Bad news on GDP vs good news on employment

Here is a macro exam question:

Q. You are the New Keynesian governor of an inflation targeting central bank. Two bits of information arrive simultaneously: GDP is lower than you expected; employment is higher than you expected. Relative to what you had otherwise planned to do, how do you respond to the news? Do you loosen monetary policy (because of bad GDP news), or do you tighten monetary policy (because of good employment news)?

I think I would tighten. Taken together, the two bits of news tell you that labour productivity is lower than you had expected it to be. Take an extreme case: suppose labour productivity unexpectedly halved. Each worker produces half as much output as before; each unit of output takes twice as many workers as before. To keep inflation at 2%: will employment need to double; or will output need to halve; or something in between? Well, there's a labour supply curve out there somewhere, and it probably isn't very elastic. Doubling employment would mean an extremely big increase in real wages to get twice as many people willing to work, and it would be a very strange (though not theoretically impossible) halving of average labour productivity that would be compatible with a very large increase in equilibrium real wages. It is much more plausible that the new equilibrium requires employment to stay roughly the same, and output to halve. So if GDP news and employment news give conflicting signals, the central bank should respond to the employment news and not the GDP news.

Or, you could break the question up into two parts: A. What does the news tell you about whether the current actual rate is above or below the natural rate? B. What does the news tell you about whether the natural rate has increased or decreased? You could answer that the news tells you that the actual rate has been below the natural rate, but that lower productivity growth implies that the natural rate will be lower in future than in the past.

It would be nice to wait for confirmation from the CPI news. And ultimately, that is what inflation targeting central banks eventually do. But you don't have the luxury of a crystal ball that tells you what the CPI will be. You can only target the forecast of CPI. And it would be foolish to ignore the information contained in the GDP news and employment news.

Exam questions always oversimplify the real world. But the real world question facing the Bank of Canada next week looks quite a bit like that exam question. Here's Statistics Canada. (One complication is we have April GDP data but May employment data. A second complication is that the data are never perfect, because GDP data get revised, and employment data is from a sample. And the employment news isn't really very good, but it's not bad like the GDP news.)

And remember that bit in the question about "Relative to what you had otherwise planned to do,...".

Comments

[Sandwichman]

If I was the New Keynesian governor of an inflation targeting central bank, all day long I would biddy-biddy-bum.

[Sandwichman]

(relative to what I had otherwise planned to do)

[Stephen Gordon]

I also think a negative productivity shock is how we should be looking at the most recent data. The drop in resource prices has sharply reduced our terms of trade, and that's best interpreted as a negative supply shock.

Inflation has been creeping up over the past few months, too. Just what we'd expect from a negative supply shock.

[Nick Rowe]

Stephen: that makes sense to me. But I keep on forgetting how to get my head around your beer and pizza example.

[Stephen Gordon]

How about as an oil price shock with the signs reversed because we're an oil exporter?

[Robert]

Does it matter in NGDP terms? Aren't I just swapping some real GDP for a bit more inflation? Or was that your point?

[Nick Rowe]

Stephen: Take an extreme case. Suppose we produce only oil, export all of it, and import apples. Then the world price of oil falls, relative to apples. If our output of oil stays constant, our GDP measured in oil stays constant, but falls measured in apples.

[Nick Rowe]

Robert: I'm assuming the central bank targets 2% inflation, not NGDP. (This is not a post about why NGDP targeting is better than inflation targeting; at least, I don't think it is.)

[Stephen Gordon]

The drop in GDI has been pretty large. For GDP, I'm also thinking of transition costs associated with a structural shift.

[Chris J]

Nick, I followed the link tobthe Stats Can page. We are badly missing the inflation target. 0.9 < 2.0.

Using short words for a non economist, surely the central bank goal ought to be to get inflation up. Yes or no?

[Nick Rowe]

Chris J: yes, it is targeting 2% inflation. But it takes time for its current actions to have much effect on inflation. So what matters is not last year's (May 2014 to May 2015) 0.9%, but where inflation will be (say) July 2016 to July 2017. And what it needs to do now to ensure the answer to that question is "2%". Was that 0.9% just a temporary blip due to the drop in oil prices, which may not continue to drop? And how the news on GDP and employment affects the answer to that question.

[david]

You should indicate your concerns over labour productivity in your May press release (remaining appropriately ambiguous in your medium-term intentions), then circulate a memo to suggest that the Labour Department to ease off on programmes pushing the marginal low-income into the workforce, and to Citizen & Immigration to loosen up on highly-skilled immigration visas.

