How eco-efficient are crop farms in the Southern Amazon region? Insights from combining agent-based simulations with robust order-m eco-efficiency estimation

Summary

Agricultural production plays an essential role in food security and economic development, but given its direct links within the environment, it is also an important driver of environmental degradation. It has become essential to not only produce more crops but doing it while maintaining or reducing the respective environmental impacts. A promising method for evaluating production efficiency is the nonparametric eco-efficiency analysis, which compares the economic value added against a composite environmental pressure indicator. This article proposes a novel method of evaluating the eco-efficiency scores, which does not depend on field survey data, but rather on multi-agent simulations. We present the first estimates of eco-efficiency for crop farms in the Amazon and Cerrado biomes in Brazil, identify regions and farm profiles that could be the focus of targeted interventions, and evaluate whether eco-efficiency scores could be improved using an alternative scenario. We combine a biophysical model with bioeconomic agent-based simulations to mimic land-use decisions of real-world farms. We then estimate the efficiency scores with an enhanced order-m estimator that conditions the efficiency estimates on explanatory variables, thus producing robust efficiency measures. Our simulations reveal that there are indeed differences in eco-efficiency estimates between macro-regions in the federal state of Mato Grosso. According to our simulations, the Southeast exhibited the greatest occurrences of inefficiencies, followed by the West macro-region. In our life-cycle inventory, sunflower cultivation had the lowest levels of environmental pressures. However, when evaluating it in a prospective scenario of infrastructure development, we could not observe a positive impact on efficiency. By using efficient computational methods, we replicate our simulations many times to create robust estimates that are more representative than a single field survey. In addition, our novel method combines simulated farm data with eco-efficiency analyses, allowing ex-ante impact evaluations where policy interventions can be tested before their implementation.