 bootstrap replicates were run. The boot. ci () function was applied to derive the adjusted bootstrap percentile (BCa) confidence intervals [84,85]. Both the robust confi-dence intervals and the bootstrapping confidence intervals were evalu-ated at 97.5% (see Section S2, Supplementary Electronic Material). Finally, for the regression analysis we use the svyglm () from survey package. This allows for the incorporation of inverse probability weighting on data used on the regression analysis.

3. Results

3.1. Sample characteristics and food security patterns

Table 3 contains some of the main characteristics of the study groups. Overall plantation workers in the sugarcane areas have very low land ownership and fertilizer compared with the other study groups in their respective areas. Results show large diversity in fertilizer use across areas, with farmers in Dwangwa (Malawi) (especially sugarcane farmers) using very large amounts of fertilisers compared with the rest of the study groups. Crop diversity shows some variability between study groups in the two sugarcane areas, while there is almost no difference between study groups in the jatropha areas. Finally it is also worth noting that in term of assets respondents in sugarcane areas (and espe-cially those engaged in sugarcane production) report more assets than respondents in jatropha areas.

Fig. 2 presents the mean indicator scores for each study group and Fig. 3 the significance of the difference between indicator scores. The results are rather variable between sites as described below. The Sup-plementary Material contains more details including group distributions across different food security levels and detailed descriptive statistics for the study groups (Tables S1-28, Supplementary Electronic Material). Table 4 outlines the fraction of respondents falling in different levels of food insecurity in terms of the HFIAS.

In Dwangwa (Malawi) groups engaged in biofuel production gener-ally have (with few exceptions) significantly higher food security in



7

terms of the FCS, HFIAS and MIHFP scores compared to their respective control groups, with the best performers being irrigated smallholders, followed by rainfed smallholders and plantation workers (Fig. 3). Only few households involved in sugarcane production fall below the “poor” and “borderline” FCS thresholds (Table 4). Irrigated and rainfed sugar-cane growers report the lowest HFIAS scores, indicating better food security compared to other groups (e.g. 90.8% of irrigated and 76.0% of rainfed sugarcane smallholders report secure food access in terms of HFIAS) (Table 4). Surprisingly, the only contradiction was the relatively high HFIAS and MIHFP scores for plantation workers, which were not significantly different from the levels reported by control groups. Feedback received during the dissemination of these results to local stakeholders suggested that this might be due to the low wages offered to low-skilled workers, which might not be sufficient to sustain their families for the entire month. However, this is anecdotal evidence that needs to be tested in future studies.

In Tshaneni (Eswatini), workers in the large RSSC sugarcane plan-tations, workers in community plantations and irrigated sugarcane

farmers reported the highest mean FCS in this order. No members of these groups scored in the “poor” food consumption threshold (Table 4). Workers in the RSSC and community plantation also report very low levels of HFIAS compared to the control groups, but only RSSC planta-tion workers report very low levels of MIHFP compared to other groups (Figs. 2 and 3, Table 4). With the exception of the FCS, sugarcane smallholders have reported almost similar indicator levels with the control groups. These findings possibly reflect the lower reliance of RSSC plantation workers on agriculture for their livelihoods considering the much lower land ownership, fertilizer use and livestock ownership (see Table 3), in an area characterized by dry climate and poor soils (even for irrigated farmers).

In Mangochi (Malawi), jatropha smallholders and non-growers report almost similar mean FCS and HFIAS scores (Fig. 3, Table 4). Finally, in Buzi (Mozambique), though workers at the jatropha planta-tion (Niqel) report relatively higher food security in terms of the FCS, HFIAS, and the MIFHP scores than their control groups (particularly the close control), the differences are not significant (Fig. 3, Table 4).

Table 3 Summary of farming variables.

Project Respondent type Land ownership (ha)

Fertilizer (kg)

Livestock ownership (TLH)

Asset ownership (HDAI)

Crop diversity (all crops, number)

Crop diversity (only food crops, number)

Dwangwa Plantation workers (Illovo)

0.7 37 0.3 54.1 3.7 2.6

Irrigated sugarcane farmers

4.9 281 0.7 127.6 6.1 4.1

Rainfed sugarcane farmers

6.3 782 1.2 132.9 5.9 4.0

Control group, close 1.7 105 0.2 44.6 4.8 3.8 Control group, far away 3.5 145 0.5 58.2 5.6 4.5

Tshaneni Plantation workers (RSSC)

0.1 1 0.6 118.8 3.1 2.1

Plantation workers (Community)

2.6 21 3.3 116.7 3.8 2.8

Community plantation farmers

1.6 46 4.0 174.3 4.8 3.6


Buzi Plantation workers (Niqel)

3.6 0 1.6 90.3 4.9 3.9


Mangochi Jatropha growers 2.0 44 1.1 54.1 5.4 3.5 Control group, close 1.6 35 0.7 47.4 5.1 3.4

Fig. 2. Means and confidence intervals for food security indicators. Note: High FCS scores denote high levels of food security. High HFIAS and MIHFP scores denote lower levels of food security.



8

3.2. Determinants of food security indicators

Regression analysis results identify the factors that drive the levels of food security indicators in each area (Tables 5 and 6). In Dwangwa (Malawi), involvement in irrigated sugarcane production was found to be a strong predictor of the FCS (β = 10.251, p < 0.01) and HFIAS (β =− 0.826, p < 0.05) (Tables 5 and 6). Fertiliser use and income are also found to be statistically significant predictors for the FCS and HFIAS, albeit having small effect. Interestingly market orientation has a highly negative and significant effect for the FCS (β = − 10.551, p < 0.01). In Tshaneni (Eswatini), involvement in plantation work was found to be a strong predictor of the FCS (β = 9.370, p < 0.01) and HFIAS (β =

− 1.430, p < 0.1). Asset ownership, and income also have a positive and statistically significant effect on FCS and HFIAS (albeit small).

In Mangochi (Malawi), crop diversity, dependence on ecosystem services for food, and asset ownership have a positive and significant effect on FCS (β = 3.561, p < 0.05; β = 0.365, p < 0.1; β = 0.086, p <0.05 respectively). Market orientation and engagement in jatropha production have respectively a significant negative effect to food secu-rity in terms of the FCS (β = − 8.847, p < 0.05), and the HFIAS (β =2.793, p < 0.05). Factors that significantly improved the HFIAS include gender of household head (male) (β = − 2.307, p < 0.01), jatropha experience (β = − 0.384, p < 0.05), market orientation (β = − 4.011, p <0.01), assets (HDAI) (β = − 0.019, p < 0.01), and land size (agricultural)

Fig. 3. Statistical significance of differences in mean indicator scores. Note: The differences between groups are calculated after the Inverse Probability Weighing (IPW). Significance level: *** = 1%, ** = 5%, * = 10%, ns = not significant.

Table 4 Food security status of study groups.

