Sunday, February 1, 2009

Media Trims GDP stats

According the BEA's advance estimate, the fourth quarter GDP growth rate of -3.8% was better than the -5.4% pundits expected (although worse than I expected).

Many of the media responded by finding a spending segment that grew and declaring that it should not be counted. They choose inventory accumulation, saying that "fixing" that makes it -5.1%.

Their GDP trimming is incorrect, both as statistical and economic practice. Even if it were correct to trim, they have highly exaggerated the results.

A. If you want to eliminate inventory accumulation from GDP growth, then you should eliminate it from the pundits' forecast too. After all, the pundits were trying to predict all of GDP, not just the parts of it that the media would find relevant ex poste.

B. As a matter of economic theory, there is not a "problem" that inventories accumulated:

  1. Carrying an inventory takes some financing. The fact that businesses are doing so tells me that they have financing -- yet more evidence against the common wisdom that we are experiencing a terrible credit crunch.

  2. Nobody is forcing business to accumulate inventory: businesses CHOSE to do it. Maybe it's their judgment that they will make more money selling the items later rather than dropping price now.

  3. I am not aware of statistical evidence that high inventory accumulation forebodes low GDP growth in the future. Professor Nunes kindly send a scatter plot of historical measures of inventory accumulation and subsequent GDP growth:

[source: Professor Marcus Nunes]

The horizontal axis measures inventory accumulation as a percentage of GDP. The vertical axis measures real GDP growth in the next quarter. I don't see any obvious negative relationship.

C. Even if there were on obvious negative relationship, it would not tell us much about 2009 Q1 because 2008 Q4 inventory accumulation was not particularly unusual. Inventories were de-accumulated in an amount of .3% of GDP in Q3. That changed to accumulation of 0.1% of GDP in Q4 -- neither of these are notable deviations from the center of Professor Nunes' data.

That this trivial line in the GDP statistics makes headlines proves the media's (and by deduction, their readers') desperation for bad news.

1 comment:

dWj said...

Anecdotally, inventory accumulation is usually unintentional. When I've played around with datasets in the past, I've found that about half of inventory accumulation is typically given up soon thereafter, though I never looked at as much data as Professor Nunes offers, and what I did could not be called rigorous in any sense. What I also found was that attributing the other half of inventory accumulation to consumption rather than investment seemed to reduce the variance of both series, suggesting to me the same thing as the anecdotes: they represent anticipated consumption that didn't turn up. (This also was done crudely.)

Actually, what I was playing with was the line from table 2 of the bea release, i.e. the contribution of the change in inventories to GDP. Since inventories are a stock and their accumulation is a flow, I believe what I was looking at was the second derivative of inventories, divided by GDP and annualized. In any case, that number from table 2 last quarter the worst since Q4 of 2005.

I wouldn't be the least bit surprised if about half of the deviation from forecasts that itemized such things was the inventory figure, but I have no data to back that up.

Finally, one should obviously be consistent. If we're excluding even half of inventory accumulation from the GDP figure, the negative quarter from 2007Q4 is slightly positive instead of slightly negative. I don't recall the Times drawing this to our attention (though I wasn't paying a whole lot of attention to the Times at the time).