Are the Basic Principles Still Relevant?

by: Marc Gerstein

Why did Lehman (LEH) fail? Why are we having a financial crisis? Is it overuse of exotic derivatives, structured products, and so forth, as many commentators have been saying? I doubt it. Exotics or not, if you lend a ton of money to people who can't pay it back (or take over bad loans originated by others), you're going to suffer. That is so in 2008. It was so in 1998. It was so in 1558. It will probably be so in 2528. It's just that simple. All else is noise. Using those thoughts as a jumping-off point, I decided to take a closer look at what is and isn't working in stock selection to see if I could get a sense of whether we're in some sort of brave new world, or whether deep down, the situation resembles the lending decision in that the basics still govern.

New or old world?

Describing the financial meltdown as a matter of imprudent lending is not 20-20 hindsight. Just consider one simple factor, the existence of the no-income-verification loan. What does that tell us? How can any rational person see that as anything other than financial suicide? The mystery is why that and other comparably shoddy practices persisted for so long. But that's a topic for another day. At this time, I'm more interested in noting how this seemingly modern crisis is actually quite classic and in considering whether basic common sense principles remain relevant elsewhere, specifically, in stock selection.

Challenging times for "stock pickers"

Anybody who has tried to pick stocks lately using anything other than tea leaves knows that times have been tough, not just because prices are falling but because so many once-strong stock-selection factors and models seem to have lost their effectiveness. Has stock selection entered a new era? Or, should we recognize that even the best of strategies go cold every now and then, and patiently stand pat?

As a matter of principle, I hate to ever tell anyone to stand pat, because often, that's interpreted as a prescription for complacency. Although it's common nowadays to poke fun at the phrase "it's different this time," the world does, in fact, evolve. So however much continuity we think we enjoy, when we look closely, we see that each period really does have a unique flavor. But there's a difference between fine tuning to stay fresh, versus whipsawing one's strategy trying to dogmatically chase the last thing that was seen to have worked in the latest period.

Assessing the basic strategies

I ran some tests using a kitchen-sink multi-factor ranking system I created on based on more than 50 factors across four equally-weighted styles: growth, value, quality and sentiment. I don't suggest this is the ultimate in stock-picking. There's much room for debate regarding use of so many factors. And equal weighing was used in all sub-categories in order to maintain a deliberately naive flavor. But it can be useful in helping us get a general sense of what kinds of factors are and are not associated with relatively strong stock performance over different periods.

Here's the result of a back test of the system covering a long time frame, from 3/31/01 through 9/15/08.

Click to enlarge

Figure 1: The full ranking system, 3/15/01 - 9/15/08

The leftmost red bar depicts the annualized performance of the S&P 500. The next dark blue bar depicts the performance of the stocks ranking in the bottom five percent. The rightmost bar shows the annualized performance of the best five percent etc.

It's interesting to see that such a simple system performed pretty well over the long time period. Perhaps this explains why so many investors who used rules-driven strategies did well. It seems that the approach in general is what worked, and that one need not have used any particular "magic formula."

Figures 2 and 3 provide close-ups of more recent periods.

Click to enlarge

Figure 2: The full ranking system, 9/15/07 - 8/15/08

Click to enlarge

Figure 3: The full ranking system, 8/15/08 - 9/15/08

The model performed poorly in the latest month, but interestingly, for much of the past year, it performed reasonably well. The latter seems a bit surprising since stock-picking woes were prevalent well before August 15th. It seems likely that something else is beneath the surface.

The growth component

The factors used in this section of the model are:

  • Trailing 12 month EPS growth
    • Relative to all stocks and relative to industry peers
  • Year-to-year EPS growth in the latest quarter
    • Relative to all stocks and relative to industry peers
  • Five year EPS growth
    • Relative to all stocks and relative to industry peers
  • EPS acceleration (the latest quarterly comparison compared to the trailing 12 months, and the trailing 12 months compared to the latest five year growth rate) relative to all stocks

Figure 4 shows the long-term back test result for the Growth component.

Click to enlarge

Figure 4: Growth, 3/31/01 - 9/15/08

It's not perfect, and not quite as good as the results we saw for the overall model, but on the whole, it's respectable.

Figures 5 and 6 paint a different picture of the more recent past.

Click to enlarge

Figure 5: Growth, 9/15/07 - 8/15/08

Click to enlarge

Figure 6: Growth, 8/15/08 - 9/15/08

The latest month was a mess, with the growth factors listed above having been essentially useless in pointing the way toward potentially better-performing stocks. Figure 5, which covers the preceding 11 months, was also generally poor, although the model was at least able to keep us out of the very worst performers. Above the negative extreme, though, it offered little assistance.

