Advertisement

Food Security

, Volume 9, Issue 6, pp 1219–1235 | Cite as

The integration of smallholders in agricultural value chain activities and food security: evidence from rural Tanzania

  • Luitfred Kissoly
  • Anja Faße
  • Ulrike Grote
Original Paper

Abstract

The integration of smallholders into agricultural value chains is considered an important pathway for raising the welfare of farmers, including their food security. Distinct from literature that has mainly dwelt on smallholder integration in high-value and export-orientated agricultural value chains (AVCs), we focus on domestic, traditional AVCs, which are relevant to the majority of smallholders. Using primary household data from Kilosa and Chamwino districts in rural Tanzania, we examined the nature and extent of smallholder participation in traditional AVC activities and their associated welfare effects, focusing primarily on household food security. Cluster analysis was used to explore different smallholder livelihood activities and the extent of participation in traditional AVCs while propensity score matching and inverse probability weighted regression adjustment approaches were employed to analyze food security effects of various AVC activities. Results revealed that smallholders participate at varying levels in different AVC activities and their integration in traditional AVCs plays an important role in improving food security. Whereas other studies analyze only the impacts of participation in single AVC activities, we show the relevance of assessing the effects of multiple AVC activities on food security. Findings highlight the importance of promoting policies that enable effective vertical and horizontal integration of smallholder farmers into traditional AVC activities for enhanced food security and improved livelihoods.

Keywords

Smallholder agriculture Integration Traditional value chains Food security Tanzania 

Introduction

Smallholder agriculture is an important driver of food security, employment and poverty reduction for rural households (IFAD and UNEP 2013). This is especially the case in most countries in sub-Saharan Africa. In Tanzania, for example, about three-quarters of the population are employed in smallholder agriculture and about 70% of the population lives in rural areas where food insecurity is prevalent because of reliance on low-productivity subsistence farming (World Food Program 2012; World Bank 2014). In recent decades, agricultural systems have continued to change due to rising incomes, demographic shifts, urbanization and globalization (McCullough et al. 2008; Barrett et al. 2010; Jayne et al. 2010). These changes affect not only modern agricultural value chains (AVCs) but also traditional ones employed by the majority of smallholders. With the ongoing efforts to raise productivity and promote commercialization (Jayne et al. 2010), smallholder farmers are increasingly integrated in AVCs (through input and output markets), in activities such as the procurement of inputs, crop production, post-harvest handling and selling of products.

Integration in AVCs through the participation of smallholders in various AVC activities is seen as a potential pathway to raising the food security and welfare of farmers (Mitchel et al., 2009; Barrett et al. 2010; Bellemare 2012; Fischer and Qaim 2012). Such benefits are realized through increased productivity, market access, and reduced transaction costs, among other factors (Taylor and Adelman 2003; Minten and Barrett 2008; Barrett 2008, Jaleta et al. 2009). However, risks such as exclusion from the value chains and exploitative relationships among smallholder farmers and other participants in the value chain may undermine the welfare of smallholders (Sivramkrishna and Jyotishi 2008; Wiggins et al. 2010). Welfare outcomes of participation by smallholders in AVCs are, therefore, increasingly determined by the terms and relations of participation (or exclusion), given the local context (Challies and Murray 2011).

Recent studies have mainly dwelt on welfare effects, including household food security of smallholders, who have mostly not been integrated into modern AVCs, which are largely concentrated on high value and export-orientated crops (Challies and Murray 2011; Barrett et al. 2012; Bellemare 2012). Little attention has been paid in the literature to smallholder integration into traditional AVCs. This is also the case for Tanzania. Distinct from other studies, this paper focuses on traditional AVCs and analyzes not only individual AVC activities, but also a combination of these, to take into account the complex interrelation of various AVC activities and their associated spillover effects.

The objectives of the paper are, therefore, to (1) explore the livelihood activities of smallholders and their participation in traditional AVCs in rural Tanzania, (2) analyze the food security effects of participation in individual traditional AVC activities, and (3) compare the effects of individual AVC activities, and combinations of these, on household food security. The rest of the paper is organized as follows. The next section reviews the literature and presents a conceptual framework. The third section presents the data and methods. Empirical results and the discussion are given in sections four and five, followed by a summary of findings and conclusions.

Literature review and conceptual framework

In agriculture, value chains constitute a set of linked actors and activities that bring an agricultural product from production at the farm through to final consumption, with value being added at each stage (FAO 2005). Such a value chain can involve a series of activities such as production, processing, storage, transport and distribution. AVCs can be traditional or modern. According to FAO (2005) and McCullough et al. (2008), modern AVCs have well-established vertical and horizontal linkages among input suppliers, producers, processors, output suppliers and retailers, and have a consolidated supply base. In contrast, traditional AVCs are less coordinated among participants, involve a large number of producers and retailers, and transactions are mainly governed through spot markets. Although both types of AVCs coexist, grades, standards and various forms of contracting characterize modern AVCs while traditional AVCs are largely informal and lack enforceable quality standards (FAO 2005; Arias et al. 2013).

Smallholder participation in AVCs has gained attention owing to its relevance to rural development and the reduction of poverty (Barrett et al. 2010). According to KIT et al. (2006), smallholders can participate in AVCs in two ways: (1) vertically through undertaking different activities such as crop cultivation, post-harvest handling, storage and marketing of their produce and (2) horizontally, through collective action in farmer groups or cooperatives (see Fig. 1). However, several factors inhibit the effective participation of smallholders in AVCs. These include limited households’ productive assets (such as land, livestock and labour), inadequate production technologies, geographical constraints, and institutional constraints such as inadequate access to credit and insecure land rights (Barrett et al. 2010). Consequently, low productivity and less marketable surplus impair effective smallholder participation in AVCs.
Fig. 1

Forms of participation by smallholder farmers in the agricultural value chain (AVC) (Authors’ construction based on KIT et al. 2006)

Empirical literature reveals varying levels of smallholder participation in AVCs. For smallholders’ participation in input markets, literature documents a low usage of improved inputs in smallholder agriculture and weak linkages between smallholders and input suppliers, especially in informal, traditional AVCs (Crawford et al. 2003). In modern AVCs, where horizontal and vertical linkages are more pronounced, smallholders have better access to improved inputs and extension services (Barrett et al. 2012). With respect to participation in output markets, empirical studies suggest that there is still limited smallholder market participation (Barrett 2008; Barrett et al. 2010). For example, Barrett (2008) observes that in Eastern and Southern Africa, a small proportion of crop producers participate in the staple food-grains market and such participation is associated with their asset holdings and access to market. Concerning horizontal coordination, smallholders still face a lack of effective farmer organizations and cooperatives along with high transaction costs to successfully participate in AVCs (Wiggins et al. 2010; Fischer and Qaim 2012).

Welfare effects of smallholders’ integration in AVCs

The integration of smallholders in AVCs is considered an essential pathway to raising smallholder welfare and hence promoting rural development and the reduction of poverty (Mitchel et al., 2009; Barrett et al. 2010; Bellemare 2012). Linking smallholders to markets is assumed to improve welfare and increase their utility (Taylor and Adelman 2003). With market access, farmers can produce goods for which they have a comparative advantage and thus produce a marketable surplus, which can be used to buy other goods and services (Barrett 2008).

