This study utilizes primary data from the Center, Littoral, and West regions of Cameroon. Cameroon is located between latitude 2°N to 13°N and longitude 8°25°E and 16°20°E in the Central African subregion. It opens to the Atlantic Ocean in the west with a total coastline of 402 km. It is bounded to the west by Nigeria, northeast by Chad, south by Gabon, DR Congo, and Equatorial Guinea, and to the east by Central African Republic. It has a total surface area of 475,650 km which is distributed into five agroecological zones in ten geographical regions (MINFOF, 2018). The most notable zones are the Sudano-Sahelian, High Guinea Savannah, Western Highland, Monomodal Humid Rainforest, and the Bimodal Rainforest (Figure 1). This agroecological diversity permits the growth of most crops that characterize other African nations hence the name-tag “Africa in miniature” (MINEPDED, 2017). By this natural virtue over other African nations, Cameroon has become the breadbasket of Central Africa and supplies to Gabon, Central African Republic, Equatorial Guinea, and Tchad, as well as neighboring Nigeria to the West.

Our study covers the West, Northwest (Western highlands); the Littoral and Southwest (Monomodal Humid Rainforest); the Center, East, and South (Bimodal Humid Rainforest) regions of Cameroon (MINRESI, 2007; MINADER, 2012). These three agroecology zones were chosen because they have a characteristic tropical climate of two seasons (a rainy season and a dry season), which permits the cultivation of cassava (Molua and Lambi, 2006). Apart from the favorable climate for cassava production, these areas also harbor the largest markets for retail and wholesale of cassava produce (Yaoundé, Douala, and Bafoussam) and its derivatives. Furthermore, the “main” markets of these areas are interconnected by accessible roads. Thus, making it ideal for our analysis.

A three-stage sampling procedure was used. Center, Littoral, and West regions were purposively selected in the first stage based on the intensity or volume of cassava marketed per year. Using data from MINADER (2012) on cassava trade volume per region, in the second stage, we selected 15 markets as follows: Mvog-Ada, Mfoundi, Etoudi, Nkolafamba, Bafia, Bokito, Mvog-Mbi, Sandago, Nkololoun, Bonamoussadi, Mbanga, Bafoussam, Baham, Bagangte, and Bandjoun. In the third stage, we utilized information from the cassava vendor association of each region to establish a proportional sampling frame based on the relative distributions of retailers and wholesalers of cassava within our choice markets. In the end, a total of 400 vendors were studied. After cleaning, however, only 326 were identified to be complete and suitable for analysis.

A structured pretested questionnaire was the main survey instrument for the data collection. The questionnaires were administered by well-trained enumerators who were conversant with the local market “terminologies.” To ensure consistency of information, the enumerators conducted face-to-face interviews where the market agent was recorded directly. This information was further supplemented by key informant interviews with stakeholders in the various regions. This aspect involves holding open-ended interviews with randomly selected key stakeholders in the cassava vendor association, cooperatives, and the Ministry of Agriculture. These interviews were later transcribed word verbatim in French or English depending on the preferred language of the interviewee. Table 1 summarizes the proportion of wholesalers and retailers sampled per region.

Our framework is inspired by the Bain-Labini-Modigliani limit-output theory. Their theory asserts that for a given market where products are undifferentiated and there is a sequential entry of new competing firms, established firms turn to set market prices (“limit-prices”) below the market equilibrium when entry is imminent. This limit price serves as a deterrent against future entrants while assuring long run profit maximization for established firms. We postulate a static model where firms adopt a limit price to maximize expected profits over an infinite horizon prior to the time when entry is uncertain.

The approach involves a multistage game between the entrant firms and the established firms (Baron, 1973; Eaton and Ware, 1987) but its success depends on a set of assumptions:

In this light, if we consider K_n firms in the end of the (n−1) period, an established firm chose a period n with p_n(K_n) limit price conditional on the number of established firms. For any n period, k_n entrant firms may decide to enter a market by playing a form of Cournot subgame. Entrant firms use their knowledge of the limit price and the number of established firms to decide whether to play the game. On the contrary, under the same game, the established firm attempts to preserve its long-term profitability by altering its price for the next period and hence chooses a price p_n+1(K_n+1) for the period (n+1) where K_n+1 = K_n+ k_n. Thus, the profitability of established firms in period n depends on the number of entrant firms in this period. Specifically, if no firm enters the market, then (k_n = 0), K_n+1 = K_n and profit is maintained as denoted by π_n(p_n(K_n), K_n). On the other hand, if firm(s) enter (k_n > 0) at period n, then the profit of established firms decreases such that the postentry profit is given by π_n(K_n+1) < π_n(p_n(K_n), K_n) for any K_n+1 > K_n, where π_n(K_n+1) is assumed constant and independent of any pre-entry prices. The postentry profit reflects the short run reaction of the established firms to entry, while their complete reaction is to choose a price p_n+1(K_n+1) in the next period of the game such that new entrants may be discouraged or forced out of the market. This is especially true since the assumed sunk cost of entry has a bearing on the computation of their profit function. The model establishes the path that links firm profitability to market concentration in Cournot markets.

In this study, we utilize the Gini coefficient (GC), Market efficiency (ME), and Lorenz curve to estimate the structure of cassava markets from their market concentration. In concept, market concentration captures the dynamics of market players. Specifically, it depicts the extent to which market shares are distributed among firms in each market. Unfortunately, the picture is sometimes considered imperfect since it omits other factors such as entry barriers, product differentiation, and innovation (Baagyere et al., 2023). From this premise, it is recommended that most market concentration analyses be associated with other complimentary analyses, which will clarify the market snapshot.

In our context, we employ GC, ME, and Lorenz curves, as well as a multinomial logistic model, to analyze the cassava market structure. The multinomial model is used to evaluate inherent market factors which are omitted in most market concentration models while GC and Lorenz curves are used to examine the market concentration. According to the Food and Agricultural Organization (FAO, 2006), using the GC and Lorenz curve enables an evaluation of the level of inequality or equality in the sales volume (sales capacity) and realized income among the sellers within a market. This approach has also been used by Thu et al. (2020) to analyze the tuna value chain.

As in Thu et al. (2020), the basic estimation model for GC is given as follows

Where i(1, 2, 3, …..n) represents the number of firms (cassava sellers) while x_i denotes the percentage of cassava sellers in the market, and y_i represents the cumulative proportion of cassava sales for i sellers in the market. As mentioned above, GC represents the Gini coefficient. Bukar et al. (2015) and Bannor et al. (2018) assert that GC values approaching 0 indicate perfect equality and perfect competition (no difference in market shares) between market participants while median values indicate oligopolistic or structures of monopolistic competition. On the other hand, values approaching 1 indicate perfect inequality and monopolistic market structures. To better appreciate the outcome, the generated non-aggregated GCs are represented on a Lorenz curve. A Lorenz curve closer to the line of equality indicates equality among the market participants and a Lorenz curve further from the line of perfect equality indicates inequality among the market participants.

Given that our aim is to depict a clear snapshot of the cassava market, we augmented GC with marginal efficiency analysis. ME, generally, permits the measurement of intermediary value addition which is omitted in other concentration models (Haruna et al., 2012; Jeyanthi, 2018). In essence, following Jeyanthi (2018), the basic Shephard model for market efficiency is given as follows:

Assuming marketing cost is represented by total cost as

Marketing margin is represented by total revenues from sales (total value product);

Then the model could be rewritten as follows:

Where F represents the entry cost which is mostly sunk and V represents the variable cost of the firm (cassava seller). k_i and p_i represent the capacity (the quantity of cassava volume) and price per unit for every i, for every cassava seller in the market. Like GC, when values of ME approach 0 it indicates perfect competition among market participants. Whereas, when ME approaches 100 it indicates a situation of perfect monopoly. Median values equally indicate varying degrees of oligopolistic competition.

Finally, to estimate a multinomial logistic model to estimate the barriers to market entry and exit, we estimate a multinomial logistic model which is given as follows:

Where the In(.) parameter evaluates the log likelihood of selecting a desired market (either farmgate, local village market, or central market) and K is the reference group. In our case, let us assume K as a farmgate market. In that case, j = 1 if the individual participates in a central or village market over a farmgate market, and j = 0 if otherwise. The result is evaluated against x_i, which represents i^th predictor variable. β_i also represents the coefficient estimates of the i^th predictor. This approach is highly desirable where dependent variables have multiple categories (Midi et al., 2011). Moreover, the computed maximum likelihood permits the computation of various scenarios despite the size of available data, and to some extent, it is based on a strong normality assumption which makes its results more robust as compared to other multiclass models.

Table 2 summarizes the relative distribution of individual characteristics of the sampled vendors. From the table, cassava markets are dominated by young people and slightly elder people with an average age ranging from 36 years in the Center to 53 years in the Littoral region. Most marketers are in their economically active growth period. Penningroth and Scott (2019) attest that young people and economically active people are usually self-motivated and innovative thus explaining their drive to embrace new opportunities in the cassava agrifood market. The distribution indicates high household sizes with an average of nine persons per household in the pooled study area. This value is higher for the Littoral and Center regions, which might explain their relatively higher engagement in cassava marketing. The Littoral and Center regions host the economic and political capitals of the nation and have larger household sizes which might imply increased pressure on household food security (Table 2). These factors drive marketers to grab new market opportunities to improve household welfare (Agbugba and Obi, 2013; Olagunju et al., 2020). Furthermore, larger households provide available labor necessary for the efficient coordination of marketing operations.

Women were more involved in cassava marketing than men. On average, 53% of the surveyed marketers were women, while 29% were men, and 19% is missing data. This could be due to the relative skewed distribution of economic opportunities between women and men. With men having more opportunities, they are free to choose other segments and industries to be employed. This is not the case for women as such they are more likely to hop on any “neglected” market “leftovers.” Baagyere et al. (2023) highlight similar results with women in the chicken egg market in Northern Ghana.

On average, 30% of cassava marketers were married, while 26% were widowed, 19% were single, 6% were divorced, and 18% corresponds to missing data. Specifically, in the West region, the marital status showed that 67% of the marketers were married. In the Littoral region, the results showed that 45% of the marketers were married. In the Center region, the results showed that 40% of the marketers were married. According to Anjikwi et al. (2020), marriage brings stability to the activity and could lead to higher productivity. In some cases, married women benefit from their partners' networks and resources to enhance their activities. In addition, engaging in cassava marketing for a married woman might serve as a suitable household income and welfare diversification strategy (Anjikwi et al., 2020).

Cassava marketers are averagely educated. Over 98% have at least gone through the primary education which may imply most marketers can read and write. A level of education is critical for cassava marketers as it might enable them to track market trends of cassava prices, and this increases their ability to develop market strategies. Ammanuel (2020) highlights that the education of sorghum marketers in Benue State played a crucial role in their market participation. In addition, Nkamigbo (2019) observed that educated people are more enlightened, well conversant with efficient marketing of their marketable surplus and hence can reduce marketable loss.

The market power and structure of cassava markets are described based on the volume of sales from the market. The results in Tables 3, 4 show the volume of sales from the marketing of cassava. For the pooled retailers, the findings indicate that the cassava market is highly concentrated at the retail level, reflecting an uncompetitive condition. This may be the result of collusive practices in buying and selling and differences in the degree of risk involved in sourcing suppliers by different type of marketers. Although there was a relatively high income inequality in both markets, we observed that the level of concentration in retailing (0.79) was higher than in wholesaling (0.76). This finding is in line with those of previous studies (Haruna et al., 2012; Ahmed et al., 2018) which highlighted that the Gini coefficient >0.5 indicates an unequal distribution of sales among market participants.

Specifically, the result indicated that marketers in the Center region within the sales range of 650–851 constituted 96% of the total respondents and earned 25,646 FCFA (USD42.7),¹ and the Gini coefficient was 0.12. Marketers in the Littoral region within the sales range of 1254–1454 (which constitutes 37% of the total respondents) earned 17,000 FCFA (USD28.3), and the Gini coefficient was 0.48. Marketers in the West region within the sales range of 450–650 (which constitutes 65% of the total respondents) earned 46,705 FCFA (USD77.8). For the pooled sample analysis of retailers, the Gini coefficient was 0.93. This implies that the cassava retail market is perfectly competitive. That is, there are relatively many retailers in the market such that none of them had control over the largest portion of total sales.

The resulting Lorenz curve of Figure 2 depicts a similar equality scenario for retailers. The x-axis (highlighted as p) represents the percentage of market participants while the y-axis [indicated by L(p)] highlights the market share fraction occupied by these participants. The graph however shows that the level of inequality is higher for the bottom 30% of the market participants as compared to the top 70%. The kink effectively highlights this divide where we observe that the bottom 30% have access to < 10% of the market, whereas the remaining 90% of the market share is split almost equally between the top 70%. This result supports the notion of an unequal market as presented in Table 3. Precisely, it highlights a new caveat of inequality in the bottom 30% of the retail markets, which was shadowed in the marginal analysis. The kink might also suggest a market trap (barrier) that prevents agents from taking advantage of the opportunities within the agrifood market.

The results of a pooled sample of wholesalers indicate similar trends as those in the retail market but the market is segregated, and the income gap is wider between groups.

Specifically, the result indicated that marketers in the Center region within the sales range of 7,755–8,755 (which constituted 74% of the total respondents) earned 339,225 FCFA (USD565.4), and the Gini coefficient was 0.61. While marketers in the Littoral region within the sales range of 2,751–4,751 (which constituted 62% of the total respondents) earned 137,530 FCFA (USD229.2), and the Gini coefficient was 0.55. Marketers in the West region within the sales range of 2,750–3,750 (which constituted 58% of the total respondents) earned 46,705 FCFA (USD77.8), and the Gini coefficient was 0.35. This implies that the cassava wholesale market was perfectly competitive. This is reflected the existence of many small-scale producers and suppliers of the produce with the result that none of them could influence supplies by either increasing or decreasing the quantity supplied to influence price. The result of the analysis is similar to the findings of Mafimisebi et al. (2006) in the analysis of palm oil marketing in Osun state which shows a Gini coefficient of 0.47 indicating a low level of inequality in sellers' income levels.

The segregation within wholesaler markets is well depicted in the Lorenz curve of Figure 3. The curve highlights three kinks (inflection points), notably 20%, 40%, and approximately 80%, each associated with a market share. In this case, the bottom 20% corresponds to approximately 4% of the market share while the bottom 40% occupies < 20% of the market share. The remaining 60% of the market is unequally split by the top 60%. From the 80% inflection point, the top 20% controls over 50% of the market share. These results indicate a higher within market inequality among wholesalers which is far higher than what is observed with retailers despite the pooled Gini coefficient results of Tables 3, 4.

The pooled findings show that the marketing efficiency (ME) of retailers ranges between 1.42 and 6.31 with an average and standard deviation of 3.21 and 0.99, respectively. Specifically, in the Center region, findings show that ME ranges between 2.88 and 6.31 with an average and standard deviation of 4.42 and 0.73, respectively. In the Littoral region, findings show that ME ranges between 1.94 and 3.6 with an average and standard deviation of 2.76 and 0.43, respectively. In the West region, we find that ME ranges between 1.42 and 5 with the average and standard deviation of 2.90 and 0.89, respectively. The values indicate that retailers are efficient despite the challenges and unequal distribution of market shares among participants. Our findings corroborate the results obtained by Haruna et al. (2012) and Adegbite and Adejobi (2018) who analyzed tomato traders in Pwalugu in Ghana. Table 5 summarizes the results of the ME of retailers.

The pooled findings show that the ME of wholesalers ranges between 0.87 and 8.22 with the average and standard deviation of 2.88 and 1.72, respectively. Specifically, in the Center region, our findings show that ME ranges between 3.16 and 8.22 with an average and standard deviation of 4.64 and 1.15, respectively. In the Littoral region, we observe that ME ranges between 0.87 and 1.91 with an average and standard deviation of 1.34 and 0.22, respectively, and in the West region, our findings indicate that ME ranges between 1.48 and 3.62 with the average and standard deviation of 2.2 and 0.62, respectively. Wholesalers are thus efficient despite existing market inequality. This finding is in line with Ojo et al. (2016) who reported marketing efficiency of tomato marketing as low as 24.87% and 10.07% in Abuja Municipal. Table 6 summarizes the results of the ME of wholesalers. Table 6 highlights a significant mean square difference for both retail and wholesale cassava markets. The mean square difference is significant for both the retailers and the wholesalers. This highlights the implicit difference in market structures of the studied regions.

Table 7 highlights the correlates that drive the participation of market agents to engage in either farmgate, town, or central market within the sampled regions. These results should be taken cautiously as correlates only and nothing more. Specifically, our analysis excludes control for endogeneity between market drivers on market participation between agents. From the result, agents' participation in either town or central markets is significantly correlated with the regional location of the market. Precisely, being in the Littoral region increases the odds of agent participation in either the town or central market over any farmgate market. Additionally, being in the West region equally increases the chances of agent participation in a central market but decreases the odds of their participation in a village market over any farmgate markets. Possibly this could be attributed to two factors. First, these regions are the most economically active. Second, they are the most accessible in terms of road and transport infrastructure (MINADER, 2012). However, the West region is also a dynamic production basin where farmers most often turn to have contracts and agreements with other participants thus making it less profitable to participate in the local village market over selling at farmgate markets (MINRESI, 2007; MINADER, 2012).

Marketer type is also a significant driver of market choice. It turns out that being a wholesaler decreases the odds of participating in local village markets over participating in farmgates. This might be so because the marketing margins in selling in these markets might be meager especially when compared to the possible benefits of selling in central markets. Indeed, most local village markets are not spatially distant from so-called farmgate markets in most regions of Cameroon. The margin in this case is therefore not large enough to warrant farmers' interest. Being a retailer on the other hand reduces the odds of participating in central markets. This could be linked to the highly competitive nature of central markets. In some cases, such competitions necessitate participants to invest massive amounts of financial resources which might not be suitable for retailers. Notably, Pierre and Kaminski (2019) highlights that central markets are usually located at the heart of active cities in Cameroon thus contributing to higher purchasing power within these markets. As noted by Baagyere et al. (2023), being a female increases the odds of selling in central markets. An explanation might be the relatively higher agreeableness of females which aids them to understand and manage the wide range of costumer pool that might buy cassava or its derivatives.

Marital status is equally a significant driver of market participation choice among agents. It is unusual that being married does not account for any changes in the choice of participation. Contrary to Anjikwi et al. (2020)'s study, being single, divorced, or widowed contributes to market participation. Though unexpected, it is plausible that this is partly because of the labor requirement and the associated logistic challenges to the activity. Being “unmarried” increases the odds of participating in the local village and central markets over farmgates. It is exceptional that being single reduces the odds of participating in central markets. This is possibly because singles are relatively less burdened and as such have the possibility of participating in other ventures in central markets and cities.

Furthermore, we find that education plays a significant role in driving agent participation. Indeed, education increases actors' understanding of risk and margin evaluation, which contributes to better decision making in various situations Ammanuel (2020). In addition, being an indigen from either regions significantly affects the odds of choosing a town or central market over any farmgate market. This corroborates with Abdulai's (2000, 2007) study, which highlighted that tribal cooperation plays a key role in improving the market competitiveness of market actors. The income source of households and access to finance, markets, and group membership were equally found to play a critical role in improving the odds of participation among agents.

Notably, being a “part-time” trader (privately employed or a civil servant) turned out to drive the odds of participation in town or central markets over farmgates. Being privately employed, however, decreases the odds of participating in any local village market but increases the odds of participating in the central market. This is considerably logical especially given that most private institutions and central markets are located in cities. Intuitively, it is fair to assume that these agents juggle between their jobs and a secondary income source. As such they may not have the luxury of participating in farmgate markets. Interestingly, the type of cassava product or derivative in stock plays a role in the choice of market between agents. The odds of cassava chips, cassava flour, garri, and cassava leaf traders are highly correlated with the choice of market. Trading either cassava chips or garri reduces the odds of agent participation in any village market over farmgates. Similarly trading cassava leaves as vegetable decreases the odds of agent participation in any central market. This is because of the relative fragility and sensitivity of cassava leaves to abrasion and spoilage. Finally, we observe that the purchase price of cassava decreases the odds of participation of traders in either town or central markets. Increased purchase cost cuts down marketing margins thus making it less profitable to sell in distant markets rather than other farm markets. This equally corroborates the ideas of the limit price theory highlighted in Section 3.3 above.

The most challenging factors that affect cassava marketing include transport difficulties and a lack of finance or credit. Table 8 shows that 97% of vendors agree that transport cost is a major challenge, whereas 66% of vendors agree that credit accessibility usually affects their activities. This is true for both wholesale and retail markets. More specifically, vendors in the Center region indicated transportation difficulties (34%), high transportation costs (28%), too many buyers (15%), and a lack of finance or credit (13%). In the Littoral region, respondents cited the same difficulties with varying percentages as transportation difficulties (30%), high transportation costs (28%), too many buyers (16%), and a lack of finance or credit (25%) as main constraints. On the other hand, in the West region, respondents claim transportation difficulties (33%), high transportation costs (25%), too many buyers (11%), a lack of finance or credit (28%), and the risk of commodity deterioration (perishability) (12%) as main challenges. Intuitively, the high transport cost is probably due to bad roads which increased transport fares coupled with the fact that the market participants mostly self-finance their business activities through their personal savings, and as such these marketers suffer serious financial constraints which affect their trade volumes. By extension this implies that improving both variables could possibly increase net income for wholesalers and retailers.

There are some policy implications from these findings. First, this study depicts the structure of retail and wholesale markets for cassava which makes it relevant to both researchers and policy makers. To researchers, we provide a preliminary basis for which other price-based models can be conceived for sophisticated market analysis. Second, this study exposes the nature of inequality within various markets. Specifically, we highlight the within group inequality in cassava wholesaler and retailer markets. This caveat may permit policy institutions to target the most appropriate segment for various interventions. Finally, we highlight the main constraint to market participation and the drivers for agent market participation. This can help governments design appropriate motivations for driving market efficiency within the cassava agrifood system. These policies may ultimately improve the cassava agrifood system thereby improving the livelihood of millions of agents (including farmers) who participate at various nodes of the agrifood value chain.