14 May 2014

The Absurdity of Conventional GDP Accounting

Economics 29 Comments

In two weeks the BEA will announce an update on its estimate of 1q 2014 GDP growth, but I want to point out a problem with the “advance estimate” (of a dismal 0.1%) that they issued at the end of April. Here is how Forbes reported it at the time:

The U.S. economy grew in the first quarter — but very, very, very slowly. Most economy watchers blame frigid winter weather for dampening forward progress but not everyone is convinced weather tells the whole story.

The Bureau of Economic Analysis’ advance estimate of first quarter 2014 real gross domestic product shows output produced in the U.S. grew at a glacial 0.1% rate. This is growth relative to fourth quarter 2013, when real GDP increased 2.6%. Economists were anticipating growth around 1.1%.

Most of the weakness came from trade and inventories which subtracted 80 basis points and 60 basis points from overall GDP respectively. According to BEA, the slowing growth also reflected a downturn in nonresidential fixed investment growth, as well as lower state and local government spending.
[Bold added.]

The fundamental confusion in conventional GDP accounting is that it relies on spending as a measure of economic output. Yes, there is a certain logic in this; if businesses sell $15 trillion worth of stuff, then their customers must be buying $15 trillion worth of stuff. Yet it is still a very dangerous game, because it leads people to confuse accounting tautologies with economic causality: they think that the way to boost economic growth is to hike government spending, for example.

Yet things are even worse. Because policymakers and the pundits routinely refer, not to GDP itself, but to GDP growth, then “contributions to” or “subtractions from” GDP growth are actually second derivatives of the various variables involved.

For example, consider the recent GDP report. The Forbes article tells us that “inventories” subtracted “60 basis points from overall GDP.” What the heck does that even mean? How can goods sitting on warehouse shelves affect the ability of workers to take resources and produce output?

Of course, the ultimate explanation here is that the first pass at calculating GDP looks at the final spending by customers. Then, we look to see how much inventories changed during the period. For example, if customers spent $15 trillion on final goods, but during that period inventories fell by $1 trillion, then actual new production that period must only have been $14 trillion.

OK, so since inventory was apparently responsible for shaving 60 basis points off of GDP in the 1st quarter, I guess inventories across the US fell from 4q 2013 to 1q 2014, right?

Nope, that’s not right. The BEA report tells us (and note that that hyperlink will only be valid until the next BEA news release): “The change in real private inventories subtracted 0.57 percentage point from the first-quarter [2014] change in real GDP after subtracting 0.02 percentage point from the fourth-quarter [2013] change. Private businesses increased inventories $87.4 billion in the first quarter [of 2014], following increases of $111.7 billion in the fourth quarter [of 2013] and $115.7 billion in the third [quarter of 2013].”

Now we see why I was talking about second derivatives. Private inventories have been increasing in each of the last three quarters. But the increase (in dollar terms, and a fortiori in percentage terms) has itself been shrinking each quarter. That’s why inventories “subtracted” 2 basis points from GDP growth in 4q 2013, and “subtracted” 57 basis points (which the Forbes article rounded to 60) in 1q 2014.

If you want to see a more careful exposition of these ideas, and a simple numerical example to pinpoint the craziness, read my article at Mises.org cleverly titled, “Inventories Don’t Kill Growth–People Kill Growth.”

29 Responses to “The Absurdity of Conventional GDP Accounting”

  1. Transformer says:

    (Hit send by accident so posting again)

    Are you just saying:

    NGP growth consists of combined growth in final sales and growth in inventories,

    Growth in inventories slowed down so contributed to a overall slowing down in GDP growth ?

    Obviously growth in inventories can slow down and still be > 0, So I’m not seeing this as absurd. Am I missing he point ?

    • Bob Murphy says:

      Transformer I don’t like people saying “inventories slowed growth” because it makes it sound like a worker had boxes fall on him in the storeroom. But click the link to the old Mises.org article to see why I have such a bad taste in my mouth.

