Wednesday, August 11, 2010

Misinterpreting Accounting Identities

Bloomberg News journalist Bob Willis quotes UniCredit Global Research chief economist Harm Bandholz with a positive spin on today's U.S. trade deficit numbers:

"The expiration of export-tax rebates on some Chinese commodities beginning in July may also cut U.S. imports from China in coming months, helping to narrow the deficit and thus contributing to growth in the third quarter, said Bandholz."

But here he misinterprets the GDP accounting identity: Y= I+C+G+NX. Where Y is GDP, I is investments (including inventory investments), C is private consumption, G is government consumption and NX is net exports (the trade balance).

He interprets this accounting identity as an improvement in NX (the trade balance) as causing an imrpovement in Y. In reality though, an improvement in NX is just as consistent with a reduction in I, C or G as an improvement in Y. The accounting identity only says that given a certain level of I, C and G, a change in NX will cause an equal change in Y. This analysis has relevance for GDP forecasters when they have knowledge of the values of I, C and G, but in the real world an "improvement" of NX will almost always be associated with a reduction in I, C or G.

Specifically in this case, if retailers in America wanted to import more before the expiration of the expiration of a Chinese export tax rebate, then that would certainly increase the trade deficit. But that doesn't mean that it will reduce GDP, because this accumulation of Chinese goods will be associated with either an increase in I (inventory investments, as companies store goods meant to be sold later) or an increase in C (as companies sell the goods they import). For this reason, this import surge doesn't causally reduce GDP because the apparent statistical reduction overlooks that this surge raised I, C or G.

The flip side of this is that the drop in the trade deficit in the coming quarter won't causally increase GDP because the improvement in NX will be cancelled out by a drop in I if the goods were stored, or a drop in C if the goods were sold.