Q: Specifically, who is your book intended to reach?
A: The intended audience is business people. The object is to make managers understand how important good data is when making decisions. Another group is other statisticians. Unfortunately, they are a weak lobbying group. They don't have the power to initiate change in government statistics.
Q: You're calling for "broad statistical reform" of U.S. economic indicators. What does that mean?
A: Fundamentally, the concepts and the time series we use today were developed in the 1940s and 1950s. For example, gross domestic product and unemployment. All these fundamental concepts that worked well at an earlier point in time don't work well in today's society. So the reform would entail re-examining those concepts and then collecting better information to update those important time series.
Q: Would that include all monthly government economic reports?
A: For the most part there have been some modest improvements in some reports. But most series literally have not changed since they were created.
Q: You also recommend the formation of an advisory committee on economic statistics.
A: This is actually a procedure already used in many other countries. The committee would include members of the statistical and economic professions. It would also include users of the data from the private sector along with representatives of trade organizations, large and small business, and academics.
It would be a broad-based committee, standing for an extended period of time, with some limited rotation. The committee would not change frequently because there is a long learning curve with these statistics and it takes years to get things done.
Q: You also suggest simplifying government statistics?
A: Essentially, that means getting rid of conflicting information. For example, in the book we showed that the second and third estimates of GDP statistically have the same error as the initial estimate when comparing the figures to finalized estimates calculated after the end of the year. So that means we're really not adding new information with those reports.
We've become such a statistically obsessed society. And I'm saying let's eliminate the twitches that don't really mean much. There's a conflict between timeliness and accuracy. A purist would shoot for complete accuracy, but that would put the user of this data out of business. We have to opt for some sort of timeliness.
Q: You'd also like to see more public input. What would that entail?
A: The advisory committee I propose could take input from users of this data. The committee could act as a steering committee, in turn providing input to the government. As it stands now, private sector users of these reports do not advise the government on what they need or want. For example, 10-day auto sales is reported as a year-over-year percent change. It's a headline grabber but it does not mean anything analytically. You have to know what happened a year ago to make any sense out of it. Maybe there was a terrible snowstorm last year and nobody visited the showrooms.
There are other times when data is annualized, and that magnifies onetime events out of proportion. You get unreal swings in the data. That happened with the effects from Hurricane Andrew. You have to be careful. That's why I like to use four-month and four-quarter moving averages. They give you a better idea of what is happening to the fundamentals of an indicator.
Q: Your book also discusses the system of national accounts. What is that?
A: It is an improved accounting system for nations. In fact, now for the first time in history there is general agreement among nations regarding the gross domestic product of nations. A 780-page manual of concepts and guidelines will be published in January 1994. This is a golden opportunity to reform the U.S. statistical system. This system better distinguishes between production and services than the government's current accounting system. It also adds balance sheet concepts.
Statisticians and government agencies are ready to make the change. The question remains whether elected officials are willing to commit the time and money to it. They tend to be shortsighted. It would take a decade to revamp the current system.
Q: Would you like to see one government agency take control of all economic statistics?
A: That is not really an issue. In the 1970s, when I ran the government system of reporting, I argued against that. It is good to have a pluralistic system, which is probably more objective than a centralized one. And it would be easier to cut the budget of one central agency. There is really no advantage to centralization.
Q: How do U.S. economic statistics compare to other countries, such as Germany and Japan?
A: We're ahead of them. The U.S. still has the best accounting system in the world. But other countries have come a long way. For example, Canada has very good statistics. Surprisingly, the Japanese are woefully behind, but they are now trying to improve their system.
Q: Getting back to that balance between timeliness and accuracy. Do you favor postponing any reports to give the government additional time to make them more accurate?
A: No. I advocated requiring government agencies to announce a schedule of release dates then hold officials to that. People are hungry for information. There's nothing wrong with preliminary statistics. What gets confusing is when you have preliminary statistics then immediate revisions that turn everything around. Then there is no credibility to the numbers. There are almost no cases where waiting would yield enough better information to make it worthwhile.
Q. Do you view certain indicators as more troublesome than others in terms of being misleading?
A: Three or four years ago, international trade data had the biggest errors. Recently, the payroll survey has had grave problems, in terms of accounting for job creation. I guess my point is that no single data series is either good or bad. Over time, almost all series have problems as reality changes. With all the white collar restructuring, many individuals have stopped working for firms and instead work out of their homes. They are classified as consultants and don't show up on company payrolls. We should be running specialized surveys to count these people.
Q: So economic indicators by definition need continuous supervision?
A: Definitely. You don't want to be capricious. But you should be constantly probing on the fringes to find out what else is going on in society so data can be changed to account for it.
There is no such thing as a statistic without judgment. There is a value judgment in everything. Who is defined as employed requires a judgment and a precise definition.
Q: Is there much talk within the statistical community of the government collecting more information about what businesses plan to do?
A: The only forward-looking data the government collects now regards planned capital spending. That's the only forecasting-type information. For example, nobody collects projected sales figures. That would be hard. Few countries attempt to collect that sort of information. That's proprietary information that companies don't like to give out. There is also the fear that this information would be manipulated by companies for political reasons.
Q. What is your opinion of the media's reporting of economic data? Are there recurring errors?
A: There are glaring errors from time to time, but not necessarily recurring. It all depends on what part of a report the press focuses on. For example, good news tends to get played down, while bad news gets more attention. Unfortunately, I think that affects Congress. We should try to make favorable economic news more interesting. For example, a corporation laying off 5,000 workers gets a lot of play in the news while a company hiring 100,000 new workers over several quarters gets very little attention.
Q: How do you feel about the way financial markets react to economic reports?
A: I think they react perversely, if I may say so. Take, for example, a consensus forecast. It comes from calling 20 economists and getting their guess. That combined guess becomes the benchmark. There's not a single one in that group who has an independent source of information. The guess becomes the benchmark and markets react to whether the real number comes in higher or lower than the consensus of the guesses. Markets have lost track of things. They react to immediate trends rather than real levels.
Q: and A:
The U.S. government releases a monthly economic indicator, such as the employment report. Financial markets react. Fortunes are won. Fortunes are lost.
The government's method of calculating economic statistics represents, arguably, one of the best systems in the world. That does not stop private economists and statisticians from griping about it.
They say many of the indicators fail to portray current economic realities because they were designed decades ago. For example, heavy industry now accounts for much less of the total U.S. economy than it did 20 years ago. But government statistics do not necessarily reflect that.
Dun & Bradstreet Corp. recently published a book, "Statistics for the 21st Century," designed to educate the non-statistician and lay out a proposal for revamping the federal government's statistics system.
Joseph Duncan, Dun & Bradstreet's chief statistician, is co-author of the book. In a recent interview with staff reporter Dean Patterson, he discussed the current state of government statistics.
Duncan started with Dun & Bradstreet in 1982.
As chief statistician, he heads the firm's economic analysis department, which includes overseeing the firm's data bases and developing statistical products.
Previously, Duncan was the chief statistician of the federal government's Office of Information within the Office of Management and Budget. He has also served as the U.S. representative to the United Nations Statistical Commission, which he chaired in 1981.
