Sunday, April 19, 2015

The IMF on Investment since 2008

Vox today has a useful piece by five IMF economists on the behavior of business investment during and since the Great Recession. [1] From my point of view, there are three important points here.


1. The most important difference between this cycle and previous ones is the larger fall and slower recovery of private investment. This has always been my view, and I think it's an especially important point for heterodox folks to take on board because there has been such (excessive, in my opinion) emphasis on the inequality-consumption link in explaining persistent demand weakness.

This relationship between output and investment is consistent with previous recessions: 
business investment has deviated little from what could be expected given the weakness in economic activity. In other words, firms have reacted to weak sales – both current and prospective – by reducing capital spending. Indeed, in surveys, businesses typically report lack of customer demand as the dominant challenge they face.
In other words, the old Keynesian "accelerator" story explains the bulk of the shortfall in investment since 2008.


2. Historically, deviations in output and investment has been persistent; there is no tendency for recessions to be followed by a return to the previous trend. 

The blue line shows the behavior of output and investment in recessions historically, relative to the pre-recession trend. Note that is no tendency for the gap to close, as much as six years after the previous peak.

The authors don't emphasize this point, but it is important. If we look at recessions across a range of industrialized countries, on average the output losses are permanent. There is no tendency for output to return to the pre-trend. If this is true, there's no basis for the conventional distinction between a demand-determined "short run" and a supply-determined "long run." There is just one dynamic process. Steve Fazzari has reached this same conclusion, as I've written about here. Roger Farmer has just posted an econometric demonstration that in the postwar US, output changes are persistent -- there is no tendency to return to a trend.


 3. There's no reason to think that the investment deficit is explained by financial constraints. I should say frankly that the paper didn't move my priors much at all on this point, but it's still interesting that that's what it says. By their estimates, firms in more "financially-dependent" sectors (this is a standard technique, but whatever) initially reduced investment more than firms in less financially-dependent sectors, but as of 2013 investment in both groups of firms were the same 40 percent below the pre-crisis trend. If you believe these results -- and again, I don't put much weight on them, except as an indicator of the IMF flavor of received opinion -- then while tighter credit may have helped trigger the crisis, it cannot explain the persistent weakness of demand. Or from a policy perspective -- and the authors do say this -- measures to improve access to credit are unlikely to achieve much, at least relative to measures to boost demand.


Investment by sector

So these are features it might be nice to incorporate into a macro model -- investment determined mainly by (changes in) current output; a single system of demand-based dynamics, as opposed to a short-run demand story and a long-run supply-based steady state growth path; a possibility of multiple equilibria, such that (let's say) a temporary interruption of credit flows can produce a persistent reduction in output.  On one level I don't especially trust these results. But on another level, I think they provide a good set of stylized facts that macro models should aspire to parsimoniously explain. 

[1] The European Vox, not the Klein-Yglesias one.


UPDATE: Krugman today points to the same work and also interprets it as support for an accelerator story.

10 comments:

  1. " The most important difference between this cycle and previous ones is the larger fall and slower recovery of private investment. This has always been my view, and I think it's an especially important point for heterodox folks to take on board because there has been such (excessive, in my opinion) emphasis on the inequality-consumption link in explaining persistent demand weakness."

    I don't think the heterodox folks will be convinced. I know I'm not.

    The IMF says that the lackluster investment is a result of weak demand. Weak demand was the problem in the 2000s , alleviated for a period only by means of a huge debt bubble , it's a problem currently , and it's a problem going forward. This fits perfectly with an inequality/consumption thesis.

    Marko

    ReplyDelete
  2. I think it is certainly false that inequality led in any straightforward way to to higher debt; I think it is false that there has been a tendency for aggregate desired consumption to be lower at full employment than in the past; I think it's possible that higher inequality has been one factor among others contributing to weaker demand; and I think it's unlikely that it is the most important factor, and possible that it is not a factor at all.

    I realize that a lot of people believe that the inequality --> lower consumption demand story is central to recent macroeconomic history, and that to convince them I need to make a stronger empirical case than I have so far. We'll keep at it.

    ReplyDelete
  3. As one who (as you well know) is forever beating his spoon on the highchair about the inequality/consumption thesis, I'll just say:

    It may not be the most important driver of the (very convincing) weak-demand story you tell above, but it comports seamlessly with that story.

