The impact of agricultural credit on the cattle inventory and deforestation in Colombia: a spatial analysis

This study investigates how agricultural credit influences cattle production and deforestation across Colombian departments. Credit is widely recognized as a lever for agricultural development and multiple SDGs, yet its effects on poverty and welfare can be ambiguous due to transaction costs, information asymmetries, and access barriers. Colombia’s credit system is largely channeled through FINAGRO and the Agricultural Bank of Colombia, with banking presence expanding yet still concentrated in the Central-Andean region. The cattle sector is socioeconomically important but linked to environmental impacts—especially deforestation—due to extensive land use and low productivity. The research question centers on whether and how departmental access to credit relates to cattle herd size and deforestation, and whether spatial spillovers are present.

The literature documents generally positive relationships between formal credit and agricultural production across Latin America (Peru, Colombia, Brazil, Chile) and globally (Africa, Southeast Asia, China), with specific attention to coffee and cattle sectors due to their economic significance. Methods range from surveys and impact evaluations to econometric analyses (OLS, panel data, IV). Despite the spatial nature of agricultural production, few studies examine credit-production links spatially; several scholars recommend spatial approaches due to non-random territorial distributions. On cattle, productivity, and deforestation, research streams include: climate action-focused studies; reports by multilateral organizations highlighting the role of extensive cattle ranching and credit policy; and work on local climate action and policy frameworks (e.g., Colombia’s Green Growth Policy). Findings are mixed: intensification in consolidated regions may reduce deforestation, whereas in frontier areas or contexts of insecure land tenure it may increase it. There is also evidence of spatially correlated deforestation dynamics and policy instruments (e.g., Soy Moratorium, Cattle Agreements) with varying success.

Design: Department-level unbalanced panel. Two dependent variables: (i) cattle production (herd size), 2011–2020; (ii) annual deforestation rate, 2012–2019. Data sources: Cattle herd size from ICA; deforestation rates from IDEAM; credit access from FINAGRO (number and total value of agricultural credits; number and total value of cattle-sector credits) and the number of branches of the Agricultural Bank of Colombia; land use from DANE’s ENA (agricultural land; cattle land); coca hectares from the Colombian Drug Observatory; sociodemographic control: % rurality (from 2005 and 2018 censuses with projections). Econometric approach: For cattle production, compared OLS and fixed-effects (FE) panel models, then estimated spatial panel models: SAR, SEM, and SARAR using a queen contiguity weight matrix W. Spatial dependence diagnosed via ESDA (Moran’s I, Geary’s C) and LM/Robust LM tests to select model. For deforestation, tested spatial dependence; in its absence, estimated OLS and random-effects (RE) models, with Hausman tests guiding FE vs RE selection. Spatial parameters ρ (lag) and λ (error) estimated where applicable. Joint significance and collinearity checks among credit variables guided inclusion of four credit measures. Software: QGIS 3.12 (mapping), StataMP 13 (OLS/panel FE/RE), RStudio (spatial models).

