Yield curve arbitrage in the EUR swap rates market. Replicating the strategies of quantitative arbitrageurs

Abstract

OBJECTIVES OF THE STUDY:

In this thesis, I look into a hedge fund strategy known as a yield curve arbitrage, where arbitrageurs take relative value bets on interest rates. Earlier research has shown that the strategy produces favourable returns in the USD swap rates market in 1988-2004. My objective is to study whether the strategy yields attractive risk-adjusted returns and multifactor alpha in the recent period of 2002-2015 in the EUR swap rates space. I shall employ an enhanced modelling framework to implement the trading strategy. Moreover, I test the replicated strategy returns with respect to high-level and style-specific hedge fund index returns. Finally, I look into whether 'high-noise' periods in the markets coincide with large model-implied mispricing of rates. The empirical objectives of the thesis are linked to literature on yield curve formation and no-arbitrage.

DATA AND METHODOLOGY:

The dataset consists of monthly mid-market observations of constant maturity EUR swap rates for maturities of one to ten years. Also, Hedge Fund Research and Credit Suisse hedge fund index data for both high-level and style-specific indices is employed. Moreover, noise measure data by Jun Pan is employed to study the relationship of replicated returns to the level of noise. The methodology builds on Cox-Ingersoll-Ross and Longstaff-Schwartz two-factor stochastic short-rate models of interest rates. A calibration and trading algorithm is constructed based on these models to replicate the arbitrage strategy returns. Back tested trading is done explicitly out-of-the sample.

FINDINGS OF THE STUDY:

The yield curve arbitrage is found to produce attractive risk-adjusted returns and favourable return distributions. Moreover, the alpha of the strategy is statistically and economically significant when controlled by a number of commonly employed risk factors. Additionally, it is found that the replicated arbitrage strategy does not have statistically meaningful connection to neither high-level nor style-specific hedge fund indices. Finally, it is shown that high noise coincides with large model-implied mispricings when the measure of the mispricings is smoothed. No evidence is found in support of the idea that yield curve arbitrage alpha is compensation for carrying tail risk.