My Investment Model Recommends THAT? What the . . .

In my recent article presenting a micro-nano-cap screen, I warned readers that “(t)he key to screening for small stocks is to not get crazy with guru-like ideas about finding great companies. Small companies often aren’t all that “good,” at least not in terms of fundamentals compared to their larger  brethren,” and suggested modeling with an “aura of ripped jeans and dirty t-shirts, as opposed to the $1,500 suit world of large cap screening.” I was reminded how true that admonition was when I started looking at stocks uncovered by the screen in order to write about some in more detail. And apparently, other Portfolio123 users have been wrestling with this issue too, judging by a thread  saw in our forum entitled “Your model returns stocks you don’t like.” So before discussing any stocks from the screen, it seems appropriate to further explore this important issue.

dumpster fire

The Dream

Surely you’ve heard this before: Buy shares of good companies at good prices. Invest in companies with sustainable competitive advantages, or moats. And of course there are such buzz words as excellence, innovation, leadership. We want to get behind great managers (and even pretend we can evaluate them to determine that they are, indeed, great.) And then, there are today’s media-driven tele-gurus who make it seem like a an unpardonable betrayal if a company comes up snort of guidance or, heaven forbid, has a down quarter or year: We want earnings momentum, sales momentum, margin expansion and while we’re at it, debt reduction, increased dividends and share buybacks and better still, all at the same time.

No doubt about it: We investors are high maintenance people (or machines). 

The Reality

Good luck finding the perfectly investable company. It doesn’t exist, and if it did, the stock price would probably be insanely overvalued because everyone would want in. Even the bluest of the blue chip leaders aren’t perfect. That’s why all of the models I use to identify better companies are framed in relative terms (which means I could be getting really good companies, or the those that are least lousy).

Most companies are not great and smaller companies, those that have disadvantages in terms of scale and operational diversification, tend to be less great than blue chips; see my June 18th post. So idealistic rhetoric aside, we have to understand that we’re going to be investing in companies with baggage. And the smaller we go, the more baggage we have to prepare to accept.

The good news is that this is OK. We can accept any level of company or business quality if the share valuation is aligned in such a way as to make it work. In other words, if a company grades at 5 units of dismal, we can load up on its stock if it’s priced consistent with an assumption of 12 units of dismal. Conversely, a company with 84 units of excellence is a Sell or a Short if the stock is priced under the assumption of 125 units of excellence. 

Putting it all together, we’re not looking for good companies nor are we looking for bad companies. We’re looking for inefficiencies, mismatches between what the company is and what the stock price assumes it is. We can make money on a company that moves from pathetic to meh just as well as we can when good companies march further ahead.

Liking Stocks Uncovered By A Model

Back to modeling: What should you do if your model produces dumpster fires rather than the corporate gems you imagined it would uncover?

Manage Expectations

Step one is to make sure your idea of what you want is aligned with reality (as discussed above) and with what your model is really seeking. The former is a trap that can snare fundamental investors who approach the process with unrealistic expectations of corporate wonderful-ness. The latter is a trap that can ensnare quants and/or technical-analysis fans who study and test systems looking for what works without always thinking about why something might or might not work (if a model is chock full of trend-based or momentum factors, you really can’t complain if the stocks it shows you are overpriced; this sort of investing or trading requires a willingness to buy overpriced stocks in the hope of them becoming more overpriced). 

With this in mind, let’s revisit the three key fundamental screening tests I used:

    • The stock must rank above 25 in the Portfolio123 Basic: Quality ranking system
    • The stock must rank above 50 in the Portfolio123 Basic: Value ranking system
    • The stock must rank above 65 in the Portfolio123 Basic: Growth ranking system

That’s what we’re looking for; decent but not necessarily excellent in terms of historic EPS Growth; a collection of valuation ratios that on the whole is not in the worst half of the investment universe, and as to quality ratios, we don’t need good; we don’t even need mediocre nor do we even need not good. When it comes to quality, we’ll take it as long as it’s not a irredeemable garbage.