Descriptive patterns: - Credit offices and agricultural credits are concentrated in the Central-Andean region. Cattle-sector credits align with main cattle production areas. Credit volumes rose notably from 2016, peaking in 2020 (COVID-19 emergency support). Substitute portfolio credits dominated since 2017, suggesting potential concentration among larger or more collateralized producers. - Deforestation rose markedly 2014–2015 to 2017–2018 in Amazon/Orinoco departments (Vichada, Guaviare, Caquetá), aligning with cattle expansion zones. Cattle production (herd size): - Robust LM tests indicated both lag and error dependence; SARAR provided best fit for spatial dependence. - Access to credit shows mixed effects: number of agricultural credits positively associated with herd size (e.g., SARAR direct effect positive; Table 1/2), but total value of agricultural credits negatively associated (SARAR ≈ −0.305**). For cattle-specific credit, the number of cattle credits is negatively associated (SARAR ≈ −151.8***), while the total value of cattle credits is positively associated (SARAR ≈ +2.606**). Number of Agricultural Bank offices strongly positive (≈ +13,017*** in SARAR). - Land use: agricultural land use negatively associated with herd size (SARAR ≈ −0.854*), while cattle land use positively associated (≈ +0.299***). Coca hectares are not significant. - % rurality is significant (negative in spatial models). Spatial spillover decomposition (Table 2) shows impacts are predominantly direct; indirect spillovers are small and largely insignificant. Deforestation (annual rate): - No spatial dependence detected; RE preferred over OLS by Hausman. - Credit variables (counts and values; agricultural and cattle) are not robustly significant predictors of deforestation in RE models; overall no evidence of a credit–deforestation relationship. - Cattle production variables (herd size, cattle land use) and % rurality significantly influence deforestation rates; increases in cattle production are associated with higher deforestation rates. Coca cultivation also shows significant association with deforestation, indicating multiple drivers. Policy-relevant quantitative highlights (from Tables 1–3): - SARAR coefficients (approximate): total agricultural credit value −0.305**, number of cattle credits −151.76***, total cattle credit value +2.605**; agricultural land use −0.854*, cattle land use +0.299***; Agricultural Bank offices +13,017***; % rurality negative and significant. Spillover total effects align in sign with direct effects (Table 2). For deforestation RE, herd size and % rurality significant; credit variables not significant.

Findings confirm a center–periphery pattern in financial access: credit offices and disbursements cluster near Colombia’s Central-Andean core, elevating access costs for remote producers and potentially biasing credit toward larger or urban-adjacent producers. Econometrically, credit’s relationship with cattle production is significant but directionally mixed across credit measures: counts vs values and general agricultural vs cattle-targeted lines have opposite effects, aligning with literature on heterogeneous credit impacts and timing. Spatial dependence characterizes cattle production across departments, but not deforestation at this aggregation and with available variables. The positive association between cattle production and deforestation underscores that, in the context of limited adoption of sustainable intensification, herd expansion contributes to forest loss. However, credit access per se does not exhibit a direct link to deforestation, suggesting that omitted social, legal, and political drivers (e.g., conflict dynamics, land tenure, enforcement) are crucial mediators. These results align partly with Brazilian evidence where intensification can reduce or increase deforestation depending on regional consolidation and governance. Policy implications include coupling credit with environmental safeguards (e.g., no-deforestation verification, targeted sustainable intensification lines) and strengthening extension, monitoring, and enforcement to avoid rebound effects.

This study contributes spatially explicit evidence on how agricultural credit relates to cattle production and deforestation in Colombia. It shows: (i) access to credit significantly affects cattle herd size, with credit counts and values having opposite effects depending on sectoral targeting; (ii) cattle production exhibits spatial dependence, but deforestation does not at the departmental scale with the selected variables; and (iii) cattle production increases are associated with higher deforestation rates, while credit itself is not directly linked to deforestation. Policy should expand equitable rural credit access (especially in peripheral departments), integrate environmental pre-conditions into cattle credit (e.g., no-deforestation checks, sustainable intensification lines), and pair finance with technical assistance, monitoring, and enforcement. Future research should use finer spatial units (municipal/producer level), incorporate higher-quality deforestation data (e.g., satellite-based), and include social, legal, and political variables (e.g., conflict, land tenure) to better identify mechanisms and spatial spillovers.

- Spatial resolution: Departmental analysis; lack of municipal/individual data may obscure localized spatial dependence, especially for deforestation. - Unbalanced panel: Missing data for some years (e.g., land use variables not available for 2020) can affect estimates. - Measurement: Land use data from ENA are self-reported and incomplete for some departments/years; deforestation variable is an inter-period rate with limited availability (2012–2018) and may mismatch with annual covariates. - Omitted variables: Social, legal, and political drivers (armed conflict, land tenure, eradication campaigns, migration) and shocks (COVID-19) are not fully captured, possibly biasing deforestation models. - Credit allocation data may reflect institutional constraints (e.g., branch coverage, substitute portfolios) not fully modeled.