Which production factor gets destroyed in major recessions, part I?

(This is my first post at Core economics, cross-posted with ClubTroppo)

There has been much talk in the last 12 months about the relationship between macro-economic theory and explanations of the current recession. Krugman essentially dismissed most current macro theory as being delusional about the workings of major recessions. His major argument (which goes back at least to Stiglitz in the mid 1990s) is that, within Real Business Cycle theories, major recessions are at their core viewed as mass holidays.

The ‘holiday’ view of recessions essentially arises from the fact that most macro-models take a perfect-market view of the aggregate economy and boil down the complicated machinery of GDP creation into a smooth production function which usually only includes Labour, Human capital, Physical capital, and Technology. As soon as you realise that the factories have not been bombed, that no-one has shot the workers or de-educated them and that people haven’t forgotten how to use the internet, then you are by necessity forced to say that any large reduction in production must have been because workers decided to go on holiday. Of course, you can redefine any of these ‘basic’ production factors to mean ‘the rest’ in which case you can tautologically say changes in it explain everything, but that kind of window dressing is ultimately useless.

In this blog I simply want to pose the question of which additional production factor we would have to think of to augment our models with, such that we get a more palatable story of what happens in major recessions, whilst still remaining within the confines of a production function view of the economy. Let’s look at what must roughly be true of this mystery production factor X:

– It must be easy to destroy X and hard to build up. If it wasn’t true that X was easy to destroy, then one couldn’t have a major reduction in GDP because the other production factors don’t really take a hit during recessions. If it was easy to build up again, then recessions should be over very quickly and one should be able to return the aggregate economy back to the path it was on previously. We know this is not true and that it has, for instance, been argued that the Great Recession of the 1930s really only ended in the Second World War (see here for support and here for a paper with a contrary view).

– X must have something to do with the utilisation of labour. This is because we know that the utilisation of labour quite closely follows the downturn and subsequent upturn, just as in this recession the GDP downturn was very quickly translated into losses of jobs and losses in labour participation. In the latest recession for instance, the US Department of Labor estimated the gross job losses totalled 7.4 million in the first quarter of 2008, whilst the US Department of Labor figures showed a loss of 533 000 jobs in the month following October 2008, the biggest drop since December 1974. For more, see here.

– X must have an element of a negative externality about it. If this weren’t true, then one would be forced to arrive at the absurd conclusion that people knowingly destroyed their own X and accepted the huge loss of income stream associated with it. There is no believable story that would make individuals inflict the kind of income loss that we see in recessions on themselves. Hence, to some degree, the reduction in X must be due to the actions of others and the destruction of the X of others might well be due to our actions. X must therefore be two-sided in that it is not something that would arise in a Robinson Crusoe economy.

– If we take the stylised story of this financial crisis at face value and accept that things like banks can be ‘too big and integrated to be allowed to fail’, then X has to have something to do with the other production factors being ‘integrated’. It must also be relatively easy to make up stories that tie the making or destructing of X to what happens in the financial sector.

– X must make internal sense as a production factor. This means it must cost resources to build up, that investments into it are in some sense visible (even if they are not yet measured by statistical agencies), and it must actually be associated with production. Hence things like ‘trust’ or ‘confidence’ do not qualify because they don’t actually directly involve the production and sale of goods. A lack of trust makes it hard to organise production and sales, but is not itself a production factor. Trust might be involved in the cost of building X, but it doesn’t make sense to call it a direct production factor in itself. In a pure command-and-control economy for instance one, in principle, needs no trust between people at all to have a reasonably high GDP. This of course does not preclude the possibility that the start of a recession is a dramatic change in things like trust which might affect the costs of making or breaking X.

