The relationship between agricultural credit and agricultural production, and its subsequent impact on sustainable development goals (SDGs) such as poverty reduction, inequality reduction, food security, and economic growth, is well-established. However, the effects can be ambiguous. This study focuses on Colombia, a country where the cattle sector plays a crucial role in the economy, employing 1.1 million people (6% of national and 19% of agricultural employment) and receiving a significant portion of agricultural credit (24.6% between 2004 and 2014). Despite its economic significance, extensive cattle ranching is a major driver of deforestation in Colombia (approximately 60%), raising concerns about ecological conservation and climate change. The study aims to analyze the impact of agricultural credit, specifically through FINAGRO, on both cattle production and deforestation, addressing the challenge of balancing increased agricultural productivity with environmental sustainability and the achievement of multiple SDGs (UN-SDG 1, 8, 10, 3, 7, 11, 12, 13, and 15). The research uses spatial panel data models to account for the spatial dependencies inherent in agricultural practices and deforestation patterns.

Existing literature extensively documents the importance of credit for agricultural development in developing countries, generally showing a positive correlation between credit access and production. However, research also reveals ambiguous effects on poverty reduction and rural welfare improvements. Studies on credit and agricultural production in Colombia have largely focused on the coffee sector, demonstrating potential positive impacts on farmer welfare. The cattle sector, given its economic and environmental importance in Colombia, warrants specific investigation. Previous research highlights the challenges of formal credit, including high transaction costs, information asymmetries, high interest rates, and stringent requirements. This underscores the importance of geographical location of financial institutions, government support, producer associations, and diversified financial inclusion channels.

This study utilizes a departmental-level panel data set for Colombia spanning 2011-2020. Data sources include public entities such as FINAGRO (Fund for the Financing of the Agricultural Sector), the Colombian Agricultural Bank, the Colombian Agricultural Institute (ICA), and the Institute of Hydrology, Meteorology and Environmental Studies (IDEAM). Two dependent variables are employed: cattle herd size (representing cattle production) and annual deforestation rate. Independent variables encompass access to credit (total agricultural credit amounts and values, cattle-specific credit, and number of Agricultural Bank branches), land use (agricultural land, cattle land, and coca cultivation), and a sociodemographic control variable (% rurality). Both standard panel data models (OLS, FE, RE) and spatial panel data models (SAR, SEM, SARAR) are employed to account for potential spatial autocorrelation and spillover effects between departments. A weight matrix (W) defines spatial relationships, based on departmental borders. Model selection is informed by LM tests for spatial dependence. Descriptive analysis using choropleth maps visualizes data distribution. Statistical analysis is performed using Stata and R.

The descriptive analysis reveals a center-periphery structure in credit distribution, with concentrations in the Central Andean region. Econometric results, using spatial panel data models (SARAR for cattle production and RE for deforestation), show: For cattle production, access to credit has a significant but ambiguous effect, varying depending on the credit metric. The number of agricultural credits and the total value of cattle credits show positive relationships with cattle production, whereas total agricultural credit value and the number of cattle credits show negative relationships. Agricultural and cattle land use significantly influence cattle production. Spatial spillover effects in the SARAR model are primarily direct, not indirect. For deforestation, the RE model reveals significant negative relationships with cattle herd size and cattle land use; a larger cattle herd correlates with higher deforestation rates. Interestingly, coca cultivation also negatively correlated with deforestation rates, highlighting other drivers of deforestation beyond the cattle sector. No significant relationship was found between deforestation and access to credit, nor evidence of spatial autocorrelation in deforestation.

The findings partially align with existing literature on agricultural credit and production, demonstrating the significance of credit access in influencing cattle production. However, the ambiguous impact emphasizes the importance of the metric considered (number or value of credits). The spatial dependence in cattle production underscores spillover effects, and the absence of spatial autocorrelation in deforestation suggests departmental-level analyses might overlook finer-scale interactions. The negative relationship between cattle production and deforestation confirms its contribution to deforestation; however, other factors (coca cultivation, omitted variables) also play crucial roles. The lack of a relationship between credit and deforestation might stem from data limitations or the omission of variables relating to conflict or social and political factors. These findings suggest the need for further research to clarify the interplay between credit, land use, and deforestation at a finer spatial scale and incorporating social and economic factors.

This study provides valuable empirical evidence on the relationship between agricultural credit and both cattle production and deforestation in Colombia. While access to credit positively impacts cattle production, its effect on deforestation is not significant. This highlights the need for targeted credit policies that promote sustainable intensification of cattle farming and incorporate environmental safeguards. Future research should focus on disaggregated data (municipal or individual level), incorporation of socioeconomic factors, and improved deforestation data collection methods.

The study acknowledges limitations, including the departmental level of analysis (limiting the ability to capture finer-scale interactions), unbalanced panel data, reliance on self-reported survey data, and potential omission of variables related to armed conflict and social/political events (affecting deforestation). The choice of deforestation rate as the dependent variable may also be problematic, especially due to the limited availability of data.