Growth Data Are Theoretical Constructs

February 1, 2010

by Mario Rizzo

The reader of newspapers may be confused about the newly published growth figures for the US economy. We are told that the growth rate for the fourth-quarter of 2009 was 5.7%. But when the inventory effect is taken into account it was only 2.3%. What is going on?  

Consider a simpler case in which the inventories that firms hold are unchanged. Let’s us assume that firms intentionally leave them unchanged. They desire to continue the current level of inventories. At the same time, they produce 5% more stuff for the market (perhaps because they foresee a temporary increase in demand). Then real resources, including labor, must be used to accomplish this. Production has increased.  

Now consider a case in which inventories intentionally rise by 2%. Firms produce more but do not sell the output but they are expecting to sell it later.  

Put both cases together and the production of stuff will increase (roughly, if everything is proportional) by 7%.  

Fairly straightforward.  

But now imagine a world where inventories have been intentionally declining as firms use up existing production to sell, but are not confident enough in future demand to produce more real stuff. Now because of a change in expectations, these firms reduce the rate of decline in their inventories. But as a substitute for even more inventory decumulation, they produce more real stuff. 

The first part (the slower rate of inventory decumulation) is a function of what firms expect future conditions to be. In itself, it involves no current increase in real stuff. Only the associated increase in output does. The latter part of the aggregate statistic is what most of the public is thinking about – increased output produced by using factors of production.  

The other part (decreased rate of inventory decumulation — not inventory accumulation) is an expectational component. It is an artifact of the expectations of firms that future sales will be better than current sales. They could be wrong. Nevertheless, it is not about now, but, at best, about later – more exactly, the expected-later.  

I am not saying that this component of the aggregate “growth” figure is meaningless. But it is not the same thing as, say, intentional inventory accumulation. Real stuff is not being produced now to the extent that inventories are falling, but at a lower rate.  

There have been other historical cases where GDP growth had been dominated by changes in inventory decumulation followed by only weak growth. So the connection between reduced rates of inventory decumulation and future growth in real production is not automatic, by any means.  

To summarize, these data are not brute facts. They are an aggregate of empirical observations collected on the basis of a theory of what these observations portend for the future.  

Therefore, I prefer that the GDP growth without inventory change be the headline figure, with the currently-touted figure being the secondary or supplementary figure. The former is closer to what most of the public thinks the combined figure is.  

More Technical Addendum: One of the complications in the analysis of inventory changes is the distinction between intentional and unintentional changes. Inventories could rise because firms believe that sales will soon rise and they want to be prepared (intentional). Or they could rise because sales are unexpectedly low and inventories pile up (unintentional). 

Something similar is true for changes in the rate of decumulation. Suppose we are coming from a period in which firms intentionally allowed their inventories to fall because sales were falling. For a time, sales continued to fall and so inventories continued to fall at a particular rate. We can then imagine a later time when these firms believe that sales will fall, but at a lower rate. So then they intentionally allow their inventories to fall also at a lower rate. 

In the current circumstances why would firms both let their inventories fall at a lower rate and increase their production of real stuff? 

First, we are dealing right now with aggregate figures. Some firms may be letting their inventories decline at the lower rate but not increasing real output; some may be actually adding to inventories, not decumulating them.  

However, one can imagine a firm both decumulating at a lower rate to satisfy current demand which is still not that great and raising production to satisfy expected future demand (that cannot satisfied by existing inventories alone).  

If this lower decumulation is intentional that is because demand is expected to pick up relative to the declining trend of the previous months. Inventories are held in the expectation of a certain rate of sales. 

Increased production of real stuff – using resources now – is also in anticipation of a certain (higher) rate of sales. There will be an effort to restore inventories to an appropriately higher level. This is the so-called inventory “bounce.” It is a temporary phenomenon. 

(If, on the other hand, the lower decumulation is unintentional, it would be because the firms were confronted with demand falling more than they had expected. This does not seem plausible as a general phenomenon right now.)  

The key questions are: (1) to what extent is the reduction in the rate of inventory decumulation based on overly optimistic expectations? If so, inventories will be higher than desired later; (2) to what extent might the increased production also be based on overly optimistic expectations? If so, this will also wind up increasing inventories later, unintentionally.  