There is considerable empirical literature on the welfare effects of smallholder participation in modern AVCs. Evidence from high-value and export-orientated AVCs is still ambiguous. Positive and significant effects on smallholder welfare are observed, such as increased household incomes and reduced transaction costs, through contract farming, collective action and other forms of integration in AVCs (Markelova et al. 2009; Fischer and Qaim 2012; Bellemare 2012; Herrmann and Grote 2015). However, positive welfare effects are not automatic. Narayanan (2014) in a study in India, for example, found heterogeneous impacts from contract farming, depending on the type of crop. Additionally, smallholders may face considerable exclusion (attributed partly to strict standards and quality requirements) that undermines their welfare (Sivramkrishna and Jyotishi 2008; Muriithi et al. 2010).

Traditional AVCs have received much less attention, and research that has systematically analyzed the welfare effects of smallholder integration is rare. Furthermore few studies have explicitly analyzed food security effects of smallholder integration in AVCs (Bellemare and Novak 2015). Literature on individual aspects of traditional AVC activities points to heterogeneous welfare effects on smallholders. Specifically, linkages to input markets are found to be instrumental for raising productivity and subsequently smallholders’ income and food security (Minten and Barrett 2008; Kassie et al. 2014). Likewise, the participation of smallholders in effective post-harvest handling activities (such as initial processing and storage) is linked to reduced post-harvest losses and improved food security (Abass et al. 2014). Participation in output markets through commercialization is also linked to significant increases in farmers’ incomes and food consumption (Barrett 2008, Jaleta et al. 2009). However, several concerns are noted in the literature about welfare effects of smallholder integration in traditional AVCs. Smallholders participating in traditional AVCs are faced with prohibitive transaction costs arising from infrastructural and institutional constraints in both input and output markets (de Janvry et al. 1991; Barrett 2008). Also, since most smallholders sell immediately after harvest, price uncertainty is a major risk they face (Arias et al. 2013).

Conceptual framework

The sustainable livelihoods (SL) framework offers a suitable conceptual basis for understanding smallholders’ livelihoods (DFID 1999). Based on literature from SLs, the integration of smallholders in traditional AVCs can be contextualized as part of a livelihood strategy, dependent on available livelihood assets and associated with particular livelihood outcomes (Fig. 2). Therefore, participation of a smallholder household in a given set of AVC activities constitutes part of the portfolio of activities through which it secures a livelihood (Challies and Murray 2011). Given the interconnectedness of production, consumption and investment decisions in typical smallholder households (Singh et al. 1986), their livelihood activities are intricately linked with traditional AVCs and therefore determine households’ livelihood outcomes, such as income and food security (Fig. 2).
Fig. 2

Conceptual framework (Authors’ construction, based on Ellis 2000 and Soltani et al., 2012)

Data and methods

Study area

The study area comprised two regions in Tanzania; Morogoro and Dodoma (Fig. 3). Morogoro region is largely sub-humid with 600–800 mm annual rainfall and is characterized by maize, legumes, rice and sesame as main crops, with little livestock. Different areas of the region have varying levels of food security. Dodoma region is predominantly semi-arid (350–500 mm annual rainfall) with main crops including sorghum, millet, groundnut and sunflower. Food insecurity is more pronounced in Dodoma, and livestock keeping plays an integral part in the livelihoods of households. Study sites were located in two districts from the two focal regions: Kilosa district in Morogoro and Chamwino district in Dodoma. These districts were purposively selected based on their agro-ecological conditions. They represent 70–80% of the farming system types found in Tanzania (Graef et al. 2014), and thus offer ideal sites to study livelihood strategies of smallholders, their interactions with traditional AVCs and associated welfare such as food security.
Fig. 3

Study sites in Morogoro and Dodoma regions, Tanzania (Source: Trans-Sec 2014)

Three villages were selected from each district: Changarawe, Nyali and Ilakala in Kilosa, and Ilolo, Ndebwe and Idifu in Chamwino. Several criteria guided the choice of study villages, including having: 1) different market access; 2) similar climate by district; 3) rain-fed cropping systems; 4) livestock integration; and 5) relatively similar village size (800–1500 households). Apart from representing the majority of the farming systems in Tanzania, the study villages offer comparable but yet diverse agro-ecological and socio-economic conditions, ideal for analysis of smallholder integration in AVCs and their associated welfare outcomes (Graef et al. 2014).

Data and variables

Data were collected through a primary household survey in January and February 2014. Using household lists, the survey covered 900 households, with 150 households randomly selected from each of the six villages, proportional to sub-village sizes. Detailed information was collected covering linkages to input markets, the production of various crops, post-harvest management activities and participation in output markets. For household food security, information on household food expenditure, food consumption and food security shocks was collected.

The variables used in the analysis are presented in Table 1. A set of variables was used to capture the vertical and horizontal aspects of participation in traditional AVCs following the framework by KIT et al. (2006). To reflect the vertical integration of smallholders in traditional AVCs, their use of inputs (improved seeds, fertilizer and pesticides), involvement in post-harvest initial processing (e.g. sorting, pressing and squeezing),1 and share of home consumption and storage for selling were used. To assess horizontal integration, households were asked what activities they perform collectively in farmer groups and therefore collective action in producing, processing and selling were used.
Table 1

Summary of variables used in the study

Variable

Description

Mean

Std. Dev.

Household characteristics

 Gender of household head

Gender of household head (Male = 1)

0.78

0.40

 Age of household head

Number of years of household head

48.64

17.10

 Household size

Number of household members (n)

4.81

2.28

 Head primary education

Household completed primary education (yes = 1)

0.53

0.49

 Head secondary education

Household completed secondary education (yes = 1)

0.02

0.16

AVC activity variables

 Use improved inputs

Household uses improved seeds, fertilizer and pesticide (yes = 1)

0.21

0.41

 Collective action-production

Household produce collectively (yes = 1)

0.12

0.33

 Processing (excluding drying)

Household does initial processing (excluding drying) (yes = 1)

0.26

0.44

 Collective action-processing

Household process collectively (yes = 1)

0.09

0.29

 Length of storage

Average number of months household store crops (n)

1.7

2.3

 Store for selling

Household store for selling (yes = 1)

0.47

0.49

 Collective action-selling

Household sell collectively (yes = 1)

0.04

0.20

 Subsistence share

Household share of home consumption (%)

0.55

0.49

 Main cash crop producer

Whether highest share of crop income is from cash crops (yes = 1)

0.17

0.38

 Main staple crop producer

Whether highest share of crop income is from food crops (yes = 1)

0.46

0.49

Household assets and livelihood variables

 Land size owned (ha)

Size of agricultural land owned by household (ha)

1.99

2.24

 Livestock

Number of livestock owned by household in Tropical livestock unit (TLU)

0.89

4.70

 Off-farm casual employment

Household has off-farm employment activities (yes = 1)

0.33

0.47

 Non-farm self-employment

Number of non-farm self-employment business (yes = 1)

0.25

0.43

 Access to credit

Household has access to credit (yes = 1)

0.14

0.35

 Mobile expenditure

Household mobile expenditure per month (PPP US$)

8.95

24.85

 Access to market information

Number of people household head talks to about market information

3.85

7.45

 Number of crops cultivated

Number of crops cultivated by the household

2.85

1.57

Food security indicators

 Food consumption score

Household food consumption score (FCS)

48.00

15.37

 Coping strategy index

Household coping strategy index (CSI)

25.13

29.33

Locational variables

 Distance to village office

Distance in kilometers (Km) from homestead to village offices

1.81

1.69

 Kilosa

Household lives in Kilosa (yes = 1)

0.50

0.50

 Chamwino* base category

Household lives in Chamwino

0.49

0.50

To assess welfare outcomes for households, food security and income indicators were used. For food security, the food consumption score (FCS) and the coping strategy index (CSI) were estimated. The FCS captures frequency and type of food consumed by the household (World Food Program 2008). It is based on the quantity and quality of food, and weights are assigned based on specific food groups consumed. A higher FCS indicates higher household food security. The CSI captures food security indirectly through food consumption behaviour by measuring the frequency and severity of behaviours that households employ when there is not enough food or not enough money to buy food (Maxwell et al. 2014). A higher CSI score indicates greater household food insecurity. With respect to household income, income portfolios of smallholders were used. Aggregation of household income was done following Johnson et al. (1990). This includes income from agriculture, livestock, off-farm activities, non-farm self-employment and remittances. The final household income was converted from the local currency Tanzanian Shilling (TZS) to 2010-based purchasing power parity (PPP) United States Dollars (US$).

Identifying livelihood strategies

To identify smallholders’ livelihood strategies in the context of activities and extent of participation in traditional AVCs and associated welfare outcomes, an asset-based conceptual framework for choosing a livelihood strategy was used, following Brown et al. (2006). Assuming a utility maximizing household, the level of welfare y i is a function of productive assets available (A i ) and an error component (ε i ):
$$ {y}_i={f}_i\left({A}_i\right)+{\varepsilon}_i $$
(1)

Using this framework, factor and cluster analyses were then adopted to identify smallholder livelihood groupings with respect to activities and extent of participation in traditional AVCs. The former is used to reduce the multidimensionality of variables to a small number of factors that are uncorrelated with each other while the latter assigns a large number observations to a tractable number of distinct groups of observations called clusters. Variables capturing livelihood strategies (such as off-farm wage employment, non-farm self-employment, total number of livestock owned and participation in farming) and those reflecting integration in AVCs – through participation in AVC activities – such as in production, post-harvest handling and selling, are used in the analysis. Given the different scales of our variables, a two-step cluster approach was applied as it allows the use of metric, ordinal or nominal scaled variables simultaneously (Chiu et al. 2001).

Analyzing the effects of AVC activities on household food security

To empirically analyze the potential effects of traditional AVC activities on household food security, several evaluation challenges need to be addressed since smallholder farmers are not randomly assigned into different AVC activities. Smallholder farmers may self-select into different value chain activities. Also, in most cases, traditional AVCs involve activities that are interrelated and carried out together rather than as a single activity. This is addressed by estimating treatment effects of participating in traditional AVC activities on household food security through the Propensity Score Matching (PSM) approach in a binary treatment variable case (Rosenbaum and Rubin 1983). At this stage, the analysis focused on three particular activities in traditional AVCs to analyze their effects on household food security: (1) use of improved inputs, (2) storage for selling, and (3) collective action. These activities represent important aspects of vertical and horizontal integration in traditional AVCs.

To implement PSM, in a binary treatments case, treatment effects of individual value chain activities (I = 1 for use of improved inputs, I = 2 storage for selling and I = 3 collective action) on household food security are were evaluated, in a binary treatment variable case. For the first treatment, we evaluated the food security effects of use of improved inputs on those smallholders using improved inputs compared to those who do not use them. The same strategy was implemented for the second treatment (storage for selling) and the third (collective action). Propensity score matching was then implemented in three steps:
  • Step 1: Probit models were first used to estimate the propensity scores for the three different treatments. For a particular treatment, the probit model was specified as follows:

$$ I={\beta}_0+{\beta}_1X+\varepsilon $$
(2)
where I is a binary variable representing a household’s choice of AVC activity and X is a set of household and farm variables relevant in the choice of participation in AVC activity. β is a vector of coefficients to be estimated and ε captures a vector of random unobserved factors affecting the choice of AVC activity. The household and farm variables were drawn from the reviewed literature. The same set of relevant explanatory household and farm variables were used for all three probit models, given considerations of having balancing properties satisfied, and because the presence of non-significant variables in the probit model does not bias the resultant estimates (Caliendo and Kopeinig 2008).
  • Step 2: The treatment group was then matched with the untreated comparison group based on similar propensity scores. Matching was done using nearest neighbour matching (NNM) with replacement and kernel matching (KM) algorithms. While NNM chooses a case in the control group that is nearest to each case in the treatment group, KM matches the treated and control cases based on kernel-weighted averages (Caliendo and Kopeinig 2008).

  • Step 3: The average treatment effect on the treated (ATT) of AVC activities on household food security was calculated. This is the difference in outcome (household food security) between the treatment and control group matched using the propensity scores generally shown as:

$$ ATT={E}_{\left(P(X)|I=1\right)}\left\{E\left[Y(1)|I=1,P(X)\right]-E\left[Y(0)|I=0,P(X)\right]\right\} $$
(3)
where Y(1) and Y(0) represents outcomes for the treatment group and control group respectively. The outcome (household food security) in our case was measured by FCS and CSI. Also, I = 1 are the treated smallholder farmers and I = 0 indicates the control group.

The PSM approach assumes conditional independence (CIA), under which the potential outcomes are independent of the treatment status once observed characteristics have been controlled for. The approach thus controls only for observed heterogeneity between treated and control groups (Rosenbaum and Rubin 1983). To account for hidden bias, Rosenbaum bounds were used to test the influence of unobserved heterogeneity (Rosenbaum and Rubin 1983; Becker and Caliendo 2007).

Comparing the effects of multiple AVC activities on household food security

As a further analysis, we compared the household food security effects of one AVC activity, and a combination of AVC activities, using an inverse probability weighted regression adjustment (IPWRA) approach in a multiple treatments case. We followed literature on the multivalued treatment variable case by Imbens (2000), Lechner (2001) and Lechner (2002). The objective was to estimate the impact of participating exclusively in one, or a combination of AVC activities, on household food security relative to participating in another AVC activity. Unlike in the binary treatment, in the multiple treatments case three treatment levels were exclusively specified as: j=0 for households who were only involved in crop production but did not use improved inputs and also did not store for selling, j =1 for households that were involved in crop production and additionally used improved inputs, j =2 for households that were involved in crop production and also stored for selling and j =3 for households that were involved in crop production and pursued both use of improved inputs and stored for selling.2 Following Lechner (2002), estimation of ATT for the multiple treatments case, is specified as,
$$ {ATT}^{j\Big|m}=E\left\{{Y}_j-{Y}_m|J=j\right\},\kern0.75em \forall m\ne j,\kern0.5em j\in J=\left\{0,1,\dots 3\right\} $$
(4)
where ATT j|m estimates the expected average effects of participating in AVC activity j relative to an alternative AVC activity m.
Treatment effects were estimated using IPWRA.3 The IPWRA approach, which was developed by Robins and Rotnitzky (1995) and by van der Laan and Robins (2003), involves three steps: Step 1: The treatment model is estimated with a multinomial logit model:
$$ p\left(I=j\right)=\alpha \left(X{\gamma}_j\right)\kern1.25em \forall j=0,\dots, 3 $$
(5)
From equation (5), the inverse predicted probabilities of the treatment \( {d}_i(j)=\frac{1}{p_i(j)} \) for j = 0 ,  …  , 3 are derived for all treatment levels. Step 2: Using linear regression of the form:
$$ {Y}_i(j)=X\beta +\varepsilon \kern1em \forall j=0,\dots, 3 $$
(6)

This is estimated by weighted least squares and the corresponding potential outcome is predicted for all households for each treatment level j. Step 3: The average treatment effects on the treated are then estimated by taking the mean difference in the predicted values for the entire treated sample.

By combining the outcome modelling approach of Regression Adjustment (RA) and the treatment modeling approach of Inverse Probability Weighting (IPW), IPWRA consistently estimates the treatment effect parameters with correct specification in either the outcome or treatment model, and hence is known as having a doubly robust property (Robins and Rotnitzky 1995; Wooldridge 2007). An important condition in using IPWRA is the overlap assumption, which requires that all observations have a positive probability of receiving each treatment level.

Empirical results

Descriptions of the integration of smallholder farmers in various AVC activities, and other characteristics of smallholders, are presented in Table 2. Smallholders participated at varying levels in different AVC activities. Despite cultivating an average of three crops, only about 21% of the households used improved inputs. With regard to post-harvest handling activities, 26% of households undertook initial processing which is largely manual, except for milling and the extraction of oil where simple technologies were applied. Households stored crops for an average of 1.7 months and about 47% stored their crops to sell later. With respect to horizontal integration in traditional AVCs through collective action, around 18% of the households pursued some of their agricultural activities in groups. Collective action was undertaken for the purchase of inputs, and processing and selling, though mainly in small and informal groups.
Table 2

Integration of smallholder farmers in various traditional agricultural value chain (AVC) activities in Tanzania

AVCs aspects

Mean

Standard deviation

Household uses improved inputs (1 = yes)

0.21

0.41

Number of crops cultivated (n)

2.85

1.57

Household undertakes initial processing (1 = yes)

0.26

0.44

Household’s average months of storage (n)

1.70

2.3

Household stores for selling (1 = yes)

0.47

0.49

Household participates in collective action (1 = yes)

0.18

0.39

Total sample = 899

Smallholder livelihoods and AVC activities

In identifying the different smallholder livelihood activities, four principal component factors were extracted from the factor analysis (see Table 8 in the appendix). Principal component factor 1 has high loadings on collective action and processing while factor 2 has high loadings on mobile phone expenditure (used to access information) and average months of storage. The principal component factor 3 loaded highly on off-farm wage employment and non-farm self-employment, while factor 4 had high loadings on livestock owned and subsistence share. The two-step cluster analysis was then conducted based on the extracted factors and binomial variables presented in Table 1. Based on Bayesian information criterion (BIC), the cluster algorithm could categorize 890 households out of 899. In a subsequent analysis, the nine outlier observations were dropped. Six clusters were identified with varying levels of activities and participation in traditional AVCs. Table 3 presents the descriptive results of the cluster analysis, together with some continuous variables from the factor analysis. Some clusters reflected high participation in AVC activities but others medium to low participation. The clusters were named based on the livelihood strategies and extent of participation in traditional AVCs.
Table 3

Descriptive results of clusters: Integration in agricultural value chain activities in Tanzania

Variable

C1 (n = 154)

C 2 (n = 94)

C 3 (n = 149)

C 4 (n = 153)

C 5 (n = 200)

C 6 (n = 130)

Production

 Use fertilizer, improved seeds & pesticides (%)

46.0

30.0

100.0

0.0

0.0

0.0

 Land preparation/weeding expenditure ($ PPP)

202.8

28.8

25.3

8.2

8.7

10.6

 Collective action in production (%)

24.0

1.0

14.0

10.0

7.0

15.0

Post-harvest

 Initial processing (%)

44.0

27.0

28.0

100.0

0.0

0.0

 Storing for selling %

58.0

41.0

59.0

45.0

0.0

100.0

 Average months stored for selling (n)

2.0

0.6

1.3

1.2

0.2

2.0

 Collective action in processing (%)

15.0

1.0

11.0

16.0

5.0

6.0

Marketing

 Subsistence share (%)

40.0

66.0

51.0

66.0

64.0

58.0

 Collective action in selling (%)

7.0

2.0

6.0

2.0

1.0

5.0

 Household capital

      

 Household monthly mobile cash expenditure ($ PPP)

14.1

8.3

9.6

5.5

5.6

6.9

 Tropical Livestock Unit (TLU)

0.25

3.90

0.54

0.52

0.30

0.33

 Land size owned (ha)

2.45

3.34

2.28

1.87

1.84

2.28

 Male household head (%)

79.0

90.0

85.0

75.0

73.0

78.0

Location

 Kilosa (%)

81.8

1.06

67.7

58.8

32.5

53.8

 Chamwino (%)

18.1

98.9

32.2

41.7

67.5

46.2

Outcome variables

Food security by cluster

 Food consumption score (FCS)

49.6a

40.5bd

44.9ab

39.7cd

35.4d

42.4bc

 Coping strategy index (CSI)

11.5d

24.2bc

23.6bc

30.9ab

36.9a

18.9cd

Income portfolio

 Total income year (US$ PPP)

1468a

1008ac

1589a

896bc

619c

1250ab

 Agriculture (%)

54ab

37c

61a

47b

31c

53b

 Livestock (%)

11bc

29a

11bc

16b

10c

10c

 Natural resource (%)

14d

21bc

12d

21b

28a

15cd

 Off-farm wage (%)

2d

3cd

7bc

5bd

11a

8ab

 Non-farm self-employment (%)

13a

8ab

4b

6b

13a

7b

 Main crop

Maize

Millet

Sesame

Maize

Maize

Maize

Post hoc Tukey test: Different superscripts (a, b and c) in a row indicate significant difference of means (p < 0.10)

Cluster 1, Staple and cash crop farming households with non-farm self-employment, had high traditional AVC participation and included households engaged in staple and cash crop farming (54% of total income) with some non-farm self-employment (13% of total income). Forty-two percent of households used improved inputs, 71% did initial processing, and 58% stored for selling, on average for 2 months. Also, households in this cluster had the lowest subsistence share (40%) and they pursued a considerable degree of collective action. Cluster 2, Livestock keeping households with subsistence farming had moderate traditional AVC participation and included 30% of the households using improved inputs despite a high subsistence share (66%). These households stored on average for 0.6 months. A distinguishing feature of this cluster was the high orientation towards livestock keeping (3.9 TLU) and high endowment with land (on average 3.9 ha). Cluster 3, Cash and staple crop farming with off-farm wage had high traditional AVC participation with high use of improved inputs (100%), significant involvement in processing (53%) and storage for selling (59%), but low collective action. Cluster 4, Subsistence farming households with livestock keeping had low traditional AVC participation with less investment in the productive stage. Despite low collective action, 45% of the households stored for selling. The subsistence share in the cluster was 66% and there was low endowment with productive assets (such as land). Cluster 5, Subsistence farmers with off-farm wage had low traditional AVC participation and was least integrated in production, post-harvest and marketing activities of the value chain, indicated by low links to input markets, short storage periods (0.2 months) and high subsistence share (64%). There was also high dependence on natural resource extraction. Lastly, Cluster 6, smallholders with off-farm wage employment and non-farm self-employment had moderate traditional AVC participation with the highest proportion of households who stored for selling with an average length of 2 months and an average 58% share of home consumption.

With respect to income and food security status of the different clusters, the results showed that value chain integration related to the food security and income status of the households. Households in more integrated clusters (Cluster 1 and 3) achieved on average a higher income and food security level (FCS of 49 and 44, CSI 11.5 and 23.6, respectively) compared to least integrated ones. Similarly, households in cluster 4 and 5 with low participation in AVCs had lower income and food security level (FCS 39.7 and 35.4, CSI 30.9 and 36.9, respectively) compared to other clusters with moderate to high traditional AVC participation.

