The Census Bureau's March 1, 2017 release estimated January construction at a seasonally adjusted annual rate of $1,180.3 billion. The reported monthly change was minus 1.0 percent, with a 90 percent sampling margin of 1.0 percentage point. Because the interval included zero, the release did not establish the direction of the total change. [pr201701][pr201612]

Public construction was different: its annualized rate was $268.7 billion, down an estimated 5.0 percent with a 1.8-point sampling margin. Public educational construction also had an interval excluding zero. Thus uncertainty about the aggregate does not mean every component was equally uncertain, and a significant component decline does not make the aggregate decline significant. [pr201701][pr201612]

The February 1 release had estimated actual construction value for the whole of 2016 at $1,162.4 billion. That is an annual total, whereas January's figure expresses a seasonally adjusted monthly pace on an annual scale. The earlier release also put December's annualized rate at $1,181.5 billion; March's release revised it to $1,192.2 billion before making the January comparison. [pr201701][pr201612]

These distinctions answer different questions: scale, period, comparison baseline and statistical uncertainty. The releases belong to one Census series, not two independent confirmations. Their sampling intervals exclude nonsampling error, and preliminary estimates can change. Census also cautioned that irregular monthly movements could require two months to identify a total trend and as many as eight for specific categories. One release therefore offers a sector signal, not a settled construction trajectory. [pr201701][pr201612]

Source documents

  1. Monthly Construction Spending, January 2017https://www.census.gov/construction/c30/pdf/pr201701.pdf
  2. December 2016 Constructionhttps://www.census.gov/construction/c30/pdf/pr201612.pdf