Such mistakes could be generated by uncertainty (“noise”) with regard to both fiscal and monetary policy or simply as a consequence of uncertainty.  If either were the case, the factors constituting the fourth-quarter’s growth will be offset to the degree necessary, once the mistakes are realized, by decreased production of real stuff.  

So we must stay tuned. But that is exactly my point. The “facts” are not what they may seem.

6 Responses to “Growth Data Are Theoretical Constructs”


  1. This is a tough post, but I think’s its essence is in the title: the “facts” of macroeconomics are theoretical constucts. When econometricians are doing their work, they are “testing” one theoretical construct against another.

    For decades, a British civil servant overseeing Hong Kong would not permit the collection of macro statistics. He knew that doing so would just lead to calls for intervention. Hong Kong remains the freest economy in the world.

  2. Bob Murphy Says:

    Mario, you say the public thinks the GDP figure is closer to the figure without inventory adjustments. But I don’t think so. I would guess most people think the headline GDP figure refers to how much stuff was produced in the period, and that is what the headline figure (with inventory adjustment) refers to, right?

  3. Bob Murphy Says:

    Mario wrote:

    The other part (decreased rate of inventory decumulation — not inventory accumulation) is an expectational component. It is an artifact of the expectations of firms that future sales will be better than current sales. They could be wrong.

    OK Mario, I’ve read your post twice now, and I think we finally have a case where I genuinely disagree with you. So one of us will come out of this looking foolish, and boy I hope it’s not me…

    If I understand you correctly, you are saying that the actual production of real goods and services didn’t really increase 5.7% from 3q to 4q, that the actual level of production only rose 2.3%, and the remainder is due to an extrapolation from getting inside the heads of inventory managers.

    If that’s what you are saying, then I think you are wrong. I believe the 5.7% figure is perfectly correct–the actual amount of new stuff coming off assembly lines, i.e. resources being transformed into finished goods and services, was 5.7% higher in 4q than that same figure was in 3q.

    This is why I hate the way the press reports the “without inventory adjustment” figure. The inventory adjustment “contributes” to GDP in the same fashion that depreciation “contributes” to it. But obviously is more machines wear out in a period, that doesn’t “cause” GDP to increase.

    It seems you are very close to agreement with me on this, but yet we are reaching diametrically opposite conclusions.

    Try it this way: What Krugman (correctly) is saying is that the increase in final demand was only 2.3% from 3q to 4q. But final demand isn’t the same thing as output, and in this post it seems you are trying to pin down the correct way to measure real output.

    If so, that’s the headline GDP figure, not the non-inventory-adjusted figure.

  4. Mario Rizzo Says:

    Jerry,

    I really do think that the enormous collection of macro statistics that we now have is an independent factor contributing to intervention. Sure, the interventionist mentality and the specific intellectual frameworks used call for such data. But once we have it the excess nervous energy of many economists will find some (bad) use for it.

  5. Harry David Says:

    Prof. Rizzo wrote:

    >>(If, on the other hand, the lower decumulation is unintentional, it would be because the firms were confronted with demand falling more than they had expected. This does not seem plausible as a general phenomenon right now.)
    <<

    Is this correct? I'd think that if a firm's inventory decumulates at a lower rate than it expected, this suggests the firm had made an error of excessive _pessism_. It anticipated rapidly declining sales and instead found sales in fact did not fall so rapidly. Am I missing something?

    Also, in your comment to Prof O'Driscoll you suggest that the stock of accumulated statistical data, though it had been developed earlier as a tool for policymakers and intellectuals then, may tend to serve others in their plans in subsequent years.

    It might be potentially useful to think of statistical data sets that government bureaus put together as part of the overall structure of capital. The data is multi-specific (the data's implicit assumptions make it useful for some inquiries but not for others); its value is contingent on the circumstances of time and place (the people in a position to make effective use of it); it's heterogeneous, in that different data sets are suited to different enterprises and plans.

    from this perspective, macroeconomic statistics are intermediate capital goods in the provision of governmental services (for good or for ill) and the statistical bureau's cognitive beliefs about, say, inventory decumulation, are part of the technology of government.

    (But I may be making a trivial observation here)


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 1,731 other followers

%d bloggers like this: