Forecasting Peruvian Blueberry Exports employing Markov Chains, SARIMA, and Log-Linear Growth

This study analyzes the export dynamics of Peruvian fresh blueberries by comparing three forecasting approaches: a Markov–Monte Carlo model, a log-linear growth mod-el, and a seasonal SARIMA model. Monthly data on export value (FOB), export vol-ume, and unit price covering the period 2012–2025 are employed. The analysis com-bines descriptive statistics, seasonal Markov transition matrices, out-of-sample fore-cast evaluation using MAE and RMSE, and residual diagnostics. The results show that log-linear growth models provide accurate short-term point forecasts for export val-ues, reflecting the dominance of a strong trend component. SARIMA models perform better for export volumes and exhibit favorable residual properties, supporting their use for uncertainty quantification. In contrast, the Markov–Monte Carlo approach yields superior performance for export prices and offers valuable insights into seasonal regimes, persistence, and risk through simulated distributions. Overall, the findings indicate that no single model dominates across all dimensions of the export chain. A combined use of deterministic, stochastic, and regime-based approaches provides a more comprehensive framework for forecasting and risk management in dynamic ag-ricultural export sectors.本研究围绕秘鲁新鲜蓝莓的出口动态展开分析，通过对比三类预测方法开展研究：马尔可夫-蒙特卡洛（Markov–Monte Carlo）模型、对数线性增长模型与季节性SARIMA模型。研究采用2012至2025年的出口额（离岸价FOB）、出口量与单价月度数据。本次分析融合描述性统计、季节性马尔可夫转移矩阵、基于平均绝对误差（MAE）与均方根误差（RMSE）的样本外预测评估，以及残差诊断手段。结果显示，对数线性增长模型可针对出口额实现精准的短期点预测，反映出出口额存在显著的趋势性主导成分。SARIMA模型在出口量预测上表现更优，且具备良好的残差特性，可用于不确定性量化分析。与之相对，马尔可夫-蒙特卡洛方法在出口单价预测上性能更出色，其通过模拟分布能够为季节性制度、持续性与风险提供极具价值的洞察。整体而言，研究结果表明，不存在能够覆盖出口链条全维度的最优单一模型。将确定性方法、随机性方法与基于制度的方法结合使用，可为动态农产品出口行业的预测与风险管理提供更为全面的分析框架。