Summarizing the different value chain activities, in Table 3 we show that households with low food security and low income (especially Cluster 5) were mainly integrated in collective action activities for production, processing and selling. These activities require only the opportunity costs of the farmers participating in groups. Here, even poor smallholders can gather market information and get help in case of problems during crop production. Cluster 4, in addition, stored for selling, enabling them to generate higher income due to higher market prices. However, farmers in both clusters did not use any improved inputs, which could have increased their productivity. Cluster 3, where all smallholders used fertilizer and improved seeds, generated the highest income, but those households also participated in collective action and storing for selling. Poor households are limited in their ability to afford improved agricultural inputs. We hypothesize that single activities can increase food security and income, but ultimately the combination of activities should result in a better food security and income situation.

Due to the many different value chain activities, we limited our statistical analysis on the effects of AVC activities on food security to three. The use of improved inputs, storage and selling and participation in collective action emerged as important activities that integrate smallholder households in AVCs. These activities also represent horizontal integration (through collective action) and vertical integration (through use of improved inputs and storage for selling) in traditional AVCs. The next section assesses how the three single value chain activities affect household food security.

Food security effects of participation in individual traditional AVC activities

In analyzing the effects of the use of improved inputs, storage for selling and collective action on household food security, we first descriptively analyzed selected smallholder characteristics and food security status based on participation in the above AVC activities (see Table 4). On average, those smallholder households using improved inputs, storing for selling and participating in farmer groups had younger and more educated household heads and were less reliant on livestock keeping and off-farm wage employment. Also, food security was generally higher (higher FCS and lower CSI) for smallholders who used improved inputs, stored for selling and participated in collective action than for those who did not.
Table 4

Smallholder characteristics based on participation in AVC activities in Tanzania

Variable

Use of improved inputs

Store for selling

Collective action

Yes

(n = 190)

No

(n = 700)

Yes

(n = 420)

No

(n = 470)

Yes

(n = 163)

No

(n = 727)

Age of household head (Years)

44.90

(16.54)

49.69

(17.12)

47.14

(16.90)

49.98

(17.19)

46.73

(17.05)

49.09

(17.10)

Gender of household head (Male = 1)

0.87

(0.33)

0.76

(0.42)

0.82

(0.75)

0.75

(0.43)

0.80

(0.40)

0.78

(0.40)

Years of education of household

5.59

(3.36)

4.10

(3.36)

4.81

(3.33)

4.07

(3.46)

4.90

(3.43)

4.31

(3.40)

Land size owned (ha)

2.02

(1.88)

1.98

(2.33)

1.96

(2.46)

2.01

(2.02)

1.59

(1.70)

2.08

(2.3)

Total number of livestock owned (TLU)

0.47

(1.62)

1.01

(5.23)

0.73

(20.4)

1.03

(6.17)

0.35

(1.39)

1.02

(5.17)

Off-farm casual employment (Yes = 1)

0.22

(0.42)

0.36

(0.48)

0.27

(0.44)

0.38

(0.48)

0.26

(0.44)

0.34

(0.47)

Non-farm self-employment (Yes = 1)

0.19

(0.39)

0.27

(0.44)

0.24

(0.42)

0.26

(0.44)

0.25

(0.43)

0.25

(0.43)

Food consumption score (FCS)

52.70

(16.03)

46.71

(14.94)

50.87

(15.42)

45.44

(14.88)

51.88

(15.77)

47.09

(15.14)

Coping strategy index (CSI)

16.79

(25.83)

27.44

(29.83)

17.62

(21.55)

31.90

(33.4)

18.02

(24.86)

26.83

(30.06)

Numbers in parentheses are standard deviations

In the second step, we used probit models to estimate influencing factors for traditional AVC participation and to calculate propensity scores. A broad set of explanatory variables was used, as drawn from the reviewed literature, since the main objective was not only to predict participation in the various treatments but also to achieve a suitable balance of all covariates for implementing PSM (Caliendo and Kopeinig 2008). Results are in Table 5. Gender of the household (in this case being male) and primary and secondary education had positive and significant effects on the probability of using improved inputs (column 1) suggesting gender and educational advantages in adoption of agricultural technologies. Household size seemed to have a significant negative effect on a household participating in collective activities (column 3). In relation to food-cash crop orientation, households that were main staple crop producers had a lower probability of using improved inputs and storing for selling (columns 1 and 2). This is plausible given the rain-fed dependent and low investment nature of staple crop farming in the districts where farming is largely subsistence. Having off-farm wage employment was also associated with a decreased probability of storage for selling (column 2) suggesting the reliance of households on off-farm income, apart from selling stored crops for food and other consumption expenditures. Access to market information showed a positive effect on collective action, indicating the role of information on households to participate in farmer groups for production, processing or selling. On crop diversification, a greater number of crops cultivated by a household was associated with an increase in the probability of storage for selling. Consistently in the three models, a household residing in Kilosa was more likely to use improved inputs, store for selling and participate in collective action, underscoring the role of better access to public services and agricultural support structures to increase integration in traditional AVCs.
Table 5

Probit estimates for participation in different AVC activities in Tanzania

 

(1)

(2)

(3)

 

Use of improved inputs

Store for selling

Collective action

Age of household head (Years)

-0.0040

-0.0038

-0.0042

(0.25)

(0.19)

(0.24)

Gender of household head (Male = 1)

0.3011**

0.1661

0.0139

(0.05)

(0.16)

(0.92)

Household size (Number of households)

-0.0167

-0.0060

-0.0716**

(0.52)

(0.78)

(0.01)

Head primary education (Yes = 1)

0.2098*

0.1155

0.0640

(0.09)

(0.25)

(0.61)

Head secondary education (Yes = 1)

0.6810**

-0.0292

0.3294

(0.02)

(0.91)

(0.26)

Land size owned (ha)

0.0110

-0.0204

-0.0248

(0.68)

(0.34)

(0.43)

Main staple crop producer (Yes = 1)

-0.3027**

-0.1946*

0.0797

(0.02)

(0.06)

(0.54)

Main cash crop producer (Yes = 1)

0.1768

0.1056

-0.1595

(0.23)

(0.42)

(0.32)

Total number of livestock owned (TLU)

-0.0129

-0.0110

-0.0169

(0.52)

(0.38)

(0.56)

Off-farm wage employment (Yes = 1)

-0.1141

-0.1973**

0.0886

(0.36)

(0.05)

(0.48)

Non-farm self-employment (Yes = 1)

-0.1884

-0.0294

0.1549

(0.17)

(0.78)

(0.24)

Access to market information (Yes = 1)

0.0071

0.0077

0.0225***

(0.29)

(0.18)

(0.00)

Number of crops cultivated

0.0474

0.1038***

0.0651

(0.32)

(0.00)

(0.18)

Access to credit (Yes = 1)

0.2381

-0.0258

0.1633

(0.13)

(0.84)

(0.30)

Distance to Village administrative center (Km)

0.0156

0.0166

-0.0008

(0.62)

(0.56)

(0.98)

Household resides in Kilosa (Yes = 1)

0.9917***

0.4760***

1.0189***

(0.00)

(0.00)

(0.00)

Constant

-1.5697***

-0.4571

-1.3253***

(0.00)

(0.08)

(0.00)

Log likelihood

-363.10

-551.76

-359.34

Chi-square

140.3***

57.51***

99.1***

***, **, *, represent statistical significance at 1%, 5% and 10% level, respectively

Standard errors are shown in parentheses

From the probit models, propensity scores were predicted and used to match the treated households with the untreated. The predicted propensity scores need to have a sizeable overlap to achieve the common support condition (Caliendo and Kopeinig 2008). Propensity score distributions for the treated and untreated groups are shown in Fig. 4 in the appendix. Results of the balancing tests (Table 9 in the appendix) show that the overall mean and median bias after matching was below 6.7%, and the pseudo-R 2 was dramatically reduced after matching. The likelihood ratio test of joint significance of the model covariates also confirmed that there were no significant differences in the covariate distributions of the treated and untreated after matching (Caliendo and Kopeinig 2008).

