October 27, 2026
Credit risk in rural credit operations in Brazil: comparison between funding sources
Lucas Mattos de Freitas; Juliana L. de Oliveira Bergamini
DOI: 10.22167/2675-6528-202601095
Article prepared by the ResumeAI tool, an artificial intelligence solution developed by the Pecege Institute focused on synthesis and writing.
Summary
The credit risk in rural credit operations in Brazil was analyzed, focusing on the comparison between loans granted with free resources and those from controlled sources. The central objective of the study was to verify to what extent the origin of resources and the cost associated with credit altered the probability of default in rural operations. The research was developed based on administrative microdata extracted from the Rural Credit and Proagro Operations System (SICOR), provided by the Central Bank of Brazil, covering the period from 2021 to 2025. Methodological procedures included extraction, treatment, and statistical analysis, with emphasis on the application of logistic regression models to evaluate the association between the type of resource and default, controlling for other variables. Descriptive results indicated an average default rate of 8.07% in operations with free resources, significantly higher than the 1.96% observed in operations with controlled resources. The incorporation of logistic regression reinforced this finding, demonstrating that, even after controlling for the interest rate, operations with free resources showed an odds ratio of 2.84 for default compared to operations with controlled resources. The findings suggested that the lower default observed in controlled resources was not solely explained by the cost of credit, but also by institutional, regulatory, and operational selection mechanisms. It was concluded that the regulatory framework of controlled rural credit played a relevant role in mitigating credit risk in the sector, promoting greater financial stability.
Keywords: Nonpayment; Financial Market; Agricultural Policy; Logistic Regression; Regulation.
1. Introduction
Brazilian agribusiness is a fundamental pillar of the national economy, contributing significantly to the Gross Domestic Product, the trade balance, and job creation. In this context, rural credit plays a strategic role, enabling the financing of production, investments in infrastructure and technology, and the commercialization of agricultural products. The financial health of this segment is, therefore, vital for the stability and growth of the sector. Historically, rural credit in Brazil has presented peculiar characteristics, marked by the coexistence of different financing sources. On one hand, there are controlled resources, originating from banking requirements and specific funds, with regulated interest rates and conditions. On the other hand, free resources operate, raised in the market and priced according to the supply and demand logic of financial institutions. Understanding the factors that influence the quality of the rural credit portfolio, especially default risk, is crucial for the formulation of public policies and risk management strategies. The architecture of Brazilian rural credit, with the strong presence of the State and sectoral financing policies, is a central element for this understanding (Bacha, 2004).
From a theoretical point of view, the discussion on credit risk in financial markets is frequently anchored in the literature on information asymmetry and credit rationing. The seminal model by Stiglitz and Weiss (1981) postulates that, in markets with imperfect information, an increase in interest rates can not only select higher-risk borrowers (adverse selection) but also incentivize riskier behaviors from borrowers (moral hazard). From this perspective, it would be plausible to assume that free resources, by allowing greater flexibility in pricing and an allocation more aligned with market conditions, would tend to generate a portfolio with lower default risk, due to more efficient credit selection. However, the presence of regulatory and institutional mechanisms associated with controlled resources may introduce additional filters for eligibility, monitoring, and follow-up of operations, which are not restricted solely to price, altering the expected dynamics of risk selection.
Previous studies have explored the determinants of nonpayment in Brazilian rural credit from various perspectives. Braga de Melo and Resende Filho (2017), for example, analyzed the influence of macroeconomic factors, such as interest rates and inflation, as well as the impact of debt renegotiations and agricultural profitability on nonpayment. Complementarily, Almeida (2016) and Pereira and Melo (2021) investigated microeconomic and climatic factors, highlighting the relevance of the producer’s income conditions, the guarantees offered, and the productive environment for the performance of credit operations. Although these works provide a solid foundation for understanding the complexity of rural credit risk, the literature still lacks an in-depth analysis that directly compares the performance in terms of nonpayment between operations financed by free and controlled resources, considering the underlying institutional and regulatory mechanisms. The preliminary empirical observation of the present study, which points to a superior credit performance for operations with controlled resources, challenges the expectation that the pricing freedom of free resources would lead to more efficient risk selection, indicating the need to investigate the factors that explain this dynamic.