Is it worth saving the production function approach at all, you might ask? Shouldnt we simply give it up as a bad job? I think it is worth saving, because the production function approach is the most obvious way to interpret GDP and growth regressions, and forms a logical basis for expanding the set of economic macro-variables the statistical agencies look for. It is also the easiest way to teach students about the macro-economy because it is nice and compact. What is hence wanted are reasonable candidates for ‘X’ and a model that convinces the profession it forms a palatable answer. We need an X to ‘save’ the production function approach to macro. Any ideas?

Author: paulfrijters

Professor of Wellbeing and Economics at the London School of Economics, Centre for Economic Performance

15 thoughts on “Which production factor gets destroyed in major recessions, part I?”

  1. One production factor that is always ignored in macro production functions is ‘raw material’. Convenient to ignore because then you can do value-added stuff but there is quite a lot of evidence that raw materials/intermediate inputs are used in different ways across countries. It’s probably not your X though.
    But do we need an X? Prices are an alternate mechanism. The production function is a physical relationship between production at the micro/firm level. During a recession the capacity to produce doesn’t really fall; as you say the factors don’t disappear. What happens is that the demand for the product falls and so the marginal product of each input falls and so firms choose to employ less of them unless their prices fall too. Of course it’s not as simple as I’ve made it sound there, but I am a mere microeconomist.


  2. Provocative first post Paul.
    I am no macroeconomist, but do we really have a production function approach to macro? The AS curve is based on productive factors (at least partly) but the AD curve isn’t. Can’t we depart from potential output from rigidities, say new-Keynesian wage rigidities, information stickiness or even misaligned inflation expectations leading to a Fisherian debt-deflation spiral? None of these really have anything to do with production functions.
    So isn’t your question more tightly targeted at saving RBC theories (or even Austrian malinvestment theories)? From what I learned of RBC, I would not miss its passing.


  3. maybe i am missing something but …real production can fall sharply in response to many unexpected shocks a big loss of wealth, a pandemic, a tidal wave, a big change in relative prices (eg an oil shock) etc

    in the current recession, a combination of mispricing of credit/risk, excess supply of houses, animal spirits etc has impacted both ss and dd


  4. Great post, Paul.
    Simon Baptiste- You also make a handsome point.

    A student of mine asked at the beginning of the semester “We in Thailand have plenty of labour; loads of capital; and our skills aren’t bad. Why does this production function say that we’ll be on the production frontier, yet despite our endowments, we’re in the middle of a recession?”

    Of course the “mass holiday” approach to labour supply is one answer—and an unconvincing one.
    What if your X took a Hicks Neutral (as in, augmenting all factors of production equally) `technology’? I think that could work if you:

    coupled it with short-term complementarity between labour, capital, and intermediary goods,
    made the labour supply indivisible, and
    make capital rents and quantities fixed at an ex ante efficient price/quantity (implying that contracts on rent, etc, are binding)

    I think you’d then get what you want.

    Picture a bicycle store. The owner smells a recession coming along (ie. a `technology shock reducing the marginal product of all inputs’), and reduces the two inputs he can-labour and intermediary goods. If labour is Hansen indivisible (people want 36 hour/week jobs), then there is involuntary unemployment.
    Unfortunately, he is bound into a 5 year lease, so he can’t actually reduce his floorspace or number of tools. So he has some unemployed capital in his business-a couple of empty storerooms, on which he is still paying rent, which was set at a previously efficient price.
    The expectations (technology) shock then turns into a lasting shock through lower wages/consumption & investment. This is the “quick collapse in output”.

    The “slow rise in output” would come through firms’ allowing capital to depreciate to the point where is becomes efficient to re-hire labour & intermediate goods. Or, for a more complex function, by allowing the inputs to become substitutable. An easy way of doing this is use a stacked CES production function where the exponents are time-varying—allowing complementarity of inputs at first, and substitutability in the medium run.