      • Transformer says:

        I think the relevant point in that Bloomberg article is that if inventories in that quarter were higher than people had planned (That is: they fell less than businesses wanted them to fall) this might be bad news as those business will produce less next quarter as a result. This seem a reasonable theory.

        When they say “with more than half of that growth — 3.4 percent — attributable to changes in inventories” I think this just means that in that quarter inventories fell by an amount that reduced GDP (based on final sales alone) by 3.4% less than it reduced it last quarter. If so then this is a bit confusing but not really absurd IMO, in the light of the overall point being made.

        • Transformer says:

          BTW: Your example in that earlier doesn’t seem to involve a year (apart from possibly year 1) where businesses might have higher inventory higher than they want to hold and so cut back on GDP as a result. I don’t think itaccurately reflect the Bloomberg point.

        • Bob Murphy says:

          Transformer, I have two responses and then I have to drop this:

          (1) Even on your own terms, it’s not that businesses incorrectly forecast what would happen with inventories. Rather, they incorrectly forecast what would happen with *sales*. It’s simply nonsensical to be saying that volatile inventory growth affects GDP.

          (2) Does this sound like Hasset is saying inventories “fell less than businesses wanted them to fall” as you were forced to describe it? He wrote:

          In other words: If we see inventories piling up, firms may need to adjust their future activity downward. On the other hand, sales sometimes jump unexpectedly, driving inventories below their desired levels. When that happens, we can expect firms to ramp up production to replenish their stocks.

          Since 1970, there have been nine quarters, like the last one, when GDP grew by at least 3 percent and inventories accounted for at least half of that growth. The history of those quarters is hardly a favorable sign of what is in store.

          Inventory spikes make for blowout quarters. In the nine quarters with such spikes, the average growth rate was 6.6 percent and the average inventory contribution was 4.4 percent, even higher than what was observed for last quarter.

          No, either Hasset was unaware that inventories actually fell, or he decided his readers were too stupid to understand second derivatives.

          To repeat (everyone, not just talking to Transformer here), the way the BEA does this, it’s not simply a first derivative, which would be weird on methodological grounds (with inventories “causing” things to happen to real output) but somewhat sensible.

          But no, what they do is say that an inventory change did such-and-such *2 quarters ago* to the real GDP growth rate 2 quarters ago, and so if we’re now reporting what happened 1 quarter ago, the “contribution” from inventories is relative to the change that inventories caused 2 quarters ago.

          • Transformer says:

            1) Inventory size and trend are clearly important in business planning. Its not just a question of sales hitting target – suppose (for example) bad weather prevents you producing as much as planned so inventories run down and lead to an expectations of a GDP spike in the next qtr?

            2) I agree that the casual reader of Hassett’s piece would come away thinking inventories had actually risen – but one example of slightly sloppy journalism (I think the overall point he is making is sound) doesn’t render the whole concep absurd does it ?

            • Transformer says:

              At the risk of being annoying: I looked at how the numbers are calculated and concluded its not that complicated once you have got your head around it:

              Say GDP consists of 2 things

              – Final sales
              – Change in Inventories

              GDP increases by 10% from $10T to $11T. This Is an increase of $1T.

              If this is made up by

              – Change in Final sales: +$800B over last qtr
              – Change in “Change in Inventory” +$200B over last qtr

              Its easy to see that “Change in Inventory” contributed 20% to GDP growth.

              (I think the confusion may arise because the change in “Change in Inventory” can fall qtr on qtr even when inventories are actually rising)

              • Transformer says:

                or 2% of the overall 10% growth.

              • Transformer says:

                Ugh, starting to lose confidence that I got this right now…Let me do a sanity check.

              • Bob Murphy says:

                Transformer, please do review all this and tell me your final answer. I liked your idea of rendering it as

                GDP = final sales + change in inventory


                GDP growth = change in final sales + change in change in inventory.

