Built upon theoretical foundations in informality, urban risk, and food governance, the model synthesizes the interplay between exogenous shocks such as climate change, COVID-19, and social unrest and endogenous structural barriers, including infrastructure limitations and governance exclusion. As illustrated in the framework (Figure 1), climate events (e.g., floods), pandemic disruptions, and social unrest (e.g., the 2021 KZN riots) act as converging stressors that destabilize informal market systems. These shocks disrupt supply chains and intensify infrastructure gaps, such as the absence of cold storage, unreliable transport, and informal trading spaces. These systemic deficiencies, in turn, exacerbate vendor-level challenges, notably income loss, spoilage of perishable indigenous crops, and mounting fixed costs.

Moreover, the framework highlights the bidirectional relationship between vendor challenges and policy/systemic barriers. Vendors are often excluded from infrastructure development, emergency response systems, and food governance dialogs thus limiting their capacity to adapt or recover. Simultaneously, worsening vendor vulnerability reinforces their marginal status, perpetuating a cycle of neglect in policy planning and resource allocation. The final node food security outcomes capture the downstream impacts on access, nutrition, and affordability. As vendors struggle to maintain income and stock, urban consumers lose access to culturally relevant, nutritious indigenous crops. The framework thus situates informal vendors not only as economic actors but also as frontline agents in urban food resilience. This model offers a systems-level perspective aligning with recent calls for integrative urban food governance. It moves beyond single-shock models by considering how multiple, compounding stressors intersect with existing inequalities, providing a more realistic lens to understand the fragility of informal food networks in cities like Durban.

This study employs a robust linear regression model to investigate the determinants of income loss among street vendors selling indigenous food crops in the eThekwini Metropolitan District. The choice of robust regression (rreg) is both methodologically and empirically justified. Preliminary diagnostic tests revealed violations of key ordinary least squares (OLS) assumptions, including the presence of heteroskedasticity (Breusch–Pagan/Cook–Weisberg test, p < 0.05) and influential outliers, which are common in small-sample, survey-based data collected from informal vendors. OLS is known to be sensitive to such violations, leading to inefficient and biased parameter estimates. In contrast, robust regression employs an iteratively reweighted least squares (IRLS) approach that minimizes the influence of extreme residuals. It begins with Huber weights to reduce the leverage of moderately deviant observations and progresses to Tukey’s biweight function to suppress the impact of extreme outliers, resulting in more stable and reliable estimates.

Although the sample size of 34 is small, it constitutes a complete census of formally registered indigenous crop vendors in the Warwick Early Morning Market. Given this bounded population and the presence of categorical predictors, we employed a robust regression model to account for potential non-normality and outliers, which are common in small datasets from informal sector research. Robust regression down-weights extreme values and provides more reliable coefficient estimates than ordinary least squares in such conditions. As shown in (42), who applied robust regression to two small-sample neuroscience datasets (n = 37 and n = 46), the method yielded stable and conservative results when classical assumptions were violated. They further argue that the use of parametric models is statistically valid for sample sizes ≥ 30 based on the Central Limit Theorem (CLT). With our sample of n = 34, this principle supports the appropriateness of a parametric approach. Furthermore (43), demonstrated via Monte Carlo simulations that robust regression performs consistently well in small samples (n = 6 to 20), especially where outliers are present. Collectively, these methodological precedents validate the use of robust regression in this study.

The use of this method is further supported by recent studies in agricultural economics. For example (44), applied robust estimation techniques to model farm income among cassava farmers in Nigeria, concluding that robust regression outperformed OLS in the presence of data irregularities. Similarly (45), used robust regression to identify factors influencing household income among smallholder farmers, demonstrating that education level was a stronger determinant of income than pesticide usage itself, highlighting the method’s utility in agricultural livelihood research. These precedents highlight that robust regression is well-suited for analyzing income data within informal, climate-sensitive agricultural markets such as those observed in this study.

The econometric model is formally specified as follows (Equation 1):

Where: TotalIncomeLossᵢ is the income loss (ZAR) reported by vendor i . The independent variables are defined, measured, and explained in detail in Table 2, which includes socio-demographic factors, structural constraints, shock exposure variables, and market/product risks. The term εᵢ denotes the error term capturing unobserved heterogeneity.

The robust regression estimates the model by minimizing the weighted residual sum of squares (Equation 2):

Where:

Y i : The observed value of the dependent variable (Total income loss) for observation i .

This approach preserves all observations while reducing the influence of outliers and ensures more reliable coefficient and standard error estimates, especially in small-sample, non-normal, or heteroskedastic datasets.

While the model includes a significant number of categorical and binary variables, this reflects the socio-economic realities of the vendor population—where characteristics such as education level, crop type, access to storage, and exposure to shocks are inherently categorical. These variables are essential for capturing the multidimensional factors influencing income loss and market participation. To address potential multicollinearity, dummy variables were constructed with appropriate reference categories, and model specification was carefully examined. Variance Inflation Factor (VIF) analysis was conducted, and no predictor exceeded the conventional threshold of 10, indicating no serious multicollinearity concerns. The use of robust regression further supports the reliability of the findings, as it accommodates violations of classical assumptions associated with small sample sizes and categorical predictors. Although the inclusion of multiple categorical variables introduces some interpretive complexity, the model yields directionally consistent and contextually meaningful insights into the socio-economic drivers affecting indigenous vendors in the post-shock environment.

As shown in Table 6 Ubhatata (Ipomoea batatas, sweetpotato) emerged as the most widely traded crop, sold by 70.6% of vendors. This was closely followed by amadumbe (Colocasia esculenta, taro) at 64.7%. Traditional green leafy vegetables such as imfino yezintanga (Cucurbita maxima leaves) and imbuya (Amaranthus spp.) were sold by 38.2 and 29.4% of vendors, respectively. Less frequently sold crops included umdumbulu (Manihot esculenta, cassava) at 5.9% and amantongomane (Arachis hypogaea, groundnuts) at 8.8%, reflecting their more specialized or seasonal demand.

Significant price variation was observed across crops. Ubhatata and amadumbe recorded the highest price ranges, R40–360 (US$2.09–18.82) and R35–400 (US$1.83–20.91) respectively, and were generally sold in 50 kg bags, indicating bulk trading. Imfino yezintanga and imbuya were more affordably priced between R5–30 (US$0.26–1.57) and typically packaged in small bundles or bags, enabling access for low-income consumers. Umdumbulu and amantongomane were sold by weight or in pre-measured batches, offering greater flexibility in pricing and packaging.

Table 6 indicates that weekend trading dominated peak sales periods for most crops, particularly ubhatata (41.2%) and amadumbe (38.2%), suggesting alignment with consumer purchasing cycles and market foot traffic. Imfino and imbuya had modest weekday sales, while cassava and groundnuts were sold irregularly, indicating sporadic availability or lower mainstream demand.

Sourcing varied depending on the crop. Ubhatata and amadumbe were largely sourced from rural farmers and highland regions via traders, reflecting relatively extended value chains. Imfino and imbuya were mostly grown by vendors themselves or acquired from local rural suppliers (as shown in Table 7), demonstrating short, informal supply loops. Cassava was obtained from smallholder farmers, while groundnuts were accessed from both farmers and wholesalers, implying wider sourcing networks.

Handling methods reflected the perishability of each crop. Imfino and imbuya lacked cold chain support and were prone to spoilage, despite being wrapped in damp material. Amadumbe required crate storage to avoid mold, while ubhatata maintained shelf life under dry conditions but could sprout over time. Cassava remained perishable and largely unprocessed, while groundnuts benefited from dry storage and long shelf stability, making them a low-risk commodity for vendors.

Consumer usage aligned with local dietary norms. Root crops like ubhatata, amadumbe, and umdumbulu were used as staple starches, commonly boiled or mashed. Imfino and imbuya were used in traditional vegetable dishes or relishes, maintaining cultural culinary relevance (see Table 7). Amantongomane (groundnuts) were consumed as snacks, added to stews, or processed into paste. Nutritional profiles emphasized the value of these crops: imbuya was particularly high in iron (8.9 mg) and vitamin A (850 IU), while groundnuts provided dense levels of protein (25.8 g) and fat (38.3 g), making them essential contributors to energy and nutrient intake (46).

