Factors influencing access to produce markets for smallholder irrigators in South Africa

Abstract

The objective of this paper was to determine the factors influencing access to produce markets for South African smallholder irrigators. Market access was measured according to whether all of the produce that was meant to be sold the previous season was sold or not. Data were collected from six smallholder irrigation schemes in three provinces: the Eastern Cape, Limpopo and Mpumalanga. A principal component analysis was performed to deal with the problem of multicollinearity, yielding six principal components from 14 original explanatory variables: physical access to the market, farmer skills, nature of access to the market, inventory of support services, ability to respond to opportunities, and off-farm income. A logit regression analysis was then performed, with the principal components as explanatory variables. The results show that physical access to the market, farmer skills and nature of access to the market are highly significant factors in influencing market access for these smallholders.

1. INTRODUCTION

South Africa's development policy focuses on economic growth, reducing income inequalities and eliminating poverty. Agriculture is central in achieving these imperatives. It can contribute to these objectives by increasing agricultural productivity and output, which will enhance the sector's contribution to national economic growth; increasing the incomes of the poorest groups in society, through the creation of opportunities for small-scale and medium-scale farmers to raise their production for their own consumption and the market; creating additional employment opportunities in agriculture; and improving household food security through expanded production and a more equitable distribution of resources (National Department of Agriculture, 2000).

It is estimated that two-thirds of South Africa's smallholder irrigation schemes are dedicated to food plots for subsistence purposes, and that 200 000−230 000 rural black people are dependent for a livelihood at least partially on such schemes (Perret, 2002). Several policies have been introduced to improve the status of small-scale agricultural production in South Africa. Makhura and Mokoena (2003) argue that, despite these policies, small-scale farmers in South Africa still find it difficult to access resource and commodity markets, to obtain agricultural inputs and consumer goods, and to sell their produce. Rural people in Africa, especially the poor, often cite this problem as one reason why they cannot improve their living standards. This lack of access to profitable markets is a major reason why even those farmers who can produce a surplus remain trapped in the poverty cycle. All too often farmers are forced to sell to the buyer of convenience at whatever price that buyer dictates (International Institute for Tropical Agriculture, 2001; Heinemann, 2002; Magingxa & Kamara, 2003).

No development can be expected in large areas of South Africa, and no upliftment of the rural poor can occur in the absence of significant improvements in the marketing set-up serving the people in the rural areas (Van Schalkwyk et al., 2003). In this regard, the importance of improving smallholder access to produce markets is increasingly being realised by those concerned with development. The main objective is to find a profitable outlet for the surplus produce.

A paper by Matungul et al. (2002) provides empirical evidence for the impact of transaction costs on crop income earned by communal farmers in two regions of KwaZulu-Natal. The hypothesis that the variety and quality of marketing channels is determined by transaction costs, which in turn influence the level of crop income, was tested using a block-recursive model. Transaction costs are seen as a primary determinant of crop income level. Greater improvement in marketing channels, which indirectly reflects low transaction costs, has a positive influence on crop income. Three key determinants with statistically significant coefficients – the improved quality of marketing channels, the size of allocated arable land, and off-farm income – had a positive impact on crop income. The significant impact of the regional dummy on crop income implies that the fertility of the study area also affects crop sales. The context of this study is very relevant because, according to Killick et al. (2000), the most important market for the poor is access to their own local market.

Against this background, the objective of the present study was to investigate factors that influence market access for smallholder irrigators in South Africa. The paper is intended to contribute to the body of knowledge about smallholder farmers and fill in knowledge gaps that may exist and thereby help inform policy decisions. The study was conducted in six smallholder irrigation schemes in three provinces of South Africa: the Eastern Cape (Melani, Qamdobowa, Roxeni and Somgxada), Limpopo (Sepitsi) and Mpumalanga (Hereford). The Eastern Cape and Limpopo provinces are where most of the smallholder irrigation schemes are found in South Africa. These are also classified as the poorest provinces in the country (Satistics South Africa, 2003). The selection of schemes was conducted taking the requirements of the research project funding institution into consideration and also to ensure representivity of the sample. Table 1 provides some basic information about the selected irrigation schemes. This paper provides a more detailed account of output market access in the schemes studied and complements previous studies on market access (for example, Makhura & Mokoena, 2003; Van Schalkwyk et al., 2003).

Table 1: Summary of scheme information

Scheme	Scheme size (ha)	Number of farmers involved	Water source	Type of irrigation
1. Hereford	160	33	Loskop Dam	Sprinkler
2. Sepitsi	13	90	Olifants River	Flood
3. Melani	14.5	16	Beanfield Dam	Sprinkler
4. Somgxada	12	14	Tyume River	Sprinkler
5. Roxeni	46	51	Roxeni Dam	Sprinkler
6. Qamdobowa	48	6	Tyume River	Sprinkler

The rest of the paper is structured as follows. Section 2 presents a review of relevant literature on market access and factors that influence it, Section 3 presents the methodology followed in analysing the data, Section 4 supplies the results of the analysis and the final section concludes.

2. FACTORS AFFECTING MARKET ACCESS

Access to markets is critical in allowing new farmers into the mainstream because it is considered one of the most important determinants of their success (National Department of Agriculture, 2000). In the absence of an effective marketing system for their products and inputs, farmers do not have either the opportunity or the incentive to become productive (Van Schalkwyk et al., 2003). The process of integrating the emerging farmers should thus not be viewed in a narrow context of only allocating land and water, but in a broader perspective that embeds the access to these resources in an overall economic framework that includes access to markets, credit, extension, and so on.

In their dealings with the market, smallholder farmers find themselves at a major disadvantage. Many do not understand the market well, how it works and why prices fluctuate, they have little or no information about market conditions and prices, they are not organised collectively, and they have no experience of market negotiation (Freeman & Silim, 2001; International Institute for Tropical Agriculture, 2001; Heinemann, 2002). Poole et al. say:

Smallholder farmers in SSA face a range of marketing problems among which informational constraints are widely acknowledged but little researched. Smallholders often do not have timely access to salient and accurate information on prices, locations of effective demand, preferred quality characteristics of horticultural produce, nor on alternative marketing channels. Producers often experience a weak bargaining position vis-à-vis traders. Information that is available to rural communities may not be equally distributed, and smaller scale producers and those distanced further from the market are more disadvantaged. (1998:7)

According to the International Fund for Agricultural Development (2003), the issue of market access can be considered according to three dimensions: physical access to markets (distances, costs, etc.), the structure of the markets (the asymmetry of relations between farmers, market intermediaries and consumers), and producers' lack of skills, information and organisation (their understanding of the market, prices, bargaining, etc.).

Killick et al. (2000) define ‘market access’ as the processes by which people access markets and the nature, efficiency and costs of these processes. According to them, market access is determined by the following: (1) information about product availability, attributes and prices, including the frequency, quality and cost of this information; (2) information about counter-parties to transactions, as trustworthiness is critical if payment is not instantaneous or checking of quality is costly; (3) the extent of confidence in market conduct−for example, how well markets are regulated (voluntarily or by government); and (4) the physical costs of accessing the market, which are a function of the quality of infrastructure and the organisation of the transport sector, and the actual prices found in the markets in which people transact. Moreover, the relevant literature often cites the following additional factors as determinants for smallholder market access: access to credit, availability of extension services, level of organisation, relevant training, and the farmers' socio-economic conditions (for example, Kherallah et al., 2000; Makhura & Mokoena, 2003).

According to Van Schalkwyk et al. (2003), the established commercial sector and the areas in which commercial agriculture preponderates are served by a sophisticated agricultural marketing system with infrastructure supporting agricultural production and marketing. In general, farmers engaged in small-scale agriculture have limited access to factors of production, credit and information, and markets are often constrained by inadequate property rights and transaction costs (Lyne, 1996). According to Morrison et al. (2000), transaction means the activities that allow or constrain transformation activities, and transaction costs are the costs associated with the transactions that are necessary for transformation to take place.

Even though some of the issues that determine and limit small-scale farmers' access to commodity markets in South Africa are related to policy, most are location specific, farmer specific and/or commodity specific, and some are more pertinent to some provinces than others (Makhura & Mokoena, 2003). In addition to the problems listed earlier, Makhura and Mokoena include farmer discrimination and lack of institutional responsibility in the list of small-scale farming problems in South Africa. They argue that some farmers experience discrimination through their products being traded last and this affects their prices. They also argue that there is a lack of responsibility regarding ensuring market access for small farmers, with various government departments and organisations bouncing the responsibility between themselves.

A report by the National Agricultural Marketing Council (1999) provides pertinent information on the marketing problems for smallholder producers. The report is based on an investigation into market access for disadvantaged market participants, with the ultimate goal of being a commercialised, efficient agricultural system. It summarises South African smallholders' problems as follows:

• As most of the farmers do not have their own means of transport, they rely on contractors, taxis or neighbours and some expensive hired transport because of relatively small quantities of produce. These means are sometimes inaccessible themselves because of the poor roads network in most rural areas.