                That does two things:

                (a) Makes the conventional approach seem understandable.

                (b) Still allows for the nutty news reports and glib interpretations of the BEA reports which were angering me.

              • Transformer says:

                (In case anyone cares)

                I looked at the numbers:for last quarter:

                A. % GDP change=0.1

                B. $ GDP change=4.3B

                C. $ change in change in inventory=24.3B

                By taking (C/B)*A you get 0.57 ( which is what the BEA says is the contribution to the quarterly change in GDP). This is consistent with my understanding described above that I again think is correct.


              • Bob Murphy says:

                Great this is cool Transformer. I think I will write this up for my next Mises CA post.

              • Transformer says:

                It seems non-intuitive that inventories can increase in this quarter by the same amount they increased in the previous quarter but add zero to this quarters GDP growth rate – but I validated that in excel and it checks out.

              • Bob Murphy says:

                It seems non-intuitive that inventories can increase in this quarter by the same amount they increased in the previous quarter but add zero to this quarters GDP growth rate – but I validated that in excel and it checks out.

                Are you saying that’s necessarily true, or just that it could conceivably be true? Because now I’m getting all mixed up again.

              • Transformer says:

                What I find non-intuitive is this:

                Sales = 7
                Increase in inventory = 1
                GDP = 7 + 1 = 8

                Sales = 10
                Increase in inventory = 1
                GDP = 10 + 1 = 11

                Between Q1 and Q2 GDP increases by 3 (11-8).

                This 3 is entirely made up of the increases in sales (10-7).

                So even though inventories increased by 1, this doesn’t count as part of increased GDP (because the change in the change is the same).

                Am I missing something here ?

              • Bob Murphy says:

                No you’re not missing anything. Suppose spending on final consumption were held fixed at $900 billion, and spending on inventory were held fxed at $100 billion. Then GDP would be $1 trillion forever, period after period. Every year, the economy would create $1 trillion in new output, of which people would consume $900 billion and $100 billion goes into bulking up inventories.

                So GDP growth is 0% forever, even though inventories grow by $100 billion each period.

                Do you agree?

              • Bob Murphy says:

                And in fact, if I’m understanding it correctly now, your result is general. If spending on inventories (which is the same as inventory growth) is unchanged, then inventories can’t “contribute to” or “subtract from” GDP growth, right?

              • Transformer says:

                I see why I found it non-intuitive now:

                I was thinking of inventory as a stock but that is irrelevant as far as GDP is concerned.

                All that matters is the flows into and out of that stock and if the flow stays constant then GDP can be constant too , even if the stock of inventory is growing bigger.

              • Transformer says:

                And yes to your question.

  2. Major-Freedom says:

    Murphy, great post (and Mises article. I especially liked the 2010-2012 example in the table).


    So inventories have been on a significant (real) rise recently.

    Just want to point out that inventory accumulation can be, and often is, the result of textbook ABCT malinvestment. When the prices of capital goods and consumer goods rise at a sufficient enough rate (meaning not necessarily absolutely, but relative to costs of storage) what otherwise would not be profitable (accumulating inventory, hold, and then wait for price to rise before selling) becomes profitable. Resourves get devoted to accumulating inventory for the purposes of profiting off the mere holding of it as the prices rise sufficiently relative to costs).

  3. Michael Tew says:

    I see how the way discussions about GDP are framed in a way that leads to sloppy thinking about the subject. But isn’t this more a problem with bad reporting and a muddled understanding of what GDP is?

    It seems to me that a way to salvage the discussion is to be precise about what we’re talking about when we talk about changes to GDP. Say decelerations in growth instead of making it sound like there were absolute decreases. Unless, of course, there’s something inherent about GDP that creates the confusion, which is possible. I can’t help but think, though, that a measurement of production in an economy is a useful concept, despite the many flaws of GDP.

