**Introduction**

This is intended to be the first in a series of posts which will

quantitatively examine RuleMaker (and other) metrics using statistics

and other mathematical tools and relationships. The primary goal of

this analysis is to answer questions such as:

o Do the RM metrics embody "quality" that the market recognizes?

(I already have statistical data from the RM ranking spreadsheets

that suggests *yes*)

o If so, how is that quality recognized... Valuation? Which

valuation method?

o What is the "relative quality" of each metric? For example, is

1 point of Net Margin equivalent to 1 point of Gross Margin? Or

is it closer to 1.856 points? Can 3 points of CashKing Margin be

traded for .05 points of Flow Ratio, and still have equal quality?

o Assuming the questions above are answered, can a mathematical

model of the "valuation vs. quality" relationship be constructed?

And can this model then be used to identify companies which may be

currently undervalued or overvalued based on the metrics.

o Can the model constructed above be backtested, and if so, how well

does it perform in a strategy that tries to buy RuleMakers at

attractive prices, keeping them until they either fall from

RuleMaker grace or become too overvalued to have a reasonable

chance of continuing to achieving the 2x5y objective.

**Some Important Assumptions!** (*My* assumptions... yours may differ)

o There are MANY MANY forces at work in the market.

o Different forces affect different companies and sectors. "Future

hope" is probably more relevant than "net margin" for many

biotechs. I haven't found the "future hope" line item in anyone's

financial statement yet; trying to quantify how some biotech

companies behave as a function of RM metrics is probably futile.

This will influence the types of companies included in this study.

o SOME of the forces at work are a function of the fundamentals of

the business, and will have a long term impact on the stock price.

Examples include margins, free cash flow, and management

integrity.

o SOME of the forces at work have little to NOTHING to do with the

business, and will have only a short term impact on the stock

price. Examples might include reaction to a product recall

(Tylenol!!... Firestone???), a poor call made by an analyst, or a

false merger rumor.

o A mathematical analysis can not totally replace other Due

Diligence work. (The folks on the Mechanical Investing Board, as

well as some Technical Analysts may disagree, and they may in fact

be correct under certain circumstances. That's OK with me. There

is a LOT of good brainpower at work in those camps that I will not

pretend to fully understand. My assertion here is that using ONLY

math and TA will be inferior in the LONG TERM to a strategy that

ALSO uses some intelligent subjective analysis, like looking at

management integrity, pending litigation, and accounting

procedures when you suddenly find that "too good to be true"

undervalued company.)

o The "answer" (a.k.a. holy grail, epiphany, motherlode...) is

unlikely to pop out quickly and obviously. At best, I am hoping

to find relationships that will help highlight potential fertile

starting points for further investment research. It is also

hoped that as appropriate, the data can be used to dampen

generalizations that are not universally true concerning certain

metrics, so I can stop wasting my time computing them.

o The "answer" will not only be hard to find, but pieces of it will

also change over time. I wish I had started this a little over a

year ago, just before we learned the truth about bubbles.

**Methodology**

To try and answer some of the questions posed above, I began studying a

group of 160 companies (See Below) with respect to their RuleMaker

(and other) metrics approximately 3 months ago. For each company, I

maintain a spreadsheet onto which I enter data from each quarterly or

annual report. The spreadsheet then calculates all of the RM metrics,

plus a few others that are frequently encountered in business analysis,

such as inventory turns, cash conversion cycle, ROA, ROI, and ROEIEIO.

All "fundamental" data is calculated on a Trailing Twelve Month (TTM)

basis, since many companies have business models that exhibit an annual

cyclicality, and also because I am more interested in how the market

reacts to longer term trends. Speaking of "trends", the data also

includes some data on the prior TTM, so that statistical analysis of

both current *position* and *direction* can be done for certain metrics.

The RM Data from each individual spreadsheet is then transferred to a

larger spreadsheet, which also includes "transient" items such as

price, forward earnings estimates, forward growth estimates, and

analyst recommendations. On a monthly basis, this data is "frozen" in

time and saved as a separate data file for future analysis. This

allows me to go back in time for instance to March 2, 2001 and see all

of the RM data that was currently known on that date, as well as the

price, forward projections, and valuations that were current on that

date. This will be used in the future to backtest various hypotheses.

Statistical analysis to uncover underlying relationships between two

different parameters is performed by using a "statistical calculator"

which can accept a set of "x-parameters" (say net margins for the 160

companies) and a set of "y-parameters" (say Price to Free-Cash-Flow

ratio for the same 160 companies) and perform a regression analysis to

identify the "best fit" line that passes through these 160 x-y pairs.

By examining the equations derived for these lines, as well as

parameters like "correlation" and "covariance", you can begin

to assign real live numbers to the "value" the market assigns to

certain metrics. You can also begin to assess whether the parameters

are strongly or weakly related. At this time, no plans are in place to

explore non-linear solutions to achieve higher accuracy. I'm just an

80/20 kinda guy that way.

