Getting statistical revisions right

• Pages: 200-201

Page 200

Revisions have the potential to make or break the reputation of a statistical office. If they are not explained properly, confusion, even distrust, may result. "Revisions can be particularly sensitive if statistical agencies handle them in an unprofessional manner," Carson observes. For this reason alone, revisions matter-a lot. But data revisions have also come to the fore for three other reasons:

New work on methodology has prompted revisions in many countries. As countries adopt new international standards, major revisions often follow (see box below).

The lack of international guidelines on revisions policy is complicating the work of regional organizations. According to Carson, "there is a need, for policy purposes, for data that cover the whole region," but regional statistical offices are finding it difficult to compile such data because each country has its own approach to revisions. As a result, regional data often do not match countries' own data, raising questions about credibility. Constantly revising the regional data is not an option either. If regional aggregates are revised every time a country revises, "it is changed so often that it's hard to know what the numbers mean."

Organizations such as the IMF need to distinguish between bona fide data revisions and suspect ones. If a country has a loan with the IMF and knowingly submits false data, it becomes an instance of "misre- porting." This can lead to various sanctions against the country, including having to repay the loan early, and, in serious cases, being banned from further lending. Revisions can sometimes be taken as a clue to suspect data.

Wait for perfect data?

Of course, different users have different needs, and most statistical organizations recognize this. "One of the jobs of the statistical agency is making the tradeoffs that keep everyone optimally happy," Carson says. Policymakers, investors, international organizations, and the media strongly emphasize timeliness. Politicians tend to tell statisticians to "put together the numbers, tell us what their strengths and weaknesses are, so that we can be aware, but go ahead and give them to us-don't wait for complete source data to make them perfect." In contrast, academics typically say, "We want to do research on the definitive numbers, and we will cheerfully wait until all the source data have come in." Financial...