Making Value Work

By now, many (or most or all) of you know I don’t advocate using any factor in isolation. An example I often use is Value, where I derive the ideal P/E as E / (R-G) where R is risk/quality and g is growth.

I did a quick back of the envelope empirical study to test some of this.

I started with the PRussell3000 universe.
I used a 5-years ago as of date (1/30/13)
I screened for stocks that back then Ranked above 90 for “Basic: Value”

So as of 1/30/13, all of these stocks could have been considered good value plays. And since the ranking system I used (it’s available as a pre-set on p123) is pretty generic, I figure any other ranking system would produce a list with very heavy overlap.

I then measured performance over the next 5 years; well, actually, since this is spit and chewing gum, I roughed it out by fast forwarding to today and computing close(0)/close(1200) assuming 240 trading days in a year.

After knocking out 16 stocks that didn’t have full 5-year histories, the average % return for the group was 63% and the median was 54%. But the maximum was +397% and the minimum was -91%. Remember, this range is only for stocks that were ranked above 90 for value. That’s a pretty big range and one I’m sure many have wrestled with as models were taken from backtest to live trading.

I then ordered the 90+ Value Rank 2013 stocks into five buckets based on the future 5-year returns that were actually delivered. For each bucket, I computed (as of today) the 5-year rates of sales and EPS growth tat actually occurred. Here are the results:

Bucket # Avg % Return Avg 5Y Rev Growth Rate Avg 5Y EPS Growth Rate
1 +171% +10.4% +20.0%
2 +98% +7.2% +15.4%
3 +56% +9.9% +8.3%
4 +19% +11.6% +3.1%
5 -32% +7.0% +3.0%

What does this mean for p123 users?

Obviously, this is not anything you can literally apply because I’m matching returns with KNOWN future growth rate (we see sales is meh but EPS growth is important). But it conforms the strategic roadmap I’ve been advocating. When it comes to value, don;t just look for good value. You also need clues that allow your models to make reasonable assumptions about future growth prospects. The difference between a successful and unsuccessful value strategy has little and probably nothing to do with the value rank you use. It probably depends more on how good a job you do (1) creating a pre-qualified list of companies against which you’ll apply your value ranks and/or (2) how good a combo rank you develop; one that effectively combines value with growth.

How do you model for clues to future growth? That’s the challenge. If I knew the answer, I’d be rich enough to own the world. But I suggest ways to move models in the right direction is to focus on clues provided by fundamental capacity for growth (returns on capital, etc.) and sentiment measures that let you tap into investment community expectations such as estimate revision, projections and good solid technical analysis. What about historical growth rates? I’ve had so-soi success at this, but it’s fertile ground for testing (Are growth rates persistent? If not, what might you discover and measure that might allow you to decide some growth rates are likely to be more persistent than others?)

I have just launched this DM: 40-Russell3000 Dividend-Growth Stocks: Moderate-Reverse-Cap-Weighted

Basically this model selects Value stocks which also pay a reasonable dividend. The moderate-reverse-cap weighting gives higher weights to the smaller stocks than the bigger ones. I used the Position Weight Formula in the new rebalancing module.

This is a 40-position model which produced a simulated return of about 26% with very low turnover of about 170%, without market-timing. It is interesting that a 20-position model produces similar returns with 130% turnover.

Equal weight produces much lower returns of about 17% annualized. So there is a good case made for reverse-cap-weighting.

I don’t think that anybody would want to subscribe to a 40 position model, but I thought it is a good demonstration of what can be done with the Position Weight Formula.

I agree with the premise that you can’t look at value alone. However, I’m not sure that your results show that value is the prevailing factor. In fact, it’s showing that it was growth that gave you the better returns and that there was in fact a linear response to it (i.e. stocks with the highest growth rate gave you the best returns)

What would the results show if you divided the original pool of stocks into 5 buckets according to their growth rates in 2013 and looked to see what happened to them in time?

Or what would the results show if you followed the bottom 10% of the value stocks and looked at their returns according to their future growth rates?

Edit: or further yet, if you took 10% of random stocks from the original pile and looked at the return based on their growth rate?


Thank you for your thoughts on differentiating undervalued securities from “value traps”.

I think your explanation is a good reason for incorporating “quality” factors with value. In fact, I think that this factor interaction underlies the selection of variable in Fama-French’s 5 factor model which incorporates investment aggressiveness (i.e., asset growth) with profitability (operating return on book equity). Quality factors tend not to work as well on their own as when they are paired up against factors which attempt to make sense of investor expectations. I.e, a low P/B or P/E does not portend that security is cheap, but rather than expectations are low.

And since investor beliefs can be fallible, it makes sense to try to determine cases where incorrect beliefs are more likely than not.

