Paul McCulley and David Rosenberg like to look at core capex orders as a broad economic indicator. Prior to today, these orders have shifted to a negative trend twice in the past twenty years following multi-years of growth. Both of these trend changes in capex orders occurred at the beginning of the only two multi-quarter declines in S&P500 operating earnings in the last 20 years and two 50% peak to trough market declines. Today, orders have shifted to a negative trend for the third time in 20 years and it looks like S&P500 earnings will decline y/y in 3Q12 after posting many quarters of growth.  The market is within 1% of its 52-week high and valuations are near the levels associated with important prior market tops.

I started looking at this to get the big picture view of a broad industrial company that my quant short screen is identifying. My screen looks at micro data. Things like cash flow, accruals, earnings revisions, insider selling, valuation, etc. Perhaps the micro data was picking up something afoot in the macro data?

I note with the opening paragraph that there were two shifts to negative trends in U.S. core capex order growth rates* before today’s. But let’s start with a little broader definition, as I did when first looking at the data, and define a trend change as a shift from any zero or positive monthly growth to three consecutive negative monthly y/y growth rates. This picks up all the major periods of declining orders. The data is in a table at the end of this report. Here are the beginning months of the major negative trend shifts using this definition:

1. December 2000
2. September 2007
3. October 2008
4. June 2012

Looking at the first three instances, by the time we knew the trend changed, the market had already fallen from the prior 6-month peak by 20%, 6% and 37%, respectively, on its way to further declines. Today is an outlier. The market is up 2% since we found out at the end of September that the order trend turned negative beginning in June 2012.

The order weakness in October 2008 could be viewed as a continuation of the soft patch that began in September 2007. Let’s say that we only want to look at trend changes according to the definition above and those trend changes that have occurred after a string of multi-year growth in orders. Defined this way, the trend change might capture major turning points better and therefore be a bigger surprise to the market. Under these criteria, there are only three periods in the last twenty years that this has occurred. These three trend changes started in:

1. December 2000
2. September 2007
3. June 2012

You’ve probably already noted that the two very major market peaks of the last twenty years, in March 2000 and October 2007, occurred near the first two trend changes in capex orders. While the market was down a lot from its peak by the time you knew about the order change trend of December 2000 (you knew at the end of March of 2001), the market still had another 33% decline to go from the end of March ’01 before it hit bottom in October 2002. From the end of December 2007, when you knew capex order growth had changed for the worse starting in September 2007, the market declined 54% until it hit bottom in March of 2009.

Capex orders have made major negative trend shifts only when S&P500 earnings start major declines. It makes some sense. Companies tend to cut back on investment when they think earnings could decline. At the same time, companies are probably unlikely to cut investment a lot when they think the earnings decline will be mild. Do companies cut back only when they see a big shift in their business coming? Right now orders are down 10% from their recent seasonally adjusted 52-week high(data not shown in table; using a lookback window of 12 months to calculate declines from peaks requires seasonal adjustment). This has happened three other times in twenty years: February 2000 (within one month of the S&P500 peak), December 2000, October 2008 and more recently, July 2012. In September 2007, although the trend had shifted negative, orders were still only down by 3-4% from recent seasonally adjusted highs, but this was a big change from the prior string of growth rates.

There are only two periods of consecutive quarters of y/y declines in operating earnings in the last twenty years. These periods of declining earnings started with the same quarters as the two shifts to a negative trend in capex orders:

1. December 2000
2. September 2007

It looks like September 2012 quarter's earnings will decline after a string of growth. Will this be the start of a multi quarter earnings decline?

What about valuation?  Those familiar with Robert Shiller’s CAPE (http://www.econ.yale.edu/~shiller/data.htm) will note that today’s valuation of the S&P500 has been associated with other major market peaks in 1901, 1937, and 1966. After these peaks in the market it took investors decades to recoup their investment in the S&P500 in real terms. Only the stock market bubbles of 1929 and 1999 peaked at significantly higher CAPEs. Do you think we are entering an environment for equities like 1929 or 1999? Or does it feel more like the 1901 and 1937 market rebounds that occurred after very significant recessionary periods in the U.S.?

There are other economic indicators that portend a decline in the market, not the least of which is real hourly earnings growth. Finally, John Hussman has done extensive work on prospective market returns and his estimate for prospective market returns is in the most negative 0.5% of historical observations.

Putting it together, we have a market that is up and near its 52-week highs despite a negative trend change in capex orders that has occurred only twice in the last twenty years. Both of these trend changes in capex orders occurred at the beginning of the only two multi-quarter declines in S&P500 operating earnings in the last 20 years and two 50% peak to trough market declines. Finally, valuations are very high by historical standards. Perhaps this is a reasonable time to be short stocks.

Risks

There are all kinds of risks here. Prior correlation is not causation, though I do believe there are sound economic reasons why this data should continue working. This data fits with my bearish position and therefore bias. Timing a short on the market is very difficult and I’m probably early, as usual. History doesn’t repeat itself. The sample size is small.

This could be like July-Aug 1998. Capex orders hit their low point in July, which we knew at the end of August. But at this point the market had corrected by 20%, so there was less opportunity to short. Today the market is within 1% of its 52 week high. Also, in 1998 capex orders chopped around and there were a few quarters of declines, but not three in a row. Also, the capex order data is weaker today than in late 1998. It’s twice as weak in fact. In 1998, seasonally adjusted capex orders declined only 5% from the high point in the prior 12 months whereas today orders are down about twice that in July and August 2012 form the recent peak in December 2011. As in 1998, we have the Fed trying to engineer loose monetary conditions, but this attempt seems rather much longer in the tooth now than back in 1998.

*Not seasonally adjusted year/year growth in total U.S. new orders of nondefense capital goods excluding aircraft from http://www.census.gov/cgi-bin/briefroom/BriefRm . The idea to look further into U.S. core capex orders came from David Rosenburg’s chart showing this data versus U.S. total private employment with periods of recession shaded.

Data available with less than one month lag. September 2012 data is due October 25^th.