Is Real Business Cycle Theory Dead?

No. Real business cycle theory is alive and kicking.  If we write Y=a*F(K,L) and call “a” technology then an RBC theory is mostly about how fluctuations in “a” change output.  Amusingly, Brad DeLong calls this the great forgetting theory of recessions and indeed it is hard to see how we could forget about technology, thus reducing output in some periods.  But this view takes the term technology too literally.

I am in my office every day (L=1), my computer is here every day (K=1) but my output and thus my productivity fluctuates.  Why?  It’s not that I forget how to use STATA.  Some days, however, a reporter calls and distracts, another day I need to tidy my office, on other days creativity just doesn’t strike.  In short, everyone recognizes that at a micro-economic level productivity fluctuates a lot so why should macro productivity follow a smooth process?

In fact, there is a standard answer to that question which is the law of large numbers–spread idiosyncratic productivity shocks across many firms and in the aggregate volatility will be low.  In an important paper, The Granular Origins of Aggregate Fluctuations, Xavier Gabaix takes on this answer with a simple but important point: large firms matter.

In the United States, for example, sales of the top 100 firms account for about 30% of GDP.  (The share is even larger in most other developed economies.)  In fact, we know from my GMU colleague, Robert Axtell, that firm size follows Zipf’s law.  As a result, large firms get larger in larger economies so that firm-level productivity shocks do not disappear in the aggregate even in large economies.

Gabaix shows theoretically that combining idiosyncratic shocks and Zipf’s law for firm size can produce significant fluctuations in GDP.  Empirically the difficulty is to distinguish aggregate shocks from firm-specific or sectoral shocks.  Using one plausible, but no doubt debatable decomposition, Gabaix shows that idiosyncratic shocks to the top 100 firms can explain about one-third of aggregate volatility.

The bottom line is that Gabaix has opened the way for a much richer real business cycle theory in which real shocks can be identified and tied to specific firms and through transmission mechanisms these real shocks can affect the aggregate economy.


"I am in my office every day (L=1), my computer is here every day (K=1) but my output and thus my productivity fluctuates. Why? ... Some days, however, a reporter calls and distracts,"

...then L does not equal 1.

"another day I need to tidy my office,"

...then L does not equal 1.

"on other days creativity just doesn't strike."

...then your production function may be stochastic, but not necissarily on a.

The fact that there are business cycles doesn't really imply for me that anyone has a good theory about it. We have history and variations, yes.

I watched Blogginheads yesterday, in which Alex Rosenberg & David Levine do their bit on Economics as Science. I thought the bit they missed (or what Alex R. should have put to David) was that as science develops predictions should coalesce. We shouldn't be satisfied when one or other economist says that answers, theories, equations, or predictions are "better." We should also see more agreement.

Based on my blog subscriptions, argument still prevails, especially about policy, often divided along ideological lines. We aren't seeing the agreement that a solid, scientific, theoretical framework would provide.

Been awhile since I was in grad school, but your description of the underlying premise of RBC theory is fundamentally unsatisfying. The notion that there are idiosyncratic risks that can aggregate up to a nontrivial shock at the macro level seems to imply a large degree of randomness in the macro business cycle. Now, this may be how the economy really works, but if this is the case then economists have no business even bothering to model the macroeconomy because it is essentially driven by noise. I mean seriously, can't we do better in understanding the macro economy than to say, "Oops, those guys at Exxon were just too lazy this quarter"???

Aaron K,

I don't know why you would say that L != 1 when Alex is cleaning his office or talking to reporters. Well, actually I do know why but I think you're confused.

In a simpler example, say we have a laborer who produces usually 800 widgets per day or 100 widgets per hour (assuming average 8 hour day). One day, he starts day dreaming and only produces 400 widgets. Some people would say that his hourly productivity level would drop in half. You want to say that he only "actually" worked 4 hours and that his productivity remained unchanged.

Unfortunately, labor productivity is just the amount of product that a laborer produces in a given amount of time. That laborer (like Alex) still worked 1 day, its just that he did not accomplish as much because he was distracted.

If we took your approach and tried to figure out when people were "actually" working, calculating average labor productivity would be much much harder than it is.

Large firms have the LoLN operating *inside* the firm. You may get a call from a reporter, but if you work for, say, Microsoft, every worker at Microsoft won't get a call from the same reporter on the same day. The productivity of a large firm depends on the productivity of a large number of employees aggregated together and except on the day of the company picnic, a lot of the variations in individual employee productivity will be independent.

The LoLN also strikes again over longer time periods - your productivity may be down today, but it won't necessarily be down tomorrow. In fact, many productivity reducers are negatively self-correlated over time. You may need to tidy your office today, but then you probably won't need to tidy it again tomorrow. The company picnic comes but once a year, not every day for several years until it brings a recession. Most people can only go to four unfeigned grandparent funerals and then they have run out of grandparents. Etc.

note to self: read post carefully before commenting

I'm unclear on how one distinguishes forgetting from writing down or from recalculation Can we name names and work out some examples? I propose:
1. When Lehman fails and any customers who traded with Lehman no longer have access to the same liquidity pool and thus pay higher bid asks for their smart, safe, plain vanilla transactions.
2. When a bank cuts their mortgage issuance staff by 50% and then has trouble training workers quickly enough to issue new, Fed supported mortgages
3. When a manufacturer opens a new plant with a totally new design and all their previously skilled workers are now unskilled
4. When a regulator bans a major form of securitization and all the traders/investors/ etc involved now have no reason not to forget the associated skills

Everybody who just read this post should also read this post:

Where does the gap in GDP come from? A large share is due to the auto industry and the buidling sector, right? But is it really decreased productivity that causes the drop in output in these sectors?

Probably not. There is a drop in demand, maybe a demand shock, something which does not exist in RBC models. This is why they are irrelevant and useless in light of current events.

BTW: What is productivity on an aggregate level? In order to compare produced apples with produces oranges, one has to weight them by prices. Thus, aggregate productivity cannot be something exogeneous as assumed by RBC theory.

I don't think "large firms" is the right explanation, but rather "correlations". Some correlations exists because of policy choices of large firms, some because of sectoral coordination, some because an individual's creativity may be linked to another's. The (strong) law of large numbers does not apply if the variables being summed are correlated.

Alex - "The bottom line is that Gabaix has opened the way for a much richer real business cycle theory in which real shocks can be identified and tied to specific firms and through transmission mechanisms these real shocks can affect the aggregate economy."

Super. I trust you have done that for the latest economic excursion and will be reporting the detailed verification soonest.

Chris's comment is wonderful because it shows how the coarse-graining could be done continuously with scale from individuals up to markets of the largest firms, and from there to the macroeconomy. The limits on the accuracy of the models are the flow of data about individual behavior and the ability to process it.

I think the correlations that Chris is talking about are best visualized as a time-dependent frequency spectrum of activity.

LoLN assumes uncorrelated (or slightly correlated) observations and shouldn't be over-applied to highly correlated situations. If the captain turns the ocean liner north, all the passengers move north.

Sure, but all the action at the macro level has to be occurring through those correlations then. If one manager makes a bad decision at a firm, the firm's competitors will rush in to exploit that weakness and maybe attempt a takeover of the firm. You have to explain the correlations in output shocks among firms and in this realm, RBC has no advantage over other theories.

Interestingly, in the financial world we do have a situation pretty close to every firm making the same lousy decision at the same time (to lever up and go long on real estate). Yet RBC cannot tell us why a negative shock in the financial sector should spread to other sectors.

Of course RBC is OK. There was a Great Forgetting of Technology Shock- in the Credit Assessment industry.

A different human-activity measure, Internet traffic patterns, were also long expected to show that kind of aggregation. Finally, though, somebody got around to checking on it, and it turned out to be wrong. What turns out to be true, instead, is that similarly-patterned randomness persists from micro to macro scales.

So the defense of RBC is that business cycles are driven by more or less noisy fluctuations (like not feeling creative or responding to a reporters call, but for firms) that aggregate somewhat because of relatively large firms. Seriously? That's your defense? And what about all the evidence in plain sight of housing bubbles and financial meltdowns?

The RBC is a very interesting theory and I tend to think that it's also true. Looking back over the fluctuations in business trends and development of technology we will see that indeed, this theory applies. If a pattern could be generated, there could be some interesting predictions related to business development in the future.Voip Phone Systems

Comments for this post are closed