« What is the link between unionisation and inequality? | Main | Population Apocalypse and Apocalyptic Populations »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Nick, a nice discussion of the evolution of your thinking. And it all sounds very sensible to me.

First, just to be clear, I have nothing against NGDP targeting. I just think it's oversold, especially in present circumstances (that is, if one subscribes to the view of AD shortfall owing to nominal rigidities, then PLT should be working fine and to the extent that NGDP is still below trend, this must be due to something else).

Second, as far as your "barking dogs" are concerned, I have two things to say: (a) some sensitivity analysis with your trend-cycle decompositions would be nice (start dates, in particular); and (b) there were other barking dogs in the room, namely the behavior of asset (real estate) prices (and not all barking dogs imply a role for monetary policy intervention -- there is a fiscal authority out there too.)

A very honest post that clarifies better than any other the concepts and logic supporting NGDPLT. The only thing that is a slight worry is the over-reliance on a single recent crisis to support these ideas.

I'd like to know that NGDPLT will also be optimal in the next crisis which may not resemble the financial, demand side problem we had last time.

What if the next shock is a supply one, a war that chokes oil supplies or a worldwide drought. What would NGDPLT do in a situation of "stagflation"? What would it do in situations of international imbalances, if every country adopts it or if only some countries adopts it? What would be the impact on currency wars?

This doesnt seem to address the original question:

"Seems to me that they are just asking for more price inflation and wishfully hoping that some of the subsequent rise in NGDP will take the form of real income.

Tell me I'm wrong (and why). "

In your link to the older post where you point out some graphs from Gordon you make the claim that the goal of IT is to prevent recession. It seems like your whole move away from IT is simply based on the fact that if the goal of IT is to prevent recessions and we get a recession then IT has been proved incorrect simply by contradiction.

But I think your assertion that the goal of IT is to prevent recessions is incorrect and I would like to see some citations from IT supporters who make this claim. The purpose of IT was to reign in the inflation of the 70's and 80's (before my time so I am vague on the exact dates). A specific type of recession, an inflationary recession.

Now the argument is being floated that because we are in low inflationary times we can move to NGDPT because it is not going to set off runaway inflation. Assuming this is correct (we dont need to get into a debate about austrian paranoia about hyper inflation) then the success of NGDP targetting is predicated on its ability to solve the problems faced of IT AND avoid potential pitfalls specific to the models of NGDPT.

This is why I do not support the move to NGDPT, its untested. The economy is too complicated to say that because IT didnt work its opposite, NGDPT, will. It seems to me that even if we can avoid runaway inflation the NDGPT models wont properly model the effects that a negative interest rate will have on the capital structure of the economy. I dont think the good old Ramsey, Kass, Koopman models I learnt in grad school can explain extended bouts of ZLB let alone what happens when the interest rates go negative. We still do not know if negative interest rates will be a panacea or if they will just get the economy's wheels spinning as the indicators look good on paper but dont translate into real wealth for labour outside the finance and banking industry.

In the past five years we have already seen a huge shift from capital investment to financial speculation as the ZLB incentivizes investors to leverage as high as possible and throw their high frequency algorithms into overdrive as they try to gain from every marginal speculative opportunity that is available. Even if we dont get run away inflation there is nothing that says that NGDP isnt creating asset bubbles that will dwarf what we saw in the housing bubble of 2008.

Like the original questioner I do not think that NGDPT will lead to SUSTAINABLE increases in real income and I do not think you or the other NGDPT supporters have shown us that we are wrong and why. Japan has doubled down with the NGDPT of Abenomics and it doesnt seem to be doing them any good at the moment.

David: thanks!

I understand you are on the fence re NGDPLT, just as I was. I think the biggest difference between us is that I'm more confident than you are that there are big nominal rigidities (even if I admist I don't understand them very well).

Yep, I was surprised to see that inflation (and the price level) didn't fall much at all for very long in the recent recession in Canada (especially if we look at core). Because theoretically (given my prior theory) a pure negative demand shock should have caused all three dogs to start barking in unison. And I don't know why. Supply shock, with the US recession being that supply shock? Very flat SR Phillips Curve? Some noise in the SR Phillips Curve? Longer lag than I thought? Some sort of Schelling focal point at 2%? Dunno.

