Tim Duy: Third Quarter GDP, Part II

Tim Duy weighs in on the controversy over the auto production figures in the latest GDP release:

Third Quarter GDP, Part II, by Tim Duy: A bit of a controversy is simmering with regards to the jump in auto production as reported in 3Q06 GDP report.  According to Bloomberg:

An unexpected increase in auto production last quarter was a statistical fluke that will be reversed, making current U.S. economic growth even weaker, according to a former Commerce Department economist.

The report, in part, leads David Altig to conclude that:

Even an optimist would have to admit that the fourth quarter begins with a few pretty big challenges.

While Nouriel Roubini takes a more aggressive stance:

This mismeasurement of motor vehicle production in Q3 is highly suspicious coming about ten days before the US mid-term elections. It is also highly suspicious as it is not clear how the Bureau of Economic Analysis (BEA) at the Department of Commerce could have made such a gross mistake when seeing an alleged 26% increase in auto production that was patently at odds with many facts. During Q3 all the major US automakers - Ford, GM, Chrysler - announced production cuts for both Q3 and Q4. So, how could the folks at BEA argue and estimate that production went up by a whopping 26%?.. We thus expect BEA to provide a rapid clear and open explanation of this gross mismeasurement.

Admittedly, I was surprised as well – but I quickly thought to myself “oh, so prices fell.”  Didn’t seem like a big deal, so I was caught off guard by the controversy. 

It is worth it to step in and help the BEA – this is not an issue of “gross mismeasurement.”  This is an issue of some confusion about the NIPA accounts themselves.

Truth be told, the BEA is not helping itself.  From the Guide to the NIPAs:

Gross domestic product (GDP) (1–34), the featured measure of U.S. output, is the market value of the goods and services produced by labor and property located in the United States.

From this definition, it is reasonable to conclude that GDP measures output.  But, as I explain every time I teach principles of macroeconomics, in practice the BEA does not measure output.  Instead, the BEA measures demand:

GDP is measured as the sum of personal consumption expenditures, gross private domestic investment (including change in private inventories and before deduction of charges for CFC), net exports of goods and services (exports less imports), and government consumption expenditures and gross investment. GDP excludes intermediate purchases of goods and services by business.

For instance:

Personal consumption expenditures (PCE) (1–15) measures goods and services purchased by U.S. residents. PCE consists mainly of purchases of new goods and of services by individuals from private business.

In the background, the BEA is making use of the “national income equals national output” identity:  If it was purchased, we must have had the income for it, which implies that we produced the output in the first place.  Still, in practice, I am careful about using GDP as a measure of output, simply because it is calculated by measuring demand (purchases), not supply.

The 3Q06 “mismeasurement” of auto production is a perfect example – the supposed automobile “production” component of PCE is a measure of final sales, not production:

Sales estimates (PCE and PFI) for new cars and trucks are prepared using unit sales, allocations by sector, and estimates of average expenditure per vehicle…

In converting from nominal to real, a decline in prices yields a positive increase in real sales, even if nominal sales stay constant (or even fell).  As a demand side concept, this is not a problem.  If the price of automobiles falls relative to other goods in my basket, I am unambiguously better off as my budget constraint increased.  The real quantity of aggregate goods and services I can consume is greater.  No mismeasurement.  One, however, has to be careful of the supply side interpretations.

In practice, I tend to think of GDP as a demand side indicator, particularly when examining quarterly data.  In any event, I find it disingenuous to accuse the BEA of manipulating the data for political purposes – the confusion is readily cleared up once you understand how the data are constructed.

Update (by MT): Brad DeLong responds.

Tim Duy responds to Brad:

Update (by Tim Duy): Sorry for the delay – I had to shift to household production to compensate for a sick child.

Brad DeLong responds:

Now I am confused. Real GDP is (a) real final demand, plus (b) the change in business inventories. If cars were sold more cheaply in Q3 than anybody expected--and thus if the same flow of money spent on auto purchases resulted in more cars sold--then the extra cars sold must have come out of inventories, right?

Auto companies didn't ramp up production in Q3, did they?

Which serves to remind me that swimming with the big fish means you risk getting bitten now and then. Fair enough – some clarity can’t hurt, especially since Brad is getting to an important point. Note that I never claimed US auto production increased. My point is simply that the reported rise in auto production in the GDP accounts – as measured by final sales – is not necessarily inconsistent with the reported decline in US auto production. And it certainly does not imply that the BEA is manipulating the data for political purposes.

Let’s disregard the price decline, and just switch to physical units. In the comments, spencer points out that auto and light truck sales rose from 16.17 to 16.63 million units in Q3. Did in fact those sales never really happen because the Big 3 cut domestic production? 

No – at least two channels can be operating. One, as Brad (and spencer) points out, is inventory depletion. Which gets to the point that although something may get counted as final sales in 3Q, that doesn’t mean it was actually produced in 3Q. It could have been counted as an inventory accumulation in a previous quarter, and is subtracted in the current quarter. If the BEA underestimated inventory draw for this quarter, they can fix it on a revision – nothing nefarious need be occurring.

Moreover, just because a car is counted as a final sale in the US in 3Q does not mean it was PRODUCED in the US at all, in any quarter. In this case, the value is subtracted from GDP in the imports category. Note the large negative contribution due to rising imports (although I can’t say this was due to autos). Aggregate data can sometimes mask sectoral shifts unless you are looking for the shifts. It is not the BEA’s fault that Detroit chooses to make cars Americans are less interested in purchasing.

Getting back to the price decline, a decline in prices causes a statistical increase in real output. No big surprise – this is the way the data is computed. There may be an issue of whether the seasonal variations in prices have been accounted for properly. I have mused on this with other data. But I try to remember that the BEA has very qualified economists who spend hours studying the minutiae of these data, including the seasonal effects.