**How the Companies were chosen**

As hinted at above when discussing market forces, it is my belief that

if you want to study the affect RM metrics have, you should probably

select companies for which RM metrics are at least partially relevant.

Therefore, all the companies that have ever had an RM analysis

performed on the RM company board served as the *starting* point. Then,

since a full RM analysis requires at least a passing glance at

competitors, a significant number of competitors to RuleMaker companies

was added. I also included some of the "biggest" companies in the

sectors where RM's were found, independent of whether or not they were

viewed as a direct competitor to one of the RM companies. Since these

last inclusions have now "polluted" my list with companies that may not

be even close to looking like RM's, it will be necessary (eventually)

to perform some of the statistical number crunching with two sets of

data... sometimes with all companies, and sometimes with only those

that have "*RuleMakerness*", to be defined later.

Now since the data is already polluted, I added some companies that

have decidedly un-rulemakerness attributes, just to serve as a test

bed for the hypothesis that different types of companies and sectors

do indeed behave differently when looking at their metrics. Finally,

I eliminated companies that require unique analysis. (All financial

companies or others that have "*financial looking*" financial

statements). The complete list of tickers for companies included in

the study is shown at the end of this post. The subset that are deemed

to have "RuleMakerness" will have to be identified at a later date,

since at this time I am still waiting for or analyzing about 30 annual

reports for the period ending 12/00.

**First Analytical Results:**

(I bet you can hardly wait, right?)

My first results to share have to do with a decidedly un-Foolish

metric, the "Analyst Recommendation". On March 2, 2001, I took the

first "data snapshot" for the 160 companies. On March 31, 2001, I took

the second "data snapshot". Looking only at the analyst recommendation

and price on these two dates, I set out to answer the following 3

questions:

**1) How did the stock price for the period 3/2/01 to 3/31/01 perform**

as a function of the analyst recommendations on 3/2/01?

Well, as most of you are probably aware, analyst recommendations can be

found on the "Research" Stock quote page of Yahoo!, expressed as a

numerical value between 1 (Strong Buy) and 5 (Strong Sell). The number

listed is simply the average value of all the ratings by all the

analysts. So if there were two analysts covering XYZ company, and one

analyst rated the company a "Strong Buy" (Value=1) and the other had it

rated a "Hold" (Value=3), then the average recommendation would be

(1+3)/2=2 or "Buy".

So to answer the above question, I began by simply feeding my

statistical calculator 160 x-values corresponding to the 160 average

recommendations. Then, I computed the percentage price change for each

company according to the formula

% Price Change = 100 * (((Price on Mar31)/(Price on Mar2))-1)

Spreadsheets are marvelous, as I only had to type this formula in once,

and was then able to replicate it for the other 159. Anyway, I was now

armed with 160 y-values, which I promptly fed to the calculator using

the copy and paste utility between the spreadsheet and the web page

where the calculator resides. I then hit the "submit" button, and

almost immediately was looking at a scatterplot of the 160 x-y pairs

along with two "best fit" lines, one which assumed x was the

independent variable, and the other which assumed y was the independent

variable. And what did the equation of that line tell me? Well,

hopefully everyone here remembers their basic algebra (or was it

geometry?) in which a straight line can be expressed in the form

y = mx + b

where the dependent variable y can be calculated for any independent

value of x, assuming you know m (the slope of the line) and b (the

y-intercept of the line).

For the Analyst recommendations and price changes of our 160 companies,

the best fit line came back as:

% Price Change = (6.25 * Analyst Reccomendation) -21.06

(y) = (m) (x) (b)

Those familiar with looking at these types of equations will

undoubtedly notice that there is a positive slope to the line. In

other words, the higher in the positive direction the analyst rating,

the higher in the positive direction the price changed. **But WAIT!!**

Isn't a higher rating a *less* attractive rating? (1 = Strong Buy, 5 =

Strong Sell). Yup, as the data played out for *this* group of stocks in

*this* time period, the better results (statistically speaking) were

achieved by going exactly counter to the analyst recommendations.

March was yet another down month, so on an absolute basis, most stocks

did poorly.... but making a table of various values from the above

equation, it is easy to see that the statistics definitely favored

the least favorite analyst picks:

Analyst Recommendation I % Change in Price

___(AVERAGE)____________I____(AVERAGE)______

I

1.0 (Strong Buy) I -14.81

1.5 I -11.69

2.0 (Buy) I - 8.56

2.5 I - 5.44

3.0 (Hold) I - 2.31

Once again, these are statistical averages only, which are very similar

to the "average family". It doesn't actually exist. And just because

we have discovered some sort of "relationship" between these parameters

does not necessarily mean that the relationship is one of "Cause and

Effect". It is higly unlikely that a lousier rating CAUSED improved

performance. A more plausible explanation might be that the analysts

were expecting a certain sector or group to do well (such as tech),

and the opposite occurred.

