How’s that for a sophisticated sounding title? What it describes is actually far simpler than it sounds, and if you bear with me, I’ll explain this foolishness. Its a favorite Wall Street error, as well as a pet peeve of mine.

What "Single vs. Multiple Variable Analysis" means:  due its inherent complexity, Market behavior cannot be explained or predicted by merely looking at just one thing — a single variable. If it could, than you would be able to pick that factor — Earnings, GDP, Interest Rates, Sentiment or what have you — and perfectly forecast what’s gonna happen next. Even a moderately accurate set of directional predictions would have obvious value.

But you cannot. At least not on a regular and consistent basis over a very long time. Given that the computing horsepower on Wall Street rivals NASA, you woulda thought that THE single variable, if it exists, would have been found already. But it hasn’t. That’s why some of the "cuter" analytical discussions are worthless (or worse!) to investors: Which party controls the White House, the Superbowl victor, and as of lately, Earnings.

The reality is that not only is market behavior a function of multiple variables interacting, they do so imprecisely — meaning, the exact same (or at least very similar) data sets at different times sometimes produce different results. Its not just one variable, but many dozens.

In a lot of ways, Markets exhibit behaviors quite similar to those predicted by Chaos theory — dynamic, non-linear, sensitive to initial conditions, etc. (for those who may be interested, James Gleick’s Chaos is a marvelous primer on the subject). But quantum physics aside, I think we can all agree that the market is
a terrificially complex mechanism for digesting an unholy spectrum of all too many data points. Merely controlling for one single variable over time is a surefire long term money loser.

The reason I bring this up is that the latest refrain from the chattering perma Bulls has been something like this: 

"Earnings are good this quarter, better than expected, and therefore, the market’s going higher."

That’s not only a gross oversimplification, its simply an untrue statement. This is not a new subject for us, but rather is something we have discussed repeatedly over the past few years.   

And one of our favorite commentators on the subject, Mark Hulbert, addressed this very subject (again) recently:

"Over the past 80 years, faster earnings growth has reliably been accompanied by a more sluggish market – except when earnings were falling out of bed and were more than 25% below year-earlier levels."

Why is that? As we discussed yesterday, these things do not occur in a vaccuum. As earnings heat up, that pressures the Fed to (re)act. By the time they are raising rates, you can assume we are rather late in the cycle:

"Faster earnings growth puts more pressure on the Federal Reserve to
raise interest rates. And in a head-to-head contest over which factor
has greater impact on the stock market, interest rates usually trump
earnings growth.

The 13.6% earnings growth rate that Thomson Financial is now projecting for the first quarter falls in the middle of a category associated with an average S&P 500 gain of 5.8% annualized – about half the market’s long-term historical growth rate. Because there is a wide range in the actual returns of the quarters that fall into this category, however, the S&P 500′s actual return during 2005′s first quarter – minus 2.6% — is well within the confines of the historical record.

The projected market return for the current quarter would be only slightly higher if Thomson’s projection for the second quarter — 7.2% growth — is accurate. The category into which this would fall is associated with an average annualized return since 1924 of 9.4%, which is still below the market’s long-term average."

Counter-intuitive, to be sure, but the data backs it up. As someone else once said: "If it were all that easy, we would all be filthy rich."

>

Source:
Rates usually trump earnings growth
Mark Hulbert
MarketWatch: 12:01 AM ET May 3, 2005 
http://www.marketwatch.com/news/story.asp?dist=&param=archive&siteid=mktw&guid=%7B2A736569%2D1C64%2D461F%2D9B19%2DC6BCCD83746D%7D

Category: Earnings, Investing, Markets

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

7 Responses to “Single vs. Multiple Variable Analysis in Market Forecasts”

  1. spencer says:

    Since WW II the correlation between the change in the market and the change in earnings is -0.002.

    In other words it is perfectly random, so even if you have a perfect forecast of earnings growth it is of no help in telling you what the market will do.

  2. steve says:

    It stands to reason that a complex model would be more descriptive but that is loosing sight of the real evaluative criteria,the R square. An R square is pretty descriptive, so a very high R square would say your dependent variable is very predictive. Different disciplines use different R squares but anything above .9 would seem very good. I am a novice suffering investor and I would appreciate it if someone could give me between three and five variable which correlate with positive and negative price movement.

  3. J Kunken says:

    I could not agree more…
    the same is true with the over-emphasis on interest rates. Nevertheless, it looks as if the butterfly that flapped its wings in Beijing did not evoke an aggressive rate hike on Tuesday.

  4. marku says:

    I am most of the way through Mandelbrot’s new book ” The Misbehavior of Markets”. It’s a contrast between the bell curve distribution of market returns that conventional financial anaylsis assumes, and the real behavior, which has “fat tails” and dependencies. It’s been a fascinating read (and scary– real market behavior is much more risky than conventional analysis assumes)

    Steve- In the social sciences, you’d never see an r squared as .9. That is rare even in engineering, my field. In fact most social scientists get excited about r squareds that would get me laughed out of a design review.

  5. spencer says:

    Steve — in doing economic time series analysis it is often posible to get 0.9 correlations because the data all seems to move together. But to test if it is really a good correlation the first thing you need to do is look at how the rates of change correlate. That is the mistake most people make about the market and earnings. They correlate earnings with the market and get high results just because both tend to rise over time. But if you look at the change in earnings and the change in the market you get a very different picture.

    If you look at the correlation between the change in the market and the change in the market PE you normally get 0.8, or better correlations depending on exactly what period you look at and how you smooth the data. So now you are only left with the problem of determining what drives PEs.

  6. This Week’s Carnival Of The Capitalists

    This week’s COTC is up at A Penny For Your Thoughts. For those with a Finance bent (or is that, “for those who are financially bent”?) there are a couple of outstanding finance pieces in this week’s group:

  7. This Week’s Carnival Of The Capitalists

    This week’s COTC is up at A Penny For Your Thoughts. For those with a Finance bent (or is that, “for those who are financially bent”?) there are a couple of outstanding finance pieces in this week’s group: