Since the beginning of financial markets, speculators have attempted to find a method for determining future prices. Difficult as this endeavor is, the reward – wealth without work – is great enough as to tempt many folks to try. A speculative environment, the existence of which is the subject of another day, provides the present conditions for a vast array of parasites to sell their wares as “experts” to unsuspecting customers. Such are the conditions of today’s markets where newsletters abound, along with a wide range of advisories, websites, email solicitations, and conventional mail flyers.
Throughout this time, there have also been attempts to model the financial markets by using algorithmic models. For the most part, these attempts are as successful as those for spinning straw into gold or making chicken salad from chicken excrement. Yet even with this in mind, when an analyst claims to have an algorithmic model that provides value, it draws significant attention in the financial community, including mine. This attention is piqued when the analyst also brings to the table a dubious analysis of my own commentary’s accuracy as well as subjective criticisms lodged at some professionals whose opinions I value and respect.
It is in the spirit of unbiased evaluation and good science that the stock market models published on the internet by the CXO Advisory are evaluated here today. As stated at the top of the CXO advisory website, the CXO Advisory’s charter is to, “present models, research summaries, analyses and reviews designed for objective, unique and concise value to experienced investors and traders -- a modicum of actionable conclusions filtered from a very noisy environment. (The CXO) default approach is to challenge any and all conventional market wisdom with analytical skepticism.” The models which CXO present and update almost daily are those linked near the top of their homepage. The models are known as the Real Earnings Yield (REY) Model and the Reversion-to-Value (RTV) Model which are used to project the future price of the S&P 500 index. Within the REY model there are two sub-models which rely on two different inflation parameters as inputs.
The specifics of the models are explained with some degree of both clarity and ambiguity on the CXO website. The important inputs for the REY model consist of the current price of the S&P 500 (S&P), the S&P future operating earnings projections, and the rate of inflation derived by a non-seasonally adjusted published Consumer Price Index (CPI). There is a REY model that is based on core inflation (REYC) and one that is based on total inflation (REYT). Do these models work? Are they valuable? According to the graphics and text on the CXO site and reproduced hereon, the answer seems at first blush to be a resounding “YES.” Here’s the graphic correlation of the REY models versus the S&P 500 from January of 2005 to the present.
Image cannot be displayed
A pretty good fit. Yes? Indeed, CXO states,
“In summary, using the E/P-inflation rate gap to establish an investor/trader-required "real earnings yield" for stocks, with inflation rate volatility muted, generates a pretty good fit between modeled and actual behaviors of the S&P 500 index.
A simpler method, the CXO’s RTV model predictions are based upon future S&P earnings projections and a reversion to the mean rationale of price to earnings ratio. Here’s the correlation between the RTV-modeled data and actual S&P.
Image cannot be displayed
Again, there is a real good fit for the RTV model as well, and CXO agrees and pats themselves on the back accordingly! CXO states,
“In summary, using the average P/E to calculate a value-supported level for stock prices generates a reasonably good fit between modeled and actual behaviors of the S&P 500 index over long periods.”
CXO has modeled the stock market using not one, but three (3) different algorithms. May this be the tool used successfully to make money in the stock market? If it could, then CXO would be providing much more than “a modicum” of actionable conclusions. Rather, they would be producing a means of achieving wealth without doing much work. This would be the easy four step process to making money using the CXO models:
Step 1. Turn back the clock to January 2005.
Step 2. Buy the S&P 500.
Step 3. Turn the clock up to Halloween 2007.
Step 4. Sell the S&P 500 (Go short if you are an aggressive investor and cover today.)
You can see that the “rub” is that as good as the correlation is, it only works going backward in time – a difficult task to apply profitably for most speculators and investors not practicing late trading. The reason the models work when looking backward is that the S&P earnings parameter is an actual value and not a Wall Street analyst projection. Using Wall Street earnings forward projections as the CXO models do, is a more difficult endeavor.
However, one can attempt to use the CXO models to determine whether and when to buy the S&P. As of today, the REYT model published on the CXO site infers buying the S&P today (at about 1380) and selling it on December 31, 2008 (at 1580). Similar gains are projected by the other CXO models. Not a bad gain appears to be in store for investors who utilize the CXO models!