The apparent contradiction between market theory, which suggests better risk allocation in free pricing environments, and empirical observation in Brazilian rural credit, where controlled resources exhibit lower default rates, demands in-depth investigation. Understanding whether this difference is explained by factors beyond credit cost, such as institutional and regulatory arrangements, is fundamental for the development of more effective agricultural and financial policies. Thus, the central objective of this study is to analyze credit risk in rural credit operations in Brazil, focusing on the comparison between loans granted with free resources and those from controlled sources, in order to verify to what extent the origin of resources and the cost associated with credit alter the probability of default in rural operations, controlling for other relevant variables.
2. Material and Methods
This research adopted a quantitative approach, using administrative microdata to analyze credit risk in rural operations in Brazil. The study focused on the comparison between loans granted with free resources and those from controlled sources, seeking to verify how the origin of resources and the cost associated with credit impact the probability of default. This methodology allowed for an in-depth investigation of the institutional and regulatory mechanisms that may influence the quality of the rural credit portfolio, as per the central objective outlined in the introduction.
The study’s context covered the Brazilian rural credit market, focusing on operations registered in the Rural Credit and Proagro Operations System (SICOR), provided by the Central Bank of Brazil (BCB). The analysis period comprised the years 2021 to 2025. This temporal scope was strategically selected to observe the most recent dynamics of rural credit, in a post-pandemic scenario characterized by significant changes in financial conditions and the cost of credit, minimizing interferences from atypical periods and concentrating the analysis on a comparable horizon.
The primary unit of analysis consisted of individual rural credit operations. The population of interest encompassed all rural credit operations registered in SICOR and Proagro during the period from 2021 to 2025. The sample was composed of all operations for which it was possible to extract and integrate complete data from the SICOR databases, after applying selection criteria that ensured the clarity and consistency of the information, especially regarding the classification of the resource source, as detailed in the data treatment steps.
The data collection instruments were the databases of the Rural Credit Operations and Proagro System (SICOR), maintained by the Central Bank of Brazil. The research was built from the integration of two main SICOR databases, which are complementary to each other. The first database, named “balances”, provided periodic information on the stock of operations, including average daily balance, outstanding balance, balance on the last day of the period, and the operation’s status, essential for identifying default.
The second base, referred to as “basic operation by state”, contained the contractual characteristics of the operations. These characteristics included the resource source, the nominal interest rate, the total effective cost, the federative unit, the borrower category, the credit program, the insurance type, and the financed amount. The joint use of these bases allowed for the relationship, in the same analysis, of the contractual characteristics defined at the time of credit granting with the information observed throughout the execution of the rural credit operation, providing a comprehensive view of the rural credit lifecycle.
The research execution steps began with the individual import of annual files for the period from 2021 to 2025. Subsequently, these files were consolidated into a single database through vertical stacking, standardizing the record structure. Next, the two main SICOR databases were integrated using a “left join” type of join, employing the fields X. REF_BACEN and NU_ORDEM as composite keys. This strategy ensured the preservation of observations from the balance sheet database and the incorporation of corresponding contractual information.
After integration, a variable selection step was performed, keeping only those considered necessary for credit risk analysis. From the balance base, variables such as ANO_BASE, MES_BASE, VL_MEDIO_DIARIO_VINCENDO, VL_ULTIMO_DIA, VL_MEDIO_DIARIO, and CD_SITUACAO_OPERACAO were preserved. From the contractual base, X. REF_BACEN, NU_ORDEM, CD_FONTE_RECURSO, CD_TIPO_SEGURO, CD_EMPREENDIMENTO, CD_PROGRAMA, VL_JUROS, CD_ESTADO, VL_PARC_CREDITO, VL_PERC_CUSTO_EFET_TOTAL, and CD_CATEG_EMITENTE were maintained, focusing on information directly related to the operation’s behavior and the borrower’s profile.
A crucial data treatment involved converting the total effective cost variable (VL_PERC_CUSTO_EFET_TOTAL). Originally, this variable was in text format, containing percentage symbols and decimal commas, which made its direct use in statistical models unfeasible. To correct this inconsistency, the percentage symbol was removed, the comma was replaced by a point, and the content was converted to numeric format. This procedure allowed the variable to be interpreted as a continuous measure of credit cost, essential for subsequent econometric analyses.
The classification of resource sources also required specific treatment. Based on the guidelines of the Rural Credit Manual, the codes for the CD_FONTE_RECURSO variable were grouped into two main analytical categories: free resources and controlled resources. Free resources comprised operations financed by market sources, such as free banking resources, LCA at free rates, free rural savings, and free BNDES. Controlled resources, on the other hand, covered operations subject to mandatory direction, subsidy, equalization, or specific rules, such as the National Treasury, bank requirements, constitutional funds, FUNCAFÉ, equalizable BNDES, and controlled rural savings. Observations without clear classification were excluded from the analytical sample to avoid ambiguities.
The study’s dependent variable, the occurrence of default, was constructed from the variable CD_SITUACAO_OPERACAO. Operations classified under codes 9, 11, and 12, which correspond, respectively, to loss, collection, and active debt, were classified as defaulted. Based on this rule, the binary variable INADIMPLENTE was created, where the value 1 represented the occurrence of default and the value 0 represented other situations. Additionally, a historical variable per contract, named teve_inadimplencia, was created to indicate whether the operation had defaulted at any point during the analyzed period.
For the econometric modeling stage, a specific analytical base was constructed containing a single observation per contract. This reduction was carried out by identifying duplicate records based on the X. REF_BACEN key, keeping only one line per operation. Subsequently, only observations with complete information for the variables essential for modeling were preserved, including resource type, base year, federative unit, producer size, total effective cost, and default history. Categorical variables were converted to factor format, allowing their proper use in logistic regression.
Econometric estimation was performed using logistic regression models, a technique appropriate for binary dependent variables, such as nonpayment. Logistic regression allowed estimating the probability of the event of interest occurring as a function of a set of explanatory variables. The general formulation of the model expressed that the probability of nonpayment depends on the vector of observable covariates. In practical terms, this type of model indicated how each characteristic of the operation was associated with an increase or decrease in the chance of nonpayment, providing insights into the determinants of credit risk.
Three main logistic regression models were specified. The first, a basic model, considered only the resource type to measure the raw difference between free and controlled resources. The second model incorporated the total effective cost, allowing verification of whether the resource type effect persisted after controlling for the interest rate. The third, a complete multivariate model, included resource type, interest rate, base year, federal unit, and producer size, aiming to estimate the net effect of the resource type, while simultaneously controlling for temporal, regional, and structural heterogeneities.
The logistic regression results were presented in the form of odds ratios, obtained by exponentiating the estimated coefficients. This format facilitated the interpretation of the effects, where values above 1 indicated an increase in the chance of default/ non-payment and values below 1 indicated a reduction. In addition to econometric modeling, the research developed complementary descriptive analyses. Default/ Non-payment statistics were calculated by type of resource, annual evolution, return to normal course, heat maps by resource source and interest rate, default/ non-payment by issuer category, and predicted probabilities by type of resource, offering a broader interpretation of the results.
All data processing and analysis were performed in the RStudio environment, using the dplyr, ggplot2, broom, and tibble packages. These resources were employed in all stages, from data import and consolidation to processing, modeling, and visualization. The use of these packages and the documentation of the complete pipeline in the script developed for the research ensured the reproducibility of the analytical procedure and the traceability of the executed steps, guaranteeing the integrity and transparency of the methodology applied to the dataset.
3. Results and Discussion
The results obtained from the analysis of administrative microdata from the Rural Credit and Proagro Operations System (SICOR) revealed distinct credit risk patterns between operations financed by free resources and those with controlled resources in Brazil, during the period from 2021 to 2025. The initial descriptive analysis demonstrated a substantial difference in average default, with free resource operations showing a rate of 8.07%, significantly higher than the 1.96% observed in operations with controlled resources. This finding is crucial, as it contradicts the initial hypothesis that greater pricing flexibility and market logic associated with free resources would lead to more efficient risk selection and, consequently, lower default rates. On the contrary, empirical evidence suggests that the regulatory environment and institutional mechanisms inherent to controlled resources play a preponderant role in mitigating credit risk.
The persistence of this default differential throughout all the analyzed years, as illustrated in Figure 1, reinforces the idea that it is not a punctual fluctuation, but a structural pattern in the behavior of the portfolios. Figure 1 clearly shows that, in each year from 2021 to 2025, the proportion of defaulted operations was consistently higher in the free resources segment. This systematic behavior challenges the simplistic view that regulatory intervention in the rural credit market, by imposing administered or limited interest rates, would necessarily distort capital allocation and increase risk. On the contrary, the lower nonpayment in controlled resources points to the effectiveness of other credit control and selection mechanisms that operate in this segment.
Source: Original research results
Figure 1. Percentage of non-payment by source of funds (2021-2025)
The literature on information asymmetry and credit rationing, such as the seminal model by Stiglitz and Weiss (1981), postulates that an increase in interest rates can attract riskier borrowers (adverse selection) and incentivize moral hazard behaviors. However, the present study suggests that, in the context of Brazilian rural credit, regulation associated with controlled resources may introduce additional eligibility and monitoring filters that overcome the potential negative effects of administered rates. Bacha (2004) had already highlighted the role of the State and bank reserve requirements in the architecture of rural credit, indicating that controlled resources are not just “cheaper credit” but rather operations inserted into a more restrictive regulatory regime. This theoretical perspective is fundamental to interpreting the descriptive results, which point to a superior performance of the regulated segment.
The analysis of the relationship between interest rate and nonpayment, when observed in isolation for the two groups of resources, did not prove conclusive, as indicated by complementary descriptive analyses. A heatmap of nonpayment by resource source and interest rate band, although not graphically detailed at this point, revealed that the complexity of the interaction between credit cost and nonpayment risk transcends a simple linear correlation. Instead, credit performance appears to be shaped by a combination of factors, including the nature of the resource source and the intrinsic characteristics of borrowers and operations. This finding suggests that the pricing of risk in rural credit is a multifaceted phenomenon, where the interest rate is just one component, and not necessarily the most determining one in all contexts. For policymakers and financial institutions, this implies that the management of rural credit risk requires a holistic approach, which considers not only the cost of capital but also the selection and monitoring mechanisms and the socioeconomic characteristics of producers.
Additionally, the issuer category was shown to be associated with default behavior. Complementary descriptive analyses, which examined the number of defaulters by issuer category, indicated that the largest contingent of defaulters was concentrated among small rural producers, followed by medium-sized ones, and, in smaller proportion, by large producers. Although the absolute number of defaulters may be higher among small producers due to their greater representation in the database or specific vulnerabilities, this data must be interpreted with caution. Studies such as those by Almeida (2016) and Pereira and Melo (2021) highlight the importance of microeconomic and climatic factors in rural non-payment, which can disproportionately affect small producers. However, the subsequent logistic regression analysis will refine this understanding, revealing the relative risk of each category. For public policies, this initial result underscores the need for rural credit programs that consider the specificities of each producer size, with support mechanisms and risk mitigation adapted to their capabilities and vulnerabilities.
Logistic Regression, used to estimate the probability of default, provided a more robust basis for interpreting the findings. In Model 1, which considered only the type of funding as an explanatory variable, operations financed with free funds showed an odds ratio (OR) of 2.48 for nonpayment compared to operations with controlled funds. This means that, in a raw comparison, the chance of nonpayment was almost two and a half times higher among operations based on free funding sources. Figure 2 illustrates this probability difference, showing that the probability of nonpayment for free funds is substantially higher than for controlled funds. This result reinforces the descriptive observation and solidifies the idea that the origin of the funding is not a neutral factor in credit risk.
