Siegel Responds to Criticisms on Stock Market Data

Professor Jeremy Siegel has responded to Jason Zweig’s criticisms of the data he’s used for long-term stock performance.

The problem Zweig highlights is how few stocks comprise the study Siegel relies on. Zweig notes that Siegel ignores 97% of stocks that were trading in that time frame—and most of those stocks he uses were blue chips. All those ignored dud stocks would certainly have lowered the 10% per year number.
Zweig also finds that Siegel raised the average dividend yield from the 1802 to 1870 period from 5% to 6.4%. That’s a huge increase and it alters the long-term results very significantly.

Siegel responds by pointing to research by Bill Goetzmann and Roger Ibbotson. I happen to be familiar with that data and I still think Zweig’s larger point holds.
The research data collected by Goetzmann and Ibbotson is impressive but I see it as only a start. For example, for months between 1815 and 1834, the data set comprises only a handful of stocks. It’s often less than 20 stocks and sometimes less than 10. The dividend series begins in 1825 and some years it includes less than 20 stocks. I don’t think we should rely on research with so little data.
The problem is that the stock market wasn’t close to a market in the sense we regard it. In fact, common stocks were viewed as similar to bonds. The shares would trade around par value and each year you’d find out what the dividend was. The idea of constant capital gains is a 20th century notion.
Siegel writes:

Researchers agree that the biggest source of uncertainty in early stock data is the dividend yield, which was not always reported. As a result, G-I formed two series of dividend yields, one assuming that those stocks for which they could not find dividends had zero dividends (3.77%), and another which uses the dividend yield of those stocks for which they could find dividends (9.27%). They conclude “The true dividend return to a capital-weighted investment in all NYSE stocks is undoubtedly somewhere in between these two extremes.” My dividend yield, which the article claims is unrealistically high, is 6.4%, actually less than the midpoint of their two estimates.

Pointing out those two data sets that are so far apart is precisely the problem. Selecting a number between them doesn’t help. Sure, 5% and 6.4% are within the bookends, but the difference between the two is extreme, especially when compounded over decades. The fact is that Siegel is basing his arguments on very thin data sources.

Posted by on August 5th, 2009 at 9:57 am


The information in this blog post represents my own opinions and does not contain a recommendation for any particular security or investment. I or my affiliates may hold positions or other interests in securities mentioned in the Blog, please see my Disclaimer page for my full disclaimer.