Table 8 reports the Variance Inflation Factors (VIFs) for all explanatory variables to assess multicollinearity. All VIF values were below the conventional threshold of 10, with the highest being 3.09 (for Age Category). The mean VIF was 1.87, indicating a low level of multicollinearity among the predictor variables. This suggests that the regression coefficients are not distorted due to strong correlations among independent variables.

The robust regression analysis presented in Table 9 examines the influence of various demographic factors, market shocks, and structural constraints on the total income loss experienced by indigenous crop street vendors following the COVID-19 pandemic, the 2021 KwaZulu-Natal unrest, and the 2022 floods. The model explains approximately 51.6% of the variation in income loss (R-squared = 0.5160), indicating a good fit. Several variables, including education level, age category, and renting in the market, show statistically significant effects on income loss among vendors.

The negative coefficient (−R2,970.91) suggests male vendors experienced marginally lower income losses than females, though this effect was significant only at the 10% level (p = 0.098). The 95% confidence interval spanning from -R6,541 to R599 includes zero, indicating uncertainty about the true effect size. The marginal significance (p < 0.1) warrants caution in interpretation given the small male subsample (n = 5). Each additional household member was associated with R699.43 higher income loss (p = 0.003), with the narrow 95% CI excluding zero. This implies 99% confidence that the true effect lies between -R1,127 and -R272 per member. The result mirrors findings by (47) in East Africa, where larger households exhibited greater financial fragility during COVID-19, likely due to heightened subsistence pressures.

The regression results reveal a statistically significant and positive association between education level and income loss (β = R4,827.49; p < 0.001), with a 95% confidence interval ranging from R3,422 to R6,233. Stall renters endured R11,696.40 greater losses (p < 0.001), with the 95% CI (−R15,010 to -R8,383) confirming high precision. This aligns with (48), where fixed costs amplified vendor vulnerability during West African food crises. The CI width (R6,627) suggests moderate variability around this large effect. COVID-19-affected vendors lost R4,045.47 more income (p < 0.01), with 95% confidence the true effect lies between R1,470 and R6,621. The CI’s positive range confirms the shock’s severe, unidirectional harm.

Flooding caused R3,986.45 additional losses (p < 0.01), with 95% CI bounds (R1,697 to R6,276) excluding zero. The narrow range (width = R4,579) indicates precise estimation, supporting (28) on climate shocks’ disproportionate effects on perishable crop vendors. Storage-constrained vendors faced R2,501.41 higher losses (p = 0.061), but the 95% CI (−R128 to R5,130) includes zero, suggesting marginal significance. While the point estimate aligns with (29) on post-harvest losses, the CI’s inclusion of null effects necessitates cautious interpretation. Perishability drove R3,651.53 greater losses (p < 0.05), with 95% CI (R761 to R6,842) confirming significance. The CI’s lower bound >0 confirms robustness. Perceived quality decline led to R7,197.85 higher losses (p < 0.001), with 95% CI (−R10,011 to -R4,385). The entirely negative CI strongly supports (49) on quality-price linkages in informal markets. The effect’s magnitude underscores consumer sensitivity to perceived crop integrity.

The annual stall rental fee of ZAR 530 (≈USD 28.19) reflects a degree of institutional support and municipal subsidy that enables traders to access formal trading platforms within the Early Morning Market. Such access is vital, as it provides physical visibility, consumer foot traffic, and perceived legitimacy factors that enhance both income generation and consumer trust. However, these low entry costs may mask the high cumulative costs of daily operations, particularly for vendors with limited capital and no formal financing mechanisms. This echoes findings by (27), who emphasized that while market stall allocations and informal space tenure remain affordable, the true economic cost of informality lies in operational volatility, such as unpredictable storage access, ad hoc labor expenses, and supply chain disruptions.

Vendors spend ZAR 50 (USD 2.66) weekly on weekday storage and ZAR 40 (USD 2.13) over weekends, incurring monthly expenditures that quickly outstrip the annual stall rental cost. These costs are not merely financial; they reflect a systemic infrastructure gap in post-harvest preservation, which contributes directly to produce spoilage, revenue loss, and market instability. This is particularly concerning given that 82.35% of vendors reported that indigenous crops “spoil fast,” confirming that inadequate cold storage undermines the shelf life of perishable crops such as imfino and imbuya. This aligns with (60, 61), who stressed that investment in cold storage infrastructure is essential for reducing post-harvest losses and stabilizing income for small-scale traders of perishable indigenous crops. Without structural solutions, recurring storage expenses act as a tax on fragility penalizing traders who are already vulnerable to climatic and market shocks.

