News Beta - A New Measure of Risk & Stock Selection

An investor evaluating a given stock for a portfolio will use market beta to gauge its potential volatility in relation to the market. The concept of a (market) news beta provides a way to measure the responsiveness of individual stock prices to a different kind of benchmark - market sentiment. While beta from CAPM only considers market returns, news beta tries to capture the underlying factors that drive prices, as typically reported in financial news. These behavioral factors may be captured by calculating a market sentiment index from news; something which I have previously covered in multiple blog postings.

Based on the news beta measure, it's possible to divide a stock universe into three groups; namely positive, negative, and zero news beta stocks. A further ranking can be done within each group using for instance the R-squared of the regression model, a measure that represents the explanatory power of the model.

For a company to be classified a positive or negative news beta company, its stock price must move with or against the market sentiment benchmark in the time period of the previous regression window. Several factors are likely to impact the relationship between news betas and stock performance, translating into an expectation of potential momentum or reversal type effects. Whether positive or negative news beta stocks are likely to outperform the market may depend on the degree of which the market is dislocated from the news as probed by a market sentiment index; or by economic cycles or regimes.

Figure 1 depicts the annualized log-return of a long/short strategy presented in my latest research paper; back-tested over a 10+ year period (Nov. 2000 through Sep. 2010). The strategy is based on a three months sentiment aggregation window and a one month investment horizon going long positive news beta stocks, and short negative news beta stocks. As can be observed, the strategy would have been profitable in 9 out of 11 years making it an interesting stand-alone signal or an overlay to an existing model.

Figure 1: Out-of-sample annualized returns for the L/S 10 and L/S All news beta strategies covering the constituents of the S&P 500. News betas have been calculated based on a 6 month regression, a 3 month sentiment aggregation window, and a 1 month investment horizon.

If you’re interested in the full study, please requested it here.