Or else, as you know, you'll tighten in deference to government policy. You'll survive an election, you're an independent central bank. They won't. If they send back a memo politely declining, then you get the knives out, but you might not have to get to that.

Philosophically, it's your job to target inflation, but it's also the government's job to engineer the marginal labour productivity equilibrium; they might not realize they need to do something (or otherwise you'll do something). It's not clear whether it's your role to validate their decisions (or lack thereof), or whether it's their role to validate yours, but in a de facto sense the central bank gets to decide whether there'll be a recession, and causing a recession so that the government takes the hint only damages your political independence.

[Mark]

I think you're right if you're restricted to only employment. If I could choose a single labour indicator I would choose the unemployment rate.

Practically, I believe the BoC weighs the (1) UR, (2) wages and (3) employment together to estimate whether the labour market is tighter/looser than expectations, with much more emphasis on (1) and (2). Best if all taken together given the unreliable nature of the LFS.

[jonathan]

I think you are correct, though one must be careful when thinking through general equilibrium.

The only step in your reasoning that I am unsure of is that, since (presumably) C falls, households will increase labor supply, and so an increase in employment may not require an increase in wages.

You are implicitly assuming that any outward shift in the labor supply curve is small. I think this is reasonable, particularly under the (plausible) interpretation that the negative productivity shock is temporary. In this case, C demand wouldn't fall that much, which fits the observed fact that Y doesn't fall 1-for-1 with the fall in Z. This would imply only a small increase in labor supply, so that the increase in labor demand would raise wages, leading to inflation and implying that the central bank should raise r.

Here is another way of answering the question: Given that this is a temporary fall in income, households will try to borrow to smooth consumption. This is a reduction in desired savings, and so in the flexible price equilibrium r should rise. The Fed is trying to replicate flexible prices, and so the Fed should raise r.

Here is a third way: w = MPL at the natural rate. MPL has fallen, and so w must fall. Assuming upward-sloping labor supply, N must fall when w falls, and to accomplish this r must be increased (to push down demand, and therefore N).

(I think all these arguments are equivalent.)

[Nick Rowe]

jonathan: Yep.

Here's perhaps the simplest way to think of it: Put N (employment) on the horizontal axis, and Y (output) on the vertical. Draw a production function, Indifference map, and competitive equilibrium at the tangency. Now shift the production function downwards, and look for the new tangency.

Like the diagram in this old post, only shifting down not up.

W/P will probably fall. (But if you try very hard to rig it right, you could get W/P to increase.)

N could rise or fall (depends on income vs substitution effects, as you say). Thinking historically, we haven't seen a big change in N as the production function has shifted up over the last few hundred years.

Is it possible to rig it so that Y actually increases???

The benchmark is that Y falls, and N stays about the same.

"Here is another way of answering the question: Given that this is a temporary fall in income, households will try to borrow to smooth consumption. This is a reduction in desired savings, and so in the flexible price equilibrium r should rise. The Fed is trying to replicate flexible prices, and so the Fed should raise r."

Good point. I had missed that one. I was implicitly assuming it's permanent, but it probably isn't.

(But what's this "Fed" thing? This is a Canadian blog!!!)

[Steve Ambler]

There has been a 50% drop in the relative price of an output that represents roughly 3% of Canadian GDP. That's a big supply shock. I think a 1.5 percentage point drop in potential output would be a conservative estimate since the output of the petroleum sector should go down along with its demand for intermediate inputs from the rest of the economy. Problem: the Bank's methods for guesstimating potential output are mechanical and produce a time series for potential output that is too smooth when there's a big supply shock like this. I predict that they will calculate a bigger change in the output gap than is warranted.

[Nick Rowe]

Steve: that sounds plausible to me, and is a nice clear way to think about it. But what happens to the GDP deflator?

[Reading your name makes me feel guilty, about not yet replying to your email.]

[Kathleen]

If Labour Productivity is defined as Actual Hours Worked divided by GDP then I would expect a fall in GDP to yield a rise in Labour Productivity for a constant Actual Hours Worked. Similarly an increase in Actual Hours Worked (i.e. increase in employment) for a given level of GDP would yield a rise in Labour Productivity. So I think you are describing an increase in Labour Productivity.

If you look at Canadian vs US productivity you will see US productivity rising when the US is in recession and Canada is not: 2000-2002.

[Kathleen]

Except that Labour Productivity is defined as GDP/Actual Hours. :)

[Nick Rowe]

Kathleen: yep! The reciprocal. Ah well, you fixed loads of my math mistakes in the past, but not this time (I think)!

[SvN]

Um....do any of you stop to consider published work on the reliability of initial GDP estimates vs those on employment? Or work on the problems of real-time measurement of productivity? (Yes, this discussion is killing me. Why do you ask?)