For the statistically non-challenged, here are some detailed stats

behind the data:

Mean Analyst Rec = 1.95 Mean % Price Change = -8.87

Covariance = 1.05 Correlation = .18

Regression (Price on rec) = 6.25 Regression (Rec on Price) = .005

For the statistically challenged, (which includes myself), I think the

most important number above is correlation = .18. Correlation can vary

between 0 (No correlation) and 1 (Perfect correlation). The value of

.18 would indicate a fairly weak relationship, in which other factors

are clearly working. This should be no big surprise. As we proceed

through analysis of the various RM metrics, I expect that no single

parameter will have a large correlation factor to valuation, since so

many parameters are simultaneously at work.

OK, now on to the next Question...

**2) Is there a relationship between "Change in Analyst Reccommendation"**

and "Change in Price"?

We frequently see complaints on the boards about analyst downgrades

causing a stock price to tumble. Clearly, there is an implied cause

and effect relationship here.

But we also see cases where a company will issue a profit warning,

which simultaneously has two outcomes: The stock price drops, and

a host of analysts issue a downgrade. Here, the analyst downgrade

is the *result* of something happening to the company (and its stock

price), not the *cause*.

OK, whatever! To answer question 2, I fed the calculator with x-values

consisting of "change in analyst recommendation" and y-values of

"% change in price". Depending on which perspective is appropriate,

2 equations can be derived from the best fit line for these x-y values.

If % Price change results from an analyst upgrade/downgrade,

the relationship was:

% Price Change = (14.56 * Analyst Rec Change) - 6.98

Solving the above equation for a couple of specifics, we find

An analyst downgrade of 1 full point caused a price drop of 21.54%

No analyst change resulted in a price drop of 6.98%

An analyst upgrade of 1 full point caused a price rise of 7.58%

If on the other hand an analyst upgrade/downgrade results

from some company news and resulting price change, the relationship is:

Analyst Rec change = (% Price Change + 6.98)/14.56

Solving for a couple of specifics again, we get

A price change of +10% resulted in 1.16 points of analyst upgrade.

A stable price resulted in 0.48 points of analyst upgrade.

A price change of -10% resulted in 0.21 points of analyst downgrade.

Once again, for the statistically non-challenged,

Mean analyst change = -.12 Mean Price change =- 8.87%

Covariance = .80 Correlation = .24

Regr'ion d(Price) on d(rec)= 14.56 Regr'ion d(rec) on d(Price)= .004

The correlation of .24 is still fairly weak, but somewhat stronger

than the correlation in the previous data.

Finally, I looked at one last potential relationship:

**3) Is there a relationship between the change in price from 3/2/01**

to 3/31/01 and the absolute analyst rating on 3/31/01.

This might be rephrased "Do analyst ratings correlate with short term

relative strength?"

Without belaboring the mechanics any further, this set of data had a

covariance of only 0.24, and a correlation of only 0.04, indicating

virtually no relationship.

OK, enough fun with numbers for today.

**COMING SOON:**

What is the relationship between the RM metrics "Gross and Net Margins"

and the valuation metrics "P/E ratio and P/FCF ratio"? (PEGs, YPEGS,

Benjamin Graham Indices, and other valuation techniques can be built

up from these by the "interested student".)

I promise the next post will be shorter, since I won't have to explain

so much background info.

Ralph

Company Tickers of included Companies (Disclosure: I own the bold ones)

ABI ABT ABV **ADBE** ADI ADP AEOS ALTR AMAT AMD AMGN ANF **AOL** APCC AVNT

BARZ BBBY BC BGEN BLCCY BMCS BMY BNG BRCD BRCM BUD

CA CAH CCE CCL CHKP CHRT CIEN CMVT COH COST CPB CREE **CSCO** CSG

DA DCX DELL DEO DG DIS DNA DRI

EAT **EBAY** EDS ELN ELY EMC ERTS

**F** FD FMX

GE GENZ GIS GLW GM GPS GUC

HAS HD HDI HINKY HNZ HON HWP

IBI **IDTI** IFX IMNX **INTC** INTU ISCA IVGN

**JDSU** JNJ JNY

KM KO KSS

LIZ **LLTC** LLY LOW LSCC LTD LUV LVMHY

MACR MAT MAY MCD MCHP MDT **MEDI** MERQ MMM **MSFT** MSIM

**NOK** NOVL NT NTAP NTT NYE

OAT **ORCL OSI**

PALM PAYX PEP PFE PHG PMCS POC PZZA

QCOM

RATL RCL RL RTRSY

S SBC SBUX SEBL SGP SLE SNPS STM SUNW **SYMC** SYY

T TGT TJX TLB TM TOM TOY TSCO TSM TXN TYC

UN

V VFC VZ

WMT WWY

XLNX

**YHOO** YUM