Merrill Lynch today announced the development of the ML Alpha Surprise Model index, which tracks the performance of a basket of stocks selected from the S&P 500 using the firm’s proprietary Alpha Surprise Model.

The Alpha Surprise Model is based on the observation of a general correlation between Merrill Lynch fundamental equity analysts’ out-of-consensus earnings estimates and future stock performance. Specifically, when Merrill Lynch equity analysts have forecasts that are significantly above the consensus estimates, these high conviction ideas have historically been predictive of above-market stock price performance.

The model focuses on those companies for which Merrill Lynch equity analysts have a high degree of conviction in the direction of earnings estimates. Starting with these high conviction stocks, it then identifies those that appear to be undervalued, and generates a list of stocks that are attractive based both on the earnings expectations and valuation.

“For the Alpha Surprise methodology, we rank stocks selected from the S&P 500 using a 75%/25% weighted combination of our Merrill Lynch vs. Consensus Earnings Surprise Model (the growth or “Surprise” component) and our Dividend Discount Model (the value or “Alpha” component),” explained Savita Subramanian, quantitative strategist at Merrill Lynch, in a release. “The ‘Surprise’ component identifies stocks where Merrill Lynch analysts have earnings forecasts that are significantly above consensus estimates, and the ‘Alpha’ component identifies those stocks that are inexpensive.”

The firm reports that the ML Alpha Surprise Model index has historically offered a higher level of return and a lower level of risk than the S&P 500. It has generated 11.4 percentage points annual total return since December 1999 vs. the S&P 500’s 1.2% over the same period, and the volatility of returns has been lower than that of the S&P 500.

“The insight of the model is demonstrated by its track record, and the philosophy behind it is simple: we seek to capture the most undervalued high conviction calls from our analysts using an objective, quantitative process,” said Subramanian.

The ML Alpha Surprise Model index is constructed as an equal-weighted basket of the Alpha Surprise Model’s stocks, which are selected each month.