  5. I think this is a really interesting question, and the suggestions that have been made have been pretty successful at honing in on where the answer lies. As Simon B. and Matt C. mentioned, demand is key here. However, this then raises the question (from the production function point of view): what caused demand to fall, when it is a function of income, which is output, which comes from production? One answer to that , as J. Savage points out, is “expectations”: demand depends on these, and these are notoriously fickle. In fact, as Matt C. points out, these can lead to sunspot equilibria (at least, under certain conditions). There has to be some truth in this line of reasoning.

    I would also like to suggest another line which, I also believe, may contain some truth, ans which also reflects some of the comments made above. At the risk of trying your patience, I would suggest two other factors, which are interrelated. The first has to do with networks: supply chains of intermediate products and professional relationships that are difficult to build up, and which can be destroyed relatively easily through failing links. Through this process, small shocks to one link can bring down entire chains, at least, in the short run.
    The other factor is rent-seeking. If the regulatory environment is not set right, this can lead to labor resources being diverted from productive use to purely redistributive use (or, even worse). Moreover, homo economicus, being who he is, will mine this vein if the returns are high. They are typically high in areas where things seem so complicated that only the best and brightest can see the light and garnish the ill-gotten gains. This can pull the B&Bs away from where they are needed: in those supply chains. With the draining of this talent, eventually, the day comes when a link fails…


  6. PS: I should mention that this line of argument draws on the work of Debasis Bandyopadhyay, who explicitly inserts the fraction of agents devoted to non-rent-seeking behaviour in the production function, as a component of TFP. This corresponds roughly to Paul Frijters’ X.


  7. Good first post.
    I don’t think its a production driven effect though, the capacity to produce is still there. Its almost certainly demand driven.
    Steve Keen would say that “X” is debt driven demand. Every now and then it takes a hit as people worry about it (or rates go up) then they gradually forget and keep borrowing.


  8. Let me first respond to the ‘demand’ explanations. Demand is undoubtedly a valid perspective, but it sits very uneasily within a production function approach, unless one redefines production to mean ‘sold and consumed output’ and translates demand factors somehow into a production factor of its own. Ian King’s suggestions of supply chains comes very close to that kind of view. Ian, care to elaborate a bit more on your supply chain idea, i.e. do you think they fit all the criteria (assuming you agree with those) and can you think of any way in which one could say that major recessions indeed had disruptions in them? Savage, let’s take your point further, which I think has some obvious truth to it. Suppose indeed someone can be stuck with capital because of prior contracts. The logic of the productoin function approach is that that capital is a resource that would costlessly and immediately find another user who would be prepared to pay for its lease. Hence in order to make your argument ‘stick’ you have to invoke other processes into the equation, or else somehow capture the difficulties of ‘re-connecting’ a piece of idle capital into a new production factor.


  9. In the supply-chain approach, X would represent, somehow, the efficacy of the network, which may grow over time, but which can be subject to (sometimes catastrophic) failure when particular links fail. The probability of link failure is a decreasing function of the number of “best and brightest” (B&B) employed in that link. When rent-seeking (as an alternative occupation) has high returns, (say, due to a lax regulatory environment) this reduces the fraction of B&Bs that seek employment in the link structure. Thus, poor quality regulation leads to more rent-seeking, and a lower (expected) value of X (and, so, lower TFP).

    Looking at the criteria:

    This X fits the criterion of easy to destroy and hard to build: destroying one link can cause a more widespread failure of many links, which are costly to re-build.

    It also fits the criterion of having something to do with (skilled) labor.
    The negative externality criterion is met by the fact that a B&B that chooses to switch from link-building to rent-seeking will have a negative (expected) effect on all of the others in the link structure that are connected to his link.

    The “integration” criterion is clearly met by the link structure itself.
    The “production factor” criterion is met by the fact that the links are costly to build and maintain.
    Modelling all of this, of course, is tricky. But I would imagine that the serious computational guys (and gals) wouldn’t have much trouble with this.