So what might we expect going forward? Have basic growth factors such as those enumerated above really gone by the wayside? Or might we treat the recent past an aberration?

Actually, what we've been seeing lately, is the sort of thing we see over and over again when business trends turn. Data-driven models necessarily use historical numbers. Stock prices reflect future expectations. If transitions are expected, then we would see good stock performance associated with bad historical numbers and vice versa.

Lately, We've been coming to grips with a meltdown in finance, and as expectations manifest themselves on Wall Street, we often see weakness in shares of companies with perfectly fine reported growth rates as investors sell based in anticipation of problems that are yet to be published in the financial statements. We also have a lot of bouncing around in areas like energy, where earnings prospects jump back and forth rapidly. Ditto the many groups influenced by energy and resources in general.

Growth is a widely-used investment theme so the experience depicted in Figures 5 and 6 were probably present in many other models constructed using this general approach and may explain why so many individual models failed to perform as well since 9/15/07 as the Figure 2 might suggest.

Does this indicate that these models should be abandoned? Market veterans have seen these transition-induced performance falloffs before and we'll continue to see them in the future. Ultimately, though, the question you'd need to ask yourself is this: Will rapidly changing expectations such as we've been experiencing lately become the new long-term norm, or will these transitions eventually coalesce into some sort of new longer-lasting trend?

Historical precedent favors the latter. That doesn't prove it will happen again. But one who wants to abandon growth models in general would be taking a very radical position, to wit, that trends are dead and that rapid transitions will follow quickly one after the other indefinitely. Speaking for myself, I believe history will, for the gazillionth time, repeat and that we will at some point settle into a new trend. It may not be as good as other trends we've seen in the past. But Figure 2 and others below show that these models are able to help us cope with weak periods too. The key is that we do return to some level of normality. I think that is within reach.

The Value Component

This is based on:

  • P/E computed with reference to trailing 12 month EPS
  • PEG ratio computed with reference to estimated EPS and the projected growth rate
  • Price/sales
  • Price/book ratios
  • Enterprise value to trailing 12 month EBITDA
  • Enterprise value to estimated EPS

As we see in Figure 7, over the long haul, this component performed reasonably well.

Click to enlarge

Figure 7: Value, 3/31/01 - 9/15/08

Figures 9 and 10 show that value modeling fared poorly for most of the past year, but interestingly, worked pretty well in just the most recent month.

Click to enlarge

Figure 8: Value, 9/15/07 - 8/15/08

Click to enlarge

Figure 9: Value, 8/15/08 - 9/15/08

And as with growth, we can recognize what's going on. Finance, a sector with traditionally low valuation metrics, got pounded as the crisis unfolded. Hence stocks with low valuations performed poorly. And with commodity prices being volatile, so too have been commodity-related company earnings and share valuations. But as I described last week, portions of the banking group rallied nicely. That, and jumps in beaten-down energy-sensitive groups, probably had much to do with the strength we saw from 8/15/08 - 9/5/08 in stocks which had lower valuations.

Again, the recent poor model performance seems to relate to specific transitions occurring at current points in time. As to whether value is permanently dead, we return to the same core question we asked in connection with growth: Will rapidly changing expectations such as we've been experiencing lately become the new long-term norm, or will these transitions eventually coalesce into some sort of new longer-lasting trend?

The quality component

These are factors that are important to the Warren Buffetts of the world, but often don't get much play elsewhere in the investment community. All these items are ranked relative to all stocks and relative to industry peers:

  • Trailing 12 month and five-year return on assets
  • Trailing 12 month and five-year return on investment
  • Trailing 12 month and five-year return on equity
  • Trailing 12 month and five-year operating margin
  • Trailing 12 month asset turnover
  • Trailing 12 month receivables turnover
  • Trailing 12 month inventory turnover
  • Latest current ratio
  • Latest quick ratio
  • Trailing 12 month interest coverage
  • Latest total debt ratio
  • Latest Altman Z-score.

Figure 10 shows that over the long haul, the market hasn't rewarded companies that excel here, but that it had been quite willing to punish companies with low ranks.

Click to enlarge

Figure 10: Quality, 3/31/01 - 9/15/08

For the difficult 9/15/07 - 8/15/08 period, Figure 11 shows that this component of the model performed better.

Click to enlarge

Figure 11: Quality, 9/15/07 - 8/15/08

What we see here is a more-or-less typical situation. When times are tough, we usually experience a flight to quality, with stocks ranking highly in factors such as those used here tending to be favored.

One might expect that to have persisted into the most recent month. But Figure 12 shows it didn't work out exactly in that manner.

