This study adopted a systems approach, incorporating the perspectives of small producers but also of other actors in the system dealing with smallholders in different roles.

Value chain actors include broiler producers, feed producers, traders/retailers and importers. Additional stakeholders can be broadly categorised into, policy-makers, academic experts, and professionals providing services to producers (such as veterinarians or other professionals providing extension services and training for farmers).

Smallholder farmers are defined, according to the classification provided by the South African Poultry Association (SAPA) as those producing fewer than 40,000 birds per cycle (SAPA, 2016), and include producers housing as little as 100 birds (Louw et al., 2017).

Smallholders or “small producers,” both as used in the literature and as defined by SAPA, refers to a heterogeneous group of actors with differing situations and characteristics (Cousins, 2010). Independent producers within this category, however, are examined together based on the challenges they face related to their exclusion from vertically integrated systems and mainstream value chains. The current approach could be complemented by further analysis of sub-groups of producers or of explicit funds allocation to specific sub-categories. Previous studies have distinguished between medium-small (under 40,000 birds) and very small producers (around 1,000) (Louw et al., 2017). Although there is no formal cut-off, we will refer very small producers or simply small producers as those with an output of under 5,000 birds per cycle, while medium or medium-small producers are above this threshold and closer to the upper limit within the under-40,000 category.

Contract growers are excluded from this category as we consider them to be part of the vertically integrated system. We will interchangeably refer tosmall holders or small and emerging producers. The term emerging farmers is generally used in South Africa to refer to relatively small, independent farmers aiming to grow and become established commercial enterprises (Mabaya et al., 2011). Broiler production, unlike indigenous chicken farming, is generally commercially oriented, even for very small farmers, so we do not talk about “subsistence” farmers as a category.

Twenty-two (22) semi-structured interviews were carried out with 28 stakeholders in the broiler sector. This study obtained ethics approval from the Royal Veterinary College's (University of London) Social Science Ethical Review Board (URN SR2018-1624) and through the SOAS (University of London) Ethical Review of Research Projects Procedure (23/April/2019), as well as being covered by the ethics approval granted to the University of Kwa-Zulu Natal, Humanities & Sciences Research Ethics Committee, Protocol number HSS0287/018 for the SHEFS project that this study is part of. Written informed consent was obtained from participants, before interviews were recorded, professionally transcribed verbatim and independently coded by two researchers (SCG-D and KQ). Initial sampling was based on a preliminary systems framework and additional interviewees were contacted by snowballing. Interviews were carried out in Johannesburg (mainly policy makers, SAPA representatives, experts and other national-level actors) and in Kwa-Zulu Natal (small broiler producers, feed producers, traders, providers of training for small farmers and veterinarians and experts). Two interviews were carried out online, using teleconferencing software (Skype).

Participants included representatives of the SA poultry association, the Department of Agriculture Forestry and Fisheries, the Department of Trade and Industry, feed producers, importers, one large contract grower, small informal traders, and professionals who deliver veterinary services and training to small producers. In addition, we interviewed different types of small producers: one medium-small producer, one small registered cooperative run by men, and one small cooperative, run by women (SAPA, 2018). Interview topics included overall perceived trends in the sector and perceived challenges for specific actors and the role of policy in driving the change in the sector and in mediating the identified challenges, as well as the relevant policy actors involved. In addition, questions were asked regarding nutrition security impacts, environmental factors, food safety, and animal health. Although the topics were common across interviews, the order and emphasis on different themes varied to focus on the most relevant issues for each participant. For example, stakeholders involved in exporting or importing were asked specifically about tariffs and import-export regulations, while other stakeholders would only be asked about this if it was identified by them as a relevant driver of changes, challenges or opportunities. Interviews varied in duration, lasting approximately 45 mins and including about 20 questions.

Information from interviews was complemented with a policy document review. The document search and analysis was concurrent with the interview design, data collection and analysis phases. The aim of the document review was to identify policies which can be relevant to the support, livelihoods and performance of smallholder broiler farmers, which were then classified into broad policy areas, including corporate standards, public health, foreign trade, food security and agricultural development. The stated aims of each policy in relation to smallholders were then identified, assessing whether the corresponding policy documents mention smallholder support explicitly as an aim, and whether they make special provisions for smallholders. This analysis was used to frame the study within the existing policy context, increasing its relevance to decision-makers, advocates, and practitioners.

