Bernstein Research analysts Vadim Zlotnikov, Ann Marie Larson and Charles Clavel provide insights into the “live” 12-month U.S. Bernstein Alpha Model and its year-to-date performance.
The 12-month U.S. Bernstein Alpha Model
The model has a modular structure and rests on four core components, based on the source of expected alpha:
- Stock selection
- Industry rotation
- Regime timing
- Factor timing
The stock selection component segregates between under- and out-performing stocks. The industry rotation component does the same, but for industries. The regime timing toggles between high or low volatility stocks, as well as high or low beta industries, depending upon the model’s inbuilt risk aversion signal. The factor-timing component acts as a filter for the stock selection component by switching off certain factors (such as price-to-book, price-to-sales or price momentum) when these indicate the likelihood of adverse performance.
Interestingly, the model allows for the component-wise breakdown of aggregate performance. The process of doing this is a complex, step-by-incremental-step computation that segregates individual performance until the aggregate model performance is constructed. Bernstein run this process on a monthly basis.
Overall performance
The model outperformed a universe of the top 1,500 stocks in terms of market cap in nine out of eleven months since the launch, as shown in the graph below.
On a cumulative basis, the out-performance for the long/short basis is +9.3 percent and long only basis is +6.1 percent.
Component-wise performance
The year-to-date return from a long/short strategy, broken down between components is as follows:
YTD Cumulative = +7%
Stock Selection = +3.4%
Industry Rotation = +4.3%
Regime Timing = -1.9%
Factor Timing = +1.3%
This Model versus Traditional Methodology
Bernstein also ran a comparison of the new Alpha Model with a previous 12-month traditional model as well as a current 1-month traditional alpha model.
Result: The new 12-month model outperformed both the traditional models for both long and long/short strategies.
Stock selection analysis performance
The stock selection analysis performance was broken down by industry and it is interesting to note that good returns were obtained from Household & personal products, materials, commercial and professional services, telecom services, and capital goods. However, it was not very effective in banks, food & staples retail, media, autos & components and energy.
Mid and small caps universe
The model was significantly effective and out-performed in mid and small caps sectors, but less so on the top 500 large caps.
