Does convertible arbitrage risk exposure vary through time?

We model the returns of the convertible arbitrage strategy using a non-linear framework. This strategy has generated long periods of positive returns and low volatility, followed by shorter periods of extreme negative returns and high volatility, associated with market upheaval. We specify a smooth transition regression model to assess performance, a class of model particularly suited to this type of strategy as it allows gradual transition between risk regimes. We show that in alternate regimes, the strategy exhibits relatively high (low) exposure to risk factors and alpha is high (low). The evidence reported can account for abnormal returns demonstrated in previous studies.