Second, when an SFES does detect a price impact, it reflects confounding effects that are unrelated to the alleged fraud.
These problems help explain why the SFES methodology is applied so infrequently in peer-reviewed research.
In Part I, we explain why the SFES as typically applied in securities litigation has low statistical power, in the sense that it cannot detect price impacts reliably unless they are large.
Figure 3 illustrates why the SFES will miss detecting that price impact with high probability.
So the SFES can detect very large price impacts reliably, but it cannot detect smaller price impacts reliably.
In one review, power in the SFES context ranges from 5% to 17%, depending on assumptions of effect size and standard deviation.
6 million for the SFES to detect it, assuming there are no confounding effects pushing it toward significance.
But even the total price move ($300 million plus $200 million) would be statistically insignificant in an SFES because it falls below the approximately $900 million detectability threshold for decile 10 firms.
The Plaintiffs submitted a report of an expert criticizing the SFES as "highly suspect," but the court was unconvinced:
It may be helpful to understand why low statistical power does not plague the MFES like it does the SFES.
The matter is much more difficult in an SFES because there is no averaging away of confounding effects.
0%, but the SFES had no power to detect that impact on its own; -2.