Click to enlarge

Figure 12: Quality, 8/15/08- 9/15/08

The market continued to punish the worst of the worst. Then, the next 15 groups performed generally, if inexactly, in line. Interestingly, though, shares of companies in three of the top four categories faltered. Without further investigation, it's hard to say exactly what happened here. It may be similar to what we saw above: shares of financial and resource firms whose latest numbers still look good reacting to expectations of diminished prospects.

In any case, three significantly aberrant sub-groups out of twenty in a single month is not, in my opinion, enough to suggest that quality factors are no longer viable going forward. (and for what it's worth, recall that this is a somewhat naive model built for study purposes; I've seen investment-oriented quality models that, over the long haul, look better than what we saw in Figure 10.)

The Sentiment Component

This is based on the following factors:

  • Revision in the EPS estimate for the current quarter
    • Relative to industry peers and relative to the entire database
  • Revision in the EPS estimate for the current year
    • Relative to industry peers and relative to the entire database
  • Revision in the EPS estimate for the next year
    • Relative to industry peers and relative to the entire database
  • Revision in the long-term growth rate estimate
    • Relative to industry peers and relative to the entire database
  • Average analyst recommendation relative to the data universe
  • Change in average analyst recommendation relative to the data universe
  • Insider buying relative to industry peers
  • Short covering relative to industry peers

One might almost describe this as the anti-Buffett component. And contrary to what many in that camp might expect, it performed quite respectably over the long haul, as we see in Figure 13.

Click to enlarge

Figure 13: Sentiment, 3/31/01 - 9/15/08

The sub-groups don't line up quite as well in the upper echelons of the rank order (except for particular strength at the very top), but on the whole, the results suggest that avoiding the lower half was a good idea in that time period. And as with quality, I have seen less naive models where even the upper half line up better than they do here.

Figure 14 shows that the sentiment factors performed generally, if still imperfectly, well in the 9/15/07 - 8/15/08 period.

Click to enlarge

Figure 14: Sentiment, 9/15/07 - 8/15/08

Figure 15, which depicts the latest month, is very interesting.

Click to enlarge

Figure 15: Sentiment, 8/15/08 - 9/15/08

This component, which is based mainly on estimate revision, reversed badly.

I suspect this has much to do with pricing volatility in the energy-commodities-resources area. Analysts may have been slow, or at least slower than equity traders, to keep up with the rapidly changing environment.

Sentiment-based models would be diminished if, indeed, analysts have really lost it. Actually, losing it wouldn't be enough. They'd have to be so bad as to resemble the once-famous odd lot indicator (a market timing signal based on the notion that small investors are all wrong all the time). Undoubtedly, many will argue that this is the case, that analysts are the new odd-lotters. But if you're tempted to accept this, consider first the track record of whoever is making the argument and make sure they're speaking based on fact, as opposed to emotion. Many of the sell side's harshest critics had no inkling that analyst-based data could be as useful as we've seen over the longer (post Eliot Spitzer) period.

Once again, a case can be made that we're dealing with aberrations relating to specific point-in-time events, rather than a full-scale reversal or even deterioration in the viability of using data-based rules to point toward stocks likely to perform better in the near future.

Adding it all up

I recognize that the foregoing does not have the force of mathematical or philosophical proof. That sort of certainty is not possible. My goal has been to break the big picture down to more manageable parts and to spotlight the sorts of beliefs one would have to hold in order to believe we're in a new era in which basic fundamental equity evaluation principles have become irrelevant.

Obviously, if the test model were designed differently, the back test results would change. But this example seems broad enough to demonstrate that the difficulties stock pickers have been experiencing are explainable in terms of unusual period-specific events, as opposed to a substantive and lasting change in the way investors look at stocks.

So it appears to me that as with the classic principle that it's better to lend money to people who can repay, it also seems to hold that aberrations aside, we're still likely to be better off if we own reasonably valued shares of companies with demonstrated records of being able to grow earnings, with track records that look good in terms of balance sheet and profitability, and which are generally gaining stature in the eyes of the investment community.

None of this diminishes the pain being felt right now, especially since the ongoing aberrations have taken a while to unfold. But we've seen before that prudent attention to the basics has worked well over time and given the here-and-now nature of the events we've been experiencing, it seems reasonable to assume that past precedent remains relevant. Remember, the here-and-now events I speak of refers to the trendless transitions we've been seeing, not how good or bad things are. The basic analytic principles can adapt if need be to a decade of high oil and/or bad bank performance. The problems are the dramatic week-to-week or month-to-month swings, and it's these I believe will eventually pass.