Monday, August 20, 2018

More on the New U.S. GDP Series

BEA's new publishing of NSA GDP is a massive step forward. Now it should take one more step, if it insists on continuing to publish SA GDP.

Publishing only indirect SA GDP ("adjust the components and add them up") lends it an undeserved "official" stamp of credibility, so BEA should also publish a complementary official direct SA GDP ("adjust the aggregate directly"), which is now possible. 

This is a really big deal. Real GDP is undoubtedly the most important data series in all of macroeconomics, and indirect vs. direct SA GDP growth estimates can differ greatly. Their average absolute deviation is about one percent, and average real GDP growth itself is only about two percent! And which series you use has large implications for important issues, such as the widely-discussed puzzle of weak first-quarter growth (see Rudebusch et al.), among other things.

How do we know all this about properties of indirect vs. direct SA GDP growth estimates, since BEA doesn't provide direct SA GDP? You can now take the newly-provided NSA GDP and directly adjust it yourself. See Jonathan Wright's wonderful new paper. (Ungated version here.)

Of course direct SA has many issues of its own.  Ultimately significant parts of both direct and indirect SA GDP are likely spurious artifacts of various direct vs. indirect SA assumptions / methods. 

So another, more radical, idea, is simply to stop publishing SA GDP in any form, instead publishing only NSA GDP (and its NSA components). Sound crazy? Why, exactly? Are official government attempts to define and remove "seasonality" any less dubious than, say, official attempts to define and remove "trend"? (The latter is, mercifully, not attempted...)

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.