Evaluating the Impact of Agricultural Credit Access on Smallholder Maize Farmers’ Productivity in the Northwest Region of Cameroon

Abstract

Food insecurity and poverty persist in Sub-Saharan Africa, largely driven by low agricultural productivity linked to limited credit access. This study investigates the effect of credit access on the productivity of smallholder maize farmers in Northwest Cameroon. Using cross-sectional data from 404 interviews and an endogenous switching regression model to address self-selection bias and unobserved heterogeneity, we show that access to credit significantly improves maize yields. However, credit access remains limited and is negatively influenced by farmer experience and group membership, while extension services and agricultural training enhance access. We recommend targeted financial support and capacity building to improve credit uptake and boost productivity in the maize sector.

1. Introduction

Agriculture underpins many Sub-Saharan African economies, including Cameroon, contributing significantly to both employment and food security. It serves as a source of foreign exchange earnings and a gateway to poverty alleviation for many less developed nations. The sector remains the major promoter of economic growth and development, employing approximately 70% of the rural population of Cameroon. Despite the agricultural sector’s importance, productivity remains low due to the lack of access to finances used to purchase inputs required to boost production and improve yields [1]. This low productivity could be due to the underutilization of large areas of arable land, leading to a steady increase in food insecurity [2]. Therefore, higher agricultural productivity brings about sustained industrial development.

Agricultural credit plays a critical role in enhancing farm productivity by alleviating liquidity constraints and enabling timely access to essential inputs such as fertilizers, hired labor, and improved seeds. According to Becker’s production theory [3], output depends on both the quantity of inputs and the efficiency of their use. In this context, credit facilitates the optimal application of inputs, thereby leading to higher productivity. Since smallholder farmers often experience irregular and seasonal income flows, credit acts as a financial bridge that allows them to pre-finance input purchases before harvest [4]. Without adequate access to credit, farmers may underinvest in crucial inputs, resulting in lower yields. This is particularly true for input-intensive crops like maize, where credit access helps mitigate cash-flow constraints and supports investment in productivity-enhancing technologies [5].

In Cameroon, maize production fell from 2.25 million metric tons in 2017 to 2.1 million in 2022, despite projections that it would rise to 2.3 million in 2023. Average yields are 2.5 t/ha cultivated in semi-intensive zones and 4.5 t/ha in large-scale farming [6]. Maize, primarily produced by small-scale subsistence farmers, is essential for food security, income, and raw materials for various industries. However, low productivity persists due to limited access to modern inputs, technology, extension services, and agricultural credit. Thus, improving credit access in Cameroon could enhance maize production by facilitating investments in modern farming techniques and inputs.

Although maize remains a low-productivity crop in Cameroon, existing agricultural credit policies are largely underdeveloped and ineffective. For instance, the weak implementation of the National Strategy for Inclusive Finance, which aims to promote financial inclusion, has limited actual access to credit due to poor infrastructure [7]. According to Business in Cameroon [8], formal credit institutions continue to prioritize large agribusinesses over small-scale farmers. This gap has led to the expansion of microfinance institutions like MUFID into rural areas, although these institutions often face high default rates and limited capital and are therefore reluctant to lend to farmers. As noted by Bime [7], ongoing insecurity in the Northwest region and high interest rates have further constrained both access to and repayment of agricultural credit.

Research conducted in the Northwest Region of Cameroon has consistently shown that limited access to credit significantly hampers maize productivity among smallholder farmers. For instance, Wenda et al. [9] reported that approximately 55% of maize producers in the region lacked any form of credit, and that access to microcredit was positively associated with maize output, based on a generalized least squares model. Similarly, Manu et al. [10] identified access to loans, education, gender, and extension services as key socioeconomic factors influencing the adoption of improved maize varieties. These region-specific findings underscore the critical role of both formal and informal credit mechanisms in enhancing maize productivity in the Northwest. While numerous studies have examined agricultural productivity in various regions, there is limited empirical evidence on how credit access enhances maize farming specifically in the Northwest Region of Cameroon. Previous research has primarily focused on technology adoption, the impact of government projects, technical efficiency, and profitability analysis of maize [11,12]. Given the persistently low maize yields and productivity in the region, this study seeks to fill a critical gap by highlighting the role of agricultural credit in enhancing maize productivity.

Recent literature on determinants of credit access in rural areas has identified a wide range of influencing factors, yet further review is warranted to comprehensively capture the institutional and structural variables shaping these outcomes. For instance, the land tenure system plays a critical role in determining access to formal credit. In many Sub-Saharan African countries, including Cameroon, land is customarily held and often undocumented, making it ineligible as collateral in formal credit markets. Studies have shown that farmers with clear land titles are significantly more likely to access loans due to the reduced risk perceived by lenders [13,14]. However, insecure tenure discourages long-term investment and weakens farmers’ bargaining power. This dimension is particularly relevant in regions where customary land practices dominate, and formal land registration systems are weak or inaccessible [15,16].

Moreover, credit access is intricately linked to the development of rural factor markets—particularly labor, land, and input markets. In underdeveloped rural economies, thin or missing markets often compel farmers to rely on informal lending or social networks, which can perpetuate inequality and limit productive investment [17,18]. For example, where labor markets are fragmented or land markets are non-existent, the incentives for lenders to provide agricultural loans diminish. Additionally, segmentation within rural credit markets, such as the distinction between microfinance institutions, cooperatives, and formal banks, results in varied interest rates, lending terms, and eligibility criteria [19,20]. Comparative studies across countries such as Benin, Ghana, and Kenya have shown that while informal financial institutions may increase outreach, they often fail to provide the scale of credit necessary for transformative agricultural investment [21,22].

Cross-country comparisons further illuminate how institutional capacity, regulatory frameworks, and political stability influence rural credit access. In countries with stronger legal enforcement, clearer land titling systems, and integrated rural development strategies, smallholder farmers are more likely to benefit from formal credit systems [23,24]. In contrast, conflict-affected or politically unstable regions often see a collapse or severe distortion of rural financial services. Thus, a more systematic review of existing literature that integrates these dimensions—land tenure, rural markets, and institutional context—is essential for understanding the nuanced barriers to credit access and for designing effective, context-sensitive interventions [25,26].

This study explores the impact of agricultural credit access on maize farmers’ productivity in the Northwest Region of Cameroon to determine whether improved credit access can increase maize production and yields, thereby enhancing the livelihoods of smallholder farmers. To rigorously assess the determinants of credit access and the relationship between credit access and maize productivity, this study employs an endogenous switching regression model (ESRM), accounting for the heterogeneity of farmers’ conditions. Access to agricultural credit is influenced by a combination of socioeconomic and institutional factors. Previous studies have identified key determinants such as education level, income, gender, farm size, collateral availability, interest rates, membership in farming organizations, and age [22,27]. For example, male farmers often have better credit access because they are more likely to own assets used as collateral. High income and larger farm size reduce perceived lending risks, thereby improving access to credit. Education enhances farmers’ awareness and understanding of financial products, facilitating credit uptake. However, as Okurut et al. [28] noted, high transaction costs, lack of collateral, and limited financial literacy remain significant barriers in many rural contexts.

Institutional and infrastructural limitations also play a critical role. Proximity to financial institutions, membership in peasant cooperatives, and access to market infrastructure such as roads significantly affect credit accessibility [21]. Atieno [29] emphasized that stringent collateral requirements and high transaction costs systematically exclude small-scale farmers from formal financial markets. In Cameroon, Wenda et al. [9] identified price volatility in maize markets and poor transportation infrastructure as additional constraints that indirectly limit farmers’ ability to access and repay loans. Despite these challenges, improving credit access can substantially enhance productivity, food security, and rural livelihoods.

Several Sub-Saharan African studies have analyzed the impact of agricultural credit access on farmers’ productivity, employing rigorous econometric techniques such as ESRM and efficiency analysis. For example, Assouto & Houngbeme [21] used an Endogenous Switching Regression Model to assess the effects of credit access on maize farmers in Benin, finding that those with credit access achieved an average yield increase of approximately 40.07%, significantly higher than the counterfactual scenario. This study highlights the importance of selection bias correction when evaluating credit’s impact on productivity. In a Cameroonian context, Tchamba [30] analyzed the technical efficiency of rural farms using a stochastic frontier approach and found that access to credit significantly boosted technical efficiency among maize-producing households. Similarly, in Nigeria, Awotide et al. [31] used an Endogenous Switching Regression (ESR) model to correct for selection bias and potential endogeneity/heterogeneity and found that farmers with access to credit achieved 28% higher yields compared to non-access counterparts.

Despite the breadth of literature on credit access and productivity, few studies have rigorously examined these issues in conflict-affected regions like Northwest Cameroon using advanced econometric approaches. This study addresses that gap by employing an ESRM to control for selection bias and unobserved heterogeneity. By focusing on smallholder maize farmers, the study contributes both to the academic understanding of credit–productivity linkages and to policy discussions on improving rural finance systems in fragile contexts.

