Benchmark of a Dollar Neutral Strategy

Question

A dollar neutral strategy invests the same amount of money long and short without accounting for the volatility (risk) of either side. Depending on volatility you either end up positively or negatively correlated with the market.

A market-neutral strategy eliminates the correlation to the market by hedging the long side with an equally risky (=same volatility) short side. E.g. A hedge of a stock portfolio with a short position on the S&P500 Future. The size of the short position is chosen in a way that the resulting strategy doesn’t correlate with the S&P any more.

I have a dollar neutral strategy that is positively correlated with market but the value of net beta is not zero. What should be the appropriate benchmark for this strategy? Are there any methods to construct a portfolio that can be a benchmark for this strategy ( like regressing market returns with dollar neutral etc) ?

Answer 1

The initial passage you quoted https://www.quora.com/Whats-the-difference-between-market-neutral-and-dollar-neutral-strategy (by Martin Fröhler on Quora) explains the difference between dollar-neutral and market-neutral, but it is slightly garbled (the word "volatility' should be replaced by 'beta').

As to your question, the benchmark for a portfolio with a given beta (in your case slightly positive) is given by the CAPM equation. In your case it will be slightly different than the Risk Free Return, but not much, i.e. $R_F+β(R_M−R_F)$ for the time period in question, with Beta small but positive.

Answer 2

Please let me be sure I have this right. Your strategy is positively correlated with market; but the beta is non-zero. IS the beta positive or negative? if negative, your previous statement cannot be true; you have to mis-measuring something somewhere (happy to help)!

If your beta is the right sign but <>1, then that just reflects the imperfect ability of models to predict reality! Beta is just the relative volatility of your output over input times the correlation of the model. Correlation will always be <1 unless you claim perfect predictive powers over the future :-) So beta <>0 becomes a measure of the volatility/variation of the your inputs here... that's the "signal" (relevant or not) tha you're chasing here...

very best, DEM