• Assembly and storage points for the farmers' produce are unsuitable or non-existent.

• Produce has to be transported over long distances to reach the National Fresh Produce markets network.

• In some places the road infrastructure is poor, or there are no roads, especially in provinces such as KwaZulu-Natal and the Eastern Cape.

• There is a lack of market information and the means to disseminate such information, which are critical for small farmers' survival in the increasingly competitive marketing environment.

3. METHODOLOGY

The preceding discussion displays a complex set of factors that influence access to markets by smallholder irrigators. Analysing the influence of these factors can be further complicated by the problem of collinearity. This is a situation where the explanatory variables (in this case factors affecting market access) are highly correlated, and collinearity can lead to biased parameter estimates.

It is understood that the problem of market access for small-scale farmers takes many forms. It may be the case that the produce is not sold to the buyer of choice, and sometimes not at the producer's price of choice. However, for the purposes of this exercise the proxy for the dependent variable was developed on the basis of a response to a question about whether all of the produce that was meant to be sold the previous season was sold or not. The variable therefore only includes whether surplus produce that was set aside for marketing found a buyer. This separated the farmers into two groups: those who sold all their produce and those who did not. This yielded a discrete variable with values of either zero for no or one for yes.

The continuous explanatory variables include extension (number of visits in the past year), creativity, entrepreneurship, distance (to the nearest market), farmer age and off-farm income. Discrete explanatory variables include infrastructure (availability), information, credit, market association, local market availability, and transport and training. To analyse the data, a decision was taken to run a logit model because the dependent variable is dichotomous. This exercise measures the strength and direction of the relationship between the dependent variable and the explanatory variables in the equation. However, this exercise was impeded by a multicollinearity problem within the explanatory variables.

On the basis of this outcome, it was therefore decided to condense the related variables into fewer unrelated clusters or principal components. This technique, known as principal component analysis (PCA), is used to reduce dimensionality of data while retaining as much information as possible (Jolliffe, 1986). But further tests had to be carried out on the data to ensure that PCA was the right way to proceed. When the data were first run on a PCA tool, the indicated Gleason−Staelin redundancy measure (phi value) was 0.327. This is an indication of how inter-related the variables are. A zero value means that there is no correlation among the variables, while a value of one indicates perfect correlation among the variables. However, this coefficient may have a value less than 0.5 even when there is obvious structure in the data, so care should be taken when using it. Further, the Bartlett's sphericity test had a probability value close to zero. This test is used to test the null hypothesis that all correlation coefficients are zero. It is advised that if a probability value greater than 0.05 is obtained, then PCA should not be performed on the data. Therefore there was sufficient confirmation to carry out PCA.

After the data had been initially run through, a number of outliers were noticed in the data by inspecting the T ². These were values of T ² that are significant at the 5 per cent level. T ² measures the combined variability of all the variables in a single observation. These observations were removed from the data as they could possibly distort the interpretation of the results. After this exercise, the number of observations was reduced from 121 to 104.

Several methods are used in choosing the number of principal components in PCA (Jolliffe, 1986). One is to use the scree plot of the eigenvalues. The scree plot can be likened to a cliff with rubble at the bottom. Different people apply different methods here, some dropping the factors represented by the ‘rubble’ and some propose taking all factors up to the cliff plus one factor at the beginning of the rubble. This method is highly subjective and can cause different people analysing the same data to get different results. Secondly, a certain percentage of variability to be accounted for can be set. The idea is therefore to keep enough factors so that this variation is achieved. Usually, this cut-off percentage is used as a lower limit. In other words, if the number of factors do not account for at least 50 per cent of variance, then the whole analysis is aborted. The third method involves using a cut-off value of 0.7 when PCA is conducted with correlation matrices. In other words, all factors with corresponding eigenvalues less than 0.7 are dropped from further analysis. However, this should be done with caution and taking into account the largest eigenvalue. If the largest eigenvalue is close to one, this method can cause useful factors to be dropped.

In this study, the number was decided by leaving out components with corresponding eigenvalues of less than one. This is the rule of thumb when conducting PCA using a correlation matrix, because an eigenvalue corresponds to the number of variables in a factor and the sum of the eigenvalues corresponds to the total number of variables. Therefore, if a factor is less than one that means that it accounts for variability of less than one variable. The initial variables (X values) were therefore grouped into six principal components, altogether accounting for 86.25 per cent of variability. Table 2 presents the results of the exercise described above with corresponding eigenvalues after Varimax rotation, which is a process that involves redistributing the variation for the different variables between components such that each variable is more or less clustered in one component rather than being spread throughout the components. The eigenvectors remain orthogonal as they are rotated and the sum of the variances of the loadings is the maximum possible.

Table 2: Principal components and eigenvalues after Varimax rotation

Variable	PC1	PC2	PC3	PC4	PC5	PC6
Infrastructure	−0.895
Information			−0.845
Credit
Extension
Market association			0.896
Local market availability
Transport availability
Training				0.903
Creativity		0.950
Entrepreneurship		0.931
Distance	0.850
Farmer age					−0.887
Transport cost	0.924
Off-farm income						0.972
Eigenvalue	3.633	2.745	2.044	1.445	1.18	1.027
Percentage variability	25.95	19.61	14.60	10.32	8.43	7.33

4. RESULTS

4.1 Derived principal components

The first principal component, PC₁, accounts for 25.95 per cent of variability, while PC₂ through to PC₆ account for 19.61−7.33 per cent, as can be seen in Table 2. For purposes of clarity, values of factor loadings below 0.7 on either direction were dropped from the analysis. Analysing the direction (sign of component loading) and strength (value of component loading) of the relationships between the components and the initial variables (as shown in Table 2) allows us to interpret the results as follows.

Distances to the nearest central market and the concomitant cost of transporting the produce are arguably the biggest contributors to the difficulty smallholders have in accessing markets. Both these have a positive relationship with PC_1. This could therefore be interpreted as physical access to the market. It is no surprise therefore that this has a strong negative relationship with infrastructure as its unavailability makes physical access to the market difficult.

PC₂ has a strong positive relationship with creativity and entrepreneurship variables. It could be interpreted as an index for farmer skills. It concerns the ability to realise, create and seize market opportunities. These attributes are critical for success in accessing the output markets. PC₃ has a strong positive relationship with the market association. It could be interpreted as the nature of access to the market – in other words, whether the farmer accesses the market as an individual or as part of an association. There is also a strong relationship with information. To a large extent, information about grades, standards and quality requirements for various markets is not available to the smallholder. However, because mostly smallholders sell to the informal market these become largely irrelevant. That could explain the negative sign for information in this component.

In PC₄ we notice a strong positive relationship with training. This is one of the critical components of the framework of services that should be available to smallholder farmers if they are to be able to compete successfully at the market level. This component could therefore be interpreted as indicating an inventory of available support services. In PC₅ we see a strong negative relationship with farmer age. As farmer age increases, it becomes more difficult to respond to opportunities, including accessing the local market. Age can, to a large extent, also affect the response to modern innovations in farming practices. Therefore this component could be interpreted as representing the ability to respond to opportunities. Finally, PC₆ displays a strong positive relationship only with off-farm income. It can therefore be interpreted as representing off-farm income.

4.2 Influence of the derived principal components on market access

After the principal components were estimated, a regression analysis was performed to study how the estimated components influence smallholder access to market. For this exercise, the market access variable was again chosen as the dependent variable. The previously described estimated components served as explanatory variables, with their values being the component scores (the score of each variable within the component). The component scores are scaled such that they have a variance of one and a mean equal to zero.

The following logit model was used in this study to determine the influence of the six principal components identified earlier on smallholder irrigators' market access:

where P represents the probability of a small irrigator having access to the market, y is market access, and χ _i are the set of explanatory variables determining smallholder irrigators' access to market. The model was chosen because of the dichotomous nature of the dependent variable (i.e. market access). The logit regression results are presented in Table 3.

Table 3: Parameter estimates of logistic regression

Variable	Coefficient	Standard error	T ratio	Probability
Constant	0.528	0.34889	1.5132	0.133
Physical access to market (PC1)	−2.478	0.71307	−3.4749	0.001***
Farmer skills (PC2)	0.695	0.29674	2.3417	0.021**
Nature of access to market (PC3)	−1.45	0.34556	−4.2031	0.000***
Inventory of available support services (PC4)	−0.272	0.21050	−0.12902	0.898
Ability to respond to opportunities (PC5)	−0.258	0.27400	−0.094009	0.925
Off-farm income (PC6)	−0.384	0.27702	−1.3852	0.169
***Significant at 1 per cent, **significant at 5 per cent, *significant at 10 per cent.