  10. Paul, your criticisms are interesting.

    In real life, we know that most capital which becomes unemployed in an expectations shock is not immediately re-rented by another party. Taking the example of the bike-shop again, though they’ve fired some workers, and bought fewer bikes to sell, they still have to rent the couple of extra store-rooms. Even if the store-rooms are put up for lease, some other firm must have a higher marginal product of those rooms than the rental price.

    What if we thought of the probability of any given firm being able to substitute away from that capital (and so rent it out) as a stochastic process, where the probability of being able to lease out the unused capital in increasing in capital-market efficiency/flexibility?

    From the practical-modelling point of view, I think that by imposing a Calvo-timing-of-sorts on the elasticity of substitution of capital to labour/intermediaries (with heterogeneous firms) could give you the desired effect.

    This would mean that after an expectations shock came along, and firms had fired their workers, firms would, with some probability, receive offers on their unused capital. I’m not sure whether the capital hirer would have to be endogenised, but I suspect not.


  11. The production function gives the equilibirium level of production, but a recession is fundamentally a state in which the economy is out of equilibrium. This is why the function does not give the correct answer during a recession – you must wait for equilibrium to return. J Savage has touched upon this with the unused storerooms – that capital is currently not being used optimally, but if we wait long enough, eventually it will be (and the recession will be over).
    You could trivially “fix” the production function by introducing factors alpha, beta, and gamma, each in the range 0..1, that represent the effective proportion of Labour, Human capital and Physical capital that are in optimal use (where 1.0 represents the equilibrium state). This seems only marginally better than the “useless window dressing” approach you mentioned, though.
    For a rigorous approach that should actually give you some predictive power (or at least better explanatory power), you would need to take a leaf from the Engineer’s book and move from static analysis to dynamic analysis.


  12. Ian,

    any ideas on how to measure your candidate (which is close to ‘doctorpat’ candidate on clubtroppo)? Also, supposing you are right for a moment, how would this X be destroyed during a recession, i.e. what decisions lead to this and are those decisions optimal. If they are, what do they get out of it and how does the externality work?

    kme: the production function apporach does not need to be interpreted in an equilibrium framework at all: you can simply interpret it as giving you the total output given production factors at any point in time (in recession or not), but if you do that then the set of production factors needs to be large enough to fit what you observe output to do. That is what the post attempts to explore. The Keynesian argument you raise (i.e. recession is a state of disequilibrium) is one where you can think of knowledge limitations, etc., that mean markets dont clear in the usual way. The minimum-level production function approach abstracts from most of that, which to some extent gives it its power. Dynamic equilibrium is another issue entirely.


  13. Hi Paul,

    Measuring X directly would be difficult under this interpretation, because it represents something relatively abstract: the vibrancy of the supply chain network. However, according to this story, it should be negatively correlated with the fraction of agents who choose careers rent-seeking. This is, in principle, easier to measure — maybe the fraction of university students who choose accounting. ?

    Recessions may be partially caused by a thinning out of the links, which makes the whole system more susceptible to (partial) collapse when one link fails (due to, say, the only truly competent person in a small organization choosing to leave). This decision may be privately optimal, but may have significant negative externalities — both within the firm and outside it (the firm’s customers, who rely on the firm to supply them with their intermediate goods).

    I’m making this stuff up on the fly, in answer to your questions — so it’s not really a coherent theory yet.

    Also, please note the views here are not necessarily those of the FEC. Accountancy is actually a splendid career, full of interesting opportunities. ?


  14. Hi Ian,

    its an interesting idea to equate the density of ‘links’ with the absense of a large number of rent-seeking opportunities. Intuitively, the direction could go either way, i.e. ask yourself when you need more links? When things are more specialised and complicated. When are there more opportunities for rent-seeking? In similar situations: when there is so much specificity that one can corner and maintain asymmetric information, incorrect beliefs.
    I like your notion of individually rational decisions leading to a thinning of networks as an externality. Do you have any examples of this last recession in mind?


Comments are closed.