Project Respondent type Household Food Insecurity Access Scale (HFIAS) Food Consumption Score

Secure (%)

Mildly insecure (%)

Moderately insecure (%)

Severely insecure (%)

Acceptable (%)

Borderline (%)

Poor (%)

Dwangwa Plantation workers (Illovo) 65.7% 12.7% 11.8% 9.8% 88% 10% 2% Irrigated sugarcane farmers 90.8% 7.1% 1.0% 1.0% 92% 8% 0% Rainfed sugarcane farmers 76.0% 12.5% 5.2% 6.2% 83% 16% 1% Control group, close 61.9% 12.4% 11.3% 14.4% 71% 25% 4% Control group, far away 64.1% 8.7% 14.1% 13.0% 67% 26% 7%

Tshaneni Plantation workers (RSSC) 57.7% 29.9% 12.4% 0.0% 95% 5% 0% Plantation workers (Community)

26.6% 30.3% 30.3% 12.8% 88% 12% 0%

Community plantation farmers

17.6% 30.6% 35.3% 16.5% 88% 12% 0%

Control group, close 14.1% 22.8% 30.4% 32.6% 72% 21% 8% Control group, far away 20.4% 26.5% 31.6% 21.4% 75% 22% 3%

Buzi Plantation workers (Niqel) 94.4% 1.1% 1.1% 3.3% 80% 18% 2% Control group, close 89.4% 1.1% 3.2% 6.4% 72% 16% 12% Control group, far away 89.9% 3.0% 1.0% 6.1% 84% 16% 0%

Mangochi Jatropha growers 52.1% 19.8% 8.3% 19.8% 72% 26% 2% Control group, close 56.1% 20.4% 5.1% 18.4% 84% 13% 3%



9

(β = 0.250, p < 0.01) (Tables 5 and 6). In Buzi (Mozambique), the main factors that significantly increase FCS are education (β = 3.806, p <0.01) and duration of plantation employment (β = 1.001, p < 0.1), while the total amount borrowed for food was negatively associated with the FCS (β = − 0.052, p < 0.1). Market orientation and assets ownership significantly reduce the HFIAS, but the overall effect is rather small.

3.3. Impacts of involvement in biofuel feedstock production

Figs. 4 and 5 outlines the impact of involvement in feedstock pro-duction on the main outcome variables, i.e. food security indicators. Below we report the ATT (Fig. 4) and the ATE (Fig. 5 calculated through inverse probability weighting (IPW). Section S2 and Tables S29-S34 in the Supplementary Material contains more extensive results.

In Dwangwa (Malawi), the involvement in plantation work increases the FCS by 5.9 points compared to close control (61.9 vs 56.1; p < 0.1); and by 7.3 points compared to the far away control (61.9 vs 54.7; p <0.01) (Fig. 4) (Table S29, Supplementary Material). The involvement in irrigated sugarcane can increase the FCS by 13.8 points (69.6 vs 55.9; P < 0.05) and reduce the HFIAS by 2.5 points (0.4 vs 2.9; P < 0.01) compared to the close control group. Similarly involvement in irrigated sugarcane farming increases the FCS by 15.1 points (69.6 vs 54.5; p <

0.01) and reduces HFIAS by 1.3 points (0.4 vs 1.7; p < 0.01) compared to the far control group. Involvement in rainfed sugarcane farming significantly reduces the levels of the HFIAS by 1.9 points (1.0 vs 2.8; p < 0.01) compared to the close control (Table S29-30, Supplementary Material). These findings strongly suggest that in Dwangwa, involve-ment in smallholder feedstock production greatly improves food secu-rity, while involvement in plantation work does not seem to have a strong effect. When comparing different smallholder feedstock produc-tion practices, irrigation use significantly increases the FCS by 10.7 points (69.6 vs 59.0; P < 0.01), and significantly reduces the HFIAS by 0.6 points (0.4 vs 1.0; P < 0.05), compared to rainfed production (Fig. 4) (Table S29-30, Supplementary Material).

In Tshaneni (Eswatini), involvement in irrigated sugarcane produc-tion significantly increases the FCS by 7.6 points (60.4 vs 52.8; P <0.05), and reduces the HFIAS by 2.7 points (10.1 vs 7.5, p < 0.1) compared to the close control group, though significance for the latter is low. Similarly, paid employment in sugarcane plantations has a signif-icant positive effect on food security for RSSC workers and to a lesser extent for community plantation workers. Employment at the RSSC plantation improves the levels of both food security indicators; FCS in-creases by 18.6 points (70.1 vs 51.4; p < 0.01) and HFIAS decreases by 8.6 points (1.7 vs 10.3; p < 0.01), in comparison to the close control. The

Table 5 Regression results for the Food Consumption Score (FCS).

Variables Dwangwa Sig Tshaneni Sig Buzi Sig Mangochi Sig

(Intercept) 40.26770 *** 41.76668 *** 39.50077 *** 40.67760 *** Age − 0.05317 0.00638 − 0.05574 − 0.05642 Gender 0.28627 − 0.98116 2.04835 1.24329 Education - head 2.14163 *** 1.08786 * 0.09522 0.77398 Education - highest 2.24153 ** − 0.05025 3.80578 *** 0.38890 Household size − 0.17442 − 0.56808 * − 0.26845 − 0.45285 Biofuel farmer (binary) 1.80062 0.00000 NA − 5.69128 Biofuel worker (binary) 4.18906 9.36982 *** − 2.84526 NA Biofuel experience − 0.00280 0.21474 1.00123 * 0.30807 Fertilizer amount 0.00020 ** 0.02255 0.00000 0.01872 Irrigation usage (binary) 10.25110 *** 4.48065 * NA NA Food crop diversity 0.77447 1.26776 − 0.01659 3.56106 ** Market orientation − 10.55121 *** − 0.03038 − 0.11150 − 8.84732 ** Ecosystem services 0.01228 0.04295 0.24446 * 0.36575 * TLH 0.05296 − 0.04193 − 0.13750 − 0.56617 HDAI 0.00866 0.02164 ** 0.02387 0.08603 ** Land size − 0.23369 0.07495 0.42060 − 0.52188 Total borrowed for food 0.00303 − 0.00415 − 0.05225 * − 0.01091 Total income 0.00042 *** 0.00047 *** − 0.00015 0.00033

Note: Significance level: *** = 1%, ** = 5%, * = 10.

Table 6 Regression results for the Household Food Insecurity Access Scale (HFIAS).

Variables Dwangwa Sig Tshaneni Sig Buzi Sig Mangochi Sig

(Intercept) 5.37829 *** 6.80511 *** 1.35639 7.69095 *** Age − 0.00874 0.07676 *** 0.01190 − 0.02511 Gender 0.10969 − 0.09111 0.33459 − 2.30681 *** Education - head 0.12715 − 0.37348 0.02410 − 0.15105 Education - highest − 0.79722 *** 0.15544 − 0.33063 − 0.42441 Household size 0.13443 * 0.09418 0.09326 0.32925 * Biofuel farmer (binary) − 1.26472 *** 0.00000 NA 2.79270 ** Biofuel worker (binary) 0.12373 − 1.43018 * − 0.84836 NA Biofuel experience 0.00572 − 0.12630 *** 0.11640 − 0.38362 ** Fertilizer amount − 0.00003 ** − 0.00372 0.00000 − 0.00439 Irrigation usage (binary) − 0.82583 ** 0.90646 NA NA Food crop diversity − 0.19664 − 0.51482 * − 0.11173 0.05864 Market orientation − 0.20624 0.12325 *** − 0.08991 * − 4.01089 *** Ecosystem services 0.03386 0.04861 ** 0.01342 − 0.09921 TLH 0.03041 − 0.08471 0.02922 0.00072 HDAI 0.00036 − 0.00406 * − 0.00518 ** − 0.01917 *** Land size − 0.02754 0.09720 *** − 0.02607 − 0.25019 *** Total borrowed for food 0.00381 *** 0.00126 ** 0.01858 − 0.01546 Total income − 0.00002 * − 0.00025 *** − 0.00002 0.00023

Note: Significance level: *** = 1%, ** = 5%, * = 10%.



10

positive food security outcomes are higher compared to employment in community plantations where FCS increases only by 9.4 points (60.0 vs 51.6; p < 0.01) and HFIAS decreases by 3.3 points (5.1 vs 8.4; p < 0.01). All of the above suggest the strong positive effect of involvement in sugarcane production on food security.

For the jatropha sites, the results suggest that involvement in smallholder-based production (Mangochi, Malawi) and paid plantation employment (Buzi, Mozambique) do not have a significant positive

effect on food security (Fig. 4).

4. Discussion

4.1. Synthesis of findings

Involvement in sugarcane production seems highly likely to increase household-level food security for plantation workers and feedstock

Fig. 4. ATT and bootstrapping confidence intervals for study indicators, Note: Significance level: *** = 1%, ** = 5%, * = 10%, ns = not significant: ( ) represent significant effect which is not robust based on bootstrapping confidence intervals.

Fig. 5. ATE and bootstrapping confidence intervals for study indicators, Note: Significance level: *** = 1%, ** = 5%, * = 10%, ns = not significant: ( ) represent significant effect which is not robust based on bootstrapping confidence intervals.



11

smallholders (Figs. 3 and 4). For smallholder producers the food security improvement is quite substantial in Dwangwa (Malawi) (for both irri-gated and rainfed farmers), and to a lesser extent in Tshaneni (Eswatini). This reflects other studies that have identified the positive socioeco-nomic effects of engagement in smallholder sugarcane production in these sites [e.g. 52,58,86]. Access to irrigation seems to have a sub-stantial positive effect on food security, as evidenced by the significantly better scores of irrigated farmers in Dwangwa (Malawi) compared to rainfed farmers (Fig. 4).

Waged employment in sugarcane plantations has different food se-curity outcomes in the two sites. In Tshaneni (Eswatini) plantation employment seems to have a strong significant positive effect, while in Dwangwa (Malawi) it has significant effect, however, for the close control the significance of the effect is not robust enough when considering bootstrapping confidence intervals (Fig. 4). This lack of strong food security was alluded to during expert workshops in Dwangwa, where participants outlined the generally low-paid and pre-carious working conditions in the local sugarcane plantation. This is in stark contrast with the situation in the large-scale plantation in Tshaneni (Eswatini), where workers have reported significantly better scores for both indicators.

On the other hand, engagement in jatropha production seems un-likely to improve household food security in both sites (Figs. 3 and 4). Studies in smallholder jatropha settings have identified that jatropha production in hedges is a secondary livelihood activity, which aims to provide some additional income with low land investment [42,57,87]. Although households engaged in paid employment in the Niqel planta-tion in Buzi (Mozambique) score better for both food security indicators (Fig. 3), we cannot conclude that this is due to their involvement in waged plantation work (Fig. 4). Studies have suggested that the received income from employment in jatropha plantations is often small and precarious given that jatropha projects are prone to collapse [88,89].

Depending on the site, we identify different factors driving the levels of the food security indicators but some interesting patterns emerge for some variables. In most cases, indicator levels are driven more strongly by the dummy variables that indicate the type of involvement in biofuel feedstock production (or lack of it), with this effect being more pro-nounced in the two sugarcane areas, rather than the jatropha areas (Tables 5 and 6).

At the same time, with few exceptions, all plantation worker or feedstock smallholder groups have substantially higher incomes than their control groups (Table S3 and S24, Supplementary Material). This is to be expected, as formal employment and other income-generating opportunities are quite scarce in the study areas [86]. However, despite being in many cases significant the absolute effect of income on the food security indicators tends to be low (Tables 5 and 6). On the contrary wealth indices such as domestic assets or livestock ownership (depending on the area) tend to have more significant effects to the different food security indicators in most sites.

The above suggest that it might not be the actual income levels that influence the performance of the food security indicators, but other as-pects such as income stability/regularity, access to credit, technical assistance, stable market for selling sugarcane and in-kind benefits from plantation employment [58,86,89,90]. In this sense involvement in feedstock production seems to provide a tightly linked “package of benefits” that contributes positively to food security. Such livelihood benefits have been identified in several studies of different industrial crops in SSA, including sugarcane [40,91].

In line with a priori expectations, irrigation use has a significant positive effect on the food security of irrigated farmers in Dwangwa, when compared to rainfed farmers (Figs. 3 and 4). This reflects studies that have identified the importance of irrigation for sugarcane produc-tion in SSA [40,52], but this is one of the most clear-cut example of its importance for increasing the household food security of sugarcane smallholders.

However, we have to note that despite the possible positive food

security outcomes of biofuel feedstock production at the local level, this might have differentiated impacts on food security at the national level. Although the assessment of impacts at higher levels was beyond the focus of this study, there are various publications that have explored biofuel’s food security impacts at the national level. For example studies using computable general equilibrium (CGE) models in Ghana [27], Tanzania [92,93], Malawi [94], Ethiopia [95], and Mozambique [96, 97] have found very diverse food security outcomes of biofuel produc-tion at the national level due to various interlinked factors such as crop diversion, improved access to food due to income/employment gener-ation, and increased food crop productivity through spillover effects. However, as these macro-level studies have focused only on some of the possible mechanisms mediating food security outcomes in SSA [46] and the fact that they use secondary data of (sometimes) low granularity, it is difficult to ascertain the possible food security outcomes at the national level in a truly comprehensive manner, see Ref. [27] for an exception. This makes this an important research gap (see also Section 4.2).

4.2. Policy and practice recommendations

Our results suggest that engagement in biofuel feedstock production has quite differentiated household food security outcomes depending on the crop, adopted production practices and engagement type. Generally, sugarcane workers and smallholders tend to be significantly better off in terms of food security than their control groups (Figs. 4 and 5). On the other hand, even though jatropha smallholders and plantation workers experience some economic benefits, these benefits do not necessarily materialise in improved household food security. These differentiated outcomes of biofuel crop production need to be seriously considered when developing incentives and policies to promote different biofuel feedstocks and other industrial crops in SSA. It suggests that one size- fits-all strategies are likely to be counter-productive, paying significant attention on local circumstances.

At the time of study, jatropha was a relatively new and unproven crop that subsequently largely collapsed across SSA [34,38,39]. Even in the fully operational projects we studied, the small observed benefits were most likely due to the low obtained yields and the underdeveloped market conditions. In fact, both jatropha projects used unimproved jatropha strains and had limited market options to channel their pro-duce, with the jatropha smallholders receiving little-to-no direct finan-cial benefits [57]. Eventually, despite being the last remaining operational projects in their respective countries, both jatropha projects collapsed. In this sense, policy support for untested crops such as jatropha should be conditional to dedicated longer-term efforts in the past or future to (a) improve yields, (b) cut production costs, and (c) create enabling market conditions [34]. The above would require at least many improvements in the basic production models used for jatropha in SSA (i.e. minimal input [38]) including optimal irrigation [98], targeted fertiliser use [99] and improving germplasm and/or introducing improved jatropha varieties [100]. At the same time there should be comprehensive efforts to improve large-scale land acquisition processes, as many of the past approaches were linked to loss of land tenure, social conflicts, and the eventual collapse of many jatropha projects [38].

On the other hand, for mature and tested industrial crops such as sugarcane, it is important to consider that different models of production can have widely different outcomes on farmer decisions and household food security. For example, in Dwangwa irrigated smallholders were given land and were asked to farm it, while rainfed smallholders simply re-allocated to sugarcane parts of their family farms [59]. In Tshaneni, farmers were asked to consolidate their lands for sugarcane production as a pre-condition by the government of Eswatini to develop the irri-gation infrastructure [58]. In Tshaneni and less so in Dwangwa the irrigated sugarcane smallholders have become owners and equal part-ners in these community plantations. Essentially these community plantations operate as private enterprises, with this model that receiving



12

some success in different parts of SSA [40,91]. In both areas sugarcane production is mostly irrigated, but admittedly the irrigation is rather inefficient. Improving the water use efficiency of sugarcane for example through innovations in irrigation [101] and/or improved agronomic practices [102] would allow attaining the high current yields, but at the same time reduce increasing water requirement [36], which is critical considering the effects on climate change on water resources in the re-gion and how the sugarcane sector in both Malawi and Eswatini was affected severely by the 2014–2016 El Nino event [103,104].

The nuances of the different types of involvement explained above need to be taken seriously into consideration when aiming to promote in SSA biofuel (and more generally industrial) crops as a rural development strategy with poverty alleviation potential. It would be necessary to identify innovative production models that can increase crop yields (and income/profits to producers), without putting them in a disadvanta-geous state. The focus should be to develop attractive and strong packages of technical support, institutional coordination and stable policies that could offer sustained benefits for those involved, as our results indicate that it is these packages that most likely have a strong effect on household food security. At the same time, our results also indicate that various factors affect food security, having a different effect depending on the site. As a result adopting a “one size fits all” mentality can possibly have counter-productive effect.

Furthermore, although not analysed directly in this paper, the posi-tive outcomes of feedstock production also depend on how the demand for the final products is articulated through policy frameworks. Without a sustained national or regional market, there is a real risk that biofuel feedstocks (and other industrial crops) will undergo boom-and-bust cycles, much like jatropha [34,38,39]. This may end up negatively affecting the investors and local communities involved in the production of these crops, as they invest their land, labour and agricultural inputs. Securing demand over a long period of time through dedicated inno-vation efforts would be particularly important for safeguarding the possible positive food security outcomes of involvement in feedstock production.

Finally we should note that our study has explored only one of the different sustainability impacts of biofuel production outlined in Section 1, namely food security. If biofuels are to become a truly sustainable renewable energy alternative there is a need for broader understanding not only of individual impacts, but also of their interactions. One important element would be to conduct deep impact assessment studies such as the one outlined in this paper in tandem with studies that explore the energy output and environmental performance of biofuels. Insights from techno-economic analyses [105,106], exergoeconomic analysis [107,108], exergoenvironmental analysis [109,110] exergy analysis [111,112], and different nexus approaches [12,113] can offer particu-larly valuable insights on whether biofuel production is truly sustain-able. A second important element is to understand better possible interventions that can enhance the sustainability of biofuels such as biorefineries,1 both in terms of their energy and environmental impacts, as well as in relation to socioeconomic impacts such as employ-ment/income generation and food security. For example there is a burgeoning research on sugarcane biorefineries [114–116], with very few studies from SSA (mainly from South Africa [117,118]). Similarly, research on jatropha biorefineries is much more limited with only a few mainly conceptual studies to date [119,120], with the research from SSA being practically non-existent at this stage. Such research gaps would be necessary to be tackled in future studies if biofuels are to become a truly sustainable renewable energy alternative in SSA.

5. Conclusions

In this study we explored the local food security impacts of engagement in biofuel crop production in southern Africa. We explore these outcomes for different types of engagement (i.e. smallholders, plantation workers) in four sites of biofuel crop production in Malawi, Mozambique and Eswatini, which reflect the main biofuel crops and modes and production options in southern Africa (and SSA more broadly). The results suggest that the food security outcomes are usually better in sugarcane projects rather than jatropha projects. However there is high heterogeneity between projects, and also between groups within the same project.

Such differentiated outcomes should be seriously considered when developing incentives and policies to promote different biofuel feed-stocks in SSA. When it comes to jatropha projects, there should be co-ordinated efforts to improve thoroughly the production models moving beyond minimal input models, for example by including optimal irri-gation, targeted fertiliser use, improving germplasm and/or introducing improved jatropha varieties. These should be complemented through the creation of strong markets, and improving large-scale land acquisi-tion processes, as a means of avoiding negative impacts linked to the loss of land tenure, the emergence of social conflicts, and the eventual collapse of jatropha projects. Conversely, for sugarcane projects there should be efforts to improve water use efficiency, possibly through irrigation innovations and/or improved agronomic practices. These will be critical to ensure the currently high sugarcane yields without over-exploiting scarce water resources, which beyond environmental benefits would reduce the vulnerability of sugarcane production to droughts.

However, such interventions should reflect the prevailing agroeco-logical, socioeconomic and institutional context as one size-fits-all strategies are likely to be counter-productive. By paying significant attention to the local contexts and the needs there is a higher chance of ensuring that biofuel production can fulfill its intended development potential without compromising local food security.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This work was supported by the Ecosystem Services for Poverty Alleviation Program (ESPA; grant number: NE/L001373/1). AG, BSB, MPJ and RDL also acknowledge the support of the Japan Science and Technology Agency (JST) for the Belmont Forum project FICESSA.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.rser.2021.111875.

