人工智能与汇率可预测性研究(英)

AI and Exchange Rate PredictabilityAmin IzadyarImperial College Business School, Imperial College LondonEmail:a.izadyar23@imperial.ac.ukJuly 3, 2025AbstractI revisit the exchange rate disconnect puzzle, first documented by Meese and Rogoff(1983), using generative artificial intelligence (AI) to forecast currency returns based oneconomic fundamentals. Using ChatGPT and DeepSeek, I analyze a comprehensivedataset of economic data releases for major currency pairs and measure the funda-mental strength of each currency. These AI-powered fundamentals exhibit significantcross-sectional predictive power. A simple trading strategy that goes long currencieswith strong fundamentals and short currencies with weak fundamentals generates aSharpe ratio exceeding 0.7 per annum. The excess returns of this strategy remainsignificant after controlling for traditional currency factors. To mitigate concerns oflook-ahead bias, I run multiple exercises to ensure that predictability stems from AIreasoning rather than memorization. Finally, I explore the potential sources of pre-dictability and find evidence that the Taylor rule framework, generally used by centralbanks to set interest rates, is a key mechanism connecting exchange rates to economicfundamentals.Keywords: Foreign Exchange, Return Predictability, Large Language Models, ChatGPT,Artificial IntelligenceJEL Classification: C53, F31, F37, G12, G15.1IntroductionThe ability of economic fundamentals to forecast exchange rates remains elusive, since mod-els based on fundamentals are often outperformed by a simple random walk, a phenomenonknown as the “exchange rate disconnect” puzzle (e.g., Meese and Rogoff, 1983). Althoughthe recent literature has identified a few economic variables that appear to have predictivepower, the answer to this empirical puzzle remains unresolved (e.g., Mark, 1995; Engel andWest, 2005; Rossi, 2013).Against this backdrop, the emergence of artificial intelligence (AI) offers new opportunitiesto re-examine this puzzle. AI’s recent advancements have enabled it to solve problems onceconsidered too complex or data-intensive for traditional methods. Motivated by these devel-opments, I leverage AI’s reasoning power and proficiency in handling large datasets to studya comprehensive dataset of economic data releases, covering over 500 indicators from 1996to 2024 for major economies. This study aims to find a link between exchange rates and eco-nomic fundamentals, thus enhancing our understanding of price discovery in the largest anddeepest financial market in the world. To preview my results, I find evidence that AI-derivedfundamentals can predict future exchange rate returns and the most important predictorsare Inflation data, Employment data, and Broad economic activity indicators.Using large language models (LLMs) like GPT-4o and DeepSeek-V3, I analyze a large datasetcomprising realized values, previous figures, and consensus forecasts of key economic datareleases, such as GDP reports, employmen