Source: Original research results
Figure 2. Probability of default by type of resource
The practical implication of this finding is that financial institutions and policymakers cannot treat rural credit operations homogeneously, regardless of the resource source. The existence of such a pronounced difference in the probability of default, even before controlling for other variables, suggests that the mechanisms for granting, monitoring, and collecting are intrinsically different between the two types of resources. For institutions operating with both modalities, this implies the need for distinct risk models and portfolio management strategies, adapted to the regulatory and market characteristics of each source. For policymakers, the initial result already points to the effectiveness of the regulatory framework of controlled resources in risk containment.
The robustness of this finding was further evidenced in Model 2, which incorporated the total effective cost (interest rate) as a control variable. Even after adjusting for the interest rate, the effect of the resource type remained statistically relevant and, notably, intensified, reaching an odds ratio of 2.84. This means that operations with free resources continued to present a probability of nonpayment almost three times higher than operations with controlled resources, even when the cost of credit was taken into consideration. Figure 3 visualizes this result, demonstrating that the difference in the probability of nonpayment between resource types persists and is accentuated after controlling for the interest rate.
Source: Original research results
Figure 3. Probability of default by type of resource, controlling for interest rate
This is one of the study’s most significant findings, as it directly refutes the idea that lower nonpayment in controlled resources would be merely a reflection of lower or subsidized interest rates. The persistence and intensification of the resource type’s effect, even after controlling for cost, strongly suggests that other factors, of an institutional and regulatory nature, are the true risk mitigators. This aligns with the literature that emphasizes the role of non-price mechanisms in credit allocation and risk management in markets with information asymmetry. The regulated environment of controlled resources, with its requirements, directives, and greater documentary discipline, appears to impose additional eligibility and monitoring filters that result in a higher quality portfolio, regardless of the interest rate charged.
From a practical point of view, this result has profound implications. For banks and credit unions, it means that simply comparing interest rates between free and controlled funds is not enough to assess risk. It is imperative that institutions understand and incorporate into their risk models the differences in the selection, monitoring, and credit recovery processes that characterize each type of fund. For policymakers, the finding suggests that the value of controlled funds lies not only in the interest subsidy but in the accompanying regulatory structure, which acts as an effective screening and risk management mechanism. Therefore, any reform in the rural credit system must carefully consider the maintenance or improvement of these institutional mechanisms.
The Model 3, the most complete adjusted logistic regression, confirmed the robustness of the association between the type of resource and default, even after simultaneous control for interest rate, year, federation unit, and producer size. Operations with free resources showed an odds ratio of 2.25 (95% CI: 2.21–2.29; p < 0.001), indicating that the probability of default in these operations is more than double that of operations with controlled resources. This result concludes the discussion on the origin of resources, confirming that the institutional and regulatory arrangement associated with controlled resources is a determining factor in mitigating rural credit risk.
The total effective cost variable (interest rate) showed an inverse association with nonpayment (OR = 0.99; 95% CI: 0.987–0.989; p < 0.001), suggesting that increases in the interest rate are associated with a *reduction* in the probability of nonpayment. This result, counterintuitive at first glance, is a crucial point for debate. It can be interpreted in light of credit selection and rationing mechanisms. In markets with information asymmetry, financial institutions may charge higher interest rates to borrowers perceived as riskier, but at the same time, they may apply stricter selection criteria, require additional guarantees, or implement more intensive monitoring for these operations. Thus, operations with higher rates may paradoxically be granted to borrowers with a better intrinsic risk profile or who are subjected to greater scrutiny, resulting in lower observed nonpayment. This finding directly challenges the premise that higher rates always imply greater risk, as suggested by Stiglitz and Weiss (1981) in their adverse selection model, by introducing the complexity of screening and monitoring mechanisms that can operate in parallel to pricing.
The practical implications of this inverse association are significant for risk management. Financial institutions should not interpret a higher interest rate as an automatic signal of increased default risk. Instead, they should investigate the underlying selection and monitoring mechanisms that may be associated with these rates. For policymakers, the result suggests that the interest rate, by itself, is not a linear indicator of credit risk and that credit policies should consider the interaction between pricing, eligibility criteria, and monitoring strategies.
A significant temporal effect was also observed in nonpayment. In relation to the reference year, there was an expressive increase in nonpayment in 2022 (OR = 1.80), 2023 (OR = 2.94), and 2024 (OR = 2.63), indicating an aggravation of risk over the analyzed period. In 2025, although still above the reference year, there was a relative reduction (OR = 1.36), suggesting a possible inflection in the trend. This temporal dynamic can be attributed to macroeconomic and conjunctural factors that affect the payment capacity of rural producers, such as fluctuations in agricultural prices, adverse climatic conditions, and variations in production costs (Braga de Melo & Resende Filho, 2017; Pereira & Melo, 2021). The COVID-19 pandemic and its economic consequences, followed by periods of high inflation and increased global interest rates, may have contributed to the increase in risk in the initial years of the analyzed period. The subsequent economic stabilization or recovery may explain the inflection observed in 2025.
For financial institutions, understanding these temporal effects is crucial for the development of dynamic risk models and for the implementation of provisioning and portfolio management strategies that consider economic and agricultural cycles. For policymakers, the results highlight the need for counter-cyclical policies and support mechanisms that can be activated during periods of greater economic or climatic vulnerability, in order to protect the stability of rural credit and producers’ income.
Regarding the spatial dimension, high heterogeneity was observed among the federation units. Several states presented a lower probability of nonpayment in relation to the reference category, notably Santa Catarina (OR = 0.19), Espírito Santo (OR = 0.17), and Amazonas (OR = 0.13). This regional variation is detailed in Table 1, which presents the estimated probabilities of nonpayment by UF and resource source.
Table 1. Probability of default by State and Resource Source
|
State |
Controlled Resource (%) |
Free Resource (%) |
|
AC (ref) |
3,6% |
8,1% |
|
AL |
3,4% |
7,6% |
|
AM |
0,5% |
1,1% |
|
AP |
3,6% |
8,1% |
|
BA |
2,1% |
4,7% |
|
CE |
3,0% |
6,7% |
|
DF |
0,6% |
1,3% |
|
ES |
0,6% |
1,3% |
|
GO |
1,3% |
2,9% |
|
MA |
3,5% |
7,9% |
|
MG |
1,0% |
2,2% |
|
MS |
2,0% |
4,5% |
|
MT |
2,3% |
5,2% |
|
PA |
3,4% |
7,7% |
|
PB |
2,7% |
6,1% |
|
PE |
2,9% |
6,4% |
|
GDP |
3,2% |
7,2% |
|
PR |
1,0% |
2,3% |
|
RJ |
2,2% |
5,0% |
|
RN |
2,8% |
6,3% |
|
RO |
2,6% |
5,9% |
|
RR |
1,6% |
3,6% |
|
RS |
1,0% |
2,3% |
|
SC |
0,7% |
1,6% |
|
SE |
1,4% |
3,2% |
|
SP |
1,3% |
3,0% |
|
TO |
2,4% |
5,4% |
Source: Original research results
Spatial heterogeneity can be explained by differences in regional productive structures, resilience to climate shocks, economic diversification, and the effectiveness of public policies and credit programs implemented locally. States with lower default rates may have more robust production chains, greater market access, or more effective technical and financial assistance programs. For policymakers, this finding underscores the importance of regionalizing rural credit policies, adapting them to the specific realities and needs of each federative unit. For financial institutions, regional analysis is essential to refine risk pricing, allocate capital more efficiently, and develop credit products that meet the particularities of each region.