We’re depending on very small size to carry us a very long way. We’re not ignoring the fundamentals but are instead consigning them to a supporting role.

Is that really a smart thing to do, to be willing to accept below-average situations just because the companies are small so long as we’re not dealing with complete basket cases? In the post about the screen, I showed a lot of test data suggesting an affirmative answer. Today, I’ll take it a step further and compare the strategy I presented with a couple of alternatives that had the potential to come up with “better” micro-nano situations.

Table 1: Testing Alternative Versions of Nano/Micro Cap Model – 10 Yrs.

Micro-Nano Benchmark
Original Model Alt. 1 Alt. 2 IWC
Minimum Rank
Thresholds in 
Screening Rules
   Quality 25 80 80
   Value 50 50 80
   Growth 65 65 80
Annual % Return 25.09% 10.69% -1.90% 9.68%
Standard Dev. 27.52% 30.13% 34.74% 21.67%
Max Drawdown -57.70% -59.57% -74.03% -55.81%
Beta 1.02 1.00 0.92
Annual Alpha % 15.44% 2.95% -5.95%
Avg. 4 wk periods
   All 1.78% 1.20% 0.89% 0.90%
   Up markets 5.71% 4.44% 4.48% 4.71%
   Down Markets -4.18% -3.73% -4.55% -4.87%

I said money could be made by investing in not-so-good situations and I wasn’t kidding. The strategy was to make size the primary theme and as a supporting theme, to refrain from seeking excellence (or anything within hailing distance of it) and focus instead on weeding out the dregs of the dregs. That was the idea. The tests support if. And that must be remembered when we look at individual names that turn up in the list. Whether there are ways to bring Quality and/or Value to the micro-nano world with a different kind of model is a topic for another day; we just know it’s not a major part of what we’re doing right now.

Understand How Data Relates to the Real World

Step two is to understand the myriad of oddball ways a data series or factor can behave. This has nothing to do with accuracy. It has everything to do with the intense and at times seemingly infinite variability of all things human and of all things measured in the realm of human activities, including, but definitely not limited to, fundamental data. Here are some simple examples of how accurate numbers can give us distorted information:

    • A write-off that drives earnings to accurately-reported but analytically useless low levels and also causes P/E ratios to look much higher than a correct valuation analysis would assume;
    • Earnings that are boosted sharply by gains on sales of assets that cause growth rates to compute well above analytically useful levels and which also make P/E ratios look appetizingly low, even though from an analytical standpoint, that’s not so;
    • Acquisitions or divestitures that can sharply increase or decrease sales growth rates to levels that have no relationship to what investors should be assuming going forward.

Suggestions that such items, which are entirely legal and proper (and mandatory) under accounting rules, can lead investors astray is not something I, a 21st Century analyst made up due to inspiration from the newfangled metrics invented by purveyors of overhyped garbage of the sort we saw in the late 1990s and early 2000s. What I’m saying has a much better pedigree than that. It’s all over Part V (Chapters 31-41) of the Graham & Dodd Security Analysis classic. Rookie analysts have been drilled with this for generations and (hopefully) still are being taught this way.

Here’s an even odder real-life example from the companies that passed the micro-nano cap screen about which I wrote. It’s Finjan Holdings (FNJN). It owns and licenses a portfolio of cyber-security patents. 

One of the requirements of the screen was a score of at least 65 in the Portfolio123 Basic: Growth ranking system. The rank as of this writing is 71.01, which is a pass. Here are the scores for each of the ranking system’s components (PYQ = Change in Current Quarter from Prior Year Quarter; TTM = Trailing 12 Months):

    • EPS Growth PYQ: 63.99
    • EPS Growth TTM: 83.00
    • EPS Growth 5 Yrs: 35.61
    • Sales Growth PYQ: 96.87
    • Sales Growth TTM: 96.19
    • Sales Growth 5 Yrs: 21.16
    • EPS Acceleration Recent: 38.37
    • EPS Acceleration Longer Term: 83.30
    • Sales Acceleration Recent: 68.55
    • Sales Acceleration Longer Term: 23.34

Given those individual factor scores and the weighting (which you can see here), there’s no reason to doubt that in terms of historic Sales and EPS growth, FNJN is exactly what the ranking system suggests it is; not great across the board but on the whole, good enough to be classified in the top 35% of the investment universe in terms of growth.

Now, let’s take a look at the quarterly Sales (in $ mill) and EPS figures for the past few years. 

Table 2

FY Qtr 1 Qtr 2 Qtr 3 Qtr 4 Total
Sales
12/14 0 0 5 0 5
12/15 0 1 0 4 5
12/16 2 7 1 8 18
12/17 25 2 0 23 50
12/18 65
EPS
12/14 -$0.08 -$0.15 $0.03 -$0.18 -$0.38
12/15 -$0.19 -$0.12 -$0.21 -$0.05 -$0.57
12/16 -$0.05 -$0.20 -$0.20 $0.17 -$0.28
12/17 $0.69 -$0.29 -$0.16 $0.41 $0.65
12/18 $0.93 $0.93

Data from Compustat via Portfolio123

Licensing intellectual property is a nice business, but unless a company is incredibly large with a large IP portfolio and/or an incredibly large customer base, it’s apt to be spectacularly variable from one quarter to the next as deals are inked, payments are made, etc. The way Finjan’s numbers worked out happened to be such as to produce a score above 65 on the Growth ranking system. Going forward, however, the pattern can just as easily go the other way. Figure 1 shows the history of FNJN’s Growth rank.

Figure 1

fnjn growth rank

FNJN may be a perfectly good investment. Ditto the many tiny biotech companies that may have Sales, EPS and Growth rank patterns that look just like FNJN. Personally, though, this sort of investing isn’t my thing.

Could I have screened more carefully to have eliminated oddball patterns such as this? Possibly. But the potential variations are infinite and a more extensive set of screening criteria designed to eliminate FNJN may allow in, or open the door to, who knows what other potential oddballs. I do try to reduce the likelihood of craziness by using many factors (statisticians hate doing this, using lots of factors likely to be highly correlated with one another, but a portfolio of factors expressing a single idea, as opposed to a single factor, mitigates the risk of extremes much the way a portfolio of stocks mitigates that sort of risk). But try as we might, all we can do is reduce the probability of oddballs. It’s unrealistic to assume we can achieve perfection in their elimination.

Evaluating Stocks In Context

When I look at an individual company/stock called to my attention from this model, I’m going to think about the benefits of small size, which are (i) the greater potential for the stock to be mis-perceived by the market given the lesser degree of attention these situations typically attract, (ii) the risks associated with smallness — fixed-cost burdens and lesser business diversification — to see if the risks are more than I can tolerate, (iii) the potential for small firms to grow more quickly given the smaller bases from which they start, (iv) and the potential for these firms to fix whatever ails them given the lesser internal bureaucracies that typify these situations, balanced against potentially less deep talent pools. I know going in there will be baggage, that the gurus are likely to be thumbs down and that earnings may, when I look, be trending down. What I’m looking for is the potential for the market to later price the stocks under expectations of movement in the right direction.

In terms of value, I know from the screening rule I used that whatever shortcomings the stocks may have, they’ll at least rank above the lower half of the universe. I also always know in the back of my mind that Value is not everything. It’s not correct to say P = V (Price = Value). We must also be aware of Noise; P = V + N.

To put the role of Noise in perspective, I’ll quantify it by first, assuming that Value is NOPAT (Net Operating Profit After Tax) divided by the cost of capital. (I refer to this as conservative standstill value since it allows for no growth — analogous to the way a fixed-income security works and where value is interest divided by market rate — and consigns growth expectations to the realm of Noise.) The percent of market cap that is not attributed to this measure of value is, what I call %N, the percent of market cap (or stock price) attributable to Noise. Noise is usually positive, can be negative if the market is not even taking full account of standstill value (whether because investors expect NOPAT to decline going forward or the market is just-plain not getting the story).

Here is the percentage of Noise for each of the 20 stocks uncovered by the micro-nano screen:

ARTX, 74%

ASYS, -24%

AUDC, 67%

AXTI, 66%

CVGI, -27%

DHX, 15%

DX, -22%

EVC, -335%

FNJN, -256%

JILL, -28%

NOA, 40%

NVFY, -40%

NWY, 68%

ORN, 89%

PFMY, 134%

PRTS, 39%

PTN, -1%

RRD, -251%

SND, 49%

SORL, -158%

Right away we see some potentially interesting things. We already covered the basis for FNJN’s -256% noise allocation. Given the up and down unpredictable nature of the company’s licensing business, it’s obvious the Street expects down comparisons as future quarters compared to those with big receipts; standstill value is not expected to be sustained. I haven’t looked yet at EVC (%N is -335%) in depth but my ten-second glance at this media firm that’s big in the Latino market showed me it didn’t get to broadcast the current World Cup; further inquiry is needed, though, before I can come to a conclusion about whether the stock is now aligned with the business.

This is a running start on how to evaluate stocks you see in a screen, how you need to set a framework based on reality and the expectations and goals you put into the screen.

But . . .

Even After All This, I Still Despise That Stock With Every Fibre Of My Being. Now What?

Here’s where you have to make a choice. Are you going to follow the model as a whole, or are you going to edit it (look at the stocks and manually eliminate those you hate), or are you going to use it as idea generator assuming all your ultimate decisions will be based on judgment.

There is no right answer for all investors at all times. This must be a personal decision.

Speaking for myself, I started out using screens entirely as idea generators. I never expected to buy everything on the list. That changed back in my Multex-Reuters days when manual backtesting (ugh!) suggested the possibility of doing something more. But I was reluctant to go whole hog. I shifted from idea generator to use of the whole screen subject to personal editing of results.

But over time, I found, much to the wounding of my ego, that my edits did not result in improved performance. In other words, my ability to outsmart my screens turned out to be less than I had envisioned.

When I got to Portfolio123 with its seriously sophisticated model-building capabilities, I moved into fully-automated investing in which I bought and sold what and when my models told me to buy and sell and have been happy with the results. When, a few years ago, I had been doing a low-priced stocks newsletter with Forbes, I felt compelled to go back to my model-and-edit mode, but again, as i turned out, my editing did not add to performance. So now, for the serious portion of my equity investing, I’m all in on the models. I don’t even look to see if I like or dislike individual companies. I download tickers from Portfolio123 into Excel and upload into my account at FolioInvesting.com. (Anything else I do with stocks is consigned to the realm of fun money.)

So if I were running this screen as an automated strategy, I’d have bought Finjan without even knowing what the company is about and without realizing it’s the sort of stock I’d wouldn’t buy on my own except for fun (assuming I researched the heck out of the company and like what I see, which I may or may not wind up doing).

Is this dangerous? Everything in the stock market can be dangerous. The key to managing danger is to do things thoughtfully. I always recognize dogs will slip into my models and design them with that in mind and set the number of positions such as to mitigate their impact. That’s why I usually like to have at least 10 positions, and typically more. I want the portfolio as a whole to average out to the qualities I envision for the model and as with a stock portfolio, I know I can handle clunkers as long as they aren’t so numerous or impactful as to mess up the average. And if they do mess up they average, that’s an indication I need to re-think the model.

Anyway, that’s how I do things. You can choose differently. But whatever you choose, choose knowingly. 

 

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s