Risk Premia Strategies - Credit Markets
The Art of Selectively Harvesting Attractive Risk Premia
  • Credit markets offer three main opportunities for risk premia:
    1. Filtered credit carry programs that are essentially long investment grade and high yield companies and implemented so as to reduce drawdowns
    2. Systematic momentum strategies for credit indices based on trend-following and mean-reversion techniques
    3. Long/short CDS strategies based on credit spread valuation models, based on the estimation of the fair value of credit spreads based on fundamental and market indicators


Carry strategies, or long credit strategies, consist in either purchasing bonds or selling CDS protection, which is a standard premium strategy. A buyer of CDS protection will receive a credit spread in compensation for the default risk of the reference entity he takes a position in.
For a bond holder, the credit spread may be defined as the additional yield an investor receives for a security with credit risk over the risk-free yield.
The credit spread of an instrument is, in general, significantly higher than the expected value of the potential loss and offers a risk premium reward to the investor. It can be broken down into three main components:
  1. The default premium, or the compensation that an investor receives for expected losses associated with a default
  2. The risk premium, or the compensation for uncertainty on default losses and realized losses potentially exceeding expected losses
  3. The liquidity premium, or the compensation for the lower liquidity of an instrument compared to other instruments with the same credit exposure
Long credit has been one of the better performing strategies in recent years. Moreover, the search for yield in a current low interest rate environment, coupled with low default rates has boosted investors’ appetite for high yield names in credit markets.


QMS Advisors' credit momentum strategy makes use of a standard signal processing technique known as the Kalman filter. This statistical technique allows for splitting the credit spread time series into two components, a momentum and a mean-reversion component. Trading rules are defined depending on the level of these two components.
    • The trend system is based on an adaptive indicator using four trend indicators from one month and three month regressions of the log spread and of the mean reverting components against time. If the slope is positive, it is a buy protection signal; if it is negative, it is a sell protection signal. The retained trend signal is the signal from the indicator which has performed the best in the past month. If all four trend indicators show negative performances in the past month, we sell credit index protection to monetize the carry. The mean-reversion is based on the assumption that the variable is mean-reverting around zero and the strength of the mean reversion increases when it moves away from zero. When the mean-reverting variable is high (z-score >1.3) we sell credit index protection, when it is low (z-score <-1.3) we buy credit index protection. The signal can be interpreted as a statistical stop gain on the trend-following strategy as the mean reversion is activated after a particularly strong widening (or tightening) of credit spreads.

As one would assume, the above momentum strategies performed best when spreads move sharply in one direction, either on the upside or on the downside and the performance is more limited when the market is range-bound.


These sources of risk premia cover strategies based on credit spread scoring models that are designed to generate alpha performance in credit markets. They aim at estimating the fair value of credit spreads based on fundamental and market indicators. Their principle is to implement long/short CDS strategies based on credit spread valuation models that buy (sell) protection on the 20 tightest (widest) names relative to the model.

    • Usually, names with CDS spreads above 1,000bp (high default risk) or with bias correction higher than the spread level (model signals lacking reliability) are excluded.
SelectionFile type iconFile nameDescriptionSizeRevisionTimeUser