I read Mike Nadel's recent article, "The Strong Often Get Stronger!" with some interest. Though I took some issue with Mike's choice of timeframe in the comments, Mike rightly pointed out that this was simply the timeframe imposed upon him by the timing of his original blog post. Nevertheless, I found the main thrust of Mike's article-that stocks considered "overvalued" tend to continue increasing in price as the market rewards excellent underlying enterprises-to be somewhat disturbing. As evidence, Mike cited the experiences of a collection of stocks he refers to as the "Dividend Growth 50", which he stratified by relative valuation at the time of the group's inception. In this group, companies that were relatively "overvalued" 21 months ago had actually outperformed the companies that were relatively "undervalued"-by several percentage points! If this were true universally, then the logical takeaway would be that one should buy companies trading at premiums to their fair value estimates, and avoid the bargain-basement losers.
There is, of course, some intellectual appeal to this. We can look at the experience of dividend-paying stocks, and find that stocks that offer the highest yields tend to be the ones that cut their dividends soon thereafter (with the stock prices soon to follow). We can look to the experience of quality stocks, where we find that stocks that maintain high-quality tend to trade at premiums to the market for extended periods of time, so long as the quality of their business remains undiminished. It also seems reasonable to assume that-assuming the market is truly efficient-companies whose stock prices have fallen may actually be facing significant long-term headwinds and are therefore best avoided.
This last point, of course, this precludes the possibility of the stock market being irrational, which, of course, it often is in the short-term. One of everybody's favorite Buffett (read: Graham) quotes is the famous saw about "weighing machines":
"In the short run, the market is a voting machine but in the long run, it is a weighing machine."
I hypothesized that what Mike had stumbled upon might be result of the voting machine, and not the weighing machine. Put another way, I suspect that selection bias or an insufficiently long timeframe may have played a central role in his results, and that if similar results could be derived using a more robust data set with a longer timeframe, then the idea would have more merit.
To begin with, I started by selecting a group of stocks considered by Morningstar to both have "wide moats" and be undervalued in late 2010. The timeframe was chosen to have a reasonable amount of time to work with (about 5 years, or more than double the amount in the Mike's article), and the focus on "wide moat" stocks was to ensure that the companies at least had been somewhat vetted by Morningstar's analysts to be quality firms. (That not all stocks on the list are currently considered wide moat stocks speaks to the difficulty of this enterprise.) I collated about 75 stock names from a Morningstar Advisor October/November Edition (flip to the "Undervalued with Wide Moats" section).
From these names, I downloaded Zacks consensus analyst price histories from Quandl.com for each security, going back to 2010 all the way to present day. I also downloaded weekly closing price history for each stock. Due to some errors in the data set (either price history or consensus history, usually the latter), 12 stocks were excluded from inclusion, leaving 63.
I then calculated a quarterly price-to-fair value ratio for each of these remaining stocks for each of the quarters for which there was data. To wit, here is the quarterly price versus consensus fair value history for 3M (NYSE:MMM) during this period:
Figure 1: 3M Price, Fair Value, and P/FV ratios. Note that this doesn't catch the last leg up in price these past few months.
And similarly, for General Dynamics (NYSE:GD):
Figure 2: General Dynamics Price, Fair value, and P/FV ratios. All data from Zacks.
From the aggregate data, I derived the quarterly time period for each stock where the stock was most overpriced (highest P/FV ratio) and most underpriced (lowest P/FV ratio). Finally, for each quarter, I calculated the price compounded annual growth rate ("CAGR") from that point in time to the present.
The data show a strong tendency for stocks to underperform when trading above analyst consensus fair value estimate, and for them to overperform when trading below it. The median peak price-to-fair value ratio was 1.047, with a standard deviation of 0.073, and the median trough price-to-fair value ratio was 0.746, with a standard deviation of 0.079. Median CAGR from point of peak price-to-fair value was 2.45%, and median CAGR from trough peak-to-fair value was 13.48%. Median CAGR for the entire time period was 5.41%. Relative to consensus estimates, these stocks appeared most expensive ; stocks appeared cheapest in this time period generally in late 2011 to early 2012.