    Just as macro models should incorporate and address the "stylized facts" and explanations you lay out above, they should also (far more frequently?) incorporate the rather straightforwardly arithmetic Marginal Propensity to Spend Out of Wealth/Income effect.

    (I think that representative-agent models, by their very nature, are incapable of doing so -- you can't model concentration/distribution of wealth/income across a single agent. But would be interested to hear otherwise.)

    ReplyDelete
  4. Just as macro models should incorporate and address the "stylized facts" and explanations you lay out above, they should also (far more frequently?) incorporate the rather straightforwardly arithmetic Marginal Propensity to Spend Out of Wealth/Income effect.

    I agree that all else equal, a higher share of income at the top will imply lower desired consumption at a given level of aggregate income. In the real world, though, all else is never equal, and changes in income distribution may or may not be an important factor in any given shift in the consumption function. In the case of recent US history, I think the effect is largely swamped by (1) increased consumption of status-goods by the rich and (2) and increased share of social consumption. (The second factor is hardly discussed since the fact that an increasing share of consumption is collective doesn't fit any prevailing political narrative.) We also shouldn't ignore demographics.

    ReplyDelete
  5. I think that representative-agent models, by their very nature, are incapable of doing so -- you can't model concentration/distribution of wealth/income across a single agent.

    Agree. But this is like finding another reason why the moon isn't made of green cheese.

    ReplyDelete
  6. Stylized facts and vacuous interpretations
    Comment on ‘The IMF on Investment since 2008’

    Obviously your post lacks an underlying coherent theory. To make a long argument (2015) short, the correct employment equation for the investment economy is given here:
    https://commons.wikimedia.org/wiki/File:AXEC46.png

    The equation says that employment L increases with
    • investment expenditures I,
    • an increasing expenditure ratio rhoE (=C/Y),
    • an increasing factor cost ratio rhoF (=W/PR),
    under the condition of product market clearing if price P and productivity R in the consumption and investment good industry as well as distributed profit Yd remain unaltered in the period under consideration; it decreases in the opposite case. The testable equation explains unemployment.

    The income distribution affects the expenditure ratio rhoE. The dependency of the average expenditure ratio from wage income and distributed profit and their different expenditure ratios is given with eq. (30) in (2014).

    The employment equation above accounts coherently for both the effects of investment expenditures and distributional changes (and a bit more).

    It is important to note that distribution theory crucially depends on profit theory. It is very probable that your profit theory is false (again see 2014) and therefore your interpretation of the IMF data is pointless.

    Egmont Kakarot-Handtke

    References
    Kakarot-Handtke, E. (2014). The Profit Theory is False Since Adam Smith. What
    About the True Distribution Theory? SSRN Working Paper Series, 2511741:
    1–23. URL http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2511741.
    Kakarot-Handtke, E. (2015). Essentials of Constructive Heterodoxy: Employment.
    SSRN Working Paper Series, 2576867:

    ReplyDelete
    Replies
    1. Gerard Dumenil once told me something very wise:

      "The error of idealism is to confuse the production of theoory with the exposition of theory; the latter builds up logicallly from the simplest concepts, while the former cannot. Exposition can follow a deductive method, but the original development of theory cannot."

      Right now, I am trying to develop a theory. That is different from the exposition of an already developed theory. When you are trying to produce theory, you can't start from a formalization with axioms and so on. You have to start from a concrete reality.

      Delete
  7. "income distribution may or may not be an important factor in any given shift in the consumption function"

    I think more in terms of the "spending" function than the consumption function. C vs I is somewhat secondary. Also spending relative to wealth vs income; different measures might suggest more or less "important." Social spending: very good point.

    I have to go back to that Fazzari paper now, and your post on it...

    ReplyDelete
    Replies
    1. You are right that the C vs I distinction is arbitrary and probably not that important for the questions we are interested in. I still hope to convince you that the expenditure-income relation matters in a way that the expenditure-welath relation does not -- not because income necessarily plays a bigger role in expenditure decisions than wealth does, but because income-expenditure makes a closed loop with the possibility of positive feedback in a way that expenditure-wealth does not. From a macro standpoint, the point isn't that income affects expenditure more than wealth does; it is that income affects expenditure AND expenditure determines income, whereas the wealth-expenditure link is more one way.

      Delete
  8. I absolutely loved reading this post. It is informative, completely comprehensive and straight to the point.. intraday tips

    ReplyDelete