The Commerce Department's Bureau of the Census reported today that sales of new one-family houses in March 1996 were at a seasonally adjusted annual rate of 672,000, 7.6 (+/-9) percent below the revised February rate of 727,000.

This is another one of those flaky housing statistics for which the uncertainty in the monthly change is larger than the change itself. The complete government report explains that these statistics are estimated from sample surveys, and are subject to sampling variability and errors of response and nonreporting. The range given for the percent change is the 90-percent confidence interval. When that range contains zero, as it does this month, it is unclear whether there was an increase or decrease. That is, the change is not statistically significant.

The government report goes on to say that changes in seasonally adjusted statistics often show irregular movement, and that It takes 4 months to establish a trend. Last month I did a linear fit through the four most-recent monthly data points; but, that didn't work this month because the points were too scattered.

So, I went all the way back to the start of last year and did a linear fit through the 15 monthly data points since then. The data was still pretty scattered; but, the slope of the line indicated an average annualized rate of growth of 13.3 percent in sales of new one-family houses over that period. In subsequent months I'll perform a similar analysis and we can see how the long-term rate of change varies from month to month.

Byline: Lafferty (MF Merlin)