The Northwest Region was selected for this investigation because it is the fourth poorest region in Cameroon, where agriculture is vital for the population’s livelihood. Most residents engage in subsistence farming due to limited resources to carry out intensive agriculture. Approximately 69% of the country’s maize yield is produced in this region, making it a representative site for the study. Ultimately, this study aims to contribute to policy discussions geared toward enhancing agricultural productivity, particularly maize productivity, while promoting financial inclusivity and addressing barriers in Cameroon’s rural credit market. In summary, improving access to affordable and well-structured agricultural credit remains a key policy lever for enhancing smallholder productivity and rural development outcomes.

2. Method and Data

2.1. The Endogenous Switching Regression Model (ESRM)

Various models are used for impact evaluation, including propensity score matching, which we opted not to use due to its limitations in addressing selection bias, as it only accounts for observable characteristics [32]. In contrast, ESRMs effectively address both endogeneity and self-selection bias from observable and unobservable characteristics.

ESRMs have been increasingly applied in agricultural economics to address selection bias when farmers self-select into treatment groups, such as access to credit or technology adoption. For example, Diagne and Zeller [33] applied an ESRM to estimate the impact of credit constraints on household welfare in Malawi, showing how self-selection significantly affects observed outcomes. Similarly, Asfaw et al. [34] used an ESRM to assess the productivity impact of modern agricultural technologies among Ethiopian farmers, emphasizing the model’s ability to account for heterogeneity in adoption decisions.

In the context of agricultural productivity, ESRMs are particularly suitable because farmers’ access to credit or technology is typically non-random and influenced by observable and unobservable characteristics. Standard ordinary least squares models fail to control for this selection bias, leading to inconsistent estimates. By contrast, ESRMs jointly estimate the selection and outcome equations, allowing researchers to disentangle the true effect of interventions (e.g., credit access) on productivity while accounting for endogeneity and regime-specific responses.

To address the problem of self-selection bias and endogeneity, an instrumental variable is required. In our study, we use extension service and agricultural training as instruments, as they influence farmers’ likelihood of accessing credit by promoting technology adoption and financial literacy [35,36]. This model allows us to control for differences between those with and without access to credit [19,37,38]. We employed a two-step approach to estimate the impact of credit access on maize productivity: a probit model to determine credit access, followed by two regime outcome equations.

Regime 1 (Access): y_1i = β₁X_1i + ε_1i, if C_i = 1

(1)

Regime 2 (No Access): y_0i = β₀X_0i + ε_0i, if C_i = 0

(2)

In this analysis, Regimes 1 and 2 refer to access to credit and no access to credit, respectively. The variables y_1i and y_0i represent the productivity of maize farmers with and without access to credit, while β₁ and β₀ are the parameters to be estimated. X includes various explanatory variables, and ε_0i, ε_1i, and ε_i are the error terms with unknown covariance. These error terms are assumed to follow a normal distribution with a mean vector of zero and a covariance matrix represented as follows:

$$\mathrm{C}\mathrm{o}\mathrm{v}({\epsilon }_{0\mathrm{i}},{\epsilon }_{1\mathrm{i}},{\epsilon }_{\mathrm{i}})=\left(\begin{array}{ccc}{\sigma }_1^2& {\theta }_{10}& {\theta }_{1\epsilon }\\ {\theta }_{10}& {\sigma }_0^2& {\theta }_{0\epsilon }\\ {\theta }_{1\epsilon }& {\theta }_{0\epsilon }& {\sigma }_2^2\end{array}\right)$$

(3)

where ${\sigma }_1^2$ and ${\sigma }_0^2$ represent the variance of the error terms ε_1i and ε_0i in the maize productivity functions in Equations (1) and (2), and ${\sigma }_2^2$ represents the variance of the error term ε_i. ${\theta }_{1\epsilon }$ is the covariance of ε_1i and ε_i, while θ_0ε is the covariance of ε_0i and vi. θ_1ε and θ_0ε refer to the correlation term between household credit access status and the impact of maize productivity in Equations (1) and (2). Thus, the problem of selection bias does not arise provided θ_1ε and θ_0ε = 0. However, selection bias arises if θ_1ε and θ_0ε ≠ 0, leading to an inconsistent ordinary least squares estimator. Thus, maximum likelihood estimation is employed as an efficient estimator for the ESRM [38], estimated as follows:

$$\mathrm{E}({\epsilon }_{1\mathrm{i}}/{\mathrm{C}}_{\mathrm{i}}=1)=\mathrm{E}({\alpha }_{1\epsilon }{\epsilon }_{\mathrm{i}}/{\epsilon }_{\mathrm{i}}\le \beta {\mathrm{Z}}_{\mathrm{i}})={\beta }_{\mathrm{i}\epsilon }{\displaystyle \frac{\mathrm{\varnothing }(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}})}{\Phi \left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}}$$

