Today the Census Bureau of the Department of Commerce released its advance estimate of U.S. retail sales for February.

We are all consumers of goods and services. Retail sales are the monthly total of all the money we spend on mail order purchases and in stores, restaurants, car dealerships, gas stations and other retail establishments. Personal consumption expenditures account for 2/3 of the U.S. Gross Domestic Product -- the ultimate measure of the state of the national economy. So, retail sales provide important information on changes in this major component of the economy.

The monthly advance estimate of retail sales is based on a small subsample of the Census Bureau's full retail sales sample. Estimates from the advance and the subsequent full survey can differ because of the earlier reporting in the advance survey and because of sampling variability present in both surveys. Because of this, the Census Bureau has to update the monthly numbers twice before arriving at the correct number. They call the number released today the "advance estimate" for February. Next month this number will be updated with a "preliminary" number. Two months from now the Bureau will release the "final" number for this month. In other words, we won't really have a handle on the sales during February until the middle of May.

According to today's report, sales for February, after adjustments for seasonal, holiday, and trading-day differences, were $201.1 billion, an increase of 0.8 percent (+/-1.3%) from the previous month. Notice that the uncertainty in the monthly percentage change is larger than the change itself. What's more, the February change can lie anywhere between -.05 percent and +2.1 percent. Thus, we can't be sure of whether the February change was up or down.

We can get a feel for how much the monthly estimate improves after another month's data is collected by looking at today's revised figure for the change in January. Last month the change was reported to be -0.3 percent (+/-1.3%). This month, the January change was revised to be -0.1 percent (+/-0.6%). Note that the uncertainty is halved; but, we still can't be sure whether the direction of the January change was up or down.

Obviously, we need to be cautious about drawing any conclusions from recently released monthly retail sales data. A better approach would be to study changes in the data over a longer period of time, say a year. That way we would be dealing with ten final numbers, one preliminary number, and one advance number, and the error introduced by the more recent numbers would be a relatively small fraction of the total. Referring again to today's news release, we find that the reported year-over-year change from February 1995 to February 1996 was +5.0 percent (+/-1.9%). It appears that the percentage uncertainty for this number is the sum of the uncertainties for the most recent two months. While this percentage uncertainty is still large, the year-to-year change was of sufficient magnitude that we were at least able to determine that the change was positive.

We can gain additional precision in the sales measurement if we wait longer before we calculate the year-over-year value. The only example of this I could find was in the 1-30-96 report on advance retail sales for December. This release reported the retail sales change for all of 1995 to be +4.9 percent (+/-0.4%). This number contained eleven final monthly values and one preliminary value. Again, the magnitude of the percentage change was sufficiently large that we were able to determine the direction of the change.

From all of this we must conclude that, if we want to get readings on retail sales that accurately portray both the size and direction of changes, we are going to have to wait two months until all the data has been gathered together and the final number calculated.

We cannot get any useful information from an advance or preliminary estimate of the month-to-month change in retail sales unless the magnitude of that change is larger then the uncertainty inherent in making the estimate.

This is the kind of thing that Chairman Greenspan was talking about the day before yesterday when he said, "all of our data are flawed to some extent."

Byline: Lafferty (MF Merlin)