人工智能与实际汇率

Artificial Intelligence and the Real Exchange Rate

本文将人工智能变量以生产任务方式引入Balassa-Samuelson模型(巴萨模型),理论分析人工智能发展对实际汇率的影响。研究发现:可贸易品部门智能化程度提高将导致实际汇率升值,不可贸易品部门智能化程度提高将导致实际汇率贬值。本文以工业机器人安装存量表示智能化程度,利用国际机器人联盟、Peen World Table 10.0和世界发展指标等数据库构建1993—2019年72个国家和地区的平衡面板数据,实证检验发现:可贸易品部门智能化程度提高导致实际汇率升值,且结论在稳健性检验和内生性处理后依然成立;不可贸易品部门智能化程度提高导致实际汇率贬值的结论在统计上不稳健。拓展分析发现:第一,人工智能发展对实际汇率的影响存在滞后效应;第二,分行业而言,制造业智能化程度提高导致实际汇率升值,建筑业智能化程度提高导致实际汇率贬值;第三,OECD成员国可贸易品部门智能化程度提高导致实际汇率升值,不可贸易品部门智能化程度提高导致实际汇率贬值;而非OECD成员国智能化程度改变对实际汇率影响不显著。

Intelligence Artificial(AI) will be the new driving force for the world economy.The studies on the economic impact of AI have been one of the hottest topics in economics.However,there are not enough studies on how AI can exert its impact in the field of international trade and international finance.This paper studies the impact of intelligence level on the real exchange rate by introducing the artificial intelligence in Balassa-Samuelson model(B-S model).The theoretical model shows that an increase of intelligence level in tradable sector appreciates the real exchange rate,while the increase of intelligence level in non-tradable sector depreciates the real exchange rate.This paper empirically tests the theoretical hypothesis by using installed stock of industrial robot as the proxy variable of artificial intelligence,along with year' 1993-2019 balanced panel data from 72 countries or regions,sourced from the International Federation of Robotics,Penn Table 10.0 and World Development Index.The empirical results show that,an increase of intelligence level in tradable sector appreciates the real exchange rate,conclusion that holds after robustness checks and addressing endogeneity concerns.However,the increase of intelligence level in non-tradable sector depreciates the real exchange rate is statistically unrobust.The extended results show that,firstly,there is a lagging effect on the impact of intelligence level on the real exchange rate.Secondly,the increase in the intelligence level of manufacturing appreciates the real exchange rate,while the increase in intelligence level of constructure depreciates the real exchange rate.Thirdly,the improvement of intelligence level in OECD members' tradable sectors appreciates the real exchange rate,while the improvement of intelligence level in OECD members' non-tradable sectors depreciates the real exchange rate.However,the intelligence level of both sectors in non-OECD member has no significant impact on the real exchange rate.This paper has two marginal contributions,firstly,discusses the impact of AI on the real exchange rate,which enriches the literatures on the economic impact of AI and the determination of the real exchange rate.Secondly,introduces the economic modeling method of AI into the B-S model,which lays a foundation for further research in the future.