The average treatment effect on the treated of AVC activities on household food security are presented in Table 6. Both nearest neighbour and kernel matching showed relatively similar ATTs. Use of improved inputs significantly increased household food security as shown by a 3.3–3.8 increase in FCS and also associated with a significant reduction in CSI by 4.0–7.1 scores. For storage, households who stored for selling were significantly more food secure compared to those who did not. Specifically, storage for selling was associated with 3.6 to 4.2 higher FCS and a substantial 10.8 to 12.0 lower CSI. With collective action, no significant effects were found for both nearest neighbour and kernel matching.
Table 6

Average treatment effects for household food security: Binary treatment case

Variable

Nearest neighbour matching

Kernel matching

ATT

S.E

ATT

S.E

Use of improved inputs

 Food consumption score (FCS)

3.34**

1.98

3.80**

1.61

 Coping strategy index (CSI)

-7.08*

4.16

-4.05*

2.45

Store for selling

 Food consumption score (FCS)

4.21***

1.21

3.68***

0.91

Collective action

 Food consumption score (FCS)

0.26

1.72

1.70

1.56

 Coping strategy index (CSI)

-1.49

2.72

-2.05

2.39

***, **, *, represent statistical significance at 1%, 5% and 10% level, respectively

ATT = average treatment effect on the treated; S.E. = bootstrapped standard errors

Comparison of individual and multiple AVC activities on household food security

In a further analysis of the effects of AVC activities on household food security, the effects of exclusively using improved inputs or storing for selling and using both improved inputs with storing for selling were compared to households who were involved only in production. The base group (n = 400), that is households only involved in production, means that these households do not use improved inputs and storage for selling. Results from IPWRA showed that the use of improved inputs alone (n = 76), without storage for selling had a significant positive effect on household food security (Table 7). Specifically, this was associated with a 6.38 increase in FCS and a − 10.23 decrease in CSI. Comparing the effects of storage for selling, results showed that households that participate only in storing for selling (n = 306) had a higher FCS of 5.73 and a lower CSI of −14.74. Participating in both (n = 117) the use of improved inputs and storage for selling raised households’ FCS by 6.88 and lowered CSI by −16.20; a slight increase compared with participation in only one AVC activity.
Table 7

Average treatment effects for household food security: Multiple treatment case

Variable

IPWRA

ATT

S.E.

Use of improved inputs vs involvement only in production

 Food consumption score (FCS)

6.22**

2.39

 Coping strategy index (CSI)

-10.23**

5.01

Store for selling vs involvement only in production

 Food consumption score (FCS)

5.75***

1.56

 Coping strategy index (CSI)

-14.74***

4.44

Use of improved inputs and Store for selling vs involvement only in production

 Food consumption score (FCS)

6.88***

1.71

 Coping strategy index (CSI)

-16.20***

4.79

ATT average treatment effect on the treated, S.E. standard errors

***, **, *, represent statistical significance at 1%, 5% and 10% level, respectively

Robustness of the results

As a check of robustness, the ATTs in the PSM were estimated using two matching algorithms, NNM and KM, and food security was assessed using two indicators (FCS and CSI). However, the presence of unobserved factors may influence the choice of AVC activities by smallholder farmers and therefore bias the estimated treatment effects. We therefore calculated Rosenbaum bounds for hidden bias as a further robustness test, to measure how much the matched households would have to differ in terms of unobserved covariates to render the significant treatment effects invalid (Rosenbaum and Rubin 1983; Becker and Caliendo 2007). The calculated Rosenbaum bounds were 1.2 (for FCS) and 1.7 (for CSI) for use of improved inputs and 1.3 (for FCS) and 2.1 (for CSI) for store for selling. For store for selling for example, the Rosenbaum bound of 1.3 for FCS entails that even with a difference in unobserved covariates of 30%, the inferred effect on FCS would still be valid. Since ATTs for collective action were not significant under NNM and KM, the Rosenbaum bounds are not reported. For IPWRA, kernel density plots were used to assess the probability of receiving each treatment level for all observations. The kernel density plot (Fig. 5 in the appendix) suggests sufficient overlap among the treatment levels, despite a slightly left-skewed mass of probabilities for treatment level 1 and 3.

Discussion

Our results indicate considerable variation in the integration of smallholder farmers in traditional AVCs for the different livelihood groups. Poor households engage in fewer and more pro-poor AVC activities, specifically those requiring fewer productive assets. Participation in AVC activities increases from mid-income to high income households. Clusters with high integration in AVCs have wealthier households, who participate in production, post-harvest and marketing stages of AVCs and are substantially integrated into input markets through the use of fertilizer, improved seeds and pesticides. This variation can be partly explained by the available household assets, market access and geographical and agro-ecological aspects (Ellis 2000; Barrett 2008), among other factors. This suggests therefore that there are constraints for some smallholder households to integrating into traditional AVCs and opportunities for others.

Participation in input and output markets was also found to be generally low. Only a small proportion of households use improved inputs - mainly those that belong to wealthier clusters. The minimal use of improved inputs has also been noted by Kassie et al. (2014) when analyzing the use of improved agricultural inputs in Tanzania and their impacts. Similar to our results for integration in output markets, Barrett (2008) also observed that a small proportion of smallholder farmers in eastern and southern Africa participate in markets for staple food grains, and the nature of their agricultural production remains semi-subsistence rather than commercial.

In relation to the food security effects of participating in various AVC activities, our results show that the use of improved inputs and storage for selling significantly increase household food security. Use of improved inputs raises productivity and crop incomes, which are essential for household food security. Minten and Barrett (2008) also found the same effect of use of improved agricultural technologies on productivity and household welfare, including food security. Our results are also in line with those of Kassie et al. (2014) which show that the use of improved inputs, such as improved maize varieties, reduce households’ chronic and transitory food insecurity. Storage for selling is also important in hedging households against low prices during harvest seasons and thus contributes to increased household incomes and food security. Abass et al. (2014) also confirms that appropriate post-harvest management, such as during crop storage, reduce post-harvest losses and enhance food security.

A further analysis of participation in multiple AVC activities showed that households participating in both activities had slightly higher FCS and lower CSI compared to those participating in only one AVC activity. Despite the marginal increase in food security, this may suggest that integration of smallholders in traditional AVCs in multiple activities could lead to higher welfare effects than participation in individual aspects of traditional AVCs. Participation in productive, post-harvest handling and marketing activities in the value chains –with effective collective action – would ensure that significant benefits of the added value remain with the smallholder producer.

