Economic Statistics Aren't That Bad, After All
WE read in the paper, ``there are increasing complaints that statistics about the United States economy are becoming less accurate.'' The cry stems from those concerned over recent data revisions. In fact, the statistics are better than ever. Could they be even better? Of course. One way would be to revise them even more frequently.
Just what is an ``accurate'' economic statistic? Some figures are never revised; other figures are revised frequently. Is an unrevised statistic more or less accurate than a revised one? How many revisions are necessary in order to arrive at an ``accurate'' statistic?
When May and June retail sales were revised considerably upward recently, June was still below May. When July retail sales were revised significantly downward, they were still above June.
Some revisions of data are made years later. On occasions, dates on which the economy was said to have started a recession have been revised many years afterward.
Those of us who have worked closely with economic statistics for decades have had to deal with revisions until we felt it would be great if we never saw a revision again. Yet every revision is recognized as an attempt to provide a ``more'' accurate picture - not an ``accurate'' picture in the layman's sense of precisely so many apples or oranges.
Economic data for the most part are based on sampling procedures. No one is about to count the individual workers each month during the week containing the 12th day of the month. No one is about to add up the sales of every mom-and-pop store in the country.
It is today's intense scrutiny of economic data by a public largely unsophisticated in statistical techniques and details which leads to widespread indignation over misleading revisions. Then economists who should know better join the fray to point out all the inadequacies of economic data, as if the inadequacies have not been there for years. And as if responsible statisticians have not been trying to improve them for years.
High-speed computers permit data-collection agencies to process more data today than was dreamed of not too many years ago. The ``relative'' absence of revisions a while back, does not mean the data then were better. On the contrary!
Some years ago, many data were not even seasonally adjusted. New computer programs changed that. Many economic indicators were then given revised seasonal adjustment every year. Today, some indicators are given revised seasonal factors every month. Were the old data better because they were not revised as often?
The ability to revise data more frequently has led some government agencies to restrict revisions to the current year's data. To do otherwise would have them spending all their time revising back data. This does not make the unrevised data more accurate than the revised data. On the contrary!
The contention that our economic data should be more accurate is well taken, as it has been for years. Now comes word of the formation of a task force to improve the data. To the extent it provides resources that dedicated statisticians have been begging for for years, hooray! To the extent it placates those who could not care less until it became a media issue, who cares!