As per the textbook, the long term (sustainable) growth of earnings converge to reinvested return on capital (e.g., (NetIncBXor - DivPaid)/BookVal). The assumption behind this is that incremental returns on capital equal the average returns on capital. However, in the real world, we know that a) returns usually diminish with greater reinvestment, and b) incremental returns are determined by a confluence on real opportunity, luck, and skill. I think that this explains why historical earnings growth rates tend not to persist and, if anything, diminish over time.

In any case, I was hoping to get your thoughts on a means by which to differentiate growth and sustainment reinvestment. I think that this distinction could further your intent to identify factors that contribute to the persistence of future growth and profitability.

One time to consider using Basic Value by itself: when there is blood in the streets.

On a 52-week hold starting March 8, 2009, the Top 2% Basic Value stocks of the S&P 500, S&P 1500 and Russell 3000 each gained at least 450% vs 71% for SPY.

On a 52-week hold starting March 9, 2003, the Top 2% Basic Value stocks of those 3 universes gained between 111-145% vs 45% for SPY.

On a 52-week hold starting Feb 14, 2016, the Top 2% Basic Value stocks of those 3 universes gained between 68-80% vs 25% for SPY.

The “basic value” ranking system does not include any enterprise-value-based ratios. In other words, it looks only at the value of a company’s equity. Graham and Dodd made it very clear that one cannot look at equity in isolation–one must consider the company as a whole when assessing its value. That’s why the standard measure of value, DCF valuation, looks at enterprise value, not just equity. This is especially important in the present day when so many companies have bought back shares, thus reducing equity, and increased their debt load. If you put in EBITDA to EV, unlevered free cash flow to EV, sales to EV, and/or gross profit to EV, you’d get better results and would avoid a few of those value traps.

I also think it’s important to look not only at the performance of the top decile, but the ninth and eighth–and first, for that matter.

Thanks Yuval! I agree with this but I have some questions.

I first encountered EBITDA/EV in Greenblatt’s “The Magic Formula.” And Ross et. al. make similar arguments in their text “Fundamentals of Corporate Finance.” As Greenblatt points out this makes sense if you are considering buying a company. EV is one measure of how much you would have to pay to own a company. And why are you buying a company in the first place? To get the earnings (or EBITDA). The information is all there if you are going to purchase all of the company (including the debt obligation). And it is both simple and intuitive: the larger the ratio the more earnings for each dollar you invests. If nothing will change in the future just buy the company with the largest EBITDA/EV. Or maybe you prefer something a little different than EBITDA (insert your own preferred metric for discussion purposes).

If I am going to purchase a few stocks is there an additional question that I need to incorporate into my port? Do I have to try to figure out what part of the earnings or EBITDA goes to the stock holders vs the bond holders? Or does that all get factored out in some “identity” that I missed in the Ross text book or in the “Magic Formula?”

Edit: It could be a simple identity. Perhaps the reduced revenue stream to the equity holders— instead, going to the bond holders–is fully accounted for by the increased cost over and above the purchase of the stock which is incorporated in the EV. The EV in the denominator—which is greater than the market cap—may reflect this perfectly. Maybe the numerical value of this ratio is reduced by just the right amount and reflects the decreased revenue stream to the stock holders perfectly. That would be my guess at this point.

The above does leave out the question of differences in CAPEX between companies, I think. If you assume similar depreciation between the companies then CAPEX is incorporated—to a large extent—in the EV. Of course, this assumption could be very wrong in some instances. But I would be happy, for now, to understand the effect of revenue to bond holders.

Thanks for any ideas.


True. And I’ve had occasion to use other approaches, particularly EV/Sales.

I think though that when all is said and done, the most important element of value is a rational relationship between price paid for the asset (company or stock) and sound measure of wealth generated by the asset for which you’re paying. How we define price and wealth generated can be defined many ways and I’ve come to believe the definitional details are less crucial in live-money investing than the way we determine the reasonableness of the relationship.

I believe if we can model in such a way as to effectively distinguish between more or less likely to grow in the future, we’ll do well regardless of which measures of prices and wealth we choose to connect. At least that’s the hypothesis of a research effort/project that’s taking shape in my mind. (In a serious study, I’ll also have to account or control for the risk-quality factor and also consider the possibility that very high valuation ratios signal strong growth expectations - which may or may not be correct. See what happens when my favorite football team sucks and now have weekends free!!).

I thought my Rev2 value model is a failure, It is proving i am wrong.
It returns 40+% since launch; not significant draw down, it lagged performance with the bench mark for last 1 year.

then i came with value blend.
Value Rev3 - (Refined Value Rev2)
Value + RS
Value + Quality + Growth

It is too early, but last 3-6 months performance looks promising for all of my models.

Kumar :sunglasses:

In any case, I was hoping to get your thoughts on a means by which to differentiate growth and sustainment reinvestment. I think that this distinction could further your intent to identify factors that contribute to the persistence of future growth and profitability.

