Attempting to Create a "Long" Leading Index of the Economy
Those who have been following us through blog posts, media interviews and The Capitalist Pigs radio show know that we use the fact that the economy is cyclical to manage investments. The economy alternates between expansion and contraction, and periods of boom and bust. To gauge where we are in that cycle at any given time, we use indices of leading, coincident and lagging indicators.
When following the business cycle, we have a choice of indicators that fit our needs. We could follow the leading, coincident and lagging indices provided by The Conference Board, a private organization that charges a princely sum for their data, or we can employ similar series provided by the Economic Cycle Research Institute, or ECRI for short, who provides the data for free. Since the ECRI indices also happen to have a better track record at predicting business cycle turning points in our view, the choice is easy. We've chosen to use the data that has a better track record and is free.
However, there is one "problem" with ECRI index data – it's proprietary. We cannot be certain of the components that make up their three indices; they're black boxes. The Conference Board series, on the other hand, are completely transparent regarding the components of its indices. Still, simply having a good idea of what components ECRI uses is good enough, given their track record.
Sometimes, like in the current environment, it would be nice to know the components and weights used in ECRI indices. That would be particularly enlightening given that ECRI's director, Lakshman Achuthan, appeared on CNBC in October claiming that a double-dip recession was off the table. He cited ECRI's Long Leading Index as the source of his confidence. This is not to be confused with ECRI's Weekly Leading Index – a shorter-term leading indicator.
ECRI does not make their Long Leading Indicator available to the public without a subscription. Frankly, the fact that their indicators are proprietary and that some are not available without subscription is perfectly understandable and acceptable. They provide a valuable service and should expect to be compensated fairly for it. But without a subscription, we have no idea what Mr. Achuthan was looking at in his Long Leading Indicator in October. Even with the subscription, we would have no idea what components were driving the index to such seemingly optimistic levels. But we can make some assumptions and come to an approximation. That is the objective of this paper.
Before we roll up our sleeves, we should first show the value of ECRI indicators, using their weekly index as an example of the fine work they do. ECRI publishes their Weekly Leading Index (WLI) for free every Friday. That indicator is claimed to have a lead time of three months at cyclical bottoms and ten months at cyclical tops.[i] The record of ECRI's WLI is best defended by using a graph showing the predictive ability of the indicator for the last three recessions: July 1990 to March 1991, March 2001 to November 2001 and the most recent recession, December 2007 to June 2009.
As you can see, this index has accurately called the last three recessions and done so with some impressive lead times. Think about the value you could have gotten from following the business cycle in these three periods of rather painful stock market performance. The average lead time at peaks was a little more than eight months and the average lead time at troughs was close to four months.
The Long Leading Index, on the other hand, is claimed to have a lead time of about one year at both tops and bottoms. More about this index below.
We were fairly convinced that the risk of a double-dip recession had all but abated by October of this year, but for entirely different reasons. Once we heard Mr. Achuthan's comments on CNBC, we quickly ran an inventory of what we know about longer-term leading indicators to see if we could justify Mr. Achuthan's confidence. At this point, however, we should probably explain the history of ECRI.
ECRI was founded and directed by the late Geoffrey H. Moore, who spent decades working for and ultimately directing the US government's official business cycle dating organization – the National Bureau of Economic Research. In 1979, he founded ECRI's predecessor, the Center for International Business Cycle Research. Moore studied under the grandfathers of business cycle research – Wesley Mitchell and Arthur Burns – and was awarded one of the most prestigious designations in all of economics, the American Economic Association Distinguished Fellow. Laksham Achuthan and Anirvan Banerji have continued Moore's valuable work at ECRI since his passing in 2000.
This is an important point because Geoffrey Moore was much more open about his research than the current directors of ECRI. Again, that is not to be taken as a dig at ECRI – we're not sure we wouldn't run the organization in the exact same way. But the fact is that Moore published enough on business cycles for us to make an intelligent guess as to what he considered to be effective long leading indicators of the economy, and it's not a stretch to assume these indicators are components of the current ECRI Long Leading Index.
In Moore's 1990 book, he considered the following four indicators to be "long leaders":[ii]
- Bond Prices – Dow Jones
- Ratio of Price to Unit Labor Cost, Manufacturing
- Real Money Supply (M2)
- Building Permits, Housing
In 1991 Moore co-edited a compendium of recent business cycle literature in which he also contributes a chapter on new developments in leading economic indicators. In that chapter, Moore states that in his prior research he identified 15 leading indicators and "classified as long-leading those [of the 15] that had average leads of at least twelve months at peaks and six months at troughs during 1948 to 1982."[iii] He found four that met this criteria:
- Bond prices
- Real Money Supply (M2)
- New Building Permits for Housing
- Ratio of Prices to Unit Labor Costs in Manufacturing
These are the exact same four components. Given Moore's role at ECRI, together with his role in mentoring the current directors of that organization, we think it safe to conclude the current components of their Long Leading Index are very similar, if not identical, to the four components identified twice in his writings.
As mentioned above, when Mr. Achuthan pointed to the ECRI Long Leading Index on CNBC in October, we took a quick mental inventory of the indicators we suspected comprise the ECRI Long Leading Index. While bond prices had certainly enjoyed a run-up over the prior 12 months, they had started to meet resistance in the fall. Bond prices were likely flat at the time.
Prices of just about everything except houses had been going up in a soft employment sector, so it was quite possible to see the ratio of prices to unit labor costs in manufacturing flashing bullish signals. If prices exceed the costs of labor, it is assumed that manufacturers are profitable, the difference between the two serving as a proxy for profit margins.
Real money supply means the growth in money supply adjusted for inflation. Although we are reluctant to use the Consumer Price Index (CPI) as a proxy for inflation (for reasons we've talked about ad nauseum), mainstream economists prefer it to other measures. If we were to adjust money supply by the CPI, it seemed quite likely that money supply growth was accelerating faster than CPI, so it too was probably bullish. If we were to use commodity prices as our proxy for inflation, this component would almost certainly be bearish.
And that leaves the final component – building permits for housing. This would surely be a train wreck. It seemed virtually impossible that building permits would be doing anything but dropping or stalling. There was no reason to suspect this indicator was flashing bullish signals for economic growth.
So that's two potentially bullish indicators, one bearish indicator, and one "well, it depends" in real money supply – a far cry from the certainty and optimism Mr. Achuthan conveyed on CNBC. So we decided to actually see if we could construct a long leading index of our own using what we know Geoffrey Moore thought were the most effective long leading indicators. Against our better judgment, we used CPI as our proxy for inflation when constructing the index simply because it was more likely to conform to Mr. Achuthan's view. Without knowing exactly what ECRI uses as weights for these components or any statistical techniques applied to the data, our long leading index is a simple, equal-weighted index. That chart is posted below.
There are a couple of obvious problems with our construction. First, the recession of 2001 is flat out strange. Our Long Leading Index appears to reach both a peak and trough before the recession even begins. This would have been a disastrous outcome had we taken the second peak, right before the recession, as an "all clear" signal. Second, the lead times at peaks appears to be much greater than leads at bottoms, and perhaps too long to be of much use. Certainly, they are longer, on average, than Moore's original target of about a year or so. However, when used in conjunction with the shorter-term leading index, the WLI, problems in assessing the economy's likely path could have been corrected.
We should probably explain at this point that we at Butler, Lanz & Wagler do not use either index. We use ECRI's monthly coincident and lagging indices as published. Our leading index is a monthly derivative of the ECRI Weekly Leading Index. We would be remiss, however, if we did not mention that we are always looking to incorporate new tools and a long leading index is certainly one tool worthy of further exploration.
So our first attempt at reverse engineering ECRI's Long Leading Indicator does not appear as though it will ever provide any useful information on its own. Clearly, a simple equal-weighted index comprising the four indicators Moore identified will shed no light on what Mr. Achuthan was looking at in October.
So we had to run a multiple regression analysis to determine appropriate weights for the components based on their propensity to lead the economy. What we found was that bond prices and real money supply (using CPI as our deflator) were more important than the other two components and should be weighted accordingly. If we assign heavier weights to bond prices and real money supply, we get the following:
Now, we can see what Mr. Achuthan was looking at in October. More accurately, we are probably looking at a close approximation of what he was looking at. You can see a chart of the ECRI Long Leading Indicator posted at Barry Ritholtz's blog The Big Picture and judge for yourself. But keep in mind that the chart posted on Barry Ritholtz's blog goes back to 1960 whereas ours only goes back to 1987.
What can we take from all this? We did not exactly duplicate the ECRI version of the Long Leading Index as evidenced by anomalous behavior prior to the 2001 recession, and that's fine. If we were that interested, we'd subscribe to the ECRI service that provides such information. The fact is, we do not need such long lead times. The business cycle indicators we use provide ample warning. However, we were able to see that two engines that probably drive the ECRI Long Leading Indicator are bond prices and real money supply. Following these indicators can give us additional confirmation of what our shorter-term cycle indicators are telling us.
Further, we can now say that two tools we were already using - bond prices and real money supply - are effective enough that they're likely included in what has proven to be a very effective indicator for many decades of real time use. We are particularly fond of real money supply. It is only a matter of time, that when money is created, that it find its way into the real economy. Yes, bankers' willingness to make loans and borrowers' willingness to take them are important factors. But it's not likely that the money be destroyed before it is lent out.
So now take a look at the ECRI WLI above. And then take a look at the second version of the BLW Long Leading Index. What do you think the odds are of a double-dip recession now? We think it's safe to say that the odds of a double-dip recession are, at the very least, less than 50-50, and dropping every month. These odds, however, need to be tempered with a caveat – the odds apply to the chances of recession due to endogenous factors, or factors that come from within the economy itself. A weak recovery is always exposed to exogenous shocks that are not picked up by cyclical indicators. We cannot predict earthquakes, other acts of God, or terrorist attacks. The best we can do is assess the endogenous risks, and with the help of the long leading indicators, we can more confidently say that they are waning.
[i] Achuthan, Laksman and Anirvan Banerji. Beating the Business Cycle, Currency Doubleday: New York, 2004.
[ii] Moore, Geoffrey H. Leading Indicators for the 1990s. Dow Jones-Irwin: New York, 1990.
[iii] Moore, Geoffrey H. "New Developments in Leading Economic Indicators," Leading Economic Indicators: New Approaches and Forecasting Records, Kajal Lahiri and Geoffrey H. Moore, eds., Cambridge University Press, 1991.