Credit author statement

Alexandros Gasparatos: Conceptualization, Methodology, Writing – original draft, Supervision, Project administration, Funding acquisition. Shakespear Mudombi: Methodology, Formal analysis, Investigation, Data curation, Writing – review & editing, Visualization. Boubacar Balde: Methodology, Formal analysis, Data curation, Writing – review & editing. Graham von Maltitz: Conceptualization, Writing – review & editing, Supervision, Funding acquisition. Francis X. Johnson: Concep-tualization, Writing – review & editing, Supervision, Funding acquisi-tion. Carla Romeu-Dalmau: Conceptualization, Writing – review & editing, Supervision. Charles Jumber: Conceptualization, Writing – re-view & editing, Supervision. Caroline Ochieng: Conceptualization,

1 Biorefinery can be defined as the “the optimised use of biomass for mate-rials, chemicals, fuels and energy applications, where use relates to costs, economics, markets, yield, environment, impact, carbon balance and social aspects” [121].





13

Writing – review & editing, Supervision. Davies Luhanga: Methodology, Investigation, Data curation, Writing – review & editing. Anne Nyam-bane: Investigation, Data curation, Writing – review & editing. Cristiano Rossignoli: Methodology, Writing – review & editing. Marcin Jarzebski: Investigation, Data curation, Writing – review & editing, Visualization. Rodolfo Dam Lam: Methodology, Writing – review & editing. Eric Brako Dompreh: Methodology, Writing – review & editing. Kathy J. Willis: Conceptualization, Writing – review & editing, Supervision, Funding acquisition.

References

[1] Worldwatch Institute. Biofuels for transport: global potential and implications for sustainable energy and agriculture. first ed. London: Routledge; 2007.

[2] IEA. Data and statistics. Paris: International Energy Agency (IEA); 2021. https ://www.iea.org/data-and-statistics. [Accessed 25 October 2021].

[3] Renewables REN21. Global status report. Paris: REN21; 2021. [4] IRENA. Boosting biofuels: sustainable paths to greater energy security. Abu

Dhabi: International Renewable Energy Agency (IRENA); 2016. [5] FAO IDB. Biofuels and rural economic development in Latin America and the

Caribbean. Food and agriculture organisation. Rome and Washington DC: FAO) and Inter-American Development Bank (IDB); 2010.

[6] El Akkari M, Rechauchere O, Bispo A, Gabrielle B, Makowski D. A meta-analysis of the greenhouse gas abatement of bioenergy factoring in land use changes. Sci Rep 2018;8:8563.

[7] Jeswani HK, Chilvers A, Azapagic A. Environmental sustainability of biofuels: a review. Proc. R. Soc. A 2010;476:20200351.

[8] Meijide A, de la Rua C, Guillaume T, Roll A, Hassler E, Stiegler C, Tjoa A, June T, Corre MD, Veldkamp E, Knohl A. Measured greenhouse gas budgets challenge emission savings from palm-oil biodiesel. Nat Commun 2020;11:1089.

[9] Yang Y, Tilman D. Soil and root carbon storage is key to climate benefits of bioenergy crops. Biofuel Res J 2020;26:1143–8.

[10] Field JL, Richard TL, Smithwick EAH, Cai H, Laser MS, LeBauer DS, Long SP, Paustian K, Qin Z, Sheehan JJ, Smith P, Wang MQ, Lynd LR. Robust paths to net greenhouse gas mitigation and negative emissions via advanced biofuels. Proc Natl Acad Sci Unit States Am 2020;117:21968–77.

[11] Gibbs HK, Ruesch AS, Achard F, Clayton MK, Holmgren P, Ramankutty N, Foley JA. Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proc Natl Acad Sci Unit States Am 2010;107:16732–7.

[12] Rulli M, Bellomi D, Cazzoli A, De Carolis G, D’Odorico P. The water-land-food nexus of first-generation biofuels. Sci Rep 2016;6:22521. https://doi.org/ 10.1038/srep22521.

[13] Filoso S, do Carmo JB, Mardegan SF, Lins SRM, Gomes TF, Martinelli LA. Reassessing the environmental impacts of sugarcane ethanol production in Brazil to help meet sustainability goals. Renew Sustain Energy Rev 2015;52:1847–56.

[14] Savilaakso S, Garcia C, Garcia-Ulloa J, Ghazoul J, Groom M, Guariguata MR, Laumonier Y, Nasi R, Petrokofsky G, Snaddon J, Zrust M. Systematic review of effects on biodiversity from oil palm production. Environ Evid 2014;3:1–20.

[15] Achten WMJ, Verchot LV. Implications of biodiesel-induced land-use changes for CO2 emissions: case studies in tropical America, Africa, and Southeast Asia. Ecol Soc 2011;16:14.

[16] Fargione J, Hill J, Tilman D, Polasky S, Hawthorne P. Land clearing and the biofuel carbon debt. Science 2008;319:1235–8.

[17] Harris ZM, Spake R, Taylor G. Land use change to bioenergy: a meta-analysis of soil carbon and GHG emissions. Biomass Bioenergy 2015;82:27–39.

[18] Alshawaf M, Douglas E, Ricciardi K. Estimating nitrogen load resulting from biofuel mandates. Int J Environ Res Publ Health 2016;13:478.

[19] Bordonal RdO, Carvalho JLN, Lal R, de Figueiredo ED, de Oliveira BG, La Scala Jr N. Sustainability of sugarcane production in Brazil. A review. Agron. Sustain. Dev. 2018;38:13.

[20] FAO. Bioenergy and land tenure - the implications of biofuels for land tenure and land policy. Rome: Food and Agriculture Organisation (FAO); 2008.

[21] Borras Jr SM, Hall R, Scoones I, White B, Wolford W. Towards a better understanding of global land grabbing: an editorial introduction. J Peasant Stud 2011;38:209–16.

[22] Mingorría S. Violence and visibility in oil palm and sugarcane conflicts: the case of Polochic Valley, Guatemala. J Peasant Stud 2017;45:1314–40.

[23] Ahmed A, Abubakari Z, Gasparatos A. Labelling large-scale land acquisitions as land grabbing: procedural and distributional considerations from two cases in Ghana. Geoforum 2019;105:191–205.

[24] Potter L. Colombia’s oil palm development in times of war and ‘peace’: myths, enablers and the disparate realities of land control. J Rural Stud 2020;78: 491–502.

[25] Fonjong L. Interrogating large-scale land acquisition and its implications for women’s land rights in Cameroon, Ghana and Uganda. Ottawa: International Development Research Centre (IDRC); 2017.

[26] Kline KL, Msangi S, Dale VH, Woods J, Souza GM, Osseweijer P, Clancy JS, Hilbert JA, Johnson FX, McDonnell PC, Mugera HK. Reconciling food security and bioenergy: priorities for action. GCB Bioenergy 2017;9:557–76.

[27] Brinkman M, Levin-Koopman J, Wicke B, Shutes L, Kuiper M, Faaij A, der Hilst F. The distribution of food security impacts of biofuels, a Ghana case study. Biomass Bioenergy 2020;141:105695.

[28] Ahmed A, Jarzebski MP, Gasparatos A. Industrial crops as agents of transformation: justifying a political ecology lens. In: Ahmed A, Gasparatos A, editors. Political ecology of industrial crops. London: Routledge; 2021. p. 3–24.