(4)

$$\mathrm{E}({\epsilon }_{0\mathrm{i}}/{\mathrm{C}}_{\mathrm{i}}=0)=\mathrm{E}({\alpha }_{0\epsilon }{\epsilon }_{\mathrm{i}}/{\epsilon }_{\mathrm{i}}\le \beta {\mathrm{Z}}_{\mathrm{i}})={\beta }_{0\epsilon }{\displaystyle \frac{\mathrm{\varnothing }\left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}{1-\Phi \left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}}$$

(5)

With ∅ representing the standard normal probability density function and Φ the cumulative density function, the ratio of ∅ and Φ in Equations (4) and (5) is the inverse Mills ratio written below as

$${\lambda }_{1\mathrm{i}}={\displaystyle \frac{\mathrm{\varnothing }\left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}{\Phi \left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}}$$

(6)

$${\lambda }_{0\mathrm{i}}={\displaystyle \frac{\mathrm{\varnothing }\left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}{1-\Phi \left(\widehat{\beta {\mathrm{Z}}_{\mathrm{i}}}\right)}}$$

(7)

Putting Equations (6) and (7) in Equations (1) and (2) above yields

$${\mathrm{y}}_{1\mathrm{i}}={\beta }_1{\mathrm{X}}_{1\mathrm{i}}+{\sigma }_{1\epsilon }{\lambda }_{1\mathrm{i}}+{\nu }_{1\mathrm{i}} \mathrm{if} {\mathrm{C}}_{\mathrm{i}} = 1$$

(8)

$${\mathrm{y}}_{0\mathrm{i}}={\beta }_0{\mathrm{X}}_{0\mathrm{i}}+{\sigma }_{0\epsilon }{\lambda }_{0\mathrm{i}}+{\nu }_{0\mathrm{i}} \mathrm{if} {\mathrm{C}}_{\mathrm{i}} = 0$$

(9)

where ${\nu }_{1\mathrm{i}}$ and ${\nu }_{0\mathrm{i}}$ are new error terms with means of zero. To account for heteroskedastic errors in ${\nu }_{1\mathrm{i}}$ and ${\nu }_{0\mathrm{i}}$, the weighted least squares method was used [39]. Therefore, the application of the ESRM is deemed suitable for this study if the estimated covariances ${\theta }_{1\epsilon }$ and ${\theta }_{0\epsilon }$ are statistically significant, and the likelihood ratio test accepts the alternative hypothesis of endogeneity.

2.2. Data Collection

Data for this study were collected from primary sources. Primary data were obtained through a cross-sectional survey of maize farmers using a structured questionnaire that included closed-ended and open-ended questions, allowing for yes/no responses and additional explanations. We selected the Mezam division from the seven divisions in the Northwest region, specifically targeting Bamenda city and Bambili from the Santa and Tubah subdivisions, respectively. Five villages—Bafut, Mankon, Pinyin, Nkwen, and Bambili—were randomly chosen. The Northwest Region was selected for this study because it is the country’s leading maize-producing area, accounting for approximately 69% of national maize output. This makes it a highly relevant and strategic region for analyzing the relationship between credit access and maize productivity among smallholder farmers. Visits were made to randomly selected farm households in these villages with assistance from a local agriculturalist after conducting a simple listing. A total of 404 questionnaires were distributed for face-to-face interviews with respondents. The collected data were coded and converted into electronic format using Microsoft Excel for cleaning, then imported into STATA version 13 for descriptive statistics, frequency analysis, and impact assessment using the ESRM.

Table 1. Operationalization of variables used in the study.

Variables	Descriptions
Productivity	Maize farmers’ total output in t/ha of land during the 2021 maize season
Credit Access	1 = Agricultural credit granted to a maize farmer in the last four farming seasons and 0 = otherwise
Sex	1 = Male-headed households, 0 = Female-headed households
Age	Age of household head in years
Education	Number of years of schooling of household head
Labor	Number of hours of daily hired labor by the maize farmer
Income	Gross income of maize farmers obtained from the cultivation of maize during the 2021 season (measured in CFA francs)
Experience	Number of years a farmer has cultivated maize
Farm size	Total size of land attributed to maize production in hectares (ha)
Farm Organization	1 = if a farmer is a member of any farm organization and 0 = otherwise
Extension service	1 = if a farmer had visits with an extension agent and 0 = otherwise
Training	1 = if a farmer received training from any professional agriculturalist and 0 = otherwise

describes the variables hypothesized to determine farmers’ access to credit and the impact of credit access on maize productivity. The key variables used in the analysis were categorized as either binary or continuous. Binary variables included credit access (access vs. no access), sex of maize farmers (male = 1, female = 0), farm size (in hectare), membership in a farm organization (yes = 1, no = 0), access to extension services (yes = 1, no = 0), and participation in training programs (yes = 1, no = 0). Continuous variables included farmers’ age (in years), educational level (in years of schooling), labor input (measured in number of persons employed on the farm), and farming experience (in years). Maize productivity was measured as the total output in tons per hectare (t/ha) of land. Gross income was defined as the total monetary value (in CFA francs) of maize produced, calculated by multiplying the total quantity of maize harvested by the prevailing market price during the production season. Gross income was included in the model as a control variable to capture farmers’ financial capacity, which could influence their ability to invest in productivity-enhancing inputs such as fertilizers, improved seeds, and hired labor.

The sex of the farmers is included, as male-headed households are often more informed about farming practices and likely to access credit, suggesting that maize farms managed by men may be more productive than those managed by women. Age, measured in years, was included to test whether younger farmers are more open to innovations and information about credit access. Education, represented by years of formal schooling, serves as a proxy for managerial skills; educated farmers are generally more efficient in acquiring technical knowledge and making informed financial decisions. Labor is quantified in labor hours hired for maize farming, as more labor typically leads to increased production. Income is measured as the total earnings in CFA francs, with higher income levels indicating a greater capacity to repay loans, making farmers more attractive to financial institutions.

Experience is measured by the number of years spent farming maize, with increased experience likely leading to better decision-making through exposure to effective farming techniques. Farming organization is a dummy variable indicating whether farmers belong to any farming organization (yes = 1, no = 0), as membership often provides access to additional information on maize varieties and credit availability. Extension service is also a dummy variable, where farmers visited by an extension agent = 1 and those who were not visited = 0. Extension agents offer vital information on new technologies, inputs, and markets. Lastly, training is measured as a dummy variable, with yes = 1 for farmers who received professional training during the maize planting season and no = 0 otherwise.

External factors such as weather variability, pest outbreaks, or regional price shocks were not explicitly controlled in this study due to data limitations. However, their potential influence on both credit access and productivity is acknowledged, and future research should account for these factors more systematically.

2.3. Socioeconomic and Descriptive Characteristics of Respondents

The frequency distributions of respondents are presented in percentages in

Table 2. Characteristics of maize farmers who participated in the survey.

Variables		Frequency (Number of Famers)	Percentages (%)
Credit access	Yes No	148 256	36.6 63.4
Sex	Male Female	231 173	57.2 42.8
Farming organization	Yes No	154 250	38.1 61.9
Extension service	Yes No	183 221	45.3 54.7
Training	Yes No	182 222	45.0 55.0
Total		404	100

. The survey revealed that out of a total sample of 404 farmers surveyed, most never had access to credit (256, 63.4%), while approximately one-third (148, 36.6%) had access to credit. Reasons for not having access to credit were high interest rates, lack of collateral, lengthy procedures during application, and illiteracy. More than half (231, 57.2%) of the farmers belonged to male-headed households, while 173 (42.8%) were female-headed. In addition, 38.1% (154) of the farmers belonged to a farming organization, while 61.9% (250) of them did not belong to any farming group. Furthermore, more than half (221, 54.7%) of the farmers also revealed that they were not visited by extension agents, while 183 (45.3%) had extension visits. Lastly, 182 (45.0%) of the farmers received training from agriculturalists on modern methods of cultivation, while 222 (55.0%) did not receive training.

Table 3. Descriptive statistics of maize farmers’ characteristics.

Variable	Overall Mean (Std. Dev.)	Credit Access = 0 Mean (Std. Dev.)	Credit Access = 1 Mean (Std. Dev.)
Productivity (t/ha)	4.104 (0.935)	4.048 (0.890)	4.202 (1.004)
Sex	0.572 (0.495)	0.594 (0.492)	0.534 (0.501)
Age (years)	38.052 (10.344)	38.367 (10.374)	37.507 (10.305)
Education (years)	10.079 (4.607)	9.895 (4.505)	10.399 (4.776)
Labor (hours/day)	3.931 (2.219)	3.766 (2.127)	4.216 (2.349)
Gross Income (CFA francs)	670,347.854 (260,813.222)	656,486.948 (252,555.237)	694,323.476 (272,854.137)
Experience (years)	14.950 (7.535)	15.668 (7.661)	13.709 (7.168)
Farm size (ha)	0.651 (0.620)	0.675 (0.665)	0.610 (0.532)
Farm Organization	0.381 (0.486)	0.395 (0.490)	0.358 (0.481)
Extension service	0.453 (0.498)	0.414 (0.494)	0.520 (0.501)
Training	0.458 (0.514)	0.340 (0.499)	0.662 (0.475)

presents descriptive statistics for the variables used in the study. According to the survey, the average maize yield was 4.104 t/ha, with a standard deviation of 0.9, and ranged from 2 to 10.8 t/ha. On average, maize farmers were 38 years old, with ages ranging from 20 to 65. The average farm size was 0.651 Ha, highlighting the predominance of smallholder farming in the study area.

More importantly, farmers had a mean of 10.1 years of formal education with a standard deviation of 0.6 and a minimum and maximum of 0 and 17 years of formal education, respectively. Thus, some maize farmers never received any formal schooling. Meanwhile, for those who attended school, the highest level of education was university, representing 17 years of formal education. On average, 4 h of daily labor were put into maize production, with the highest and lowest number of hours being 13 and at least 1, respectively. On average, farmers had approximately 15 years of experience in maize farming, with a maximum of 37 years and 1 year, respectively. The mean farm size was 0.7 ha, with a standard deviation of 0.6. On average, maize farmers had a mean income of 670,347.9 CFA francs, with minimum and maximum incomes of 260,000 CFA francs and 2,160,000 CFA francs, respectively, and a standard deviation of 261,136.6 CFA francs.

Preliminary descriptive analysis shows that farmers with credit access tend to have higher education levels and larger farm sizes compared to those without access, suggesting potential selection patterns relevant for subsequent modeling.

3. Results and Discussion

This section presents the results from the ESRM, using productivity as the dependent variable and a series of independent variables outlined in

Table 4. Determinants of credit access and its impact on maize productivity.

Variables	Coefficient
	Credit Access (Selection Equation)	Productivity (Access)	(No Access)
Sex	−0.105	0.228	0.017
Age	−0.002	−0.002	0.008
Education	0.017	0.016	−0.008
Labor	0.027	0.036	−0.002
Income	0.125	−0.027	−0.044
Experience	−0.018	0.017	0.018 *
Farm size	−0.103	0.521 ***	0.361 ***
Farming Organization	−0.358 *	0.543 **	0.533 ***
Ext service	0.388 **
Training	0.572 ***
Constant	−0.702 *	3.581 ***	2.715 ***
σ		0.977 ***	1.010 ***
ρ		−0.524	−0.905
Wald Test	35.18
LR Test	41.84
Log pseudolikelihood	−720.644

Note: * p < 0.1, ** p < 0.05, and *** p < 0.01 denote significance at 1%, 5%, and 10% levels of significance. σ represents the square root of the residual of farmers’ productivity. ρ is the correlation between the credit access equation and the equation of productivity of credit access and no credit access, respectively.

. The estimated correlation coefficient (ρ) between the credit access equation and the productivity equations is significantly different from zero, indicating that observable and unobservable factors influence farmers’ access to credit and their productivity based on credit access. A significant correlation between the credit access equation and the productivity equation for farmers with access to credit suggests the presence of self-selection bias. Additionally, the differences in productivity coefficients for farmers with and without credit access indicate sample heterogeneity. The Wald test further confirms the significance of the regression, including the significance of the constant term.

The estimates from the Endogenous Switching Regression model yield two key insights: the determinants of farmers’ access to credit and the factors influencing maize productivity depending on credit access. Membership in a farming organization has a statistically significant negative effect on access to credit (coefficient = −0.358, p < 0.1). While this finding contradicts the results of Achille and Dewanou [21], it aligns with local realities. Interestingly, the negative and significant effect of farming organization membership on credit access may be explained by contextual constraints. Many of the surveyed farmers are located in remote rural areas where formal farming organizations are scarce or weakly functioning. As a result, membership may not reflect strong institutional support or better access to information and credit, but rather a self-selection mechanism that is not correlated with actual service provision. This could partly explain the counterintuitive result. Additionally, this variable may be subject to endogeneity, as farmers’ decisions to join a farming organization may be influenced by unobserved characteristics that also affect credit access.