Figure 3: CAGRs tended to be greater when purchased near trough P/FV ratios. All data from Zacks.
Things didn't change very much if one merely attempted to time purchases either when stocks were undervalued (any P/FV < 1) or overvalued (any P/FV > 1), though the differences were less dramatic (CAGR 6.47% for undervalued, 3.25% for overvalued):
Figure 4: A similar picture when the purchase data set is expanded to include all instances and divided into discount versus premium groupings.
Interestingly, analyst consensus fair value estimates seemed to increase at similar rates from both price-to-fair value peaks and troughs, running at 9.31% and 8.04%, respectively, with median FV growth for the whole period running at about a 7.5% CAGR (just lagging the group's overall 7.91% price CAGR for the period).
Figure 5: In general, prices tend to track along analyst fair value estimates, though discounts have varied over time.
Finally, there appeared to be a tendency for forward growth rates to converge upon relative discounts/premiums to fair value. Here, for example, is a comparison of price CAGR versus P/FV (discount vs. premium, with premiums being displayed in the negative) for 3M:
Figure 6: 3M's price CAGR largely tracks changes to the discount/premium to fair value, though it has recently come somewhat uncoupled from this relationship.
This pattern re-emerges frequently. As another example, here's Pepsi (NYSE:PEP):
Figure 7: Pepsi's growth rates also closely hew to changes in relative discount / premiums to fair value.
In both cases, it's possible for prices to run away from fair value for a while, but in general, the deeper the discount, the greater the future return-and vice versa. This pattern is seen when we look at the broader group:
Figure 8: While a period of excess performance can last for a time, declines in discounts generally yield a decline in growth rates, and vice versa.
The relationship isn't perfect, but it does appear to be there. Certainly, there is not a suggestion that within this group of stocks that premiums resulted in greater gains than discounts.
What about the possibility that this list was predisposed to pick losers? After all, the list started out as a list of "undervalued" stocks; perhaps it is possible that the results are being swayed by the bulk of less desirable, lower quality stocks that were included in the beginning (despite Morningstar's contention that these were-at the time-stocks with "wide moats").
Limiting the list to the top 25% performers-with a median CAGR over the time period of about 18.2%-- yields largely the same curve:
Figure 9: a similar pattern appears even amongst the highest-performing stocks-though again a spike is seen in these names earlier this year that seems to decouple the relationship.
The drop-off in growth rates is perhaps less dramatic starting in 2011, but-excepting the spike in the first part of this year-plays catch up starting around 2014 or so. In general, the principle that smaller discounts to fair value (or even premiums to fair value) suggest lower future growth rates remains strongly intact.
Conclusions and Caveats:
Some readers may argue that perhaps I've missed the point of Mike's article. His contention-and Chowder's-seems to be that high-performing stocks continue to perform well and may continue to trade at significant premiums to fair value estimates for some time. The implication seems to be that by waiting for a discount on the shares, investors impose upon themselves an opportunity cost in lost gains. This behavior, however, is simply the opposite of rational: there is no guarantee that the long-term growth profile of a stock will continue to improve as the price increases, and in fact there is strong evidence to suggest that, over longer periods of time, compounded growth rates for purchases done at significant premiums to fair value are inferior to those made when shares are trading at a discount. Far better, I think to purchase with a margin of safety. And despite this article being based upon a list of initially "undervalued" stocks, this pattern seems to hold true even for top performers. Finally, this analysis ignores other methods of valuation, though it is hoped that these methods are captured in analysts' estimates.
This isn't meant to be a knock on Mike or his article. He raises an interesting question and his data are indeed suggestive. But I would argue that a more comprehensive scope finds that his conclusions are not as concrete as his article suggests: that buying at a discount still very much matters, and that investors should resist the urge to plunge into the market at sky-high prices based upon the notion that prices will only continue to get higher.
Disclosure: I am/we are long GD.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.