Traders and investors alike are in a constant search for any information that will help predict the markets or the economy.
The rewards are so great that we can easily push too hard and reach too far for an answer. In this article I will show how one of the most commonly cited recession arguments is mistaken.
To illustrate, please consider the "Mystery Chart" below.
In the data series depicted by this chart there are two events–a sharp decline in the series and then a very specific result. For this reason, the sharp decline is a "trigger." At the end of the series we have the largest decline ever seen.
If you looked at this series, using only the information provided, your conlcusion would be obvious:
A new event has been triggered!
I shall return to the mystery series. But first, let us turn from mystery to reality.
Forecasting Economic Growth
Everyone wants to predict economic growth. There is a continuing search for leading indicators that really lead.
The focal point of a current debate is the work of the Economic Cycle Research Institute (ECRI). The ECRI is a favorite source for many, including us. Each week there is an update on the Weekly Leading Indicators (WLI). This includes both a report of the level and the growth in the index.
The developers of the index have made a major financial investment. They need to show the power of their methodology, without giving away the results. This provides an open invitation for others to use the WLI, despite the "black box" character of the methods. I will highlight the two prominent voices analyzing the data.
David Rosenberg has noted that this level of the ECRI Growth Index has led to a recession 100% of the time. In June, quoted in The Business Insider, he had this conclusion:
…we can safely say that this barometer is now signalling an 80% chance of a double-dip recession. It is one thing to slip to or fractionally below the zero line, but a -3.5% reading has only sent off two head-fakes in the past, while accurately foreshadowing seven recessions — with a three month lag. Keep your eye on the -10 threshold, for at that level, the economy has gone into recession … only 100% of the time.
Author Joe Weisenthal carefully noted that this was not the official ECRI interpretation.
Popular chart guru Doug Short keeps the ECRI WLI in the news. Each week he publishes an update with a discussion. Here is the
most recent example:
This is a beautiful chart. That is part of the problem. I love looking at Short's charts, (now added to our featured sites). I respect his desire to help us in making data come to life. Having said this, I hope he will adopt a more critical and skeptical approach. His work has rocketed to prominence, widely cited and quoted. This carries a special responsibility.
The chart above shows the WLI growth index and suggests relationships with GDP and recessions. If you
read the article, and I urge you to do so, the entire discussion relates to the numbers in the chart — what the index has done in the past, etc. There is no discussion of how the index was developed, what it includes, and why looking at the growth versus the level might be deceptive.
The Analytical Error
Rosenberg and Short are guilty of suggesting big conclusions without properly considering the underlying data. There are two specific problems:
- The WLI index has an unkown composition. It is unwise to draw big conclusions when you do not understand the relationship. The WLI index itself is, by the design of the makers, a black box.
- The growth component is some sort of smoothed annualized growth rate in the WLI. We do not know either the nature of the smoothing or the time period for the comparison.
The Pragmatic Capitalist does a little better by showing both the WLI and the Growth Index on the same chart:
How well would you do in making inferences about the WLI (red line) if you looked only at the Growth measure (blue line)?
Trying to draw inferences without knowing anything about the underlying data is crazy! It is much wiser to listen to the interpretation from those responsible for constructing and updating the index. How can anyone else think he knows enough to offer a better interpretation? Here is
the ECRI interpretation (via Barry Ritholtz).
To elaborate further, let us look at the full published history of the ECRI data showing both series on the same chart.
The long-term growth in the WLI would not be apparent to those looking only at the growth index. It is deceptive to look at the growth index alone.
Back to the Mystery Chart
One problem in the interpretation of ECRI Growth is that it is a derivative measure — it shows the change in an underlying index. Let us now look at the index I used to derive the mystery chart. Think of it this way — the original mystery chart is like the WLI growth series. It shows the change in the underlying from one period to another, a derivative measure. Here is the underlying information, based upon the speed of an automobile.
The vertical axis in this chart is speed. The vertical axis in the original chart is acceleration. The plotted data show a car accelerating from 0 to 60, cruising, and then returning to stop. The "trigger" shows the stop position. The trigger signal worked for five cases out of five — just like Rosenberg's recession interpretation based upon data that he did not really know or understand.
What about the final trigger? Well — the car is still traveling at 60 MPH!
It is dangerous and foolish to make inferences from a rate of change index without knowing the underlying.
Conclusion
Interpreting derivative indicators requires special care and experience. Doing so properly includes a careful analysis of the underlying data.
I do not trust any source that presents a "rate of change" index without more careful analysis.
Meanwhile, the ECRI index is not completely a black box. Prieur du Plessis, in a series of well-documented articles, has identified some of the critical selements and also the likely method of smoothing the growth index. Anyone interested in this subject should review his many persuasive charts.
He concludes that much of the forecast is based upon the stock market, the bond market, and the money supply.
If I told you that these three factors suggested 80% odds of a recession, would you be convinced? I didn't think so. The ECRI folks don't think so either.
Meanwhile, the interpretation of Rosenberg gets picked up by many sources. No one is equipped to respond, since he is analyzing a string of numbers that usually represent a "trigger" for a recession.
What I understand from your article is that the WLI represents the rate of change in the index. The ECRI level represents the current value of the index. From there the point is made that although the rate of change is negative, the absolute value of the index may still be positive (60 mph in your example).
The criticism made of Rosenberg is that calling a double-dip based on the rate of change is bad analysis, because the economy may actually be going +60. Implicit in this criticism is an assertion the value of the current ECRI level must represent positve or negative growth in the economy. The problem with this analysis is that there is nothing here to tie the ECRI level to the rate of expanson or contraction in the economy.
To the contrary, the article states that the ECRI level (which is the input for the derive growth rate) is a hodge-podge of some unknown raw input (stock market, bond market, money supply are three potential sources mentioned). What stands out is that the chart smells a lot like a representation of asset prices over the given period (stock market, housing). As we know looking back on 2007, highs in the stock market and housing were no predicator of positive economic growth in the period ahead.
I don’t see any contradtiction or fallacy in making some correlations between the growth rate, and future economic expansion or contraction – at least not on the basis of your aricle.
Would be interested in your comments. Thanks –
James — The mystery chart is merely an illustration. It shows why it is a mistake to use a derivative index, acceleration in this case, without knowing and understanding the underlying data. In particular it shows the fallacy of the Rosenberg inference about a few prior cases.
I explained this as well as possible in the article. Analyzing data and describing it with charts works best when you get close to the data – -understand it fully. Those who think they can make an inference from the growth series, without understanding the underlying index, seem to be reaching hard for evidence to prove a point.
Why dispute the ECRI’s own interpretation, which has so far been very good?
Thanks,
Jeff