Making Sense Out of Monthly Housing Starts

Special Commentary from the Office of the Chief Economist
by Frank E. Nothaft, Freddie Mac's chief economist
May 19, 2004

Construction starts are an important barometer of economic activity; abrupt jumps or drops in the level of new building can signal corresponding economic upturns or downturns. The large variability in starts during winter months, however, often confounds interpretation of the construction data during that season. To illustrate, housing starts fell 2.6 percent in February, but then jumped 6.4 percent in March to the third highest level since 1978.

Seasonality of Starts

The timing of housing starts has a strong seasonal pattern. During winter months in particular, frozen ground and ice/snow storms lower productivity and disrupt material supplies, causing the number of starts to be lower and more erratic. Furthermore, builders prefer to schedule housing production to meet the spring peak in home buying while minimizing inventory-holding costs during seasons when home sales are usually low. Primarily for these reasons, housing starts are generally 50 percent higher in May than in January: Over the past five years, there have been an average of 87 thousand one-family starts in January and 129 thousand in May. Consequently, about two out of five single-family homes are completed within the last four months of the year.

Because of the significant seasonal pattern to new construction, the Census Bureau publishes “seasonally adjusted” data, too. Seasonally adjusted data have been fine-tuned to eliminate the fluctuations that would appear if no account were taken of seasonal variations such as normal weather conditions and the differing lengths of months. By eliminating these effects, seasonally adjusted data can be compared month-to-month to ascertain whether the overall construction level has changed.

Variability of Starts

Despite the seasonal adjustment, construction starts remain much more variable during the winter months than they are during milder weather. Since 1960, one-family starts have moved 9 percent on average between successive winter months but only 4 percent month-to-month during the summer, as shown in Exhibit 1. Similarly, multifamily starts have more month-to-month volatility during winter than summer. This reflects, in part, the effects of abnormal weather conditions during winter, which can disrupt building activity substantially more than at other times of the year. Weather affects the timing of new construction but not the overall number of starts during a season – the latter determined by economic conditions. Thus, abnormally mild weather in, say,

Exhibit 1: Home Construction Data are More Variable in Winter

January could accelerate builders’ production schedules, leading to a large reported increase in starts during January followed by a large drop in February. Conversely, severe January weather could show the inverse pattern as builders accelerate February construction to get back on schedule.

Permit Issuance

Issuance of residential building permits also has a strong seasonal pattern and generally peaks in late spring or during summer. After seasonal adjustment, permit issuance also shows more variability during the winter, although the differences are less than with starts: in winter permits change an average of 6 percent per month, compared with about 3 percent per month the rest of the year, as shown in Exhibit 1. Abnormal weather conditions or other unseasonable phenomena may affect the lag between permit issuance and construction but have much less impact on the timing of the authorization.