区域性自由贸易协定对双边贸易流量影响的量化研究--基于机器学习方法的分析

A Quantitative Study of the Impact of Regional Free Trade Agreements on Bilateral Trade Flows--Analysis based on Machine Learning Methods

区域性贸易深度一体化是当前世界经济贸易发展的主要趋势,区域性自由贸易协定(FTA)的签订与重订是这一趋势的重要表现形式。为量化区域贸易协定对双边贸易流量的影响,区别于普遍用于外贸领域的引力模型,本文分别采用机器学习领域中的树模型XGBOOST与线性模型Ridge,选择FTA深度指标作为衡量FTA异质性的核心变量,添加协议签订国GDP等变量,对于协议生效3年内的双边贸易流量建立预测模型,探究机器学习模型在外贸领域的可用性。实证结果显示,XGBOOST模型在双边贸易流量的预测上取得了较好结果。贸易国间的双边贸易流量越大,预测结果越精准,对双边贸易流量在十亿美元以上、协议生效第一年的贸易,预测的平均绝对百分比误差约为0.077。因此,本文得到结论,机器学习模型在外贸领域中有较强的可用性,而且树模型的效果较优。

The deep integration of regional trade is the main trend of the current world economic and trade development,and the signing and revision of regional free trade agreements(FTA)is an important manifestation of this trend.In order to quantify the impact of regional trade agreements on bilateral trade flows,different from the gravity model commonly used in the field of foreign trade,this paper adopts the tree model XGBOOST and the linear model Ridge in the field of machine learning,and selects the FTA depth index as a measure of FTA heterogeneity.Core variables,adding variables such as the GDP of the countries that signed the agreement,establish a prediction model for the bilateral trade flow within 3 years of the agreement's entry into force,and explore the availability of machine learning models in the field of foreign trade.The empirical results show that the XGBOOST model has achieved good results in the prediction of bilateral trade flows.The larger the bilateral trade flow between trading countries,the more accurate the forecast result.For the trade in which the bilateral trade flow is more than one billion US dollars and the agreement is in force in the first year,the average absolute percentage error of the forecast is about 0.077.Therefore,this paper concludes that the machine learning model has strong usability in the field of foreign trade,and the effect of the tree model is better.