Wednesday, May 28, 2008

Study Claims Milberg Weiss Scheme Hurt Shareholders

Anthony Lin, New York Law Journal, May 28, 2008:

As former securities class action king Melvyn I. Weiss awaits sentencing for his role in the payment of kickbacks to named plaintiffs in shareholder suits, a conservative think tank is set to release a study purporting to show that the scheme injured shareholders.
The American Enterprise Institute Legal Center is releasing today an article by professor Michael Perino of St. John's University School of Law that takes on the argument that the Milberg Weiss kickbacks constituted a victimless crime because the payments came out of legal fees awarded to the firm and named plaintiffs had incentive to maximize class recoveries.
Examining a database of 730 Milberg Weiss class action settlements and legal fee awards, Perino compared those that were cited in the indictments against the firm and its partners and those that were not. He found the indictment cases on average actually settled for slightly less than the non-indictment cases, suggesting the kickback incentives did not improve recoveries.
On the other hand, Perino found that the legal fees requested and awarded in the indictment cases were significantly higher than those in the non-indictment cases, and also higher than those in cases handled by firms other than Milberg Weiss.
According to the report, the findings support the notion that class members were hurt by the kickbacks, as they "appear to have received a lower proportion of the settlement proceeds than class members in otherwise substantially similar non-indictment cases."
Federal prosecutors have requested a 33-month sentence for Weiss, who pleaded guilty in March. He is in turn arguing for 18 months. His sentencing is scheduled for June 2.

1 comment:

Anonymous said...

There is no merit whatsoever to the Perino article, sponsored by the American Enterprise Institute (AEI), which is a anti-regulation advocate serving special business interests.

If there was any merit to Perino's article, wouldn't the Wall Street Journal have at least looked at it and given it to a real statistician for a critique? The WSJ is certainly no friend of the plaintiffs bar in securities fraud cases.

Perhaps the WSJ did, and that's why the WSJ isn't saying anything about Perino's article. "If you can't say anything good, don't say anything at all"--as the saying goes.

In any case, Perino uses flawed statisitics and tries to artificially boost everything in his favor to try to claim that his results are statistically significant. In spite of his flawed and biased approach, he STILL ended up with results that even he admits were statistically insignificant--meaning he couldn't prove what he set out to try to prove.

So how did Perino get around this major problem? He relied on a technique known as bootstrapping. Bootstrapping might normally be considered when estimating and testing population parameters in situations where the sampling distribution is unknown. In Perino’s study, the sampling distribution is not unknown, and, in fact, Perino worked with an extensive database in which all the case details were known and thoroughly investigated. Neither the sampling distribution was unknown, nor were any assumptions violated. So bootstrapping shouldn't could not legitimately be used in the first place.

So what is bootstrapping?Bootstrapping is a method that involves repeatedly taking random samples of size n (with replacement) from the original sample and then calculating the value of the point estimate. To illustrate how bootstrapping would work, suppose we had a sample of 4 data points: 1, 3, 5, and 9, and we estimate the median from this sample as 4. Applying bootstrapping, we repeatedly take samples with replacement from the 4 data points and get the following, using 10 cycles:

1. Median (3,5,1,1) = 2
2. Median (3,9,9,1) = 6
3. Median (5,9,9,9) = 9
4. Median (5,1,9,5) = 5
5. Median (4,9,9,5) = 7
6. Median (1,3,9,5) = 4
7. Median (3,3,9,3) = 3
8. Median (9,5,3,3) = 4
9. Median (1,5,3,1) = 2
10. Median (9,3,3,9) = 6

Using bootstrapping, we have a pseudo-sample of 10 re-estimated medians (i.e., 2, 6, 9, 5, 7, 4, 3, 4, 2, and 6). The behavior of this pseudo-sample mimics the behavior of the original median of 4, as the median of the pseudo-sample is also 4. Then, the standard deviation of these 10 values is computed, 2.3, which is an estimate of the variability of the median.

This illustration uses 10 pseudo-samples drawn from the original sample of 4 values. In practice, bootstrapping normally entails a very large number of samples, such as 10,000 and is performed using a computer program.

I'm not saying that bootstrapping doesn't have its place in statistics. But I am saying that Perino hasn't established any possible foundation for using it. In fact, no statistician with any integrity could conceivably support Perino's use of bootstrapping in his paper.

So, given these facts, the "study" is merely unsubstantiated propaganda against the Milberg firm. The Perino paper has already been forgotten, as it should be, as it doesn't even qualify as junk science.