The daily fee of ZAR 50 (USD 2.66) to hire street porters highlights a common workaround among informal vendors who lack access to personal vehicles or motorized logistics. For vendors managing bulk commodities like amadumbe and ubhatata, manual transport services are indispensable but costly. Over a standard 20-day working month, this translates to ZAR 1,000 (≈USD 52.63) almost twice the annual stall rental. This aligns with urban transport and labor burden findings from (62), who found that labor costs disproportionately affect women-led food vending enterprises, due to gendered constraints on mobility and asset ownership. For women who comprise 85.3% of the vendor population in this study outsourcing heavy logistical tasks is not just convenient but necessary, yet it further diminishes net profits. Collectively, these cost dynamics expose the invisible pressures that constrain income potential and scalability within informal food systems. While indigenous crops hold nutritional and cultural value, their market success is undercut by the accumulated weight of logistical and post-harvest expenses, which are rarely accounted for in policy design. Without targeted interventions—such as subsidized storage, vendor cooperatives for shared porter services, or mobile refrigeration units—the current cost structure will continue to entrench fragility and limit the expansion of indigenous food markets.

The most widely sold indigenous crops were Ubhatata (Ipomoea batatas, sweetpotato, 70.6%) and Amadumbe (Colocasia esculenta, taro, 64.7%), both starchy root crops that form a cornerstone of Zulu diets and offer significant caloric value. These findings resonate with those of (28) who identified root and tuber crops as crucial to dietary energy security in sub-Saharan Africa due to their adaptability, affordability, and cultural relevance. The lower market representation of crops such as Umdumbulu (cassava, 5.9%) and Amantongomane (groundnuts, 8.8%) may reflect seasonal availability constraints or supply chain fragmentation. The market’s crop diversity, while narrower than national or rural inventories, reflects consumer familiarity and preparation knowledge, which are often key drivers of urban indigenous crop consumption (63). Moreover, the sale of Imfino yezintanga (Cucurbita maxima leaves, 38.2%) and Imbuya (Amaranthus spp., 29.4%), both traditional leafy greens, confirms that street vendors remain essential custodians of indigenous vegetable biodiversity in urban settings.

The findings illustrate a distinct segmentation in crop pricing and packaging. High-volume, high-value crops like sweetpotato and taro were packaged in 50 kg bags, with prices ranging up to ZAR 400 (≈US$20.91), indicative of wholesale trade to households or secondary vendors. In contrast, imbuya and imfino were sold in small bags or bundles at R5–30 (≈US$0.26–1.57), targeting low-income consumers and enabling micro-purchases. This mirrors patterns reported by (11), who emphasized that informal markets excel at customizing food access in low-resource settings through flexible packaging and price structures. These pricing patterns also suggest value chain asymmetries: while bulk crops facilitate vendor economies of scale, leafy vegetables ensure consistent cash flow, albeit with higher spoilage risk. Notably, weekends were peak trading days for most vendors, especially for bulk starchy crops, reflecting a rhythm in consumer demand and market foot traffic (64).

The results confirm substantial variation in post-harvest vulnerability, particularly among leafy vegetables. Imfino and imbuya were highly perishable and lacked any cold chain infrastructure, echoing warnings from (65), who identified the absence of post-harvest innovations as a leading bottleneck in the indigenous vegetable value chain. These crops were often wrapped in damp paper, which offers only short-term freshness retention and exposes the produce to fungal spoilage and nutrient loss under high temperatures. In contrast, amantongomane (groundnuts) were dry-stored and shelf-stable for months, representing a low-risk, energy-dense alternative with market longevity. Ubhatata and amadumbe, while less perishable, face sprouting or mold risks if storage conditions are inadequate. The evidence supports the argument for introducing affordable, decentralized storage technologies, such as evaporative cooling or crate-based ventilation systems, as proposed by (66).