But how have the CXO models behaved for investors that do not have a time machine or a crystal ball and cannot practice late trading? CXO states that this is a difficult task because the forecasts change daily, weekly and monthly as new stock price, earnings forecast and inflation forecast data become available. However, it is a simple task to compile the CXO model projections (from the past) and see how these projections compared with the S&P 500 index at the projection dates (which are now in the past). In spite of the difficulty that CXO describes, I performed a check for those CXO model predictions (taken from the CXO site) which were available from January of 2006 to the present compared to the S&P 500 index actuals. This comparison is summarized in the following link. In all, there were 158 model outputs available including the three CXO models. Of those, the respective CXO model was essentially correct only 56 times or about 36%. In addition the batch of CXO projections made last spring projected the S&P to be over 1600 at the end of the first quarter, and at this point, these predictions are unlikely to be accurate. As a result, the CXO’s model low accuracy level will likely deteriorate even further in the near future. Therefore, the correct conclusion from this simple analysis is that the CXO model accuracy record is not one that is of much practical value to investors. But that is only one (my) opinion and this may somehow be unfair criticism. Another well known student of the market criticizes the CXO models, when she states:
“… I do not think they (the models) are accurate forecasting tools….the fact that (the inputs to the model) can “divert” for extended periods shows weaknesses in the models. Weakness in the models?”
These are pretty critical words lodged at the CXO models which are a primary basis for the CXO Advisory’ existence in the first place! Who is the harsh critic? Is this person one of the highly criticized “guru’s” referenced in CXO’s “guru analysis” section?
No. Actually, the critic quoted above is none other than the founder of CXO who emailed this exact statement to me. Within the same email chain, the CXO founder stated:
“The RTV and REY models are visualizations of two aggregate stock market valuation approaches ostensibly in wide use (P/E reversion and Fed Model-like relationship). They offer frameworks for thinking about aggregate market valuation, but I do not think they are accurate forecasting tools.” (Bold emphasis added by author.)
In the foregoing quote, the CXO founder is clear in describing what the “modicum of actionable conclusions” touted in the first sentence of the site’s introduction actually is. Generally in his words (not found on the CXO site), the actionable conclusion an investor should reach is that as forecasting tools, the CXO models are weak and offer only a framework for those individuals who wish to think about aggregate market valuation. While a philosopher may find some value in this modeling as a thought provoking exercise, it is of no practical value to most investors or speculators.
To take a “high horse” position as an evaluator of others’ work while implying or inferring the CXO modeling as being of any practical value, is to join the ranks of the worst parasites in today’s market environment. To be accurate and consistent with the objectives stated at the top of their site, CXO should state the following prominently in the reference section on their models:
“These models are of no practical value for investing in the stock market because we don’t think that they are accurate forecasting tools. They only offer a framework for thinking about the stock market’s valuations. The overall accuracy of these models is 36% since January of 2006 when they were initiated as an attempt at stock market forecasting.”
If such a disclaimer is not provided, maybe there already is some truth found on the CXO site which lays in the definition of the not-often-used word modicum in the Introduction. Dictionary.com defines Modicum as follows:
"Noun. A moderate or small amount: He hasn't even a modicum of common sense."
The definition of modicum may be what CXO Advisory has to offer experienced investors in the most simple and eloquent terms.
Today’s Market
Today’s action included another ugly day in the world stock markets. The long term weekly chart has a couple of features requiring attention:
Image cannot be displayed
The S&P 500 now is at the same level from which there was a minor top in April of 2006. This week’s action, including today’s close puts the S&P 500 at a lower low than both February of '07, and August of ’07. If this technical level is surrendered, it will be bearish for the long term trend of the S&P. Yet, this week’s climatic trading volume, coupled with the volume of last week is of enough magnitude to produce a “mini-capitulation.” With options expiration tomorrow, the S&P 500 weekly trading volume will achieve its highest level ever! With an oversold market, options expiration, a key support/resistance level, and capitulation-trading volume, it would seem to fit that a short term rally may be coming. But this doesn’t change the longer term bearish predicament of the S&P, and there is a similarly strong case for additional downside action in the short term. This case is based on recent acceleration in downward price momentum.
Separate from the technical picture, one of the most dangerous tools in this year’s market is use of an algorithm that models the stock market based largely on Wall Street earnings estimates and official government statistics.
