外匯利差交易策略──G10貨幣實證

Abstract

外匯利差交易過去常常被大型機構或私募基金利用獲取超額報酬，然而自金融海嘯發生至今，利差交易的報酬蒙受巨大損失，市場關注如此套利空間是否已經消失。本篇論文的目的在於建構一個統計模型、改善傳統外匯利差交易策略，期望有效降低最大損失的機率，我們選擇市場上最廣為流通的G10貨幣做實證，原始的貨幣組合為：借最低三個利率國家貨幣來投資最高三個利率國家貨幣，並以美金做基礎計算報酬，測試樣本內和樣本外模型是否可以改善整體貨幣組合報酬。 過去有學者運用Markowitz投資組合理論到利差交易，極小化投資組合變異數，定期調整每期每種貨幣的投資比重，本篇論文希望從另一個構面看待利差交易，藉由每一期皆相等權重的投資比例，著重在每個時間點是否採取不同的策略。此外，本文利用計量上logit模型，結合數個重要的總體經濟變數成單一指標做為平倉的依據，希冀能鎖定高低利差、減少匯率波動的影響。 實證結果顯示，swap利差、利率差以及共同波動的股票指數顯著地影響利差交易平倉的機率，本篇的兩個交易策略相較傳統的利差交易策略，呈現較高的夏普指數，並減少一半的最大損失，回溯測試也顯示，2008年中後的巨大損失可以藉由模型大幅降低。

FX carry trades have been extensively carried out but have confronted huge losses ever since the financial tsunami broke out. Concerns are aroused whether such an arbitrage opportunity has lost its attractiveness. Our purpose thus is to build up a statistical model which guides carry strategies by reducing large downside risks efficiently. G10 currencies are of our interests since they are traded with high liquidity, in which we go long three currencies with the highest interest rates and sell short another three with the lowest interest rates. We provide empirical results for our model and back tests in out-of-sample periods. A group of researchers have studied carry trades by following the concept of Markowitz Portfolio Theory. They look for the optimal allocation among a basket of currencies as time goes by. In this paper, we try to view carry strategies with a different aspect: looking for timing to implement other strategies. In addition, we use the logit model to transform information from important macroeconomic variables into one signal, which is further constructed as a unwind indicator. Hopefully, the indicator could mitigate FX fluctuation risks and preserve the locked-in interest rate differential. It is demonstrated that the swap spread, interest rate spreads and equity indices comovement have statistically significant impacts on carry unwind probabilities. Two modified carry strategies generally outperform the original carry portfolio in terms of higher Sharpe ratio and weekly draw-downs reduced by half. The back test also shows carry trades suffer severely after 2008, but the loss is mitigated by our model.