Factor Models: Backtesting Experiments

©2002 OS Financial Trading System

Does exploiting factor information to build a portfolio of stocks improve performance?

Suppose our target return is 0.0169 per month. This return lies in the relative stable region of the minimum variance variance scatter plot.

Experiment 1: Traditional Markowitz

Step 1: Uncheck Factor Model and leave Allow Short Sales checked.

Step 2: Click on Calculate Portfolio

Step 3: Set Block size to equal 100, and the drop down to Optimized. Finally, click on Backtest.

Results:

The realized average return is 0.00635 (Target = 0.0169), realized volatility 0.04714 (Target = 0.0485).

Experiment 2: Factor Model

Step 1: Check Factor Model and leave other inputs as is.

Step 2: Click on the button Calculate Portfolio.

Step 3: Click on Backtest.

Results:

The realized average return is 0.01402 (Target = 0.0169), realized volatility 0.04662 (Target = 0.0216).

Clearly, experiment 2 dominates experiment 1 strictly (risk and return).

You can check the robustness of the above finding by varying the target return above and below 0.0169 and repeating the two experiments.

Note: Be sure to delete the existing return number beside Target, enter a new target return number followed by Enter (or Return) key.