However, we found that collective action had no significant food security effects. Despite the potential of producer organizations to improve access to markets and raise agricultural productivity (Shiferaw et al. 2011), the observed low cooperation among farmers means that smallholders (especially those in remote rural areas) still procure inputs, produce and sell individually. Nevertheless, some collective action remains inevitable in addressing AVC participation constraints and achieving pro-poor agricultural growth.

Summary and conclusion

The objective of this paper was to analyze the effects of the integration of smallholder farmers in traditional AVCs on smallholder welfare, most particularly household food security. Specifically, the study set out to explore the livelihood activities of smallholders and their participation in traditional AVCs, to analyze the food security effects of participation in individual traditional AVC activities, and to compare the effects of individual AVC activities, and combinations of these, on household food security.

In terms of smallholder livelihood activities and their participation in traditional AVCs, we found that integration of smallholders in traditional AVCs is relatively low. Low income households, with mainly subsistence farming and livestock keeping, were least integrated in AVCs through collective action activities for crop production, processing and selling. Mid-income and wealthier households, with staple and cash crop farming, were more integrated in multiple activities in AVCs. Regarding the effects of smallholder participation in different traditional AVC activities, our findings demonstrate that integration into input and output markets was associated with improved food security. Specifically, household food security was higher for smallholders using improved inputs or storing for selling than those not undertaking these activities. Comparing the effects of individual AVC activities, and combinations of AVC activities, we found that participating in both, that is use of improved inputs and storing for selling, translated into marginally higher food security. Despite the small increase in food security, this suggests that effective integration of smallholders in multiple activities of AVCs might lead to higher welfare effects through the retention of more benefits of the value added along the chain.

Therefore, policies that help smallholders’ access to agricultural technologies and remove institutional and infrastructural limitations are important for raising smallholder productivity and reducing high transaction costs and other barriers to input and output markets. More importantly, the design of policies to effectively integrate smallholders in AVCs needs to take into account the overall spectrum of activities in the value chain and participation by very poor farmers.

The present study highlighted the complexities associated with smallholder integration in traditional AVCs, and the importance of analyzing all relevant activities and stages along the value chain, rather than single activities or stages. For successful and beneficial smallholder integration in traditional AVCs, participation in multiple stages and activities along the value chains is important. This study analyzed only a limited set of key AVC activities and their subsequent welfare outcomes. Further research is needed on a broader set of activities, so as to capture all the relevant spillover effects on smallholder welfare. Equally important, more empirical evidence is needed, focusing on traditional AVCs.

Footnotes

  1. 1.

    In the analysis, drying was excluded from processing activity because it is considered a pre-requisite activity and mostly done in the field, or at the homestead. To capture significant value addition from crops that may be sold, activities such as squeezing, oil pressing and sorting are considered in initial processing done by small-scale farmers.

  2. 2.

    Due to the exclusivity requirement of the groups, collective action was not included in the analysis at this stage since it was not pursued exclusively, but rather jointly with improved inputs and storage for selling. For example, smallholders may have procured inputs or stored for selling in groups.

  3. 3.

    Doubly robust approaches such as inverse probability treatment of weighting (IPTW) and IPWRA have been used in a number of studies (see for example Binam et al. 2015, Chiputwa et al. 2015) analyzing multiple treatments. In the present study, the IPWRA approach is implemented by using teffects program in STATA.

Notes

Acknowledgements

This publication is a product of the project Trans-SEC (www.trans-sec.org) funded by the German Federal Ministry of Education and Research (BMBF) and the German Federal Ministry for Economic Cooperation and Development (BMZ). The views expressed are those of the authors and may not under any circumstances be regarded as stating an official position of the BMBF and BMZ. We wish to thank two reviewers for their valuable comments on earlier drafts of this paper.

