One of the nation's most eminent remodeling industry research groups has expanded and altered its popular economic forecasting report, in part because its old measurements depended on a federal report it regards as increasingly unreliable.
The changes were made to LIRA--the Leading Indicator of Remodeling Activity. It's been published every quarter since 2007 by Remodeling Futures Program at Harvard University's Joint Center for Housing Studies (JCHS). On April 21, JCHS released an updated version of this indicator (shown here) that's derived from a slew of new indicators.
One major improvement is that LIRA now tracks spending for maintenance and repairs along with its traditional numbers on home improvements. But notably, JCHS' formula to calculate LIRA also has stopped relying so strongly on a benchmark number for housing economists: the Census Bureau's monthly report on Construction Spending Value Put in Place, popularly known as C-30. JCHS instead is putting more emphasis on another federal report--the Department of Housing and Urban Development's American Housing Survey (AHS)--even though it comes out only every other year.
"In recent years, the C-30 estimates ... have become increasingly volatile and unreliable, subject to unusually large revisions," Abbe Will, the JCHS research analyst who led the rebenchmarking effort, wrote in a 41-page paper explaining the change. "... Not only are improvement spending levels about 50% larger in the AHS, the AHS also provides estimates of maintenance and repair spending allowing for a more comprehensive market size definition."
The new LIRA also factors in single-family house prices, residential remodeling permits, the Conference Board's Leading Economic Index, and the Gross Domestic Product. It no longer will consider pending home sales, the Purchasing Managers Index, the NAHB's Remodeling Market Index, manufacturers' shipment of building materials, and payrolls of residential remodelers.
LIRA is one of the few pieces of remodeling data that seeks not just to give interested parties a look at the current market but also signal its likely direction. The problem, Will says, is that even as the residential remodeling industry is near becoming a $350 billion annual industry, it "continues to struggle for timely and consistent data on current market size and trends."
For instance, while C-30 is widely relied upon, JCHS says its small sample sizes and imprecise survey design make its results unusually volatile. "The C-30 is picking up considerable noise in its estimates and not entirely reflective of actual market activity," Will wrote. For example, here's how JCHS compares C-30's reports on home improvement spending compare with another federal report on retail sales of building materials:
The Census Bureau regularly revises its monthly data and made big revisions in January, Will said, but "the underlying volatility of the C-30 due to sample size, survey design, and necessity of forecasting remains."
Just as troubling, C-30's estimates of spending are as little as only half of what the American Housing Survey says is being spent on home improvements.
Finally, the roller-coaster ride that housing has taken since 2007 led JCHS to look anew at how it weights the various economic benchmarks that it uses to figure LIRA. For instance, it said, the LIRA model used to include a financing component, but these days there's not as strong a relationship between low financing costs and remodeling activity. "Re-benchmarking the LIRA provides a good opportunity to test for other changing relationships and replace any inputs that have lost significant correlation with industry spending," Will wrote.
JCHS' increased use of AHS in figuring LIRA doesn't mean the center loves that metric unreservedly. Last year, JCHS was so concerned about problems with the latest AHS report--problems stemming from the government's brief shutdown in 2011--that it recommended this publication delay reporting the numbers. In January, the government finally issued a user's note citing differences in how the 2011 and 2013 AHS numbers were compiled.