How different are the risk model providers, part two: absolute risk

Continuing with the question from my February 4 entry "How different are the risk model providers?" I shift my definition of risk from Tracking Error (Active Risk in Aegis-speak) to Absolute Risk (Portfolio Risk in Aegis-speak). Another way to think of Absolute Risk is as the tracking error against a cash benchmark.

While both forms of risk are important, I am sure many would have shared my initial guess that the conclusions of the analysis would have been virtually identical and, frankly, not worth another post. To my surprise, this wasn’t the case.

Using the same analytical construct, let’s review the results:

On first glance, Model Y clearly predicts much higher risk than the other models. This is different from what we observed when we considered tracking error. The results are consistent across all styles and the magnitude of the difference is striking. You can clearly and easily see, “Model Y is very different.”

Model X predicts higher risk than Model Z across all style groups, though the magnitude varies, and the smallest differences are in small cap. Probably the most important question in this comparison is whether the difference is statistically significant:

The Welch’s t-test of statistical significance confirms what we can see by eye-balling the averages. So, in total, this analysis suggests the three models are indeed significantly different predictors of absolute risk.

Coming soon: When we talk to clients about risk, we always focus on magnitude and direction. So, for my final perspective on this study, we will consider the three- and 12-month change in predicted risk. Has the predicted risk changed similarly across the three models?

To receive future posts by e-mail, subscribe to this blog.