Output gap, RIP

My immediate reaction to news that the ONS is going to raise its measure of GDP was: what does that do to estimates of the output gap?

In one sense, the answer is: nothing. The ONS is not so much revising GDP as redefining it; the main redefinition is that R&D now counts as part of GDP rather than as a cost of production. This means "potential GDP" is redefined precisely as much as "actual GDP", leaving the output gap unchanged.

But even so, this episode reminds me of my disquiet with the concept of the output gap.

One problem I have with it is that the estimates of potential GDP are parasitic upon our observations of actual GDP. For example, if the economy grows faster than we thought we might think "supply constraints are weaker than I'd thought so potential GDP is higher" or we might think "investment is rising and a higher capital stock raises potential GDP." Estimates of potential GDP tend therefore to be pro-cyclical.

This is dangerous - if policy-makers pay attention to the concept - because it might mean that monetary or fiscal policy become insufficiently counter-cyclical. In recessions, the fear that potential GDP has fallen could lead to low estimates of the output gap and hence a reluctance to ease policy: this might be because a low estimate of the output gap leads to higher forecasts of inflation, or because it leads to higher estimates of the "structural" budget deficit. Conversely, in booms, an over-estimate of productive potential leads to upwardly-biased estimates of the output gap and hence insufficient tightening.

My second problem is that the concept lacks microfoundations. Ideally, we'd measure the output gap by asking every company boss: by how much could you raise output without raising prices? But bosses might not know this. "Let's cross that bridge when we come to it" would be a reasonable answer. Their ability to expand production without incurring extra costs depends upon whether they can improve efficiency by learning perhaps very small and subtle improvements. But, by definition, we cannot know today what we'll learn tomorrow. This means the notion of "capacity" is an elastic and unknowable one. As this paper (pdf) concludes "capacity is not as well defined, even in batch-oriented manufacturing."

If people on the front line don't know what spare capacity they have, then economists trying to second-guess them won't know either. Unsurprisingly, then, estimates of the output gap vary hugely: from 6% (Capital Economics) to  0.8% (Fathom Consulting). This means that uncertainty about the output gap is huge. And if the noise-to-signal ratio about anything is high, we should discount that signal.

This is especially true because even if it were precisely known the output gap might not be much use to us.

One of its main purposes has been as an input into inflation forecasts; a big gap, ceteris paribus, predicts lower inflation. One big fact, though, tells us it's not necessarily a useful input. Inflation has fallen recently, even though the output gap is thought to have narrowed - from 4.4% in 2009Q2 to 1.7% in Q4 according to the OBR.

Of course, this might just tell us that things other than the output gap matter for inflation. But it might also corroborate what proper textbooks tell us - that in an open economy, there's a range of spare capacity consistent with any particular inflation rate.

The output gap, then, is ugly in theory and useless or worse in practice. As better economists than I have said, we should throw this daft idea into the rubbish bin.

Comments

[Luis Enrique]

I'm not convinced. Really it's just a way of thinking about capacity utilitization, I reckon it's pretty hard to purge economics of that idea altogether.

[chris]

OK Luis - but what does the output gap add to our knowledge that we couldn't get from unemployment or (say) the CBI's measure of capacity usage?

[Luis Enrique]

Chris,

I take your point. I don't know about the CBI thing, but unemployment tells us something about capacity utilization, and is probably quite well observed, whereas I imagine estimates of the output gap are model based, with very large model uncertainty.

[Ralph Musgrave]

Chris,

I don’t agree. The fact is that we just s*dding well HAVE TO make a stab at estimating the output gap (or much the same thing, NAIRU).

E.g. if you were on the Bank of England Monetary Policy Committee, your job would be to decide whether the economy could take more stimulus or not. Even if you just left interest rates unchanged, that is ipso facto a claim that there is no significant output gap.

[Ben Oldfield]

Working from the supply side is useless no company is going to produce more unless it has orders. So you must look at demand and what effects it such as unemployment and rise and fall in take home pay.

The other main driver of inflation is cost of supplies such as raw materials and therefore exchange rate.

[Metatone]

I think I'm with Ralph Musgrave on this one (which isn't regularly the case.)

Output gap - or similar measures - are a significant part of the policy decision. Now I might prefer that we junk it and use unemployment as the indicator instead, but that's politically unacceptable to the economics profession. They are ideologically committed to ignoring human capital whenever possible, so output gap it is...

[Ros]

The advance of the global supply chain has reduced the output gap's relevance, esp. for manufactured goods prices.

[Mark T]

Ah, but if we don't have the notion of an output gap and the implication that monetary authorities will set policy accordingly - or more explicitly with reference to its proxy, the level of unemployment - then how are the bond traders supposed to react every first Friday of the month when the non farm payrolls come out? Remember the fun we used to have when we though the UK government was targeting the trade data? All that breathless excitement to see whether the finger in the air forecast for a monthly figure (almost certainly subsequently revised)was one standard deviation above or below the quoted number - they even kept the markets open on Christmas eve specially one year. Sadly, now the data is routinely ignored. Would you deny them their last remaining opportunity to generate a lot of unnecessary noise and trading volume? For shame.