The model has 82.7 per cent goodness of fit, and this can be considered a good result considering the complexity of the research object and the multiplicity of factors that influence market access. As shown in Table 3, PC₄−PC₆ and the constant term are less significant. PC₂ is significant at the 5 per cent level, while PC₁ and PC₃ are significant at the 1 per cent level.

In this kind of discussion, it is best to concentrate on statistical significance because, for the logit model, the estimated coefficients do not have a direct economic interpretation. The fourth component, PC₄, represents an inventory of support services that are essential to facilitate easier access to the market for smallholders, represented by training in this case. The influence of this component is not shown to be significant in the resultant regression. This could be explained by the fact that the dominant variable in this component is almost uniform throughout the farmers. A very high percentage of them have received some training in one way or another. PC₅ represents the ability to respond to opportunities based on the interactions between farmer age and perception of the availability of the local market. The results of this study show that this is not a very important factor in accessing the market. This situation can be explained by the fact that there is no significant difference between the average age of farmers with or without market access. This agrees with the South African situation where most smallholder farmers are older people – a situation that has prevailed for years, since younger able-bodied people prefer to move to the cities for more lucrative employment rather than farm in the rural areas. So this factor may not play a highly significant role. Low (1986) explores various models to try and explain this phenomenon. He argues that the household-economics approach can be useful here for answering the questions that arise. He suggests that the reason for urban migration may be not only the declining levels of marginal income from farming, but also the comparative advantage of market over non-market activities or non-farm over farm employment. The sixth component, PC₆, represents off-farm income and is also shown to be less significant in this study, perhaps because a large number (about 65 per cent) of the interviewed farmers have access to off-farm income in the form of government grants and remittances from off-farm employment; the effect on market access may therefore be unclear.

The discussion now focuses on the components that have a significant influence in the model. The first principal component, PC₁, represents physical access to the market and is highly significant. As explained earlier, physical access to market is a function of the quality of infrastructure and the organisation of the transport sector. Transport cost has always been shown to be smallholders' biggest cost when accessing the market. Omamo (1998) argues that the cost of transport even influences these farmers' choice of crop, making them forgo cash crops for less profitable crops. The findings of this research confirm that argument. The second component, PC₂, represents farmer skills. These skills are the result of positive interactions between creativity and entrepreneurship. The fact that this component has come out as significant could be tied to the prevailing agricultural marketing situation in South Africa. Until recently, marketing of smallholder produce was done mainly by parastatal organisations through agricultural product control boards. The dissolution of these boards in the wake of the liberalisation of agricultural markets calls for creativity and an entrepreneurial spirit on the part of smallholders if they are to continue marketing their produce. Finally, the third component, PC₃, deals with the nature of access to the market. This involves whether the farmer accesses the market as an individual or as part of a market association. Again, the nature of the association has always been an important factor for smallholder farmers in developing countries in general. Associations not only help small farmers access credit but also improve their bargaining power at the market. To increase their bargaining strength or market power, historically farmers have organised cooperatives to handle and process their commodities in direct competition with other firms (Clodius, 1957). The significance of this component is therefore to be expected.

As can be seen, these components fall directly within the three broad dimensions of market access mentioned above in Section 2. The first component relates strongly to the first dimension, physical access to the market; the second component to the second dimension, the role of the farmer; and the third component, to the third dimension, the structure of the markets. It should be noted that the total effect of the various components of market access depends on the extent of the variable interactions within components and also on the significance of these components in influencing market access.

The next step was to estimate the partial effects of the principal components on market access. This exercise was limited to the significant components. The estimates of the partial effects are calculated by re-scaling the estimated coefficients using a scaling factor (0.23336) produced in the regression analysis. The estimated partial effects are interpreted as follows: for physical access to market, −0.578; for farmer skill, 0.162; and for nature of access to market, −0.339. Interpretation of the partial effects is a somewhat complex exercise, because the interpretation has to make economic sense. In the first component, PC₁, the partial effect could be interpreted as the effect of a unit change in physical access leading to a 0.578 change in the probability of accessing the market, such as a group of farmers using a single large truck, rather than each hiring a small truck, to transport produce. Similarly, the partial effect in PC₂ can be interpreted as the effect of an improvement in the farmers' skill leading to a 0.162 increase in their probability of accessing the market. Farmer skill can be improved by getting more knowledge about the way markets operate, through education and training and interaction with successful colleagues. Finally, the partial effect in PC₃ could be interpreted as the effect of an R1 (ZAR 1) decrease in the cost of transaction, due to the nature of access to the market, leading to a 0.339 increase in the probability of accessing the market.