Employment estimates in most countries undergo only trivial revision. Initial GDP growth estimates undergo more substantial revision. In fact, the employment estimates are an important input into initial GDP estimates, but their importance fades as more and more reliable information comes to hand. That means that revisions to employment and GDP estimates are much less correlated than the underlying series themselves (which tend to be positively correlated over the cycle.)

I think the right answer to your exam question depends on the extent to which you think initial estimates of productivity growth tell us something about the future trajectory of the economy.

Can you cite even one paper that looks at this for Canada?
Anyone?
Anyone?
Bueller?

[Nick Rowe]

SvN: OK, fair point, which paper(s) would you recommend?

But wouldn't the answer to that econometric forecasting estimate depend on: what other information is included in the regression; what the BoC is targeting? Because, that dependence on what the BoC is targeting is, in my experience, totally ignored in (the otherwise similar) literature on whether core or headline inflation is the best predictor of future inflation. If the BoC is responding optimally to all information, to keep future inflation at the 2% target, then NO publicly available information should forecast future (2-year ahead) deviations of inflation from 2%. And if it does forecast future inflation, that tells us nothing structural; it only tells us the direction of the BoC's systematic mistakes.

[SvN]

Nick:

You could start by looking at Anderson and Kliessen's work at the FRB St. Louis on productivity measurement and FOMC decision-making.
(http://research.stlouisfed.org/publications/review/06/05/Anderson.pdf)
(http://research.stlouisfed.org/publications/review/10/03/Anderson.pdf)
(http://www.palgrave-journals.com/be/journal/v47/n1/abs/be201136a.html)
You could then look at the Kahn-Rich model being used by the FRB NY to track US productivity growth in real-time.
(http://www.newyorkfed.org/research/current_issues/ci12-8.html)
(http://www.newyorkfed.org/research/national_economy/richkahn_prodmod.pdf)
Note the big swings in the perceptions of productivity growth 2008-2011....
Of course, I'd love feedback on the latest version of my work with Jan Jacobs of U. Groningen, but just read the first two pages to get an idea of the size of the problem.
(https://www.dropbox.com/s/era6jq7sbut2wea/JvNProductivity_July2015_JMacro.pdf?dl=0)
And, sadly, none of this looks at Canadian data!

But, for what its worth, I think this body of work suggests that the New Keynesian governor of an inflation targeting central bank should probably make policy on the assumption that data tell him very, very little about current productivity growth (relative to trend, or otherwise.)

Of course, if you know of empirical work that indicates otherwise, I'd love to learn about it.

[Nick Rowe]

Simon: come on now, you have to answer the exam question! Which bit of news do you respond to: GDP or employment? Tighten or loosen?

[Frank Restly]

Nick,

If I were the governor of an inflation targeting central bank I would realize there are three options available to me - tighten, stand pat, or loosen. If I were a politician faced with an election next year, only then would I constrain myself to attacking the problem versus letting the markets sort it out.

[Nick Rowe]

Simon: OK. I've now skimmed your paper with Jan Jacobs. It looks interesting and important for LR growth theory, but how does it help us answer this SR question? OK, if we see Y down and L up, we know that one of 3 things has happened: there's a fall in productivity; there's negative noise in Y data; or there's positive noise in L data. But unless the signal/noise ratio is 0%, we don't ignore it.

[SvN]

Nick: Correct, you don't ignore it. But you need to understand how to weight the two signals to come up with a coherent answer for your Neo-Keynesian model.

Of course, understanding how to weight them correctly might require heavier machinery....
(http://www.sciencedirect.com/science/article/pii/S0304407610002526)

Did I mention that, so far as I've seen, the characteristics of data revision can actually vary quite a bit from country to country (e.g. from one national statistical agency to another) .... and that no one has the data to do this for Canada?

[SvN]

Nick: Thanks for looking at the paper. Actually, I'm thinking it is probably much more important for things like fiscal policy (e.g. what are the chances that a country's debts are not sustainable?) and public pensions (think of the social security debate in the US) than for monetary policy. But if you want to bring productivity into current analysis.....

[Min]

My answer: Not enough information to make an informed decision.

[Min]

IIUC, the argument is that reduced labour productivity is inflationary, hence, the CB should tighten (relative to previous plans). Then suppose that the supply of employed labour remains the same, but the GDP drops. Then labour productivity has reduced, and the CB should tighten. Really?

BTW, if the combination of dropping GDP and increasing employment is inflationary, isn't that an argument for hiring low skilled workers when GDP is low and inflation is, too?