Based on stakeholder interviews, agricultural development policy was identified as the main relevant policy area and is discussed in Section Knowledge, Information and Skills.

The analysis in this study is based on the methodology developed by Kim and Andersen (2012) for the creation of causal loop diagrams based on qualitative transcripts. The main analytical steps are summarised in Table 1. Detailed analysis matrices can be made available as supplementary material or upon request from the authors.

Firstly, we carry out a full thematic analysis of stakeholder interviews and a policy document review. This serves to identify the main challenges faced by small producers in the supply chain and map the policies used to address these challenges.

Secondly, the key segments of the interviews and policy documents are analysed to identify variables and causal relationships between them (Kim and Andersen, 2012). Thirdly, variables are grouped into categories and “generalised” or replaced by more general variables. Finally, variable and causal link tables are translated into a graphical representation and archetypal dynamic structures are identified (Wolstenholme, 2003). The text, tables and diagram are labelled to keep the links between them. Rather than describing every plausible variable within the system, the resulting CLD summarises the key problematic behaviours and causal links emerging from the stakeholder interviews. The resulting CLD is not the only possible formulation. Other researchers, analysing the same information, could focus on different issues and perhaps obtain a different formulation, for example disaggregating different parts of the system. However, the use of strategies to minimise bias through step-by-step model building and triangulation across different sources support the formulation of solid, policy relevant insights and theory.

Based on a full, inductive-deductive thematic analysis of the interviews, a causal loop diagram was developed to serve as a basis for future policy analysis and identification of adequate strategies for value chain improvement. Causal loop diagrams are a systems thinking tool that qualitatively represent variables (nodes) and the relationships between them (links) (Sterman, 2001). A positive (negative) causal link between two nodes is represented by an arrow and implies that a change (increase or decrease) in the corresponding causing variable will lead to a movement of the caused variable in the same (opposite) direction. CLD can represent feedback dynamics, where causes are in turn affected by their outcomes. These feedback loops can be reinforcing (R) or balancing (B). Delays are marked with a double line crossing the corresponding arrow and important feedback loops are named and labelled in the diagrams. In this study, the CLD reflects how participants in the broiler value chain view the system and, in particular, the different challenges faced by smallholders, and the links between them. Distinct dynamics are represented in the form of partial CLD, which are combined into a final CLD.

Interview quotes are marked with an interview-specific code. The analysis of background literature and policy documents, discussed in the following subsection, informs the interpretation and integration of participants' views of the system into the CLD.

The dual nature of South African agriculture (Greenberg, 2017), and of the broiler value chains in particular (Ncube, 2018), has been described in previous literature, and discussed above. The South African case has important specificities, many of which are related to the concentration of land tenure rights in the hands of historically dominant elites, up to and beyond the end of the apartheid regime (Greenberg, 2017). However, the concentration of market share in the hands of an increasingly smaller number of companies is also related to self-reinforcing market consolidation dynamics, which aggravate initial inequalities. This self-reinforcing process is depicted, in a simplified form, in Panel 1, Figure 1. These dynamics are not exclusive to South Africa, and have been described in other contexts and at a global scale: Market consolidation processes tend to reinforce themselves through economies of scale (Sexton, 2000) (which lead to larger companies being more efficient), unequal propensities to invest, and investment returns (Covarrubias et al., 2019) (where larger companies invest more, with higher benefits). This is reflected in loop (R1a) in Figures 1, 2. Other reinforcement processes are not related to increased efficiency of larger companies, but rather to the accumulation of market and political power, which lead to increased barriers to entry for smaller producers (McCorriston, 2002; Greenberg, 2017)and corporate capture of regulators (Mindell et al., 2012) (See loop R1b, Figure 2). The two linked feedback loops shown in the figure lead to what is known as a “success to the successful” archetypal behaviour (Wolstenholme, 2003). The role of policy intervention, in this context, is to counteract this reinforcing dynamic, attempting to lead the system towards a less polarised state. Although the focus of this study is on small producers rather than integrated producers, both segments are heavily inter-linked in terms of their supply of day-old-chicks, feed, veterinary services and more and these linkages are crucial to understanding the challenges faced by small producers.

Our stakeholder interviews confirmed existing literature, identifying restrictions in input access as an important challenge for medium and smallholders (Louw et al., 2017). In the following section, we unpack some key underlying mechanisms and feedbacks underpinning input access restrictions. Partial CLD are shown on Figure 1, panels 2 and 3. The feedback loops are also shown in the final CLD on Figure 2

An important barrier for emerging producers is the issue of restricted availability of day-old chicks. This was specifically identified as a barrier for independent producers wanting to scale up beyond a certain size and compete in the formal market with established integrators. Interviewees from different backgrounds focused on the restricted availability of parent stock and the exclusive franchise for the genetics of the two main commercial breeds of broiler (Ross from Aviagen and Cobb from Cobb-Ventress), which, in South Africa, are each owned by one of the two largest integrated broiler producers. Alternative commercial breeds are not widely available.

One interviewee, involved in providing extension veterinary services and training to small and medium producers, described how restricted availability of inputs combined with a lack of awareness of these difficulties to frustrate the expansion plans of emerging farmers, and how the bottle-neck can be traced back to parent stock:

The restricted availability of breeding inputs is also framed as a symptom of independent producers being “locked out” of the mainstream supply chain, as conveyed by another interviewee:

Policy makers identified the current franchising system as a challenge for small and emerging producers, pointing out that the broiler sector […] works like Microsoft. You buy your computer but you still have to buy software. IC 190624_002. Government official

A parallel State franchise, with a smaller competitor of Aviagen and Cobb-Ventress, was suggested by one interviewee as a potential solution to the issue, reinforcing the narrative of independent producers being locked out of corporate supply chains for breeding inputs. Even if farmers still had to pay, and although it would not shelter them from competition with current franchise-holders, it was suggested that a parallel franchise might significantly contribute to eliminating the existing bottle-neck:

Public subsidies for farm infrastructure (CASP, 2014), combined with unaddressed restrictions in input access, contribute to creating excess capacity. Excess capacity is generally associated to reduced efficiency, increasing average and marginal costs, as fixed costs are distributed across fewer units.

The loop structure in Panel 2, Figure 1 shows the dynamics described above by interviewees. The small market share held by small and emerging producers prompts policy actions to support them, in this case through subsidies for farm set-up and infrastructure provided as part of national agricultural development initiatives (DAFF, 2017b). These are meant to increase efficiency and market share in the sector (balancing loop B2a). The availability of public funds creates expectations of market opportunities and leads to an increase in small producers installed capacity. However, restricted access to breeding inputs (eggs, day-old-chicks, parent and grandparent stock), create a bottleneck, and lead to excess capacity, which reduces efficiency, given that fixed costs need to be distributed across a larger number of units (reinforcing loop R2a).

Feed represents around 60% of the input cost for broiler growers. Unlike the breeding stock, feed is not under an exclusive national patent for each breed. Although the feed market is dominated by integrators, there are some large independent competitors who sell primarily to independent medium or small producers. There is not, therefore, a bottleneck for feed affecting small and emerging producers of the type discussed in Restricted Access to Key Inputs and Excess Capacity Section for breeding inputs. However, our research corroborates previous literature, which has pointed towards high transportation costs and lack of bulk-buying as important factors driving up feed prices for small producers (Louw et al., 2017). In this section we discuss how, in the absence of liquidity and credit, farmers have no choice but to resort to coping strategies which can ultimately aggravate the issue of high transaction costs, undermine efficiency, and perpetuate existing challenges.

The members of a small cooperative mentioned, for example, complementing commercial feed with maize (mielies) to cut cost.

Commercial broiler breeds, which have been genetically selected for fast growth and efficient feed conversion, require commercially prepared feed in order to realise this genetic potential. Broiler chickens do not fare well if left to scavenge or if fed a non-optimised feed mix, which varies throughout the production cycle (Alqaisi et al., 2017). Mixing commercial feed with maize, therefore, as described above, might cut costs in the short term but is likely to considerably undermine technical efficiency.

A representative from a different cooperative referred to splitting planned feed purchases across several trips to manage cash-flow following an unforeseen price increase:

Given that many farmers have to travel considerable distances to obtain feed, if credit restrictions lead to additional, potentially unplanned additional trips, this can add substantially to transaction costs.

