How Early GDP Estimates Have Become More Accurate

Tomorrow morning’s advance GDP report is likely more accurate than similar reports released a couple years ago. Why is that?

This morning the Census Bureau released the advance economic indicators report, which showed that the goods trade deficit had decreased and wholesale inventories had increased in December. We can expect BEA to incorporate these statistics in their source data for the calculation of the GDP estimate that will be released tomorrow.

Two years ago, at the time of the advance (or first) GDP report for a quarter, these source data for the third month of the quarter would not have been available. So, BEA used assumptions in place of the missing data. Beginning in July 2015, the Census Bureau started producing an advance report for trade in goods in time for BEA to have data for all three months for the advance GDP report. A year later, data for wholesale and retail inventories were added to the report.

What has been the impact of these earlier data on the accuracy of the GDP estimates? A presentation at the most recent BEA Advisory Committee meeting showed a large improvement in the accuracy of the estimates for imports of goods. The improvement was measured by comparing the revisions that would have occurred if BEA had used its traditional assumptions compared to the generally smaller revisions after using the advance Census data. Exports, retail inventories, and wholesale inventories also showed improvements in accuracy, albeit less pronounced than in the case of imports. Overall, the numbers suggest that future GDP revisions may be reduced by 0.1 to 0.2 percentage point due to the availability of the advance Census data. (For comparison, the average revision without regard to sign from advance GDP estimate to third estimate is 0.6 percentage point – see the table under “Updates to GDP”)

How did these improvements take place? Several elements were involved. First, they reflect improvements that the Census Bureau has made in areas like electronic data collection and improved data processing systems, which have resulted in general improvements to both the timeliness and accuracy of the estimates. Second, they reflect collaboration between BEA and Census. The leaders of the two bureaus, together with the Economic and Statistics Administration that oversees both agencies, identified collaborative projects that would improve the functioning and quality of the overall statistical system, with a focus on improvements to GDP.

These collaborative efforts are continuing to move forward. For example, on February 17 Census plans to release for the first time ever an advance quarterly services report, which will provide estimates for select industries covered by its quarterly services survey about 25 days earlier than the full report. That means the data from the advance report can be incorporated in the second GDP estimate scheduled for February 28. Previously, BEA had incorporated these services data a month later as part of its third GDP estimate. Again, because Census was able to accelerate the release of some of its data, BEA will be able to improve the accuracy of GDP, in this case, for the second estimate.

With all of the concerns about government data, this project is a good example of government working to improve the accuracy of data at relatively little cost, simply by working more collaboratively and taking advantage of some of the improvements that have been made in data collection and processing.

Leave a Reply

Your email address will not be published. Required fields are marked *