In contrast, access to extension services (coefficient = 0.388, p < 0.05) and agricultural training (coefficient = 0.572, p < 0.01) both significantly and positively influence access to credit. These services, typically provided through government agencies or agricultural programs, help farmers adopt improved practices and become more creditworthy. Extension agents may also act as intermediaries between farmers and lenders, facilitating the credit application process.

Regarding maize productivity, farm size has a positive and statistically significant effect for both farmers with access to credit (coefficient = 0.521, p < 0.01) and those without access (coefficient = 0.361, p < 0.01). This suggests that larger farms are better able to exploit economies of scale and utilize inputs more effectively. The positive association between farm size and productivity is consistent with the findings of Abdallah [40] and Tijani [41], although it contrasts with Achille and Dewanou [8], who reported a negative relationship.

Farming organization membership also significantly improves productivity in both credit-accessed farmers (coefficient = 0.543, p < 0.05) and those without access (coefficient = 0.533, p < 0.01). This result indicates that group affiliation supports productivity through improved access to information, coordination, or shared knowledge, even if it does not necessarily enhance access to formal credit.

Among farmers without access to credit, farming experience is positively associated with productivity (coefficient = 0.018, p < 0.1). This may be due to the accumulation of practical knowledge among older farmers, who have learned from experience how to manage inputs more effectively. This finding aligns with the work of Lunduka [42].

Although farming experience did not show a statistically significant effect on credit access in the current model, its negative coefficient may suggest an underlying behavioral pattern. More experienced farmers may rely less on formal credit sources, having gained access to alternative financing mechanisms such as savings, social networks, or informal lending arrangements. With years of experience, farmers are also more likely to develop financial management skills, diversify their income sources, and build financial resilience. In the study area, informal sources of credit are often preferred due to their lower interest rates and more flexible terms, making formal credit less attractive or necessary for seasoned farmers.

While these findings may initially seem counterintuitive, they are consistent with observations from the field and align with the structural realities of the region. These findings reaffirm the important role of extension services and training in facilitating credit access. Given their institutional nature, these supports enhance financial knowledge and credibility among farmers, thereby improving loan eligibility. Farmers are typically required to participate in these programs, as they are implemented by governmental or professional agents in the region.

To evaluate the impact of credit access on maize productivity, we quantified the results using productivity differences presented in

Table 5. Productivity differences according to credit access.

	Access	No Access	Treatment Effects (C = A − B, F = D − E)	% Impact
Access (ATT)	4.19 (A)	3.70 (B)	0.49 (C)	11.7
Non-Access (ATU)	4.08 (D)	3.85 (E)	0.23 (F)	5.6
Heterogeneity Effect	0.11	−0.15

A and E are the observed productivity of maize (t/ha), and B and D represent the counterfactual expected productivity of maize.

. The findings indicate that farmers with access to credit achieved higher productivity than those without. Average Treatment on the Treated (ATT) represents the estimated productivity gain for farmers who accessed credit, relative to the counterfactual scenario in which they had not accessed credit. Conversely, Average Treatment on the Untreated (ATU) captures the productivity gain that non-credit-accessed farmers would have experienced if they had accessed credit, compared to their actual performance. Specifically, the estimated average productivity for credit-accessed farmers was 4.19 t/ha, compared to 3.70 tons for those without access. This represents an average productivity gain of 0.49 t/ha, or an approximate increase of 11.6%, relative to the counterfactual. This improved performance can be attributed to the enhanced ability of credit-accessed farmers to invest in high-quality inputs, adopt modern agricultural practices, and hire labor more efficiently. These findings are consistent with prior studies such as Diamoutene and Jatoe [25], which emphasize the positive effect of credit on input adoption and output. Conversely, Nakana and Magezi [1] observed that insufficient access to credit led to limited productivity improvement among rice farmers.

Moreover, the estimated productivity gain of 0.49 t/ha among farmers with credit access corresponds to an approximate income increase of 77,938 CFA francs per hectare, based on the average maize market price of 159,059 CFA francs per ton during the production season. This highlights the economic relevance of credit access in improving smallholder farmers’ income.

In addition to average treatment on the treated, the analysis also estimated the average treatment effect on the untreated. Farmers without access to credit had an observed productivity of 3.85 t/ha, but their estimated productivity would have risen to 4.08 t/ha if they had accessed credit. This implies a potential gain of 0.23 t/ha, or a 5.6% increase, for non-beneficiaries.