Members from a large feed production company selling directly to smallholders identified the issue as being widespread, explaining the decision to market smaller feed bags to address the combination of high prices and credit restrictions:

Aggregation of input demand was discussed as a partial solution, as it could allow smallholders to bypass local retailers who keep an important margin on the feed and deal directly with feed companies, as well as saving on transport costs, ultimately “getting bigger loads into [rural] areas and shared loads.” IC 190618_001

Although an analysis of the price fluctuations of maize and soy feed inputs is beyond the scope of this study, it is worth pointing out the potentially important role of not only local feed markets but also more distal factors, such as drought and dynamics in international grain commodity markets.

Besides the issues discussed above, our research suggested other factors which can restrict access to high-quality inputs. In particular, some interviewees mentioned difficulties in accessing affordable antibiotics, drinking water for the birds, and reliable electricity or gas for heating. We do not have enough information for a more in-depth analysis of every one of these factors, which might be highly context specific. However, such issues can also create important challenges and should be considered in the context of any intervention aiming to support smallholders.

The dynamics described in this section are represented in the causal loop diagram Panel 3, Figure 1. Small producers face liquidity gaps when available cash is not enough to cover operational costs, particularly feed costs. They then engage in various coping mechanisms (smaller and more frequent purchases of feed, mixing specialised feed with maize), which reduce short-term operational costs (B3a) but undermine longer-term efficiency, increasing transaction costs, and reducing output (R3a). Together these loops form a dynamic which is similar to a common system dynamics archetype, known as a “fix that fails” (Wolstenholme, 2003).

Another important challenge discussed in previous literature, and corroborated by our findings, is the lack of a reliable market for smallholders' output (Louw et al., 2017; von Loeper et al., 2018). Please see Figure 1, panel 3 for the partial CLD, and the full CLD in Figure 2.

Selling small to medium batches into the channels currently available for formal commercialisation is widely considered unfeasible (Louw et al., 2017), as large integrators and supermarkets are not considered to have incentives to include small independent producers into their supply chains. An interviewee who had experience in veterinary service provision to smallholders explained:

Parallel marketing systems are discussed as a potential component of a solution. These can include aggregation schemes, where a cooperative focussing on live sales which would preserve the premium that farmers obtain in the live sales market or perhaps an independent abattoir, for example, is supplied by several smallholders in a coordinated fashion. However, interviewees highlighted some challenges in this regard as well.

Lack of certainty in being able to place the full batch, as well as cash flow issues, which were also discussed in the previous section in relation to input purchases, combine to perpetuate multi-generational production systems. Multi-generational operations, as opposed to an all-in all-out approach, can undermine the biological performance of birds, making it harder to optimise the environment, and increasing the risk of disease transmission across generations (Butcher, 2002; Sharif et al., 2014). These issues are in addition to the transaction costs involved in procuring age-specific medicines and feed inputs for birds. A cooperative member described his attempts to maintain an all-in all-out system

Other interviewees, providers of feed for smallholders, emphasised the feedback loop created when credit restrictions lock farmers into multi-generational production systems:

For this sub-section see panel 4 in Figures 1 and 2 for full CLD.

Market aggregation mechanisms, especially those that focus on live sales, preserving the price advantage small producers get by targeting this market segment (DAFF, 2012; Louw et al., 2017), can help support small and independent producers.

Attempts to create market aggregation systems without addressing the issue of cash flow for smallholders, however, are likely to fail, as producers might have limited incentives to fully participate, which can ultimately undermine the viability of the aggregation system. In the words of one interviewee:

The dynamics described in this section are shown in the feedback loop diagram Panel 4, Figure 1. The existence of a liquidity gap, and the uncertainty of being able to place the entire batch, create incentives to sell small numbers of live birds mid-way through the cycle, in the informal market (B4a). This strategy locks producers into multi-generational systems of production which, as discussed above, are associated with reduced efficiency (R4b), and makes it harder to establish market aggregation mechanisms. In the event that an aggregation mechanism had been put in place, additionally, this coping strategy can undermine the financial viability of the aggregator, leading to eventual failure (R4a), reinforcing the above-described lock-in. There are various mechanisms through which on-the-side sales and over-reporting of mortality can lead to failure of the aggregation system, including through losses associated with compensation payments, or through unfulfilled orders and client loss. Loop R3a is generic and does not specify any particular mechanism. Loops R4a, B4a and R4b create a dynamic similar to the system dynamics archetype known as “accidental adversaries” (Wolstenholme, 2003), where the strategies of actor A (in this case, any small producer) involuntarily undermine actor B on whose performance or continued existence actor A's success ultimately depends. Finally, the increased risk associated with multi-generation operations selling in unpredictable informal markets can reduce their access to credit, creating an additional reinforcing loop (R4c).

Lack of business and technical knowledge has been emphasised as a challenge for smallholders both in the literature (Wynne and Lyne, 2003) and by study participants. Business skills, it was emphasised by experts and service providers working with small producers, could reduce excess capacity by aligning expectations with actual market shares for small producers. The need for the technical skills was emphasised by producers, but was not identified as a stand-alone solution, and material challenges were referred to as more important.

One informal cooperative member also referred to barriers to accessing help, related to time, and perhaps difficulty in elaborating the required documentation or perceived difficulty of accessing support schemes:

Producers spoke of useful training and knowledge transmission but contrasted this with a lack of real opportunity or real “help”. One said:

Overall, the lack of knowledge and skills could perhaps be considered a cross-cutting issue which aggravates other challenges, rather than as a separate challenge itself.

This study highlights the importance of internally coherent agricultural development policy in South African food systems. A transformative policy agenda informed by the present analysis would target key leverage points in the system (addressing input and market access, liquidity and credit) through complementary policy interventions, with the aim of interrupting the self-reinforcing dynamics that exacerbate duality in the value chain (including accumulation of market and political power, underutilised capacity, efficiency-eroding short-term coping mechanisms and the inability to move away from multi-generational broiler production systems). The resulting focus on strategic complementary intervention across key areas would be clearly distinct from “silver bullet” policy recommendations, without falling either into the category of “holistic” approaches, where it is assumed that “all types of support must be provided at once” (Aliber et al., 2010).

The coherence of agricultural and food policy in South Africa is recognised as a wider challenge, and has been identified as lacking in the contexts of national nutrition security policy (Thow et al., 2018), food safety in the retail sector (Boatemaa et al., 2019) and nutrition-sensitive agriculture (Schönfeldt et al., 2017). There have also been several attempts to adopt coordinated system-wide policy agricultural development policies which have, nevertheless, not been fully and adequately implemented (DAFF, 2015, 2019).

The findings from the present study are consistent with previous research on the topic which analyses the role of liquidity constraints as a challenge for South African smallholders (von Loeper et al., 2016) and analyses challenges for independent broiler producers in the region including lack of market access and insufficient knowledge (Wynne and Lyne, 2003; Louw et al., 2017). Our findings add to the literature on this topic, illustrating the interlinked and reinforcing nature of the mechanisms that underpin value chain duality and smallholder disadvantage and situating potential interventions within the broader South African policy landscape. Additionally, as is often the case in complex systems analysis, many small producers' behaviours that appear to be dis-functional (certain coping mechanisms, under-utilised capacity) emerge as rational within context, with small producers acting as “locally constructive agents” (Sinitskaya and Tesfatsion, 2015), which would suggest the need to address wider systemic incentives rather than directly targeting said behaviours. That implies that, while training is undoubtedly an essential component of smallholder support programmes, it would not, by itself, solve the challenges faced by small producers.

The set of constraints identified could be, in principle, addressed by a number of different policies or initiatives. This could include a combination of policy elements familiar in the South African agricultural policy environment, such as grants, loans, credit guarantees and the creation of cooperatives or other aggregation mechanisms to facilitate market and input access constraints, as well as training and mentoring programmes.

In terms of government capabilities, a policy package coherently addressing existing constraints would probably rely strongly in the areas of information management, organisational capacity and financial resources, while relying less on governmental “authority” (Hood, 1987).

This type of approach could potentially increase smallholder market shares, improving performance and income stability. The inherent unit cost disadvantages related to scale would most likely remain (Louw et al., 2017), implying that integration of small producers within the mainstream supply chain is unlikely, and smallholder promotion initiatives would most effectively target the creation and development of parallel avenues for marketing.

Changes in discourse, attitudes and causal beliefs are recognised as fundamental components of transformational change (Polanyi, 2001). However, the realisation of transformational impact is likely to depend on to what extent and how the insights from the present systems analysis are used by small producers and advocates to challenge defeatist narratives and discourses that focus primarily on smallholder inadequacies. An informed systems perspective might be particularly useful, for example, in the context of high-stakes policy processes involving negotiation between government and integrated producers (Poultry Masterplan, 2019) and where there are opportunities for small producer advocates to exert influence.

The above section has discussed the potential for complex systems approaches to provide relevant insights in the context of smallholder promotion and inclusive food systems. This includes identifying interlinked challenges, reinforcing dynamics and policy levers, as well as challenging simplistic narratives and interpretations and causal assumptions, and providing alternative discourses.

However, there are also some important challenges. Like many policy areas in agricultural development and food systems, the issue of smallholder promotion, particularly in industrialised and dualistic food value chains, is characterised by the interaction and confluence of conflicting interests (Losch, 2004). These are negotiated through complex political processes, involving stakeholders who differ not only in terms of perspective but also of goals, knowledge culture and power within the system (Leach, 2008).

Systems analysis methods, on the other hand, often place a strong emphasis on building consensus around proposed solutions, often through participatory group model-building approaches (Winz et al., 2009; Rich et al., 2015). While the inclusion of diverse perspectives is considered central, conflict of interest is acknowledged only to an extent, and often within the assumption that, although different groups might benefit to different degrees from a proposed policy, it should be possible to overcome such relative conflicts of interest as part of the system analysis process (Rich et al., 2015). Adopting a different framing, Rammelt and Leung (2017) argue for the use of systems approaches to analyse solutions that appear to be a win-win, while in fact producing winners and losers. However, negotiated outcomes, in highly contested policy areas are likely to require compromise rather than consensus. The actual transformational impact of systems analysis research carried out by external, non-commissioned researchers is then likely to depend on the use that different stakeholders make of systems-informed narratives in the context of wider policy processes.

Group model building has been shown to be highly constructive and to offer many advantages (Lie et al., 2017; Totin et al., 2018). However, the associated framings discussed above can constrain the use of systems analysis to contexts where there is the potential for a win-win or consensus-achieving solution, and where researchers are in a position to lead a consensus-building process. An exclusive focus on “consensus building” approaches also risks not fully engaging with issues of conflict of interest and complex policy processes which are as much part of the food system as more logistical aspects of supply chains.

In order to maximise the usefulness, applicability and transformational potential of systems approaches to promote inclusive food systems, would recommend careful consideration of the wider policy context within which research is taking place. The adoption of critical systems thinking as proposed by Smith (2011), and as adopted in this study is helpful. Additionally, we recommend that trade-offs between different methods of individual and group-based stakeholder participation are considered in relation to conflict of interest and pre-existing power dynamics. The most context-appropriate method or combination of methods should be chosen to ensure that relevant conflicts of interest are captured and participation of all stakeholder groups is supported. We would also suggest that food systems analysis is combined with an analysis of the policy context and relevant policy processes. Finally, we suggest the need for further research into the uptake of insights and narratives from systems analysis into wider policy processes, going beyond consensus-building as part of the research process itself.

The present study has some limitations which should be considered when interpreting the results.

Firstly, this is an entirely qualitative systems analysis and, as such, it does not allow for quantitative prediction, policy evaluation or formal scenario simulation.

Secondly, this analysis is based on individual and small-group interviews, as opposed to group model building (Van Kerkhoff and Lebel, 2006). This methodology was deemed appropriate given the context, the nature of stakeholder relationships and the positioning of the team itself as external, non-commissioned researchers, all of which have been extensively discussed earlier in this section. While the methodology adopted offered advantages in terms of guaranteeing inclusive stakeholder participation in our context, it can limit transformative outcomes through stakeholder networking which have been found to be relevant in other contexts (Totin et al., 2018).

The influence and use of insights from systems analysis in wider policy processes has been identified as a priority area for further research. Quantitative scenario simulation and evaluation of specific policy designs addressing the key system leverage points identified are also potential areas for further research in the context of our study.