Further simulations were carried out to determine the effect of each of the significant principal components on market access while holding the effect of the others constant. The results of the simulations are summarised in Table 4.

Table 4: Summary of simulation results

Component	Probability	Improvement	Percentage change in market access
Base	0.0922	–
Physical access to market	0.1166	20 per cent decrease	26.5
Farmer skills	0.1022	50 per cent increase	11
Nature of access to market	0.1055	20 per cent decrease	14.4

The base represents farmers with no market access. It was calculated taking into account the averages of the components. The conditional probability of market access for the base group is 0.09. This means that, out of 100 smallholder irrigators, nine have access to markets. The simulations reveal that a 20 per cent reduction in the physical cost of accessing the market will lead to a 26.5 per cent increase in the probability of accessing the market. A 50 per cent increase in farmer's skill and knowledge through, for example, an increase in the number of years spent in school or attending relevant training programmes will lead to an 11 per cent increase in the probability of accessing the market. Similarly, a 20 per cent reduction in transaction costs associated with the nature of access to the market will lead to a 14.4 per cent increase in the probability of accessing the market.

5. CONCLUSION

Market access is one of the most important factors for the viability of small-scale agriculture. A major reason why even those farmers who can produce a surplus remain trapped in poverty is lack of access to profitable markets. Factors influencing smallholder access to markets are widely discussed in the literature. They often form part of the complex web of factors that affects smallholders in general. This paper attempted to contribute to the debate about the factors influencing market access by analysing the effects of these factors on the market access variable. Market access can be seen as having three dimensions: physical access to markets, the structure of the market and the role of the producer.

The paper attempted to determine the extent of the influence of these factors using primary data from six irrigation schemes in three provinces of South Africa. PCA was performed due to multicollinearity among the variables, and yielded six principal components−interpreted as physical access to market (PC₁), farmer skills (PC₂), nature of access to market (PC₃), inventory of support services (PC₄), ability to respond to opportunities (PC₅), and off-farm income (PC₆). These accounted for 86.25 per cent of the market access variable.

Further regression analysis using the principal components as explanatory variables revealed the most significant components as being physical access to market, farmer skills and nature of access to market. The model had an 82 per cent goodness of fit. PC₁ and PC₃ were significant at the 1 per cent level, and PC₂ at the 5 per cent level. These results were in line with the dimensions of market access described earlier in the paper. It should be noted, however, that the total effect of a component is a function of the interactions of variables within a component. The next step was to conduct simulations. These revealed that a 20 per cent reduction in the physical cost of accessing the market and a 20 per cent reduction in transaction costs associated with the nature of access to the market will lead to a 26.5 per cent and a 14.4 per cent increase, respectively, in the probability of accessing the market. In the same vein, a 50 per cent increase in farmers' skill and knowledge will lead to an 11 per cent increase in the probability of accessing the market.

These results have implications for policy development in areas that could be targeted if smallholder irrigators' access to the market is to be encouraged. For instance, the government is already providing transport subsidies for public transport. In the past financial year this amounted to approximately R4.5 billion (Radebe, 2006). A similar approach could be explored where smallholders would be encouraged to make use of this facility and a minimal cost recovery method could be designed to ensure its sustainability. Policy interventions that reduce structural impediments to exchange−for example, public investment in improved transport and communication infrastructure in rural areas−could reduce the motivation to substitute for imports and promote specialisation that raises farm incomes, and over time could lead to a more equal distribution of wealth (Omamo, 1998). In addition, smallholders should also look for cost-effective means of transporting their produce, such as using larger and shared transport.

There is a clear need for investment in the acquisition of skills for smallholder producers to improve their competitiveness. It may be necessary to design focused training programmes that will enable them to study market trends and plan accordingly. These programmes should also enable them to identify niche markets instead of targeting markets that are already congested.

Efforts to encourage strong farmer associations should be improved. There are clear benefits in these, especially because of the size of landholdings for most smallholders in South Africa. These farmer associations could develop into marketing associations as well as viable cooperatives, which would present an opportunity for the smallholders to access the more lucrative markets that they cannot access as individuals.

Together, these aspects of market access could form a strong base on which to further develop a stronger support framework for smallholder market access. Further in-depth studies based on a wider sample could shed more light on some of the factors investigated in this paper.