A blog post from BEA Director Vipin Arora
My guess is that the standards and methods we use at BEA to put our statistics together are a bit different than people might expect. Understandably so—most people think about statistics in terms of surveys, so concepts like response rates, sample size, margin of error, and standard deviation are usually top of mind.
We do some of that. But most of our economic indicators are produced by bringing together data collected in assorted ways—surveys, scanner transactions, and more from a wide variety of data suppliers—within a rigorous economic accounting system. To quote former BEA Director Carol Carson, we “integrate and interpret a tremendous volume of data to draw a complete and consistent picture of the U.S. economy.’’
Saying it’s a tremendous volume of data is an understatement. In drawing that complete and consistent picture of the American economy, we’re bringing together hundreds of millions of data points within our economic accounting framework to produce estimates such as gross domestic product, the current account balance, and state personal income.
As you might imagine, given such large data volumes, it’s inevitable that there are inconsistencies across data sources in terms of coverage, timing, and quality. That’s where the integration Carol referred to comes in. It’s our job to ensure—using well-established standards and methods applied by subject matter experts—that the source data “hang together” to produce an accurate picture.
A few people have told me that dealing with these inconsistencies across source data is an often-overlooked quality-control service provided by BEA. I agree.
More importantly, it allows us to paint as accurate and complete a picture of the U.S. economy as possible.