[29] Hervas A, Isakson SR. Commercial agriculture for food security? The case of oil palm development in northern Guatemala. Food Sec 2020;12:517–35.

[30] Bosch C, Zeller M. Large-scale biofuel production and food security of smallholders: evidence from Jatropha in Madagascar. Food Sec 2019;11:431–45.

[31] Mingorría S, Gamboa G, Martín-Lopez B, Corbera E. The oil palm boom: socioeconomic implications for Q’eqchi’ households in the Polochic valley, Guatemala. Environ Dev Sustain 2014;16:841–71.

[32] Malins C. Thought for food: a review of the interaction between biofuel consumption and food markets. London: Cerulogy; 2017.

[33] Schoneveld GC. The geographic and sectoral patterns of large-scale farmland investments in sub-Saharan Africa. Food Pol 2014;48:34–50.

[34] Gasparatos A, von Maltitz G, Johnson FX, Lee L, Mathai M, Puppim de Oliveira J, Willis K. Biofuels in sub-Sahara Africa: drivers, impacts and priority policy areas. Renew Sustain Energy Rev 2015;45:879–901.

[35] Hodbod J, Tomei J, Blaber-Wegg TA. Comparative analysis of the equity outcomes in three sugarcane–ethanol systems. J Environ Dev 2015;24:211–36.

[36] Hess TM, Sumberg J, Biggs T, Georgescu M, Haro-Monteagudo D, Jewitt G, Ozdogan M, Marshall M, Thenkabail P, Daccache A, Marin F, Knox JW. A sweet deal? Sugarcane, water and agricultural transformation in Sub-Saharan Africa. Global Environ Change 2016;39:181–94.

[37] Ngcobo S, Jewitt G. Multiscale drivers of sugarcane expansion and impacts on water resources in Southern Africa. Environ Dev 2017;24:63–76.

[38] Ahmed A, Campion B, Gasparatos A. Towards a classification of the drivers of jatropha collapse in Ghana elicited from the perceptions of multiple stakeholders. Sustain Sci 2019;14:315–39.

[39] von Maltitz G, Gasparatos A, Fabricius C. The rise, decline and future resilience benefits of jatropha in southern Africa. Sustainability 2014;6:3615–43.

[40] von Maltitz GP, Henley G, Ogg M, Samboko PC, Gasparatos A, Read M, Engelbrecht F, Ahmed A. Institutional arrangements of outgrower sugarcane production in southern Africa. Dev So Afr 2019;36:175–97.

[41] von Maltitz GP, Setzkorn KA. A typology of Southern African biofuel feedstock production projects. Biomass Bioenergy 2013;59:33–49.

[42] van Eijck J, Romijn H, Smeets E, Bailis R, Rooijakkers M, Hooijkaas N, Verweij P, Faaij A. Comparative analysis of key socio-economic and environmental impacts of smallholder and plantation based jatropha biofuel production systems in Tanzania. Biomass Bioenergy 2014;61:25–45.

[43] Romijn H, Heijnen S, Rom Colthoff J, De Jong B, Van Eijck J. Economic and social sustainability performance of jatropha projects: results from field surveys in Mozambique, Tanzania and Mali. Sustainability 2014;6:6203–35.

[44] Rosillo-Calle F, Johnson FX. Food versus fuel: an informed introduction to biofuels. London: ZED Books; 2011.

[45] Wiggins S, Henley G, Keats S. Competitive or complementary? Industrial crops and food security in sub-Saharan Africa. London: Overseas Development Institute; 2015.

[46] Jarzebski MP, Ahmed A, Boafo YA, Balde BS, Chinangwa L, Saito O, von Maltitz G, Gasparatos A. Food security impacts of industrial crop production in sub-Saharan Africa: a systematic review of the impact mechanisms. Food Security 2020;12:105–35.

[47] Kidula-Lihasi L, Onyango C, Ochola W. Analysis of smallholder sugarcane farmers’ livelihood assets in relation to food security in Mumias Sub- County Kenya. J Econ Sustain Dev 2016;7:40–7.

[48] Dam Lam R, Boafo Y, Degefa S, Gasparatos A, Saito O. Assessing the food security outcomes of industrial crop expansion in smallholder settings: insights from cotton production in Northern Ghana and sugarcane production in Central Ethiopia. Sustain Sci 2017;12:677–93.

[49] Mwavu EN, Kalema VK, Bateganya F, Byakagaba P, Waiswa D, Enuru T, Mbogga MS. Expansion of commercial sugarcane cultivation among smallholder farmers in Uganda: implications for household food security. Land 2018;7:73.

[50] Romeu-Dalmau C, Gasparatos A, von Maltitz G, Graham A, Almagro-Garcia J, Wilebore B, Willis KJ. Impacts of land use change due to biofuel crops on climate regulation services: five case studies in Malawi, Mozambique and Swaziland. Biomass Bioenergy 2018;114:30–40.

[51] Wendimu MA, Henningsen A, Gibbon P. Sugarcane outgrowers in Ethiopia: “Forced” to remain poor? World Dev 2016;83:84–97.

[52] Herrmann R, Jumbe C, Bruentrup M, Osabuohien E. Competition between biofuel feedstock and food production: empirical evidence from sugarcane outgrower settings in Malawi. Biomass Bioenergy 2018;114:100–11.

[53] Timko JA, Amsalu A, Acheampong E, Teferi MK. Local perceptions about the effects of Jatropha (Jatropha curcas) and castor (Ricinus communis) plantations on households in Ghana and Ethiopia. Sustainability 2014;6:7224–41.

[54] Carletto C, Zezza A, Banerjee R. Towards better measurement of household food security: harmonizing indicators and the role of household surveys. Glob Food Secur 2013;2:30–40.

[55] Gasparatos A, Romeu-Dalmau C, von Maltitz G, Johnson FX, Shackleton C, Jarzebski MP, Jumbe C, Ochieng C, Mudombi S, Nyambane A, Willis KJ. Mechanisms and indicators for assessing the impact of biofuel feedstock production on ecosystem services. Biomass Bioenergy 2018;114:157–73.

[56] Gasparatos A, von Maltitz G, Johnson FX, Romeu-Dalmau C, Jumbe C, Ochieng C, Mudombi S, Balde BS, Luhanga D, Nyambane A, Lopes P, Jarzebski M, Willis KJ. Survey of local impacts of biofuel crop production and adoption of ethanol stoves in southern Africa. Scientific Data 2018;5:180186.


http://refhub.elsevier.com/S1364-0321(21)01142-4/sref1


https://www.iea.org/data-and-statistics

https://www.iea.org/data-and-statistics
























https://doi.org/10.1038/srep22521

https://doi.org/10.1038/srep22521






























































































































14

[57] von Maltitz GP, Gasparatos A, Fabricius C, Morris A, Willis K. Jatropha cultivation in Malawi and Mozambique: impact on ecosystem services, local human well-being, and poverty alleviation. Ecol Soc 2015;21:3.

[58] Terry A, Ogg M. Restructuring the Swazi sugar industry: the changing role and political significance of smallholders. J South Afr Stud 2016;43:585–603.

[59] Chinsinga B. The green belt initiative, politics and sugar production in Malawi. J South Afr Stud 2017;43:501–15.

[60] Andrew D, Ngure FM, Pelto G, Young SL. What are we assessing when we measure food security? A compendium and review of current metrics. Adv Nutr 2013;4:481–505.

[61] Leroy JL, Ruel M, Frongillo EA, Harris J, Ballard TJ. Measuring the food access dimension of food security: a critical review and mapping of indicators. Food Nutr Bull 2015;36:167–95.

[62] Perez-Escamilla R, Gubert MB, Rogers B, Hromi-Fiedler A. Food security measurement and governance: assessment of the usefulness of diverse food insecurity indicators for policy makers. Glob Food Secur 2017;14:96–104.

[63] Anderman TL, Remans R, Wood SA, DeRosa K, DeFries RS. Synergies and tradeoffs between cash crop production and food security: a case study in rural Ghana. Food Sec 2014;6:541–54.

[64] Nsabuwera V, Hedt-Gauthier B, Khogali M, Edginton M, Hinderaker SG, Nisingizwe MP, Tihabyona Jde D, Sikubwabo B, Sembagare S, Habinshuti A, Drobac P. Making progress towards food security: evidence from an intervention in three rural districts of Rwanda. Publ Health Nutr 2016;19:1296–304.

[65] WFP. Food consumption analysis: calculation and use of the food consumption score in food security analysis. Rome: World Food Programme (WFP); 2009.

[66] Coates J, Swindale A, Bilinsky P. Household food insecurity access scale (HFIAS) for measurement of food access: indicator guide. Washington DC: USAID; 2007.

[67] Bilinsky P, Swindale A. Months of adequate household food provisioning (MAHFP) for measurement of household food access: indicator guide. Washington D.C: USAID; 2010.

[68] Lumley T. Analysis of complex survey samples. J Stat Software 2004;9:1–19. [69] Lumley T. Survey: analysis of complex survey samples. 2019. https://rdrr.io/r

forge/survey/. [Accessed 23 June 2020]. [70] Wickham H. Ggplot2: elegant graphics for data analysis. New York: Springer-

Verlag; 2016. [71] Imbens GW. Nonparametric estimation of average treatment effects under

exogeneity: a review. Rev Econ Stat 2004;86:4–29. [72] Morgan SL, Winship C. Counterfactuals and causal inference: methods and

principles for social research. New York: Cambridge University Press; 2007. [73] Austin PC, Stuart EA. The performance of inverse probability of treatment

weighting and full matching on the propensity score in the presence of model misspecification when estimating the effect of treatment on survival outcomes. Stat Methods Med Res 2017;26:1654–70.

[74] Greifer N. A guide to using WeightIt for estimating balancing weights. 2020. http s://cran.r-project.org/web/packages/WeightIt/vignettes/WeightIt_A0_basic_use. html. [Accessed 23 June 2020].

[75] Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med 2015;34:3661–79.

[76] Hernan MA, Robins JM. Causal inference: what if. London: Taylor & Francis Group; 2020.

[77] Wickham H, Bryan J. Readxl: read Excel files. 2019. https://CRAN.R-project. org/package=readxl. [Accessed 23 June 2020].

[78] Henry L, Wickham H. Purrr: functional programming tools. 2019. https://CRAN. R-project.org/package=purrr. [Accessed 23 June 2020].

[79] Hope RM. Rmisc: Rmisc: Ryan miscellaneous. 2013. https://CRAN.R-project. org/package=Rmisc. [Accessed 23 June 2020].

[80] Greifer N. Cobalt: covariate balance tables and plots. R package version 4.0.0. 2020. https://CRAN.R-project.org/package=cobalt. [Accessed 23 June 2020].

[81] Greifer N. WeightIt: weighting for covariate balance in observational studies. 2020. https://CRAN.R-project.org/package=WeightIt. [Accessed 23 June 2020].

[82] Lüdecke D. Ggeffects: tidy data frames of marginal effects from regression models. J. Open Source Softw 2018;3:772.

[83] Long JA. Jtools: analysis and presentation of social scientific data. 2019. https://c ran.r-project.org/package=jtools. [Accessed 23 June 2020].

[84] Canty A, Ripley BD. Boot: bootstrap R (S-plus) functions. 2019. https://cran. r-project.org/web/packages/boot/citation.html. [Accessed 23 June 2020].

[85] Davison AC, Hinkley DV. Bootstrap methods and their applications. Cambridge: Cambridge University Press; 1997.

[86] Mudombi S, von Maltitz GP, Gasparatos A, Romeu-Dalmau C, Johnson FX, Jumbe C, Ochieng C, Luhanga D, Lopes P, Balde BS, Willis KJ. Multi-dimensional poverty effects around operational biofuel projects in Malawi, Mozambique and Swaziland. Biomass Bioenergy 2018;114:41–54.

[87] Mogaka V, Ehrensperger A, Iiyama M, Birtel M, Heim E, Gmuender S. Understanding the underlying mechanisms of recent Jatropha curcas L. adoption by smallholders in Kenya: a rural livelihood assessment in Bondo, Kibwezi, and Kwale districts. Energ Sustain Dev 2014;18:9–15.

[88] Acheampong E, Campion BB. The effects of biofuel feedstock production on farmers’ livelihoods in Ghana: the case of Jatropha curcas. Sustainability 2014;6: 4587–607.

[89] Schoneveld GC, German LA, Nutakor E. Land-based investments for rural development? A grounded analysis of the local impacts of biofuel feedstock plantations in Ghana. Ecol Soc 2011;16:10.

[90] Matenga CR, Hichaambwa M. Impacts of land and agricultural commercialisation on local livelihoods in Zambia: evidence from three models. J Peasant Stud 2017; 44:574–93.

[91] Matenga CR. Outgrowers and livelihoods: the case of Magobbo smallholder block farming in Mazabuka district in Zambia. J South Afr Stud 2017;43:551–66.

[92] Thurlow J, Branca G, Felix E, Maltsoglou I, Rincon LE. Producing biofuels in low- income countries: an integrated environmental and economic assessment for Tanzania. Environ Resour Econ 2016;64:153–71.

[93] Arndt C, Farmer W, Strzepek K, Thurlow J. Climate change, agriculture and food security in Tanzania. Rev Dev Econ 2012;16:378–93.

[94] Schuenemann F, Thurlow J, Zeller M. Leveling the field for biofuels: comparing the economic and environmental impacts of biofuel and other export crops in Malawi. Agric Econ 2017;48:301–15.

[95] Gebreegziabher Z, Mekonnen A, Ferede T, Guta F, Levin J, Kohlin G, Alemu T, Bohlin L. The distributive effect and food security implications of biofuels investment in Ethiopia. A CGE analysis. In: Berck CS, Berck P, Di Falco S, editors. Agricultural adaptation to climate change in Africa. London: Routledge; 2018. p. 31.