The consumer applications of these crops reflect their dual role in cultural preservation and nutritional support. Roots like cassava and sweetpotato are used as boiled staples, while leafy greens are integrated into traditional stews. Amantongomane (groundnuts) serve as both snacks and protein complements, underlining their dietary versatility. The nutrient highlights demonstrate that these crops can contribute to dietary adequacy. Imbuya, with 8.9 mg of iron and 850 IU of Vitamin A, can help combat micronutrient deficiencies, while groundnuts, with 25.8 g protein and 38.3 g fat, offer high energy density—a critical asset in food-insecure households. These findings support the advocacy by (67), who argue for elevating the policy status of indigenous foods as “nutrition-sensitive crops” that can bridge micronutrient gaps in urban poor communities. This study confirms that indigenous crops remain underutilized yet essential in urban food environments. Their diversity, affordability, and nutrient profile position them as strategic assets in building resilient, low-carbon, and culturally rooted food systems (28). However, their economic potential is constrained by infrastructural deficits, short shelf life, and narrow market reach.

The results show that female vendors experienced higher income losses than their male counterparts (p = 0.098), corroborating longstanding research on gender-based vulnerability in food systems. Female actors, particularly in the Global South, frequently operate at smaller scales and face disproportionate access barriers to land, capital, and formal credit, exacerbating economic exposure during crises (27). Similar findings have been reported in Zimbabwe and Malawi, where women in informal markets had less recovery capacity post-shock due to asset poverty and household care burdens.

Household size was also significantly associated with income loss (p = 0.003), with larger households experiencing greater vulnerability. While large households are sometimes thought to offer economic resilience through labor pooling, they may also increase food, and income demands during economic stress. This is consistent with evidence from East African urban food systems where large household sizes were linked to increased food insecurity and income volatility during COVID-19 lockdowns (47, 68).

The findings show that both age and education were positively associated with income loss may appear counterintuitive, as these factors are often linked to experience or adaptive capacity. However, in this context, they likely signal business scale and cost exposure. Although counterintuitive, the positive relationship between education and income loss may reflect that better-educated vendors manage larger operations, invest more in inventory, and rent formalized stalls exposing them to higher losses when market activity is disrupted. Thus, education may proxy business scale rather than resilience. This interpretation aligns with findings from informal markets in Lagos, where more experienced traders despite their skills reported larger absolute losses during COVID-19 due to their greater operational footprints (69). Thus, education in this setting may serve as a proxy for formalization and investment level rather than resilience per se.

Among the most substantial predictors was market rental, with vendors who rent stalls experiencing R11,696 more in income loss (p < 0.001). This reflects how fixed costs in informal systems, despite their modest absolute value, become untenable under supply shocks and revenue collapses. The importance of rental exposure is also supported by (70), who found that rental debt accumulation during COVID-19 forced many small traders in Harare into default and permanent closure.

The regression revealed that COVID-19 and flooding were significantly associated with elevated income loss (p = 0.004 and p = 0.002, respectively), underscoring the compounding nature of health and climate stressors on informal food systems. During the COVID-19 lockdowns, street vendors faced mobility restrictions, perishability challenges, and demand contractions. Similar disruptions were observed in Ghana and Kenya, where traders’ supply lines were fractured by travel bans, resulting in widespread spoilage and income collapse (71, 72). Flooding, as a climate-linked event, created direct barriers to market access, destroyed inventory, and damaged physical infrastructure, including informal storage spaces. These findings align with (73), who documented how flood events in informal Ugandan vegetable markets reduced vendor turnover and decimated perishables, especially among vendors lacking access to elevated or enclosed market spaces.

Post-harvest challenges were found to meaningfully shape income trajectories. Vendors who reported fast spoilage of indigenous crops incurred R3,652 more in income loss (p = 0.016), highlighting the vulnerability of leafy greens and unprocessed tubers. Similarly, limited access to storage was marginally significant (p = 0.061), but the estimated effect size suggests substantive impact. These findings resonate with (28), who argue that underutilised crops are severely constrained by lack of post-harvest innovations in sub-Saharan Africa, particularly cooling, sorting, and packaging technologies. Most notably, vendors who perceived that crop quality had decreased experienced R7,198 greater income loss (p < 0.001). This perception likely captures both physical deterioration due to weather variability and reduced consumer trust in product integrity. Prior work has emphasized that the subjective quality of indigenous foods affects not only pricing but also market re-entry rates for vendors post-harvest (74).