Regarding producer size, Model 3 revealed that medium-sized producers presented a higher risk of nonpayment (OR = 1.15) compared to the reference category, while small producers presented a lower relative risk (OR = 0.88). This result refines the initial descriptive observation, which indicated a higher *number* of defaults among small producers. Logistic regression shows that, *controlling* for other variables, the *relative* risk of nonpayment for small producers is, in fact, lower. This finding can be explained by the greater involvement of small producers in targeted credit programs and public support policies, which often offer more favorable conditions, subsidies, technical assistance, and protection mechanisms, such as the Agricultural Activity Guarantee Program [Proagro], mitigating their relative risk. Bacha (2004) had already pointed out the role of the State and sector-specific financing policies in protecting specific segments.
For policymakers, this result validates the effectiveness of support programs for small producers, suggesting the continuity and improvement of these initiatives. For financial institutions, it implies the need to re-evaluate the risk profile of medium-sized producers, who, despite not being as exposed to extreme vulnerabilities as small ones, may not benefit from the same mechanisms of protection and credit targeting, making them relatively riskier. Figure 4, a “forest plot”, summarizes the odds ratios for the main variables of Model 3, offering a clear view of the magnitude and precision of the estimated effects.
Source: Original research results
Figure 4. Odds ratios for default
In summary, the logistic regression results showed that the origin of funds is a determining factor of nonpayment in Brazilian rural credit. Operations financed with free resources showed a significantly higher probability of nonpayment when compared to operations with controlled resources, even after controlling for crucial variables such as interest rate, year, federative unit, and producer size. This robust finding reinforces the hypothesis that the regulatory environment associated with controlled resources, marked by requirements, credit channeling, and greater rigor in granting and monitoring, acts as an effective mechanism for mitigating credit risk. The lower nonpayment observed in controlled resources is not solely due to a potentially lower credit cost, but to a distinct regime of granting, framing, and monitoring that shapes the risk profile of the portfolio.
The discussion of the findings suggests that credit risk in Brazilian rural credit is conditioned by institutional and regulatory elements that transcend the mere monetary pricing of the operation. The persistence and expansion of the effect of the type of resource in the regression, even with control for the interest rate, indicate that the origin of the resources represents a broader regime of granting, framing, and monitoring. In controlled resources, the existence of requirements, specific operational rules, and greater documentary discipline function as additional selection and monitoring mechanisms, explaining why portfolios with administered or limited interest rates do not exhibit worse performance. On the contrary, the regulated environment can reduce the entry of riskier operations or impose filters that improve the average quality of the portfolio.
On the other hand, free resources, although operating with greater contractual flexibility and commercial market logic, did not demonstrate lower risk. Evidence suggests that this flexibility may be accompanied by a distinct composition of borrowers, productive profiles, or hiring motivations that increase exposure to default. Thus, freer pricing was not sufficient to compensate for the structural factors that differentiate portfolios. The inverse association between interest rate and default in the multivariate model reinforces this complexity, suggesting that higher rates may be linked to more rigorous selection or monitoring mechanisms, or to a different intrinsic risk profile of borrowers accessing these lines.
From an academic point of view, this work contributes by demonstrating that the assessment of rural credit risk cannot be reduced to a comparison between administered interest rates and market interest rates. Logistic regression showed that the type of funding sources carries its own institutional content that robustly affects default, enriching the literature on institutional economics and rural finance. In practical terms, the findings are of interest to public policymakers, financial agents, and researchers in the rural credit market. The evidence of better performance of controlled resources suggests that regulatory mechanisms and reserve requirements can play a relevant role in risk mitigation, provided they do not excessively compromise operational efficiency and access to financing.
The implications for public policy are clear: the regulatory architecture of rural credit, especially concerning controlled resources, should be viewed as a strategic asset for the sector’s stability. Policies aiming at deregulation or the reduction of reserve requirements should be carefully evaluated for their potential impacts on credit risk. It is fundamental that policymakers consider the maintenance and improvement of the selection and monitoring mechanisms that accompany controlled resources, rather than focusing exclusively on interest rate liberalization. Furthermore, the temporal and spatial heterogeneity of the results points to the need for more flexible and regionalized policies, capable of responding to the specific dynamics of each period and region.
For financial institutions, the results indicate the need to develop risk models and portfolio management strategies that clearly differentiate between free and controlled resources. The simple application of generic risk models may underestimate the risk of operations with free resources and overestimate the risk of controlled operations. It is crucial to invest in credit assessment systems that consider not only the financial characteristics of borrowers but also the regulatory and institutional context of the operation. The understanding that small producers, despite their higher number of defaults, present a lower relative risk when controlled by other variables, suggests that targeted credit programs and support policies are effective and should be maintained, with special attention to medium-sized producers, who showed a higher relative risk.
In summary, this study analyzed credit risk in rural credit operations in Brazil, comparing operations with free resources and controlled resources from 2021 to 2025. The descriptive analysis showed an average default rate of 8.07% for free resources and 1.96% for controlled resources, evidencing superior performance of the regulated portfolio. Through logistic regression, this difference not only remained but strengthened after controlling for the interest rate. Operations with free resources presented an odds ratio of 2.48 in the raw model and 2.84 in the adjusted model, while the interest rate showed an inverse association with default, with an odds ratio of 0.99. These results lead to the rejection of the initial hypothesis that free resources would present lower risk due to pricing more aligned with the market. Instead, the findings indicated that controlled resources benefit from an institutional and regulatory environment capable of improving credit selection and reducing exposure to default. As limitations, the study can still advance with the inclusion of new explanatory controls, robustness tests, and detailing by producer profile, production type, modality, and purpose of the credit. Nevertheless, the results already offer evidence that the regulatory architecture of Brazilian rural credit is a fundamental element for understanding credit risk in the sector and for formulating more effective policies.
4. Conclusion
It is concluded that the objective was achieved by analyzing the credit risk in rural credit operations in Brazil, comparing financing with free and controlled resources, and verifying how the origin of resources and the cost of credit alter the probability of default. The results showed that operations with free resources presented a significantly higher probability of default compared to operations with controlled resources, even after controlling for variables such as interest rate, year, federative unit, and producer size. This finding refutes the initial hypothesis that freer pricing of resources would lead to lower risk, indicating that the institutional and regulatory environment associated with controlled resources acts as an effective risk mitigator. The inverse association between interest rate and default suggests the complexity of credit selection and monitoring mechanisms.
Despite the robustness of the findings, the present study has limitations that pave the way for future investigations. It is suggested that future studies may advance by including new explanatory controls, such as more detailed socioeconomic variables of producers, and conducting additional robustness tests to validate the models. Detailing by producer profile, production type, modality, and credit purpose can also enrich the understanding of risk dynamics. Such in-depth analyses would allow for an even more granular and contextualized examination, contributing to the formulation of more precise public policies and risk management strategies in Brazilian agribusiness.
Bibliographic References
Almeida, C.A. de. 2016. Determinantes da Inadimplência no Crédito Rural Brasileiro. Tese de Doutorado em Economia Aplicada. Escola Superior de Agricultura Luiz de Queiroz, Universidade de São Paulo, Piracicaba, SP, Brasil.
Bacha, C.J.C. 2004. Economia e Política Agrícola no Brasil. Atlas, São Paulo, SP, Brasil.
Braga de Melo, A.S.; Resende Filho, J. 2017. Uma crítica às renegociações da dívida rural: determinantes do risco de crédito rural no Brasil. Revista Brasileira de Economia 71(1): 69-92. Disponível em: <https://www.scielo.br/j/rbe/a/7sb9fHHyPZvsXgr8xBQgZcP/>. Acesso em: 19 abr. 2026.
Pereira, R.M.; Melo, A.S. 2021. Inadimplência no crédito rural brasileiro: uma análise dos fatores macroeconômicos e climáticos. Revista de Política Agrícola 30(2): 45-60.
Stiglitz, J.E.; Weiss, A. 1981. Credit rationing in markets with imperfect information. The American Economic Review 71(3): 393-410.
Article originating from the Final Course Work of the Specialization in Economics, Investments and Banking of the MBA USP/Esalq
To learn more about the course, click here and access the MBX Academy platform