The seamless, strong positive relationship across numerous nations, crops, livestock, and years on welfare effects of commercialization highlight its importance in achieving food security. Scientific evidence on smallholder commercialization in Sub-Saharan Africa (SSA) dating back to the 1980s has attracted increasing attention due to its vital role in modernization, agricultural specialization, and systemic economic transformation in the direction of exponential and sustainable growth (Mbitsemunda and Karangwa, 2017; Abate et al., 2021). Other previous studies conducted in SSA and elsewhere indicate that agricultural commercialization increases household incomes and promotes rural development, combating food insecurity (Lifeyo, 2017; Kyaw et al., 2018). Compelling evidence from other high-quality literature (Pender and Dawit, 2007; Barrett, 2008; Sigei, 2014; Muricho et al., 2015; Mmbengwa et al., 2018) has also established the empirical regularity that achieving increased income, food security, farm productivity, and high standards of living is positively and strongly associated with commercialization. In a plurality of context, higher income resulting from commercialization stimulated increased households' dietary diversity and better nutrition security. Among these bases, researchers, government, policymakers, and other institutional entities in low-income countries are keen to promote agricultural sustainability through smallholder market participation. However, despite the attention and employing 1,000 times more people than commercial agriculture, small-scale farming systems are characterized by low household incomes due to market integration constraints (Thamaga-Chitja and Morojele, 2014; Von Loeper et al., 2016). Ubiquitous market failures in SSA put farmers at the peril of engaging in crucial cash cropping operations, mainly if this renders households reliant on unpredictable food markets to meet their food intake needs. For example, where market failure prevails, the adverse effect is primarily registered on factors such as incomes, food consumption, and nutrition, ultimately weakening household safety to insecure land tenure, policy biases, and gender equalities (Pender and Dawit, 2007). This suggests that policy must be tailored to help farmers overcome market failure/barrier traps as indicated in the intensity of market participation literature (Omiti et al., 2009; Kiwanuka and Machethe, 2016) because smallholder agriculture accounts for nearly two-thirds of the natural wealth and 80% of food consumed in developing countries, covering pastureland, and crops.

Commercializing smallholder agriculture has been a critical component of rural development and poverty eradication programs in South Africa (SA) and other African countries. These would entail strengthening the agricultural production system and fostering stable trading frameworks, aided by greater specialization of high-value-cash crops such as the sunflower. As Omiti et al. (2009) explain, market-oriented production entails system transformation based on improved production techniques, technological advances, and agricultural modernization. According to this view, profit-maximization orientation drives input utilization and output mix selections since higher marketable share outputs lead to higher-value specialized farming operations, ultimately displacing integrated farming systems. Indeed, commercialization increases the diversity of sold outputs at the national level while also encouraging specialization at local and household levels (Nwafor, 2015). The potential advantage of specialization through market integration of smallholder farmers highlights the importance of agricultural commercialization.

On the one hand, the commercialization of smallholder farms leads to increased market competition and lowers transaction costs, resulting in a progressive decline in real food prices (Jayne et al., 1996). On the other hand, if market access is guaranteed, the benefits of better commodity prices and lower input costs for commercialization are efficiently transmitted to financially deprived households (Rahman and Westley, 2001). These transformations improve household welfare (as a consumer) by lowering real food prices, resulting in increased purchasing power, while also allowing for the reallocation of scarce farm income (as producers) to other off-farm enterprises for increased income (Davis, 2006). Jaleta et al. (2009) summarize the effects of commercialization into three major categories: (a) employment and income accumulation mirroring household welfare state, (b) nutrition and health, dependent on the level of income obtained via the current level of market integration, and finally, macro-economic aspects, which are beyond household level. Thus, the positive effects of commercialization transcend any associated adverse effects despite the heterogeneous impacts at household levels.