Furthermore, the model provides insight into the heterogeneity of treatment effects. Specifically, the productivity differential between farmers who actually accessed credit and those who hypothetically would have, even under access, indicates a positive heterogeneity effect of 0.11 t/ha. This suggests that credit-accessed farmers are inherently more productive, potentially due to unobserved characteristics such as managerial ability, risk tolerance, or proximity to input and output markets.

Taken together, these results confirm that credit access significantly improves maize productivity and income, and that both observable and unobservable differences among farmers influence the magnitude of these gains. The application of the Endogenous Switching Regression model was critical in uncovering these insights by addressing selection bias and estimating valid counterfactual outcomes.

4. Conclusions and Policy Recommendations

In some developing countries, where food insecurity and poverty are significant concerns, agricultural productivity is declining due to many poor farmers’ inability to access financial credit. The main objective of this study was to determine whether credit access increases the productivity of maize farmers in Cameroon. Using the probit model, we found that farming experience and membership in farming organizations negatively affect access to credit, while extension services and agricultural training positively influence it. The ESRM confirmed that farmers with access to credit achieved higher productivity gains compared to those without. Furthermore, farmers lacking credit access could have realized even higher productivity if they had been given the opportunity to access credit.

The findings regarding the determinants of credit access and the productivity impact of access to credit suggest refinements to policy strategies to support the agricultural sector. The positive effect of access to credit on maize productivity implies that Cameroon government authorities, such as the Northwest authorities, need to include the maize sector among their priority sectors to further strengthen the financing mechanisms for producers. Furthermore, because maize is mostly produced in the Northwest region of Cameroon, every financial support given to farmers in this region will substantially strengthen food security and improve productivity. Moreover, the Grassfield Participatory and Decentralized Rural Development Project needs to effectively include maize producers in their rural infrastructure development projects to help farmers access the finance necessary to buy agricultural inputs. Alternatively, part of the finances donated by the African Development Bank to this project could be channeled to smallholder farmers to help them purchase the necessary inputs for production. In addition, the conditions of access to agricultural credit for farmers should be made more flexible by the informal credit lenders.

To enhance agricultural productivity, it is essential to strengthen technical supervisory mechanisms for farmers during extension sessions, particularly for older farmers, as the survey indicated a positive correlation between farming experience and productivity. Frequent visits from extension agents would provide these farmers with valuable information and knowledge to improve their cultivation methods. Additionally, increased agricultural training programs, such as the Program for the Consolidation and Sustainability of Agro-pastoral Counseling and PNVRA, should focus on supporting maize producers in the Northwest region, which accounts for 69% of the country’s maize production. Given that education and labor hours had no significant impact on credit accessibility and productivity gains, further investigation into these variables is necessary to understand their true effects. To enhance farmers’ access to credit and improve agricultural productivity in the Northwest Region of Cameroon, it is recommended that the government revitalize agricultural credit institutions by re-establishing specialized banks such as FUNADER, with a specific mandate to support smallholder maize farmers. These institutions should offer reduced interest rates and more flexible repayment terms tailored to the seasonal nature of agricultural production. Furthermore, establishing community-based credit cooperatives, introducing digital microcredit platforms, and implementing government-backed credit guarantee schemes could further improve the accessibility and reliability of agricultural finance, thereby enabling farmers to invest more effectively in inputs and technologies that raise productivity.

This study acknowledges several limitations. First, the use of cross-sectional data restricts the ability to capture seasonal fluctuations and to control for unobserved characteristics that do not vary over time. A more nuanced understanding of productivity and credit access dynamics could be achieved through the use of panel or time-series data in future research.

Second, many of the variables used in the analysis, such as maize yield, income, and credit received, are based on self-reported information. This raises concerns about potential recall bias and the accuracy of responses. To address this issue, future studies should consider more rigorous enumerator training and the incorporation of basic validation techniques during data collection. In addition, while smallholder farmers in Cameroon generally engage in diversified cropping systems, this study focused solely on maize in Mezam due to its regional importance. The effects of crop diversification on credit access and productivity were not captured, representing a limitation of the analysis.

While this study employed the Endogenous Switching Regression Model to account for selection bias between credit access and productivity, alternative empirical strategies could be considered in subsequent research. For example, methods such as Propensity Score Matching or Instrumental Variable regression could help assess the robustness of the results.

Finally, the geographic scope of this study is limited to the Northwest region of Cameroon, where maize is predominantly cultivated. To enhance the generalizability of the findings, it would be valuable for future studies to include other ecological zones and examine additional crops that reflect the broader agricultural context of the country.