Reporting/Market Focus from the May 2006 Edition of the SGS Newsletter.

New Orders for Durable Goods is a subset of the U.S. Census Bureau’s survey of Manufacturers’ Shipments, Inventories and orders. The Census survey M3 (no relation to the late monetary aggregate) is published monthly, with an "advance" report on durable goods usually published towards the end of the month following the survey and a week ahead of the formal M3 report.
Durable goods are those products designed to last three years or more and tend to represent investments for businesses or consumers. Historically, the patterns of annual growth in this series have provided a reliable indicator of pending economic activity, but there are a number of problems with the survey, and the loss of reported semiconductor orders in March 2002 did serious damage to series reliability. The semiconductor industry just stopped participating in this voluntary survey.
The survey of durable goods orders never has been a scientifically designed survey with statistical significance attributable to its results, and it only covers large companies with sales generally over $500 million. Somewhere between 40% and 60% of those surveyed respond in a given month, so a fair amount of guesstimation gets built into the results. Orders supposedly are net of cancellations, but that does not always happen.
In contrast, the purchasing managers new orders index includes smaller companies and gives each company’s response equal weight in the index calculations. Accordingly, the purchasing managers survey is not dominated by sporadic large and long-term airplane orders, unlike the durable goods survey.
Within GDP reporting, durable goods are reported in several categories, but the business investment (nonresidential fixed investment) account is the one most closely followed by the markets. Business investment has its parallel in durable goods among nondefense capital goods orders, but again the scope of the survey — particularly the lack of semiconductor orders — remains an issue. The problem in losing semiconductor reporting was that those orders indicated looming activity in the computer industry, a major component of GDP durable goods reporting.
In theory, business investment growth in GDP represents an estimation of 100% of industry activity, but inflation-adjusted growth has been dominated by computer sales. In turn, inflation-adjusted computer sales have been spiked artificially by government overestimation of deflation in computer prices, where deflation has been used to adjust for relative product features and quality. Accordingly there is a related upside bias built into the GDP’s business investment calculations.
As shown in the following graph, year-to-year change in inflation-adjusted durable goods orders has a good correlation with the business cycle, reflecting all the recessions of the last 35 years. The price deflator here, though, is the CPI-U. As bad as the CPI data are, they are better than the GDP deflators used for durable goods. Also, the monthly reporting is highly volatile, so a six-month moving average has been used.

The series may still work in predicting recessions, but that remains to be seen. There has been no recognized recession since 2001, which was before semiconductor orders were lost. The series is not showing a recession at the moment. The two lines in the graph reflect the change in aggregate reporting from the old Standard Industrial Code (SIC) for identifying industries to the NAFTA mandated North American Industrial Code (NAIC). How the two aggregated series could vary as sharply as they did is hard to comprehend for a stable series. Another business cycle or two will help calibrate the reliability of the surviving durable goods data in predicting major economic upturns and downturns.