    • Michael Tew says:

      I think I better understand your point after reading the Bloomberg article. Hassett talks about inventory *causing* fluctuations in output, which is just a really weird way to think about it. I think you would say that the way GDP is discussed gets people thinking that accounting definitions actually have real world impacts. Like you say in your Mises article, it’s similar to thinking that consumer spending “is responsible for” economic growth.

  4. Michael Tew says:

    Couldn’t every argument for using a non-inventory adjusted figure also be used to argue that we shouldn’t include Investment as part of GDP? Isn’t inventory just another form of the stock of capital?

  5. grief says:

    The Commcerce Department makes very clear that they report “CHANGE IN INVENTORIES”, not inventories. The problem here is you not knowing what they report, not GDP accounting. The commerce department also reports national income which nearly matches GDP (spending is someone else’s income).

    • Bob Murphy says:

      The Commcerce Department makes very clear that they report “CHANGE IN INVENTORIES”, not inventories. The problem here is you not knowing what they report, not GDP accounting.

      Actually what is relevant is not the change in inventories, but the change in the change. That’s why an increase in inventories (a positive change) in the 1q is blamed for reducing GDP growth by 57 basis points. So it’s you who doesn’t know what’s going on.

  6. Bob Murphy says:

    Let me make sure everyone knows what is going on here: You wouldn’t be able to say “the change in inventories subtracted 57 basis points from real GDP growth in the 1q” just by looking at the change in inventories from 4q 2013 to 1q 2014. There isn’t enough information to make such a determination.

    That’s why my example in the Mises.org article is so relevant. Using the way the BEA analyzes this stuff and breaks down the “contributions to” and “subtractions from” GDP growth, you could have a scenario where inventories were $0 at the beginning, $0 at the end, and yet the BEA would say that inventories subtracted from GDP growth that period.

    • Transformer says:

      In your old Mises post you have an imaginary pundit saying “Sure, the BEA and the press are running around celebrating the ostensible doubling of real output in 2012. But if you dig into the numbers, you see that fully 100 percent of the growth is attributed to the $1 trillion acceleration in private inventory investment.”

      He means that even though inventory investment was 0 in 2012 this is $1T more than in 2011 (when it was -$1T), which happens to exactly match 2012 growth. In your contrived example his comments don’t appear to make much sense (though I suppose he is making the valid point that the changes in inventory growth allowed GDP to increase with no actual increase in final spending)).

      But in the real world surely breaking out the contribution of inventory changes may allow us to make useful predictions about the future.

      – In the case of the old post: inventories were likely higher than the level that business would have liked so there will be downward pressure on future GDP growth.
      – In the later case inventories are lower (I think) than desired and will probably lead to higher future GDP growth.

      While the calculations concerning inventory contribution are a bit confusing why are they misleading ?

  7. Benjamin Cole says:

    Egads, GDP. And pollution. And crime. The City of L.A. had terrible crime and pollution 35 years ago. Now huge reductions in both. A citizen is much, much richer but it is not measured. And adjusting for inflation? Over extended time frames, the inflation adjustment begins to look nutty.
    And take two like nations, but one spends 25 percent of GDP on security guards and police and the other is crime free. Per capita GDP could be equal but not living standards, obviously.
    What is really crazy is the nutty fixation on inflation. Since it is a rough guess why is the Fed obsessed with 2 percent vs. 3 percent?

  8. Matt M (Dude Where's My Freedom)) says:

    ” The City of L.A. had terrible crime and pollution 35 years ago. Now huge reductions in both. A citizen is much, much richer but it is not measured. ”

    Perhaps it is. Presumably, reductions in pollution were due to regulations which held back industry from producing more (would be reflected as a reduction in GDP), while reductions in crime were due to increased taxation and spending on police departments (would probably be reflected as an increase in GDP, even though we all know how dumb that is).

    There are tradeoffs for everything, you know?

    “Since it is a rough guess why is the Fed obsessed with 2 percent vs. 3 percent?”

    You can say this about virtually every statistic and percentage the government concerns itself with.

Leave a Reply