My favourite econometric method is the interocular test. But yes, someone else could maybe do better with more sophisticated techniques and sensitivity analysis, but I really don't think it would change these results much.

In Canada, house prices dipped a bit, then kept on going up. So house prices recovered well before the recession. Targeting some broad index of nominal asset prices isn't an obviously daft idea (funnily enough, that was what I was thinking about in the early 1990's when IT got introduced and changed the game). But we can think of theoretical reasons why it might not work very well. And empirically, not all blips up and down in asset prices seem to match booms and recessions in general activity.

Fiscal policy: given standard arguments for Barrovian tax-smoothing, plus diminishing marginal benefits to government expenditure, I would be wary of distorting micro-optimal fiscal policy to make it do a job that monetary policy should be able to handle.

Benoit: thanks!

Yep. We only have 20 years experience with IT, and none at all with NGDPLT (unless we look at the past and say that in some periods it is *as if* they were doing NGDPLT, and even then expectations might not conform to NGDPLT, and the expected target is just as important as what actually happens). So the data alone simply can't tell us for sure what would happen. Theoretically though, we *think* we can argue that NGDPLT should handle most supply shocks better than IT. But it does depend on how you model how supply shocks shift the SRAS and LRAS curves. If they both shift left the same amount, IT does better than NGDPLT. But when I look at examples of supply shocks, it seems to me that the SRAS curve shifts left a lot more than the LRAS curve, which means NGDPLT does better. I can't think of any reasons to prefer either IT or NGDPLT in those other cases you mention. I think they are about the same.

Ian: I can only really speak for myself. If someone had told me, say around 2000, that IT was the best monetary policy, but IT wouldn't prevent recessions, and that NGDPLT would do a much better job of preventing recessions, I would have said he was nuts to prefer IT over NGDPLT, and wondered what the hell it was he saw in IT. And if he had convinced me that NGDPLT would beat IT on preventing recessions, I would have switched back then. It's like lowering the bar after the horse has failed the jump. If we call that "success", then give me "failure" instead. (OK, that's a bit too strong, because there are lots of policies even worse than IT, and IT was still quite successful compared to what went before, but we can do better.)

Every new policy proposal is "untested". IT was untested. We can only use what experience and theory we have, and try to interpolate between what we have seen before or extrapolate from what we have seen before. NGDPLT is not *the opposite* of IT. It's more like a halfway house between PLT and targeting RGDP. We have experience of IT, which is like PLT, and experience of the attempt to target RGDP.

One of the whole points of NGDPLT is to keep us off the ZLB, and that it is less likely we will need periods of negative real interest rates. That's how it works as an automatic stabiliser compared to PLT. But if we do need negative real interest rates, then we do. And it is easy to build a long run growth model in which real interest rates will need to be negative for some periods. Take an OLG model with shocks to cohort size, for example.


I am having trouble reconciling the following 2 statements:

"We can only tell the difference between those three policies when there are real (supply-side) shocks"
"Put it another way: I do not know why the IT and PLT dogs didn't bark when they should have barked"

Am I over-simplifying to believe that once you factor in real (supply-side) shocks then we do know why only 1 dog barked? In response to a supply-side shock then basic AS/AD shows how the price level may rise even as NGDP falls. Isn't this the reason why only the NGDPLT dog barked?

What do you think happens to non-financial capital accumulation once rates go negative? What effect will this have on our output? These are the questions I do not see answers to from the NGDPT supporters.

Ron: Perhaps I should have said "We can only *see* the difference between those three policies when there is some sort of shock, like a supply-side shock or the central bank making a mistake".

We can see the difference in 2009-present, because inflation and the price level stayed (roughly) on target, but NGDP fell a lot from the (implied) target. If there were some sort of supply-side shock, that would be one explanation of why only one dog barked. But there might be others. A supply side shock would be sufficient for seeing a difference, but it might not be necessary.

Ian: "What do you think happens to non-financial capital accumulation once rates go negative?"

