Sorting Out Starts

January Construction Retreats One Percent read the headline on McGraw-Hill Construction (MHC)'s February 19 press release. "January construction starts higher," Reed Construction Data proclaimed on February 11. Far from being anomalous, the difference in direction was typical of these two measures of construction spending. Census Bureau numbers usually tell yet another story. Why?         Both MHC and Reed privately collect data on new starts. They sell the information but also put out summaries in monthly press releases that purport to provide the full contract value of all new projects. MHC includes residential spending; Reed excludes it. But even in nonresidential starts, there are large differences. Reed estimated that nonresidential building totaled $14.4 billion, not seasonally adjusted, in January, an increase of 4 percent from January 2009. MHC said the January 2010 total was $11.2 billion, a decline of 21 percent. Reed pegged "heavy engineering" starts at $9.7 billion, up 56 percent from January 2009. MHC said "nonbuilding construction" totaled $8.6 billion, 8 percent less than in the year-ago month. Adding the two categories yields a nonresidential total of $24.1 billion for Reed, a gain of 20 percent from January 2009, and $19.8 billion for MHC, a drop of 16 percent.            MHC said its total was "boosted by a massive transit complex in Lower Manhattan NY (the subway and PATH train regional transportation hub), with a construction start cost estimated at $3 billion. If this massive project is excluded from the January statistics, nonresidential building would have fallen 14 percent while total construction would have dropped 10 percent." Yet Reed showed only $1.6 billion for "miscellaneous civil" projects, suggesting that its total either omitted the New York project or assigned it a much smaller dollar value. The Census Bureau will not report January data until March 1. When it does, it will show an estimate for the amount spent on new and ongoing projects in January, not the full value of new contracts. That distinction alone means Census figures tend to swing less from month to month than MHC or Reed data. In addition, Census includes major improvements and renovations to existing projects that may not be picked up by the private data collectors. Census' total for nonresidential spending in December was $49.2 billion, more than double the level found by Reed or MHC. Starts data are useful for suggesting the future direction of construction activity. But users should recognize that the data can vary by month and by source because of outsized projects and differences in dates in which projects are added to a company's data base. Census data provide the basis for estimates of gross domestic product and give a more comprehensive picture of current construction activity. All three series are subject to revision, so the first data release may not correspond closely to final estimates. However, only Census makes historical data readily available.