[96] Arndt C, Benfica R, Tarp F, Thurlow J, Uaiene R. Biofuels, poverty, and growth: a computable general equilibrium analysis of Mozambique. Environ Dev Econ 2010;15:81–105.

[97] Hartley F, van Seventer D, Tostao E, Arndt C. Economic impacts of developing a biofuel industry in Mozambique. Dev South Afr 2019;36:233–49.

[98] Vaknin Y, Yermiyahu U, Bar-Tal A, Samocha Y. Global maximization of Jatropha oil production under semi-arid conditions by balancing vegetative growth with reproductive capacity. GCB Bioenergy 2018;10:382–92.

[99] Yong JWH, Ng YF, Tan SN, Chew AYL. Effect of fertilizer application on photosynthesis and oil yield of Jatropha curcas L. Photosynthetica 2010;48: 208–18.

[100] Yi C, Reddy C, Varghese K, Bui TNH, Zhang S, Kallath M, Kunjachen B, Ramachandran S, Hong Y. A new Jatropha curcas variety (JO S2) with improved seed productivity. Sustainability 2014;6:4355–68.

[101] Froebrich J, Bouarfa S, Rollin D, Coulon C, Belaud G, editors. Special issue: innovations in irrigation systems in Africa. Irrigation and Drainage 2009, vol. 69; 2021. p. 1–208.

[102] Olivier FC, Singels A. Increasing water use efficiency of irrigated sugarcane production in South Africa through better agronomic practices. Field Crop Res 2015;176:87–98.

[103] Henriksson R, Vincent K, Naidoo K. Exploring the adaptive capacity of sugarcane contract farming schemes in the face of extreme events. Frontiers in Climate 2021;3:1–17.

[104] Mhlanga-Ndlovu BFN, Nhamo G. An assessment of Swaziland sugarcane farmer associations’ vulnerability to climate change. J Integr Environ Sci 2017;14:39–57.

[105] Mandegari M, Farzad S, Gorgens J. Recent trends on techno-economic assessment (TEA) of sugarcane biorefineries. Biofuel Res J 2017;4:704–12.

[106] Scown CD, Baral NR, Yang M, Vora N, Huntington T. Technoeconomic analysis for biofuels and bioproducts. Curr Opin Biotechnol 2021;67:58–64.

[107] Soltanian S, Aghbashlo M, Farzad S, Tabatabaei M, Mandegari M, Gorgens JF. Exergoeconomic analysis of lactic acid and power cogeneration from sugarcane residues through a biorefinery approach. Renew Energy 2019;143:872–89.

[108] Julio AAV, Batlle EAO, Rodriguez CJC, Palacia JCE. Exergoeconomic and environmental analysis of a palm oil biorefinery for the production of bio-jet fuel. Waste Biomass Valor 2021;12:5611–37.

[109] Cavalcanti EJC, Carvalho M, Ochoa AAV. Exergoeconomic and exergoenvironmental comparison of diesel-biodiesel blends in a direct injection engine at variable loads. Energy Convers Manag 2019;183:450–61.

[110] Ochoa-Garcia M, Tejeda-Lopez L, Ojeda-Delgado K, Gonzalez-Delgado A, Sanchez-Tuiran E. Computer-aided exergo-environmental analysis of biodiesel production process from microalgal biomass. Contemporary Eng Sci 2019;11: 2201–9.

[111] Aghbashlo M, Mandegari M, Tabatabaei M, Farzad S, Soufiyan MM, Gorgens JF. Exergy analysis of a lignocellulosic-based biorefinery annexed to a sugarcane mill for simultaneous lactic acid and electricity production. Energy 2018;149:623–38.

[112] Talens L, Villalba G, Gabarrell X. Exergy analysis applied to biodiesel production. Resour Conserv Recycl 2007;51:397–407.

[113] Alherbawi M, AlNouss A, McKay G, Al-Ansari T. Optimum sustainable utilisation of the whole fruit of Jatropha curcas: an energy, water and food nexus approach. Renew Sustain Energy Rev 2021;137:110605.

[114] Freitas JV, Bilatto S, Squinca P, Pinto ASS, Brondi MG, Bondancia TJ, Batista G, Klaic R, Farinas CS. Sugarcane biorefineries: potential opportunities towards shifting from wastes to products. Ind Crop Prod 2021;172:114057.

[115] Reno MLG, del Olmo OA, Palacio JCE, Lora EES, Venturini OJ. Sugarcane biorefineries: case studies applied to the Brazilian sugar–alcohol industry. Energy Convers Manag 2014;86:981–91.

[116] Junqueira TL, Chagas MF, Gouveia VLR, Rezende MCAF, Watanabe MDB, Jesus CDF, Cavalett O, Milanez AY, Bonomi A. Techno-economic analysis and climate change impacts of sugarcane biorefineries considering different time horizons. Biotechnol Biofuels 2017;10.

[117] Delval F, Guo M, van Dam KH, Stray J, Haigh K, Gorgens J, Shah N. Integrated multi-level bioenergy supply chain modelling applied to sugarcane biorefineries in South Africa. Computer Aided Chem Eng 2016;38:2037–42.

[118] Gorgens JF, Mandegari MA, Farzad S, Daful AG, Haigh KF. A biorefinery approach to improve the sustainability of the South African sugar industry: assessment of selected scenarios. In: Godfrey L, Gorgens JF, Roman H, editors. Opportunities for biomass and organic waste valorisation. Finding alternative solutions to disposal in South Africa. first ed. London: Routledge; 2018.

































https://rdrr.io/rforge/survey/

https://rdrr.io/rforge/survey/











https://cran.r-project.org/web/packages/WeightIt/vignettes/WeightIt_A0_basic_use.html








https://CRAN.R-project.org/package=readxl

https://CRAN.R-project.org/package=readxl

https://CRAN.R-project.org/package=purrr

https://CRAN.R-project.org/package=purrr

https://CRAN.R-project.org/package=Rmisc

https://CRAN.R-project.org/package=Rmisc

https://CRAN.R-project.org/package=cobalt

https://CRAN.R-project.org/package=WeightIt



https://cran.r-project.org/package=jtools

https://cran.r-project.org/package=jtools

https://cran.r-project.org/web/packages/boot/citation.html

https://cran.r-project.org/web/packages/boot/citation.html








































































































15

[119] Navarro-Pineda FS, Baz-Rodríguez SA, Handler R, Sacramento-Rivero JC. Advances on the processing of Jatropha curcas towards a whole-crop biorefinery. Renew Sustain Energy Rev 2016;54:247–69.

[120] Navarro-Pineda FS, Handler R, Sacramento-Rivero JC. Conceptual design of a dedicated-crop biorefinery for Jatropha curcas using a systematic sustainability evaluation. Modeling and Analysis 2019;13:86–106.

[121] Bridgwater A. Fast pyrolysis of biomass for the production of liquids. In: Rosendahl L, editor. Biomass combustion science, Technology and engineering. Sawston. Woodhead Publishing; 2013.











Documents

Renewable and Sustainable Energy Reviews - Gasparatos Lab