References

  1. Abass, A. B., Ndunguru, G., Mamiro, P., Alenkhe, B., Mlingi, N., & Bekunda, M. (2014). Post-harvest food losses in a maize-based farming system of semi-arid savannah area of Tanzania. Journal of Stored Products Research, 57, 49–57.CrossRefGoogle Scholar
  2. Arias, A., Hallam, D., Krivonos, E., & Morrison, J. (2013). Smallholder integration in changing food markets. Rome, Italy: FAO.Google Scholar
  3. Barrett, C., Bachke, M., Bellemare, M., Michelson, H., Narayanan, S., & Walker, T. (2012). Smallholder participation contract farming: comparative evidence from five countries. World Development, 40(4), 715–730.CrossRefGoogle Scholar
  4. Barrett, C. B., Bachke, M. E., Bellemare, M. F., Michelson, H. C., Narayanan, S., & Walker, T. F. (2010). Smallholder Participation in Agricultural Value Chains: Comparative Evidence from Three Continents, MPRA Paper 27829, University Library of Munich, Germany.Google Scholar
  5. Barrett, C. B. (2008). Smallholder market participation: concepts and evidence from eastern and southern Africa. Food Policy, 33(4), 299–317.CrossRefGoogle Scholar
  6. Becker, S., & Caliendo, M. (2007). Sensitivity analysis for average treatment effects. Stata Journal, 7(1), 71–83.Google Scholar
  7. Bellemare, M. (2012). As you sow, so shall you reap: the welfare impacts of contract farming. World Development, 40(7), 1418–1434.CrossRefGoogle Scholar
  8. Bellemare, M. & Novak, L. (2015). Contract Farming and Food Security. Working paper. University of Minesota.Google Scholar
  9. Binam, J. N., Place, F., Kalinganire, A., Hamade, S., Boureima, M., Tougiani, A., Dakouo, J., Mounkoro, B., Diaminatou, S., Badji, M., Diop, M., Babou, A. B., & Haglund, E. (2015). Effects of farmer managed natural regeneration on livelihoods in semi-arid West Africa. Environmental Economics and Policy Studies. doi: 10.1007/s10018-015-0107-4.CrossRefGoogle Scholar
  10. Brown, D. R., Stephens, E. C., Ouma, J. O., Murithi, F. M., & Barrett, C. B. (2006). Livelihood strategies in the rural Kenyan highlands. African Journal of Agricultural and Resource Economics, 1(1), 21–35.Google Scholar
  11. Caliendo, M., & Kopeinig, S. (2008). Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys, 22(1), 31–72.CrossRefGoogle Scholar
  12. Challies, E. R. T., & Murray, W. E. (2011). The interaction of global value chains and rural livelihoods: the case of smallholder raspberry growers in Chile. Journal of Agrarian Change, 11(1), 29–59.CrossRefGoogle Scholar
  13. Chiputwa, B., Spielman, D., & Qaim, M. (2015). Food standards, certification, and poverty among coffee farmers in Uganda. World Development, 66, 400–412.CrossRefGoogle Scholar
  14. Chiu, T., Fang, D., Chen, J., Wang, Y., & Jeris, C. (2001). A robust and scalable clustering algorithm for mixed type attributes in large database environment. In: Proceedings of the 7th international conference in knowledge discovery and data mining, pp 263–268.Google Scholar
  15. Crawford, E., Kelly, V., Jayne, T. S., & Howard, J. (2003). Input use and market development in sub-Saharan Africa: an overview. Food Policy, 28(4), 277–292.CrossRefGoogle Scholar
  16. De Janvry, A., Fafchamps, M., & Sadoulet, E. (1991). Peasant household behavior with missing markets: some paradoxes explained. Economic Journal, 101(1991), 1400–1417.CrossRefGoogle Scholar
  17. DFID (1999). Sustainable livelihoods guidance sheets: Department for International Development, UK. http://files.ennonline.net/attachments/871/dfid-sustainable-livelihoods-guidance-sheet-section1.pdf. Accessed 12 Mar 2015.
  18. Ellis, F. (2000). Rural livelihoods and diversity in developing countries. Oxford, UK: Oxford University Press 273pp.Google Scholar
  19. FAO (2005). Addressing marketing and processing constraints that inhibit agrifood exports. A guide for policy analysts and planners. Food and Agriculture Organization. Agricultural Services Bulletin 60, Rome, Italy.Google Scholar
  20. Fischer, E., & Qaim, M. (2012). Linking smallholders to markets: determinants and impacts of farmer collective action in Kenya. World Development, 40(6), 1255–1268.CrossRefGoogle Scholar
  21. Graef, F., Sieber, S., Mutabazi, K., Asch, F., Biesalski, H. K., et al. (2014). Framework for participatory food security research in rural food value chains. Global Food Security, 3(1), 8–15.CrossRefGoogle Scholar
  22. Herrmann, R., & Grote, U. (2015). Large-scale agro-industrial investments and rural poverty: evidence from sugarcane in Malawi. Journal of African Economies, 24(5), 645–676.CrossRefGoogle Scholar
  23. IFAD/UNEP (2013). Smallholders, food security and the environment. International Fund for Agricultural Development, Rome, Italy/United Nations Environment Programme, Nairobi, Kenya.Google Scholar
  24. Imbens, G. (2000). The role of the propensity score in estimating dose-response functions. Biometrika, 87(3), 706–710.CrossRefGoogle Scholar
  25. Jaleta, M., Gebremedhin, B., & Hoekstra, D. (2009). Smallholder commercialization: Processes, determinants and impact. Discussion Paper No. 18. Improving Productivity and Market Success (IPMS) of Ethiopian Farmers Project. ILRI (International Livestock Research Institute), Nairobi, Kenya.Google Scholar
  26. Jayne, T. S., Mather, D., & Mghenya, E. (2010). Principal challenges confronting smallholder agriculture in sub-Saharan Africa. World Development, 38(10), 1384–1398.CrossRefGoogle Scholar
  27. Johnson, M., Andrew, D. M., & Round, J. (1990). Income and expenditure in a system of household accounts –concepts and estimation. Working Paper No. 10. The World Bank, Washington DC, USA.Google Scholar
  28. Kassie, M., Jaleta, M., & Mattei, A. (2014). Evaluating the impact of improved maize varieties on food security in rural Tanzania: evidence from a continuous treatment approach. Food Security, 6(2), 217–230.CrossRefGoogle Scholar
  29. KIT, Faida MaLi, & IIR (2006). Chain empowerment: Supporting African farmers to develop markets. Royal Tropical Institute, Amsterdam; Faida Market Link, Arusha, Tanzania and International Institute of Rural Reconstruction, Nairobi, Kenya.Google Scholar
  30. Lechner, M. (2001). Identification and estimation of causal effects of multiple treatments under the conditional independence assumption. In M. Lechner & F. Pfeiffer (Eds.), Econometric evaluation of labor market policy (pp. 43–58). Heidelberg, Germany: Physica/Springer.CrossRefGoogle Scholar
  31. Lechner, M. (2002). Program heterogeneity and propensity score matching: an application to the evaluation of active labor market policies. Review of Economics and Statistics, 84, 205–220.CrossRefGoogle Scholar
  32. Maxwell, D., Vaitla, B., & Coates, J. (2014). How do indicators of household food insecurity measure up? An empirical comparison from Ethiopia. Food Policy, 47, 107–116.CrossRefGoogle Scholar
  33. Markelova, H., Meinzen-Dick, R., Hellin, J., & Dohrn, S. (2009). Collective action for smallholder market access. Food Policy, 34(1), 1–7.CrossRefGoogle Scholar
  34. McCullough, E. B., Pingali, P. L., & Stramoulis, K. G. (Eds.) (2008). The transformation of Agri-food systems. London: FAO and Earthscan.Google Scholar
  35. Minten, B., & Barrett, C. B. (2008). Agricultural technology, productivity and poverty in Madagascar. World Development, 36(5), 797–822.CrossRefGoogle Scholar
  36. Mitchel, J., Keane, J., & Cole, C. (2009). Trading up: how a value chain approach can benefit the rural poor. London: Oversees Development Institute.Google Scholar
  37. Muriithi, B., Mburu, J., & Ngigi, M. (2010). Constraints and determinants of compliance with EurepGap standards: a case of smallholder French bean exporters in Kirinyaga District, Kenya. Agribusiness, 27, 1–12.Google Scholar
  38. Narayanan, S. (2014). Profits from participation in high value agriculture: evidence of heterogeneous benefits in contract farming schemes in southern India. Food Policy, 44(1), 142–157.CrossRefGoogle Scholar
  39. Robins, J. M., & Rotnitzky, A. (1995). Semiparametric efficiency in multivariate regression models with missing data. Journal of the American Statistical Association, 90(429), 122–129.CrossRefGoogle Scholar
  40. Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41–55.CrossRefGoogle Scholar
  41. Shiferaw, B., Hellin, J., & Muricho, G. (2011). Improving market access and agricultural productivity growth in Africa: what role for producer organizations and collective action institutions? Food Security, 3(4), 475–489.CrossRefGoogle Scholar
  42. Singh, I., Lyn, S., & Strauss, J. (Eds.) (1986). Agricultural household models – extensions, applications and policy. Baltimore: The Johns Hopkins University Press.Google Scholar
  43. Sivramkrishna, S., & Jyotishi, A. (2008). Monopsonistic exploitation in contract farming: articulating a strategy for grower cooperation. Journal of International Development, 20(2), 280–296.CrossRefGoogle Scholar
  44. Soltani, A., Angelsen, A., Eid, T., Naieni, M. S. N., & Shamekhi, T. (2012). Poverty, sustainability, and household livelihood strategies in Zagros, Iran. Ecological Economics, 79(0), 60–70.CrossRefGoogle Scholar
  45. Taylor, J. E., & Adelman, I. (2003). Agricultural household models: genesis, evolution, and extensions. Review of Economics of the Household, 1, 33–58.CrossRefGoogle Scholar
  46. van der Laan, M., & Robins, J. M. (2003). Unified methods for censored longitudinal data and causality. New York, NY: Springer Publishing Company.Google Scholar
  47. Wiggins, S., Kirsten, J., & Llambí, L. (2010). The future of small farms. World Development, 38(10), 1341–1348.CrossRefGoogle Scholar
  48. Wooldridge, J. M. (2007). Inverse probability weighted estimation for general missing data problems. Journal of Econometrics, 141(2), 1281–1301.CrossRefGoogle Scholar
  49. World Bank (2014). World Development Indicators: Percentage of Population in Rural Areas (in % of Total Population). Agriculture and Rural Development. http://data.worldbank.org/indicator/SP.RUR.TOTL.ZS. Accessed June 2015.
  50. World Food Programme (2008). Food consumption analysis: Calculation and the use of the food consumption score in food security analysis. WFP. Vulnerability analysis and Mapping Branch, Rome: Italy.Google Scholar
  51. World Food Programme (2012). Comprehensive Food Security & Vulnerability Analysis (CFSVA) in Tanzania. United Nations World Food Programme.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht and International Society for Plant Pathology 2017

Authors and Affiliations

  1. 1.Institute for Environmental Economics and World TradeLeibniz University HannoverHannoverGermany

Personalised recommendations