GDP Revisionist History

You may have looked past the headlines on Q2 2010 GDP numbers and said wait a second, how can Q1 2010 GDP go from 2.7% to 3.7% in a revision?! Good question.

Here's what the BEA says:

The revised estimates, which begin for most statistics with 2007, reflect the results of the regular annual revision of the national income and product accounts (NIPAs). These revisions, usually made each July, incorporate newly available and more comprehensive source data, as well as improved estimating methodologies. In this annual revision, the notable revisions primarily reflect the incorporation of newly available and revised source data.

Ok, but look at this, they went back all the way to 2006 and revised the numbers. How can anyone, using GDP related data in their calculations, be accurate with these kind of revisions going back over 3 years?




Here is the BEA summary of revisions:

  • For 2006-2009, real GDP decreased at an average annual rate of 0.2 percent; in the previously published estimates, the growth rate of real GDP was 0.0 percent. From the fourth quarter of 2006 to the first quarter of 2010, real GDP increased at an average annual rate of 0.2 percent; in the previously published estimates, real GDP had increased at an average annual rate of 0.4 percent.
  • For the revision period, the change in real GDP was revised down for all 3 years: 0.2 percentage point for 2007, 0.4 percentage point for 2008, and 0.2 percentage point for 2009.
  • The downward revisions to the annual estimates for 2007 and 2009 reflect partly offsetting revisions to the quarters within a year. For example, for 2009, the annual rate of change in GDP for the first quarter was revised up 1.5 percentage points from a large decrease to a smaller decrease, while the growth rates for the third and fourth quarters were each revised down 0.6 percentage point. The second quarter of 2009 was unrevised. For 2008, the downward revision to the change in real GDP reflects downward revisions for the second, third, and fourth quarters.
  • For the 13 quarters from the first quarter of 2007 to the first quarter of 2010, the average revision (without regard to sign) was 0.7 percentage point. The revisions did not change the direction of the change in real GDP (increase or decrease) for any quarter.
  • For 2006-2009, the average annual rate of growth of real disposable personal income was revised up 0.3 percentage point, from 1.2 percent to 1.5 percent.
  • From the fourth quarter of 2006 to the first quarter of 2010, the average annual rate of increase in the price index for gross domestic purchases was revised down from 2.0 percent to 1.9 percent. The average annual rate of increase in the price index for personal consumption expenditures (PCE) was revised up from 2.1 percent to 2.2 percent, and the “core” PCE price index (which excludes food and energy) was revised up from 1.9 percent to 2.0 percent.
  • For the revision period, national income was revised down for all 3 years: 0.4 percent for 2007, 0.6 percent for 2008, and 0.4 percent for 2009.
  • For the revision period, corporate profits was revised down for all 3 years: 2.0 percent for 2007, 7.2 percent for 2008, and 3.9 percent for 2009.

Gets more curious. GDP from 2007-2009, all of it, was revised down.

Gross Domestic Income (GDI) was also revised down.

The percent change from the preceding year in real gross domestic income (GDI) was revised down from 0.6 percent to 0.1 percent for 2007, was revised down from a decrease of 0.4 percent to a decrease of 0.8 percent for 2008, and was revised up from a decrease of 3.2 percent to a decrease of 2.9 percent for 2009.

These revisions are just astounding. National income was revised down for 3 years in a row. Corporate profits were revised down for 3 years as well. Now corporate profits, it's possible corporations send in modified tax returns, happens all of the time, same with small business income.

Personal outlays, which includes consumption, or C, or PCE was significantly revised, for 3 years:

PCE, personal interest payments, and personal current transfer payments was revised down for all 3 years: $15.4 billion for 2007, $15.0 billion for 2008, and $79.1 billion for 2009.

This is what the BEA claims is the reason for 3 years of revisions:

The annual revision incorporated data from the following major federal statistical sources: Census Bureau new and revised manufacturing economic census data for 2007; Census Bureau annual survey of manufactures for 2008; Census Bureau annual surveys of merchant wholesale trade and of retail trade for 2007 (revised) and for 2008; Census Bureau revised monthly indicators of manufactures, of merchant wholesale trade, and of retail trade for 2007-2009; Census Bureau annual surveys of services for 2007 (revised), 2008 (revised), and 2009 (preliminary), and of state and local government finances for fiscal years 2006 (revised), 2007 (revised), and 2008 (preliminary); Census Bureau monthly survey of construction spending (value put in place) for 2007-2009 (revised); Census Bureau quarterly services survey for 2007-2009 (revised); Census Bureau current population survey/housing vacancy survey for 2009; federal government budget data for fiscal years 2009 and 2010; Internal Revenue Service tabulations of tax returns for corporations for 2007 (revised) and 2008 (preliminary) and for sole proprietorships and partnerships for 2008; Bureau of Labor Statistics (BLS) quarterly census of employment and wages for 2007-2009 (revised); new BLS occupational employment survey data for 2009; Department of Agriculture farm statistics for 2007-2009; and BEA's ITAs for 2007-2009(revised).

You're kidding me right? Going back 3 years, they didn't have accurate raw statistics?

One point is the inclusion of more services for PCE, yet they claim this quarterly survey only started in 2010. So how can a new quarterly survey, which just started in 2010, magically revise PCE back three years to such an extent?

Other reason for revisions is a claimed quality adjusted communications equipment prices with examining fixed investment, capital goods. What does quality adjusted mean? If one bought crap routers somehow that does not count as an investment purchase or ? Anyone else believe communications equipment and quality adjustments can possibly account for these large of revisions surrounding fixed investment?

Here's a good one:

The deflator for command-basis GNP -- a measure of the goods and services produced by the U.S. economy in terms of the purchasing power of the income generated from those goods and services -- has changed to the price index for gross domestic purchases, which better reflects the uses rather than the sources of income. In addition, the gross domestic purchases index is used to deflate the trade balance in calculating command-basis GDP, which was not previously published. These changes are carried back to 1929 for annual estimates and to 1947 for quarterly estimates.

What? They are deflating the trade deficit through a host of end use adjustments instead of the income?

Look at this claim of accuracy on quarterly GDP too:

The quarterly estimates correctly indicate the direction of change of real GDP 98 percent of the time, correctly indicate whether GDP is accelerating or decelerating 74 percent of the time, and correctly indicate whether real GDP growth is above, near, or below trend growth more than three-fifths of the time.

Really? Who else is grateful we get a 74% probability on whether GDP is accelerating or decelerating?

Here are the revision differences between three years of annual GDP:

Real GDP Annual Revisions
annual 2007 2008 2009
revised 1.9 0.0 -2.6
previous 2.1 0.4 -2.4

Below are the supposed final Q1 2010 component breakdowns of GDP.

  • C = +2.13
  • I = +1.82
  • G = -0.39
  • X = +1.27
  • M = -2.09

Here are the revisions to those component GDP percentage contributions:

  • C = +1.33
  • I = +3.04
  • G = -0.32
  • X = +1.30
  • M = -1.61

Private inventories for Q1 2010 GDP alone were revised 2.64 from 1.88 as point contributions to GDP percentage. Look at consumption, that is a 38% change in point percentage.

What was that phrase? There are lies, damn lies and then statistics? I think they must have meant the BEA with such revisions to such a critical economic metric as GDP.



error corrected

Mistake on revision percentage change on PCE, it's 38%, not 62%.

Just shows don't believe the

Just shows don't believe the GDP numbers. They've always been too reliant on estimation to be very useful.

Stick with the *really* low-level data from the surveys, and do even the seasonal corrections yourself. :-P

seasonal corrections

They have some statistical algorithms (X12 or something) which spin out the adjustments and they don't offer up even the formulas that I can find. One way to do it is just ignore SA completely but that's not valid either because we know there are seasonal adjustments, it's common sense to level out the year seasons and events.