That's a question you need to ask the IT and PLT guys too. And it's a bad question. Because the answer depends on *what caused* (real) rates to go negative. If it were a fall in investment demand, or an increase in saving supply, that caused the *full employment equilibrium* (natural) rate to go negative, then a monetary policy that made the actual (real) rate negative would *increase* the levels of capital accumulation, output and employment, relative to a policy which held the actual rate positive. (Draw a standard I and S diagram, with r on the vertical axis, with the I and S curves intersecting below 0%. Now ask what happens to actual I and S if r=0%. Remember that actual I equals actual S.)

Level with me now; confession time: you learned Ramsey, Kass, Koopman in grad school. Did you ever learn ISLM?

Very good and honest post. Only I think that IT and PLT dogs are even worse than you assume. It was some time ago when Tyler Cowen wrote a post where he was asking around why there is an inflation in Greece: http://marginalrevolution.com/marginalrevolution/2013/01/why-is-there-still-inflation-in-greece-model-this.html. It was a debate where many people weighted in but I came out of the debate none the wiser.

Until recently greek inflation seemed to be churning around comfortable 3% (truth to be told, it slowed a little more last year) but Greece did not go through deflation. It is pretty hard to believe that almost 30% unemployment is purely result of supply side shock. And yes, we know that NGDP is 20% below 2008 level and 40% below trend. Why is that? Can it be true that such a terrible NGDP performance is just a pure chance, that it just coincides with possibly the worst peacetime recession any modern nation had to endure? I do not feel that you can simply wave such a strong evidence away by saying that in certain class of models there is no help for Greece and possibly other nations that may yet face similar problems in the future.

PS: I am fully aware that Greece does not have control over monetary policy. But even if it had one, it is very likely that such a hypotetical CB would be very happy with its performance from the standpoint of IT and PLT.

JV: thanks!

Good point about Greece. We should probably make a distinction between LRAS and SRAS shocks. It is inconceivable that Greece's LRAS curve shifted 40% left. It is less inconceivable that a flattish SRAS curve might shift a couple of % vertically up, which would translate into a 40% leftward shift. (This doesn't explain it, but it makes it a bit less massively weird.)

I'm a Carleton grad, TA'd second year macro both terms but my ISLM is pretty rusty.

Ian: OK. Draw the I and S curves assuming that Y is at Y* (Y* is defined by the LRAS curve or the LR Phillips Curve). (Silence you Old Keynesians! I am *not* assuming full employment!) If they cross where r is negative, that means the IS curve crosses Y* where r is negative too. Call that value r* (AKA the "natural rate of interest"). If actual r > r*, then where r intersects the IS curve will be below Y*. What's happening is that Y must fall below Y*, so that the S curve shifts left (because S(Y,r)), until it intersects the I curve at actual r.

"in a world of nominal rigidities a monetary policy like that (IT) would cause undesirable real fluctuations"

not if you target a stable price level (0% inflation)

The lower zero bound is only being hit because returns across the board of all asset classes are declining as a result of insufficient medium of exchange and therefore demand is suppressed across the board. Banks need to move out of reserves into riskier assets in order to create loans and hence deposits. If they dont then the return on everything declines. Until banks lend there is no returns to chase, and until there is any attractive returns banks don't lend. We need to create a monetary policy mechanism that can circumvent the banks and break the loop.


Nick says, “given standard arguments for Barrovian tax-smoothing, plus diminishing marginal benefits to government expenditure, I would be wary of distorting micro-optimal fiscal policy to make it do a job that monetary policy should be able to handle.” My answers:

1. Obviously tax smoothing takes place, but the equivalent phenomenon takes place when interest rates are cut: everyone knows they’ll rise again in a year or two, thus the cut is partially ignored.

2. There is no reason to suppose the marginal benefits of government spending decline with increased supply any more quickly than is the case with private spending.

In fact there is no way of objectively measuring the benefits of government spending: all one can do is assume that if the electorate wants GDP split as between public and private spending in the ratio X:Y, then if GDP is actually split in that way, then the marginal benefits of each are the same.

3. Interest rate cuts pull investment forward. So that should work for a couple of years or so. But as of September 2013, rates have been low for more than a couple of years, thus I’d argue that low rates now have no effect at all. In contrast, there is nothing to stop fiscal measures working for decades on end.

How about a football match – monetarists versus fiscalists?