Pingali and Rosegrant (1995) further explain how commercialization follows a three-stage process, starting with subsistence, semi-commercial, and finally, full-fledged commercial state. From the household's perspective, subsistence agriculture aims to achieve food self-sufficiency using non-traded and household inputs. Then, in semi-commercial, the input sources and goals rapidly shift to surplus production, leveraging traded and non-traded household inputs. And as farmers attain a full commercial state, profit maximization becomes the goal when purchasing inputs from the markets. While the net household welfare benefits via commercialization are widely acknowledged, there is no universally accepted criterion in determining the levels of commercialization. Since commercialization is continually advanced in Sub-Saharan Africa for household wealth creation and rural development, there has been an uptick concern regarding “what scale to employ in quantifying its intensity.”

While the extant literature on commercialization nods mutually in delineation, scholars have employed various techniques, such as proportions of sale output, conditional on market participation decisions to examine the level (intensity) of smallholder commercialization (Kyaw et al., 2018). Other authors have used a ratio taking value from zero (no sales) to one (all outputs sold), focused on marketed agricultural outputs, indicating that farmers can be at any point of the continuum (Fredriksson et al., 2010; Abu et al., 2014; Nwafor, 2015). The latter followed the concept of agricultural commercialization of Govereh et al. (1999) as a proportion of sales outputs. Thus, commercialization can be seen as a process of transition, with the two extremities of the continuum representing entirely subsistence and fully commercial, respectively, and combinations of households in-between. This gives a better-nuanced picture of a household output market behavior and the quantity of output they choose to sell. Though the net effects are not invariably or homogeneously positive, there is significant agreement that commercialization comes with varying impacts on different social strata (wealthy and the economically challenged, young and old, landowners and landless, small landholdings and large landholdings, children, and women) under various household, organizational, and policy settings (Jaleta et al., 2009).

To date, there has been an overwhelmingly higher number of cited successes in SSA regarding the commercialization of smallholder farmers. However, the existing empirical evidence on commercialization differs widely in several dimensions, including methodologies. There are quite a lot of studies that touch on commercialization in SSA in regard to livestock, high-value cash crops, and staple food grains using various methodologies, such as supply and demand function (Ouma et al., 2010), probit regression model (Mbitsemunda and Karangwa, 2017; Mathagu et al., 2018; Yaméogo et al., 2018; Dlamini-Mazibuko, 2019), Cragg double hurdle model (Mignouna et al., 2016; Mbitsemunda and Karangwa, 2017; Dlamini and Huang, 2019; Abate et al., 2021; Hlatshwayo et al., 2021), Triple Hurdle Model (Burke et al., 2015; Lifeyo, 2017), Heckman and Tobit models (Zamasiya et al., 2012; Adeoti, 2014; Mbitsemunda and Karangwa, 2017; Kyaw et al., 2018), and instrumental variable approach using two-stage least square (Kondo, 2019). While these studies cited above have different coverage and certain commonalities, the trend is relatively explicit-farm households participate actively in markets as a function of distinct concurrent decisions and sales volume choices, both of which are contingent on market participation. For example, Heckman's two-stage model uses the probit regression model to assess household market participation decisions. Then, the inverse Mills ratio derived from the binary probit regression model is merged with other predictor variables in explaining variance in the continuous, non-zero response variable (proportion of sales) (Kyaw et al., 2018). The primary premise of the Heckman model is that a given value of the outcome variable is observed if it is greater than a specific cutoff. While this aids in reducing selection bias, it has limits since it is unsuited for purely non-negative data. Unlike the Heckman model, the Tobit model is a single estimator that defines the choice between positive and zero sales and sales volume due to market participation (Adeoti, 2014). In addition, the Tobit model stipulates that the same drivers influence both market participation and sales quantity. Although the outcome variable (sales proportion) can assume negative values, in theory, censoring prevents them from being observed. Going one step further, Cragg's double hurdle model, like the Heckman model, is a two-stage econometric regression model. The model, however, is a generalization of the Tobit model, wherein two distinct stochastic processes drive market participation and sales volume decisions (Dlamini and Huang, 2019). The significant distinction in the two-stage models is predicated on Heckman's assumption that non-participants will likely not participate in the market under any situations (Kiwanuka and Machethe, 2016). The Cragg model is a more flexible variation of the Heckman and Tobit models in a specific context (Eakins, 2016). The model can be adjusted to account for heteroscedasticity by describing the variance of the errors as a function of a range of continuous variables (Aristei and Pieroni, 2008); even so, it cannot offer a complete depiction of the response variable. Again, another critical limitation of the double-hurdle specification is that it is based on the assumption of bivariate normality of the error terms; meaning, if the normality assumption is violated, the maximum likelihood estimates of the model will be discordant (Aristei and Pieroni, 2008). Thus, due to these limitations and apparent variability at household levels, applying models mentioned above or the classical regression model to derive inferences of the average household in this scenario may mask salient information of the outcome variable.

For this reason, we invoke a quantile regression technique to examine the relationship between covariates and smallholder commercialization across the conditional proxy for commercialization distribution. Quantile regression has numerous advantages, such as allowing for heterogeneous impact, but one downside is that parameters are difficult to assess than Gaussian or generalized regression (Waldmann, 2018). Nonetheless, only a few, if any, studies in SSA have employed a quantile regression to model smallholder farmer commercialization. We are also oblivious of any formal empirical studies employing quantile regression estimates to investigate sunflower commercialization in African agriculture. Furthermore, no other study has specifically investigated sunflower commercialization, except perhaps Mathagu et al. (2018), using a binary logistic model to examine the factors that affect smallholder market participation.

With all comparative advantages of increasing commercialization through sunflower production, this manuscript examines the drivers of sunflower commercialization and the heterogeneity at household levels while expanding the literature on this subject. Taking a completely different technique, this article compares quantitative empirical evidence that offers little insight into the effects of covariates on commercialization and conjectural intuitions with quantitative experiential work that highlights critical influences of regressors on household commercialization using a more deterministic statistical approach. The study also identifies possible research gaps that merit future studies in smallholder commercialization conceptual and analytical aspects. This study considered all smallholder sunflower farmers as market participants since all respondents sold portions, while others sold all their produce to NWK market outlets. Another contribution of this article is that no arbitrary cutoff or typology was used to categorize respondents (subsistence, semi-commercial, and commercial), as the impact of the explanatory variables differs across households. The article focuses primarily on the degree of household commercialization, employing quantiles ranging from 0 to 100% output sales. We contend that understanding these drivers and the quantiles to which the covariates influence sunflower commercialization within the district is critical for policy implications.

The remainder of this paper proceeds as follows. The following section briefly presents an overview of the sunflower sector in South Africa, preceded by the analytical context and data collection pattern. Next, the sections explain the analytical model used to evaluate the impacts of commercialization, then the summary statistics. The penultimate section explains the analytical result and discussions, and the last section presents inferences and addresses the policy implication of the result.

We now turn to interpretations of the multivariate quantile analysis for effects of predictors on the commercialization variable. As discussed in previous sections, this article provides more detailed insights into the impact of variables across the commercialization continuum, from subsistence to fully commercial households. Moreover, in contrast to the OLS regression, QR offers the latitude to estimate the complete distribution of the outcome variable based on values of regressors. While QR is regarded as an annex of OLS, it is advantageous when OLS assumptions, such as normality, linearity, and equal variance (homoscedasticity), are violated. Thus, we performed Jarque–Bera and Shapiro–Wilk W tests on normally distributed data, and the results suggest that the hypothesis of normally distributed data can be repudiated for all covariates (see Table 2). To substantiate the application of the quantile regression model, an unequal variance (heteroscedasticity) test was conducted; we observed that the Breusch–Pagan test statistic is substantially different from zero–Chi(14)2 = 52.18, prob > Chi² = 0.0000. Therefore, the residuals suffer from heteroscedasticity, and we are justified in the use of QR since the null hypothesis of equal variance is rejected in favor of the alternate hypothesis (presence of heteroscedasticity in the datasets).

The empirical quantile regression offers the medium for heterogeneous impacts; while some covariates influence the response variable at various quantiles, their effect changes and varies across the continuum. Thus, to properly assess whether changes in magnitudes of estimated impacts are significantly different across quantiles, we tested for equivalence of coefficients. We find that most estimates distinctly oppose the equality of predicted coefficients for three percentiles of the continuum. The F-values reject null hypotheses of equal slopes. Note each coefficient measures changes in ratios of sunflower sold resulting from a unit increase in continuous regressors, such as household size, or changes in dummy regressors from 0 to 1 (gender), while fixing other regressors. To aid interpretation of results, we invoke Joao Pedro Azevedo's graphical routine in STATA to display how the magnitude of each regressor varies across quantiles, in contrast to OLS fixed effects as indicated in Figure 2. Each panel in Figure 2 represents a regressor in the analysis, and vertical axes represent the magnitude of coefficients while horizontal axes represent quantiles. The horizontal solid dash line represents OLS coefficients. Therefore, they do not vary across locations of x-axes and are within 95% confidence interval (dotted lines). The solid grey polygonal line displays QR coefficients and confidence intervals. The QR coefficients are plotted as lines that differ across quantiles with confidence bands. For example, it can be inferred that the effect of regressors is significantly different from OLS estimates, where QR coefficients are outside OLS confidence bands (indicated by the + symbol in Table 3). We estimated seven distinct quantiles, ranging from 0.05th to 60th quantiles. We also consider that commercial point is attained at the 60th quantile because our observations contain a lot of 100 percent sales. Thus, it will be inconsequential to estimate the model at higher quantiles. For policy implications, households at 60th to 95th quantiles sold all their sunflower produce. Similar to mean effects, the median (q = 0.50) estimates the central tendency of a dataset by comparing the statistical significance and size of median coefficients with conditional quantiles, permitting distinct impacts of regressors across the commercialization continuum (explained variable).

It is noteworthy to state that, to keep the volume of the result to a minimum, we narrowed our discussions to selected covariates. The intercept plot in the first panel of Figure 2 is self-referential. It depicts the estimated conditional quantile function of households selling lower sunflower output at lower quantiles of the commercialization continuum than households at higher quantiles. The graph shows a positive upward slope line over the conditional quantiles, in contrast to the lower quantiles, indicating a positive impact in the upper quantiles. One point of QR, as seen in panel 5, is that certain covariates with significant influence did not exhibit any inversion signs across the quantiles, i.e., the regressors did not have a positive impact at some quantiles while having a negative effect at others. The OLS estimate in Table 3 reveals, for example, that the average quantity of sunflower sold by respondents is uniquely associated with the age of the household head, market outlet, and farm size allocated to sunflower production, while other factors are held constant. The OLS coefficient for the market outlet, which is a variable of interest, is positive, indicating that increase in the use of market outlets will, on average, result in an 11.4% rise in the ratio of output sold, while other regressors remain constant.

Quantile regression plot indicates that the estimated coefficient of market outlet is positive as well, although with varying impact across the conditional distribution of the response variable, as suggested by the equivalence test F_(2,157) = 3.24, p < 0.043. When q = 50% quantile, the market outlet appears to have a minimal effect on the commercialization continuum. The coefficient of this variable tends to increase for less commercial sunflower farmers in the lower quantiles. Another point worth highlighting is that when the quantile increases, the estimated impacts decrease, implying that the magnitude is greater for less commercial households. Evidently, the QR findings suggest that the presence of formal market outlets such as NWK (main market outlet) is indeed a significant contributing factor to the success of fast-transitioning households (from subsistence to commercial). Market outlets are characterized by several driving factors such as geographical proximities, structures, i.e., power asymmetry, expertise of the producer, adherence to quality standards, and legislations (Dolan and Humphrey, 2000; Blandon et al., 2009). Comprehensive empirical studies on smallholder farmers in SSA provide indications of bottlenecks in accessing formal marketing outlets (Martey et al., 2012; Arinloye et al., 2015). By contrast, if access is not hampered, market outlets influence shares of sale outputs, as in our case. Arinloye et al. (2015), Mathagu et al. (2018), and Kassaw et al. (2019) found consistent evidence that formal market outlets positively influence smallholder market participation and sales proportions. A farmer weighs the expected utility returns against the associated expenditure, entry barrier, and other transaction costs when deciding whether to engage in a given market. The decreasing effect of this predictor at the upper quantile of the continuum is presumably due to a lesser likelihood of farmers increasing their sales ratio as they progressively attain a full commercial state.

The remaining regressors have minor impacts on certain quantiles while exerting no effect on others. This evinces discernable cluster effects, given that most of predictors appear to affect lower quantiles of the continuum. More specifically, the magnitude of these regressors is skewed toward the lower quantiles, i.e., less commercial households, apart from the farm system variable, which ignores the clustering pattern, impacting q10, q50, and q60, respectively.

The following regressors are discussed in order of statistical significance quantiles. Contrary to OLS fix effects, household size is statistically significant but negatively influenced sale portions from 0.05th to 40th quantiles. As quantile increases, the scale of the variable diminishes, but a statistically insignificant spike emerges in the middle of the conditional outcome distribution (at the 50 quantiles). Importantly, the equivalence test is statistically significant—F_(2,157) = 2.66, p < 0.073, evidencing the heterogeneous impact of this regressor on the outcome variable. The negative impact is premised on the assumption that as the number of dependents increases, households at the lower part of the commercialization continuum incur limited returns on sale outputs. Analogously, Kassaw et al. (2019) and Abate et al. (2021) find that the larger the household size, the less commercially oriented tomato farmers are in the South Gonder zone, Ethiopia. Insofar, the QR coefficient in panel 4 sloped downward from the lowest quantiles to the 40th before stabilizing, illustrating that households in this category are less market oriented, possibly owing to financial constraints. In this context, a farmer may be pushed to redistribute scarce farm income to other household demands and responsibilities, inhibiting their production and, as a result, lowering farm outputs. Compared with the average estimate, farm size has a non-monotonic impact across the conditional quantiles, with a significant influence at the lower quantile of the response distribution. The QR coefficient of this variable displays a positive upward movement from the lowest quantile up until the 50th quantiles well over the OLS confidence interval, as seen in Panel 4 of Figure 2. The null hypothesis of equal slopes is likewise rejected by the Wald test of equality F_(2,157) = 4.00, p < 0.020, confirming the differing impact of farm size on the percentage of sales volume. In certain instances, this explains that the desire to increase hectares under sunflower production indicates commercial orientation in the lower 40th quantiles. It also suggests that expansion aspirations are more heightened for less commercial farmers. At first glance, this may appear counterintuitive that the positive effects only emerged at lower quantiles. Nevertheless, we can advance some tentative interpretations. To begin with, farmers at lower quantiles are more likely to be driven by expansion plans due to inherent effects on quantities produced, therefore raising sale outputs. While as a farmer reaches full commercial status, the urge to extend the area under cultivation diminishes since differences generated would be negligible, mirroring diminishing marginal returns. Abate et al. (2021), Aku et al. (2018) and Radchenko and Corral's (2018) also find that larger farm holdings favor less commercial households. The farming system variable impacts both lower and upper quantiles. The regressor was used to proxy whether a farmer owns irrigation infrastructure or depends on rainfed agriculture. While the association between farming systems and share of output sold is statistically significant at 10th, 50th, and 60th quantiles, the impact is negative and minuscule across quantiles, meaning the regressor is marginally sensitive to commercialization. Severe crop failure, increased biodiversity loss, and land degradation experienced in the district and other dryland regions due to agricultural intensification explain the negative association between the farming system and the share of sale outputs. The effect is visible across the continuum (less and full commercial households) with and without irrigation infrastructures. This is also reflected in the equivalence test F_(2,157) = 0.34, p < 0.713 indicating a homogenous effect. Twomlow et al. (2015) bullishly point that agricultural intensification void of appropriate exogenous input factors often results in low crop yields and farmland degeneration through the synthesis of crop nutrient absorption and weak soil conservation.

Household education attainment, like most other predictors, is insignificant in OLS but with a negative statistical impact in QR. Table 3 shows that the magnitude of this variable manifested at 0.05th and 0.10th quantiles (the least commercial households) but with an insignificant reversal effect from 0.20th quantiles. Ironically, farmers at this level—especially those engaging in cash crop production (sunflower), see farming as a means rather than an end, likely due to the wide range of constraints that are associated with emerging households. Given the hypothesized effect in summary statistics, it is reasonable to establish that the educational attainment of least commercial households stimulates non-farming activities, decreasing the share of their sale outputs. The equivalence test result F_(2,157) = 4.00, p < 0.026 explicitly contradicts the homogeneity of estimated coefficients, indicating that the impact of education differs across quantiles of the conditional response distribution.

As for the gender of the household head, the equality test F_(2,157) = 1.95, p < 0.146 fails to account for the heterogeneity of coefficients, with a pronounced decreasing significant positive effect at the lower quartiles. The findings demonstrate that male-headed households are more commercialized than female-headed households, suggesting a significant impact of gender inequalities at the lower end of the commercialization variable. Again, this is expected, considering the high male dominance in the district sunflower sector. Abdullah et al. (2019), studying the determinants of commercialization and its welfare impact, similarly finds that male rice farmers had the highest commercialization ratios, highlighting gender inequalities at household levels. This corroborates the findings of Yaméogo et al. (2018) that male farmers have a higher probability of participating in markets than female household heads, mainly due to inaccessibility to traditional lowlands. On the flip side, gender inequalities in farming operations evince the challenge of a diver, including limited access to farm inputs and low outputs returns, land rights restrictions, and gender-based commodity market anomalies faced by female household heads (Rodgers and Akram-Lodhi, 2019). At the root of these drawbacks are unfair power structures, gendered norms, traditions, and a relatively static gender severance of labor found at household levels, which impedes their agricultural market participation. On the other hand, it is reasonable to infer that female-headed households are customarily weighed down by more domestic chores, causing them to allocate less time to farm operations, which is consistent with a related study by Lekunze et al. (2011), acknowledging gender disparity in provincial sunflower production. As a result, closing gender gaps will require a shift in power structures and dynamics.

Last, yet importantly, is the proxy for production and market information access. As quantile increases, the variable appears to only exert a positive decreasing significant impact at the 0.05th and 10th quantiles. The variable underscores the importance of information (production and market information) for less commercial farmers as it influences the proportion of their sales outputs. Obtaining production and market information displays a desire to improve and strengthen their farming operations, leading to higher sales. The equivalence results verified the varied impact F_(2,157) = 4.43, p < 0.013, revealing that the magnitudes differ across the conditional-response distribution. However, as households progress into a full commercial state, they are more likely to have acquired both experience and expertise from advisory services, so the residual effects diminish with more information and advice until they are entirely waned (Kostov and Davidova, 2013).

In contemporary economic assessment, agricultural commercialization is assigned a core role in rural development and wealth creation. Empirical evidence suggests that commercialization is rapidly being advocated across numerous African countries seeking to achieve the United Nations food security goal through smallholder farming systems. Using 177 observations from a single district in North West Province, South Africa, this paper empirically investigates the auxiliary benefit of the quantile regression technique as a strategic policy instrument to examine determinants and variations of agricultural commercialization. The principal finding of this study is that the degree of smallholder commercialization differs significantly across households in the study area. In addition, sunflower commercialization, at mean, is uniquely correlated with hectares cultivated, household head, and market outlet among samples. The emphasis of ordinary least square estimates is whether variation counts on average. However, this supposition may inaccurately depict the response distribution. For this reason, bootstrap QR was implemented to tackle this limitation. Besides, QR permits for descriptive assessment of covariates across the outcome distribution, resulting in heterogeneous impact. For instance, certain regressors exert different effects at various quantiles; their impact differs for different households, as suggested by equivalence tests. These insights can help policymakers and stakeholders design inclusive measures for rural sunflower farmers to increase commercialization. Typically, OLS and other models used in market participation literature are unlikely to offer this insight. The empirical findings confirm the essentiality to look beyond central tendency, location, and direction in which regressors are associated with the commercialization variable to understand the heterogeneous influence at household levels.

The findings indicate that heterogeneity in sunflower commercialization cannot be disregarded for the case of household size, market outlet, hectares cultivated, education, access to information, and gender. Market outlet and farming system variables appear to have maintained their significant impact, although with varying magnitudes across all quantiles. This does not negate the significance of the remaining covariates. The influence of the market outlet is positive, as evidenced by the QR result, exhibiting a sharp increase from lower quantiles of the conditional response distribution but a diminishing magnitude from the median quantile. The farming system variable negatively impacted lower and upper quantiles with homogeneously minuscule effects. While irrigation infrastructure is vital, we advised that agricultural intensification without appropriate complementary input factors would enervate irrigation outcomes. Finally, the remaining regressors showed heterogenic clustering patterns. Their effects centered primarily at lower quantiles of the commercialization distribution (less commercial households), indicating that less commercial households bear the full impact of these regressors, not vice versa. According to QR findings, male household heads are more market oriented than female heads, especially at lower quantiles. Gender imbalance hinders the competitiveness and contributions to the agricultural sector and the broader rural economy and social development goals of women. It is surprising that despite being ranked fifth on the Sustainable Development Goals and receiving a plethora of policy plans, it seems that attention to rural agrarian gender disparities is often casual and laissez-faire rather than being taken as a critical element of sustainable agriculture. Rodgers and Akram-Lodhi (2019) tie the marginalization of women to the complexities of scaling the ladder of high-value crops owing to farmland ownership, size, quality, and other systemic social barriers. Consequently, increasing sunflower commercialization among female household heads in the study area will require intervention that addresses systemic societal barriers impeding their participation.

To recap, the findings of this study show that exploring the ordinary least and quantile regression approach presents distinct pictures of the outcome variable and set of predictors. Furthermore, this method enables the identification of specific areas where gaps occur, resulting in valuable insights on how to tailor policies to minimize these gaps. The findings also suggest that policymakers need to recognize the existence of heterogeneity in commercialization and be wary of using a blanket or conventional macro policy approach, which, even if effective, is unlikely to achieve overarching welfare and behavioral responses across targeted households. Of course, a no victor, no vanquish approach would entail dismantling other microscale and mesoscale barriers. Those in policy space working to promote smallholder commercialization should be cognizant of potential benefits of efficacious sunflower promotion in SA and other SSA nations. Given that this study only looked at cross-sectional data from a single district municipality, a primary assignment for future studies may be to understand how viable commercialization frameworks between smallholders and market outlets have been structured so that their successful components can be mirrored and embedded more expansively into commercialization initiatives. This can follow a panel data approach involving the remaining districts or provinces. Next, in-depth studies are needed to evaluate the welfare response to agricultural development initiatives on women's empowerment in the sunflower sector. Finally, this study and the inferences reached should be viewed in light of its limitations.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by College of Agriculture and Environmental Sciences Health Research Ethics Committee (reference number 2020/CAES_HREC/006), University of South Africa. The patients/participants provided their written informed consent to participate in this study.

EA was the principal researcher for the study and took the research from inception, including fieldwork, data analysis/management, and drafting of the manuscript. ST and OO provided expert proofing and reviews of the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.