There is often a lot of discussion (often contentious) on Seeking Alpha about survivorship bias, particularly when it comes to investors who favor a DGI approach selecting companies from the list of Dividend Aristocrats or Dave Fish's CCC list. In particular, there was a lot of discussion in Chuck Carnevale's recent article entitled "Retirees Please Don't Index, You Deserve Better Than Average."
The basis of contention is that selecting potential stocks from a current list of Aristocrats with a 25-year performance record and then backtesting performance significantly overestimates the performance of their basket of stocks going forward a because their list excludes stocks that 1) did not maintain aristocrat status, 2) did not survive the period, or 3) were acquired. There certainly may be some bias imposed by selecting stocks in this manner; on the other hand, it can be argued that it's still a valid way to screen for potential candidates because a record of increasing dividends demonstrates a track record of steady performance and the companies' commitment to shareholders.
This article makes an attempt to evaluate the 1997 Dividend Aristocrats to see how much of a difference in performance would result by removing as much of the survivorship bias as possible. This study is certainly not pristine: First, I was unable to find a definitive list of aristocrats from 1997, and attempted to reconstruct it from a dividend investing book copyrighted in 1997. Second, it was difficult if not impossible to thoroughly evaluate each and every company from the reconstructed list due to lack of readily available data. So, I must add a disclaimer: I have attempted to present data as accurately as feasible given availability of data, however the results are not 100% accurate. Companies that were excluded were not "cherry picked" in order to skew the results.
After searching for a 1997 list on the internet and coming up empty, I went to a book published in 1997 entitled "The Dividend Rich Investor" By Joseph Tigue and Joseph Lisanti. An Appendix to the book provides a list of the Dividend Aristocrats at that time. Due to the time lag from manuscript to publish in the 1990s, I suspect the Aristocrat list provided in the book would have actually been from 1995 or 1996. Additionally, the authors stated: "This is not a complete list, just a representative sample." However, I went with the list published so as not to impose any additional bias or cherry pick additional companies that were not shown. The following table shows the list published in the book.
If one were to pick only aristocrats that still exist, the list would be whittled down to the table below. As shown, the firms that are still aristocrats outperformed the S&P average of 7.60% by close to a percentage point since 1997.
Refining the List
In conducting analysis for the larger group, it was necessary to remove several companies from the list, for various reasons as described below.
- Alco Standard: In January 1997 Alco Standard changed its name to IKON Office Solutions, Inc. (IKN). It is unknown if IKON was considered an Aristocrat following its name change in 1997, and I was unable to find reasonable price data, however stock prices surged when IKN was purchased by Ricoh in 2008. The Ricoh buyout paid IKON Shareholders a 33% premium over the 60-day moving average. Due to insufficient data, and the notion that shareholders would have cashed out at the time of the buyout, I did not include this company in my analysis.
- American Business Products: Bought out by Mail-Well in 2000, which changed its name to Cenveo in 2003. I didn't include this company in the analysis because the acquisition resulted in conversion to shares of a non-dividend paying stock, which a DG investor would have sold. Since it was early in the timeframe for analysis, it is assumed that the three years of performance prior to sale would have minimal effect on the results.
- CCB Financial: It appears that CCBF merged with other bank(s) before 1997. As a result, it would not have likely been on the list of aristocrats in 1997. Data on the company and mergers that took place are also sketchy.
- Hartford Steam Boiler Insurance: This company became privately held. I was unable to determine when and data about the company are very sketchy so I had to remove it from analysis.
- Mark Twain Bancshares: This company would not have been an aristocrat in 1997 as it was acquired by Mercantile Bancorporation in 1996, which was later acquired by Firstar in 1999. (As stated previously, the list published in the 1997 copyrighted book was likely 1 or 2 years old.)
- Ohio Casualty: The company was bought by Liberty Mutual in 2007; which went private after a merger with Safeco in 2008. The company had to be dropped from analysis due to the lack of data available. Additionally, it was privately held for half of the timeframe, so shareholders would have received a lump payment for their shares when that happened.
- Providian: During the period of time from 1994 to 1997, Providian was completely changing its business from a company with its core roots in the insurance industry (which it sold off in 1996), to a focus on the credit card industry. It is uncertain if it was still considered an aristocrat in 1997, but the company was drastically changing its business profile and one article I found talked about a June 1997 IPO, which makes it seem like the company was, in a sense, starting anew. Share prices rose from $10/share in June 1997 to $65 in early 1999. In 1999, the company was under investigation for deceptive marketing practices, at which the share price lost 30%. Stock prices were on a wild rollercoaster ride between 1999 and 2001, fluctuating between $30 and $60 per share. I believe dividends were cut during that timeframe, but could not verify. Eventually, the company was sold to Washington Mutual in 2005, which went bankrupt in 2008. This stock was dropped from evaluation because it is questionable whether it was considered an aristocrat in 1997. Further, it's more than likely that given the roller coaster ride in stock prices, investigation, and dividend cut, an investor would have sold his/her position by about 2000 or slightly after (and would have still had a tidy profit compared to the initial $10 price). Finally, there was insufficient data to fully analyze, unless you want to presume that a shareholder rode it out until the final bitter end when the stock was essentially worthless.
Analysis of Remaining Companies
The shortened list includes 23 companies, shown in the table below. It's interesting to note that the average annual performance of this group (8.34%) is similar to the performance of aristocrats only (8.46%), and still exceeds the S&P average of 7.60%. Even if one included Providian and used an average of 0% (assuming the stock would have been sold at break even), the average performance comes out to 7.99%, which still exceeds the S&P average.
A Few Observations and Conclusions
While this analysis doesn't completely erase "survivorship bias," I believe it reduced any inherent bias by a significant amount. In addition, I noticed a couple of interesting trends in the data.
First, the companies that were acquired (and I analyzed in the table above based on the acquiring company name) tended to underperform. Those companies, BTL, CSR, FRO, H, RBD, and SOTR have an average annualized growth rate of just 5.93%. Is their underperformance a coincidence? A number of DG investors comment that they sell shares upon announcement of an acquisition. Such a move may well have increased their performance in the long run.
Second, it's noteworthy that the performance of KO was impacted by significant overvaluation in 1997, as shown in the FASTGraph below (and Mr. Carnevale pointed out in his recent article cited above). Performance since KO finally approached its historic PE ratio in 2002 has been a much more respectable 9%, and growth since the great recession has been about 15%.
Third, it helps demonstrate the basis for another "rule" that many DG investors state that they observe, which is avoiding banks, since they seem to get themselves into some sort of crisis or quandary about once every 10 to 15 years. While most of the financial sector companies were removed from the analysis due to insufficient data, the stories presented on banks and insurance companies do indicate that financial institutions have undergone a lot of change and turmoil in the past two decades.
Finally, while the analysis above doesn't prove beyond refute that selecting stocks from a historically-based list such as the Dividend Aristocrats is "highly superior," it does seem to indicate, in this particular case at least, that the approach has a decent chance in outperforming the S&P. The actual results, as always, depend on the investor and how he or she manages their portfolio. For example, some investors may have sold shares upon announcement of an acquisition. Such a move may well have increased their performance in the long run, as the acquiring companies tended to underperform. In addition, a value-minded investor likely would not have bought KO in 1997 when it was significantly overvalued. A number of the other companies (K and PFE in particular) were also overvalued at that time.
Ultimately, performance depends on having a well thought out plan, a reasonable set of rules, and sticking to those rules. In my opinion, using the Aristocrat or CCC list as a starting point is certainly a valid approach, but the investor needs to buy when it's reasonable based on value, monitor for significant events, and adjust accordingly, just as with any other investing approach.