That’s the magic question . . . and a research topic that should keep me off the street and out of trouble for a while. :slight_smile:

Indeed. I am generally not satisfied with conventional approaches, two of which are enumerated here. However, I think a lot of the relevant information required of smarter approximations is readily available through a historical examination of the three basic accounting statements. For example, we can observe:

i, DD&A

ii. Capitalized assets

iii. Historical patterns in sales/earnings/cash-flow growth

iv. Historical patterns in reinvestment.

But what’s still missing here is a clear distinction between growth and maintenance CapEx (i.e., the amount of spending which would have been required for zero growth). While in theory DD&A approximates maintenance CapEx, we know this is moreso an accounting tool. Moreover, we can readily observe how companies can shrink (grow) if they invest above (below) stated DD&A.

Anyway, I think that deriving a better approximation of historical maintenance spending should be tractable.

Thanks David,

Similar to what I was alluding to with EV and depreciation I think. Maybe EV is a little comparable if depreciation is similar between two companies. Otherwise, not so much is your point I think (I agree).

But anyway, my question is DD&A: depreciation and amortization, I would guess. What is the other D?

Edit: Debt? Perhaps longterm debt is assumed to have gone toward CapEx?

Thanks in advance.


I’m not sure I’m answering your questions here, but this is my off-the-cuff take.

A company with low EV multiples is far more likely to be acquired–at a premium–by another company, thereby making your shares worth a lot more than face value. A lot of the big money I’ve made in the market has happened when a company I own shares in suddenly gets acquired (I made a 50% return on KTEC last week, and late last year YUME paid me a huge one-time dividend immediately prior to being acquired). That’s one reason why looking at equity multiples alone is not enough. The second reason to use unlevered free cash flow to EV is that free cash flow can be an excellent forecaster of growth PROVIDED that it’s not all spent paying down excessive debt.

Here’s an experiment. Create a ranking system with, say, 50 or 100 factors. Weight one of them at 100% and the rest at 0%.

Then create a screen on the Russell 3000 universe using that ranking system with just two rules:

rankprev(52) > 80
eps%chgttm > 15

Press the “totals” key and write down the number. Then weight a different factor at 100% and repeat.

The factors with the highest totals will be the ones that have worked the best in the last year.

Repeat for a few previous “as of” dates to weed out flukes.

Does that seem like a sensible way to model for future growth? Obviously, the factors would have to make good financial sense.

If anyone has any suggestions for improving this experiment, I’d welcome them.


Depletion. It is usually only relevant for natural resource and extractive activities.

My thoughts on EV vs MktCap:

(free cash flow to common equity) / MktCap === (free cash flow to all financial interests) / (EV + CashEquiv())

Expanding on that:

(CashFlow() - IntExp() - IncTax() - PfdDiv() - mii()) / MktCap === (CashFlow() - IntExp()*TxRt() - IncTax()) / (MktCap + MktValue_PfdEquity + MktValue_NoncontrollingInterests() + MktValue_Debt())

In p123, we cannot observe the market values for anything but common shares, so I usually opt for MktCap in the denominator. However, it is much easier to get a clean sense of free cash flow to all interests, so I also see merit in EV ratios.

Marc, if I may ask, who’s your football team? :wink:

Well, I did my little experiment, and here are the top five ways to predict future growth.

  1. Compare CurQEPSMean to the GAAP EPS for the same quarter last year.

  2. Compare CurFYEPSMean to last year’s GAAP EPS and/or NextFYEPSMean to CurFYEPSMean.

  3. Look at volume-weighted momentum. I use VMA(15)/VMA(210), but you can probably vary that.

  4. Compare the current quarter’s operating income to the same quarter last year.

  5. Look at industry momentum: Pr52W%ChgInd.

In sixth place was the best quality measure that I tested, and I tested about forty of them: the accrual ratio. Subtract operating cash flow from net income and then divide by total assets. The lower the better.

None of the other quality measures did that well. I had high expectations for free cash flow measures, but it turns out they’re not that good indicators of future growth.

My method was to look at the top 20% of the Russell 3000 for each factor and see how many of those stocks had a subsequent one-year EPS growth of 15% or higher. I used 5 data points over the last ten years.


You’ve got 4 forward look indicators based on analysts’ estimate and price momentum and 1 measure based on precedent.

Is it fair to suggest then that stock price momentum leads the fundamentals?

If so, why even bother with fundamentals?

Good question. I was looking at the efficacy of SINGLE factors. If I had combined several fundamental factors, I might have beat the momentum factors. But even if not, I wasn’t attempting to predict the price rise of a stock, just to predict the increase in EPS. It turns out, to my considerable surprise, that analysts and other market participants (the folks who buy and sell stocks enough to influence their price) are better at predicting EPS growth than any other single factor I could find. But that doesn’t mean we should pour fundamentals down the drain. After all, predicting EPS growth is only one small part of successful investing. Marc’s original question was, given that choosing stocks based only on value means choosing quite a few stocks with little growth potential, how do you best guess growth potential. I think I answered that, and that was all I was trying to do.