For practical reasons, I could never support a NGDPL target due to the problems related to GDP data i.e. it's released two months late, and it's subject to revisions. The CPI, meanwhile, is released sooner and is not revised. Policy-makers have enough issues to deal with, and I don't think they want to keep chasing a GDP level that gets revised for several quarters before finally stabilizing.

lxdr1f7: From my perspective, a 0% IT would have been even worse than the 2% IT we actually had:

1. The Bank of Canada would have hit the ZLB even harder, because it would have wanted to set rates at minus 1.75%, rather than plus 0.25%, in order to get the same real rate of interest.

2. If prices and/or wages are stickier down than up (as they might well be), a 0% IT is like putting ear plugs in the IT guard dog.

Ralph: You are still, I think, missing my point about fiscal policy. The tax smoothing argument means we don't want to vary tax rates over time. The diminishing MB of G means we don't want to vary G over time. (And if we did want to keep G/Y constant over time, that would mean *pro*cyclical fiscal policy: we would want to *cut* G in a recession.

Greg: Coincidentally, and off-topic, I was just thinking about a post on your CD Howe Google predicts recessions paper. Wondering if it means that recessions, at least in part, are self-fullfilling forecasts.

Back on topic: Just because CPI isn't revised doesn't mean it's accurate. It's still based on a sample. All it means is that StatsCan either can't or won't revise it. If we wanted to, we could tell StatsCan it isn't allowed to revise NGDP either, and that it only gets one crack at it. Or we could just target the first estimate. or we could just target the final estimate. Or anything in between.

The longer lag in releasing NGDP data, compared to CPI data, is a problem. But it isn't a big problem. The Bank doesn't target the next CPI release; it targets CPI inflation between 12 months to 24 months from now (roughly). The Bank could target NGDP at a similar horizon, if it wanted to. The only difference is that it would take 27 months instead of 25 months for the bank to learn whether it had missed its target.

The whole reason the Bank targets CPI at such a long horizon is that it is scared that if it tried to bring CPI back to target any more quickly then real output fluctuations would worsen. In other words, the Bank is already targeting some mix of RGDP and CPI for short horizons. Targeting NGDP would make that implicit target for the shorter term explicit. The Bank is already looking at NGDP data, and has to look at NGDP data, to implement flexible inflation targeting.

Greg Tkacz,
"For practical reasons, I could never support a NGDPL target due to the problems related to GDP data i.e. it's released two months late, and it's subject to revisions. The CPI, meanwhile, is released sooner and is not revised."

In addition to Nick's reasons let me add the following.

In the US the inflation target is the PCEPI. In months in which there are advanced quarterly GDP estimates released (in the US RGDP and NGDP is released simultaneously) the PCEPI for the last month of the quarter is typically released *after* NGDP.

For example the advance estimate of 2013Q2 GDP was released on Wednesdaay July 31. The PCEPI wasn't officially released until Friday August 2. It's worth noting if you're mathematically inclined its a simple matter to back out the PCEPI inflation rate from the GDP release before the inflation rate is officially released.

How does the US produce its NGDP estimates so darned fast?

It all comes down to the fact the US uses a different system for computing the national accounts than most of the rest of the world. We use the National Income and Product Accounts (NIPA) system and most of the rest of the planet uses the System of National Accounts (SNA).

There are three approaches to computing GDP: 1) production, 2) expenditure and 3) income. Have you noticed that only the US produces a measure of Gross Domestic Income (GDI)? GDI is essentially GDP measured by the income approach. Japan produces a separate income approach GDP measure but they still call it GDP. To my knowledge no other country using the SNA system produces a separate income approach GDP measure. Here is why.

In practice the production and expenditure approach GDP measures are equal. However there are differences between the income and the other two measures of GDP. In the US we release NGDP and RGDP at the same time because both are derived purely from the expenditure and production approach. GDI takes a month longer to produce, evidently because of greater delays in collecting income data. Note also that there is no separate GDI implicit price deflator.

However, SNA is very different (page 21):


“In addition, the sum of gross value added across the sectors of the economy in the SNA-based estimates does not equal GDP, because the statistical discrepancy is added to the sum of value added across sectors to arrive at GDP. The NIPAs have two measures for the value of final goods and services produced in the economy—an income-side measure and an expenditure-side measure—and the difference between these two measures is reported as the statistical discrepancy, a concept that does not arise in the SNA.”

In short what most of the rest of the planet calls “NGDP” is really equal to NGDP (i.e. nominal expenditures/nominal gross value added) *plus* the statistical discrepancy between NGDP and NGDI. Since most of the rest of the planet, like the US, produce income data a month later than expenditure and production data, it takes them a month longer to produce NGDP than RGDP.

Of course this begs the question, where does their GDP implicit price deflator, which also comes out a month later, come from? Evidently for most countries it is the quotient of NGDI and RGDP. (Yikes!)

Needless to say I am sure things are slightly more complicated than I have portrayed here, and that they vary by individual nation. But on the whole, as far as the sausage making process of constructing national accounts goes, I find the NIPA system a little more transparent and somewhat less stomach turning than SNA.

What about revisions?

Well, the US *does* revise PCEPI. Moreover for the purposes of level targeting there is no significant difference in terms of the size of the revisions.

There’s two main ways of measuring the size of the revisions of the components of national income and product accounts: 1) Mean Revision (MR) and 2) Mean Absolute Revision (MAR). For rate targeting MAR is the more appropriate measure, and in fact the MAR of inflation is usually smaller than the MAR of NGDP. However, for level targeting MR is more appropriate. And at least in the US (Page 27):

“The MRs for the price indexes for GDP and its major components are generally not smaller than those for real GDP and current-dollar GDP and its major components.”


In fact over 1983-2009 the MR for the final revision to quarterly NGDP is 0.14 whereas over 1997-2009 the MR for the final revision to the GDP deflator and the PCEPI is 0.20 and 0.12 respectively. And over time the revisions have trended downward:


So I suspect that the MR for NGDP is actually smaller than PCEPI over 1997-2009.

P.S. What about the fact NGDP is released quarterly and price indicies are released monthly? This is strictly a convention. Private forecasting firms produce monthly estimates of NGDP. There's nothing at all preventing government agencies from producing monthly estimates of NGDP.

"From my perspective, a 0% IT would have been even worse than the 2% IT we actually had:

1. The Bank of Canada would have hit the ZLB even harder, because it would have wanted to set rates at minus 1.75%, rather than plus 0.25%, in order to get the same real rate of interest.

2. If prices and/or wages are stickier down than up (as they might well be), a 0% IT is like putting ear plugs in the IT guard dog."

1. The preferences of banks are different to the preferences of the general economy. If the broader economy including non banks directly interacted with the central bank in monetary policy the demand for reserves would be alot less and the demand for other assets, repaying debt and consumption would be much higher. The failure of monetary policy is because of the imbalanced transmission mechanism.

2.In the short term many final prices aren't sticky downwards at all (electronics, interest rates, goods and services with high profit margins). These things send the index down. Of course intermediate inputs like wages are sticky in the short term. Asset prices are very sensitive to the economy and I think the inflation index should include at least housing.

The swap does not “pour money into the economy”.

The way I think of the difference between inflation targeting and NGDP targeting is that the former manages expectations about the future value of money/path of prices and the second manages expectations about the future level of income. Since the point is to avoid those transaction crashes we call "recessions", then we want to avoid having people confident about the future value of their money but worried about their future level of income. Because then demand to hold money might shoot up and spending might drop quite a lot and we might get a transaction crash.

So it is not enough to have people confident about the future path of prices/value of their money because they are not the only expectations that matter. Since maintaining the level (number and scale or price x output) of transactions is the point (whatever else money is, it is a transaction good), why not target that? Especially when debts are so sticky, giving people an extra reason to be nervous about their future level of income and making the effects of transactions crashes worse.

"So it is not enough to have people confident about the future path of prices/value of their money because they are not the only expectations that matter. "

But money is neutral so the only thing monetary policy can do is target prices I suppose. Set a neutral target (target a price level or 0% inflation) that wont affect peoples expectations. Peoples expectations will be determined by real economic factors if money is neutral.

"And if we did want to keep G/Y constant over time, that would mean *pro*cyclical fiscal policy: we would want to *cut* G in a recession."

I still can't wrap my mind around this idea. We want to *cut* G in a recession why? To create the expectation that G will rise in the future? I just don't see why reducing government spending when incomes are falling will be helpful to falling incomes.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad