What if price-level targeting were already in place?

A popular research topic at the Bank of Canada over the past few years has been price-level targeting: committing to a path for the level of core CPI, not just to its rate of inflation. The idea here is that standard inflation targeting forgets and forgives past deviations from target. But under price level targeting, higher-than-target inflation will be compensated by lower-than-target inflation in subsequent periods.

There are several good things about this idea, including

• Forecasts for long-term inflation will become much less uncertain. Under inflation targeting, errors in future forecasts are accumulated; under price-level targeting, monetary policy cancels them out.
• It induces a stronger counter-cyclical element to monetary policy. When the price level falls during a recession, a price level target will imply an increase in the central bank's target for inflation, since higher-than-average inflation will be required to bring the price level back up to its target path. Real interest rates would be lower than what they would be under an inflation target.

(There's more to the analysis than this, but this should be enough to go on for now.)

The theoretical arguments all seem pretty sound to me, but I'm not quite sure just how it would work in practice. One approach that seems plausible to me would be to continually announce a series of short-term inflation targets designed to reach the price target path in the next two years. Two years is the outside range of the lags associated with the effectiveness of monetary policy on prices, and inflation forecasts are useful to people negotiating and setting prices. I have no idea if anyone is actually suggesting such a proposal, though; corrections from people who know more about this than I (you know who you are) are welcomed in the comments.

Let's also say that instead of renewing the inflation target in November 2006, the Bank of Canada adopted a price level target that grew at 2% a year. And let's also say that the Fed decided to do the same thing. What would the counterfactual look like?

Here is a graph of core CPI for Canada, CPI excluding food and energy in the US, and the target:

After July 2007, Canada was below the path, and the US was above it. So the implied short-term inflation targets were above 2% in Canada, and below 2% in the US:

When the financial crisis hit, Canada happened to be below its target, and the US was above, but my guess is that this this distinction wouldn't have affected their policy decisions. In the counterfactual,  the Bank and the Fed would have still gone to zero interest rates.

But there would be a difference in the effectiveness of going to zero. Real interest rates would have been something like 40bps lower in Canada: the US would be aiming for a lower inflation rate to get back down to  the path, while Canada would be aiming higher.

From what I've heard, the potential weaknesses of price-level targeting revolve around credibility issues. If an economy had the bad luck to stumble into a recession while it was above the path, the short-term effectiveness of monetary policy would be weakened by the lower-than-trend short-term inflation target. There would be an enormous temptation to simply reset the path in the recession and start over again.

In Canada, this counterfactual looks pretty good: real interest rates would have been lower than what the fixed 2% inflation target gave us. But it would have been hard to fault the Fed for questioning the wisdom of sticking to a policy that kept short-term inflation targets below trend for the first two years of a recession.

Comments

[Andrew F]

But then expected inflation affects actual inflation, and the US ought to have seen inflation rise. I'm not sure we can guess with accuracy how the US economy would have responded.

[Andy Harless]

You're presuming that Canada and the US would be -- and indeed implicitly were -- using the same underlying target. According to the Philadelphia Fed's Survey of Professional Forecasters, the 10-year expected CPI inflation rate for the US was more like 2.5% than 2%, so if the Fed was pursuing a 2% target for the CPI, it was not doing so in a credible manner (or else food and energy prices were expected to rise more quickly, but I'm assuming that's not the issue). My read is that the Fed has not been significantly above target at any time in the past 3 years and would not have had to question its policy for that reason.

Moreover, a reasonable price-targeting policy would imply a set of targets going into the indefinite future, not just to a 2-year horizon. It would be foolish for a central bank to be so obsessed with its short-run target that it ignores the implications for its ability to hit other price targets in the future. In the context of a zero or near-zero interest rate, a strong case could be made for aiming above the short-run target in order to facilitate hitting longer-run targets (that is, to reduce the risk of falling into circumstances where those targets would become unattainable).

In fact, I would strongly challenge, in the context of a zero interest rate, your statement that "two years is the outside range of the lags associated with the effectiveness of monetary policy on prices." Actually, I would challenge it more generally: I personally think the lag is often longer than 2 years, depending on what channels are most strongly mediating the impact of monetary policy in a given business cycle episode. (Surely the low interest rates in 2002-2003 were still affecting the US inflation rate in 2006, as their impact was mediated largely through a gradual increase in home prices.)

But even if we confine the ordinary lag to 2 years or less, that lag is based on a central bank that controls the real interest rate. When a central bank loses control of the real interest rate, it will (in the optimistic case where it can control the inflation rate at all) need to hold the nominal rate at zero for a longer than usual period of time to make up for the fact that it can't get the real rate as low as it would want to. (I'm assuming -- based on my reading of the empirical data -- that what matters is something like the integral of the real interest rate over time, so that if the instantaneous real rate is constrained to be higher than the optimum, you need to keep it at the minimum for a longer period of time.) This "build-up period" requirement could add significantly to the policy lag, rendering it considerably longer than 2 years.

And as I said, that's the optimistic case. For the US, in my opinion, the jury is still out on whether we are in that case. And ex ante, of course, the jury hadn't even gone into deliberations yet. Surely an observer sitting in October 2008 (particularly one who had studied recent Japanese economic history) should have recognized the risk of falling into the pessimistic case and therefore the risk that the Fed's actions at that point in time would affect prices for many years to come. From that point of view, the belief that "two years is the outside range of the lags associated with the effectiveness of monetary policy on prices" is simply untenable. If a price targeting regime were to be founded on that premise, then I would withdraw my support.

[Nick Rowe]

For comparing the 2% inflation target to a 2% price level path target, over any period long enough to matter, I think it's best to look at total CPI rather than core. The Bank of Canada does ultimately target total CPI. It uses core as an "operational guide" because it believes that most changes in non-core prices are temporary.

If we look at the time-path of total CPI since December 1995 (when the 2% target began), what is surprising is that there was next to no drift away from the 2% price level path. Inflation targeting "mistakes" (ex post) were negatively correlated (mistakes should be unforecastable at a 2-year horizon if the Bank had rational expectations). But in mid 2008 the CPI was a percent or so above the path, and only returned to the path at the end of 2008 (gas prices?).

On Andy's point: I think we need to distinguish two things:
1. What is the *causal* lag of monetary policy on prices.
2. What is the *targeting horizon* at which the Bank would want to return the price level to the target path.

Under inflation targeting, the Bank uses roughly two years for both. It believes that monetary policy has its maximum impact on inflation at a 2-year lag, and it normally wants inflation to return to the 2% target at around a 2 year horizon.

Under price level path targeting, I think the Bank would choose a longer than 2 year horizon to return to the target path. Maybe considerably longer. A very small reversion to trend can cure a lot of base drift for pension planning, etc.

I don't know how it would communicate under PLP targeting. Ideally it would just announce the forecast path of the CPI level. But since people are so used to thinking in first derivatives, plus the base=100 is arbitrary anyway, it would be tricky.

[Stephen Gordon]

Nick - I just checked, and you're quite right. A price-level target that started in January 1996 is only 0.1% higher than actual CPI as of January 2006 - almost exactly on the path. After 14 years, you have to start wondering if that negative correlation is an accident or the result of an unstated goal to get back to trend.

Maybe I should have titled the post "Is price-level targeting already in place?"

[Nick Rowe]

Stephen: Did you mean "as of January 2010"? And was that total CPI?

IIRC, I once ran the regression, and it's not an accident, at least at any reasonable confidence interval. (Not that I would place any confidence in my own econometrics!)

I think PLP targeting *is* already in place. Except it's not anchored in peoples' expectations, since there's no formal commitment.

There's a paper here somewhere.

[Stephen Gordon]

Oops - yes, January 2010, and it was total CPI. It's odd how that stylised fact has gone under the radar. I think I'll do a follow-up post with that graph - it's pretty striking.

[Patrick]

The G&M has an article on this today:

http://www.theglobeandmail.com/report-on-business/economy/at-bank-of-canada-a-debate-on-revealing-less/article1497021/

From the article:

"The central bank is putting considerable effort into studying an intriguing, yet untested, technique called price-level targeting. Instead of targeting the rate of inflation, the central bank would attempt achieve a specified increase in inflation over a period of time."

[Stephen Gordon]

I saw that, but I don't understand the bit where it's being sold as a way to introduce unpredictable movements in interest rates.

[Phil Koop]

Great post and even better commentary! You guys are complementing each other well here.

However: it is not too unusual to enter an oscillating state in a stochastic control problem with long lags. The inflation rate volatility is not so high that there is a big difference between the integral of the expected rate and the expectation of the integral over the time period you have studied. So you have some way to go before you can demonstrate that level targeting is already the de facto policy.

[David Pearson]

Andy seems to be saying, "anytime an overshooting of target coincides with the incipient creation (or the continuation) of an output gap, the Central Bank should abandon its target in the short term."

Take a look at the above statement in the case where the Central Bank continually overstates trend RGDP growth. In this case, the output gap will persist over time, and if the Central Bank's long term inflation target is credible (that is, if inflationary expectations persist in the presence of an output gap), then we will continually be in the state of the Central Bank abandoning its short term target while promising higher long term inflation.

So, my question is, what happens to NGDP and RGDP in the above case? And how easy will it be for the government to run high, structural fiscal deficits when the Central Bank is indirectly, if perhaps unintentionally, monetizing them continuously through QE? Lastly, what is the impact on long term inflation expectations of a Central Bank that is seen as monetizing structural deficits on a more or less permanent basis? (BTW, its not enough to say, "well, then the government shouldn't run such deficits": they will do so as long as unemployment persists).

It seems the success of PLP, or any other nominal targeting scheme, is highly dependent on:
-the Central Bank correctly estimating trend RGDP growth
-the perceived output gap serving as the ultimate guarantor of low inflation

The challenge for targeting proponents is not to argue how their idea can work, but to argue why it won't, and if that counterargument is rigorous and yet falls short of disputing the thesis, then we could have more faith in it.

[Andy Harless]

I have to alter David's paraphrase: anytime an overshooting of target coincides with the incipient creation (or the continuation) of an output gap large enough and expected to be persistent enough that the nominal interest rate risks being at the zero lower bound for a long period of time, given the intended future path of the price level, the Central Bank should abandon its target in the short term but continue to pursue subsequent targets in accordance with the original plan. One way to minimize the risk of encountering this situation is to set the implicit inflation target high enough that nominal interest rates are unlikely to hit the zero lower bound. (In principle, 2% might even be a high enough target to make this situation fairly rare.) Also, the central bank should, in general, abandon its short-term target if the horizon for the impact of its policy actions is expected to be longer than the horizon for the target -- but really in that case it should just have a longer target horizon in the first place (and I question whether 2 years is long enough, but that's a subsidiary point).

In any case, it is not really a problem when the bank temporarily abandons its target, even if it does so based on premises that turn out to be false and even if it does so repeatedly. That is the great beauty of price level targeting as compared to inflation rate targeting. If the central bank repeatedly overestimates the output gap, the price level will drift further and further away from its target path. Unless the output gap is increasing (or forecast to increase) over time, the need to pursue future targets will eventually outweigh output gap considerations, and the bank (unless it is willing to make output gap forecasts that become increasingly ridiculous) will be forced to stop abandoning its targets. (In particular, when the divergence from target starts to imply intentional deflation, it will become increasingly implausible that the bank would need to maintain a zero interest rate, as per the italicized condition in my previous paragraph, in order to hit the target.)

[David Pearson]

It seems you are saying there is no probable forecast error (and accompanying deviation from target) large enough to create a long term problem. But the issue is not necessarily the magnitude of the error, but the cost of correcting it. The cost ofcorrecting an asymmetric monetary policy (one which continually forgives overshooting but aggressively corrects undershooting of inflation) is complex and hard to quantify.

First, there is the issue of monetizing structural deficits, or at least the perception by the market that this will occur. This perception would raise term real interest rates, which would in turn perpetuate the perceived output gap and limit trend RGDP growth. Its important to note that the interest cost of servicing the Federal debt held by the public reached its lowest point in the last period: roughly 2.5%. What will happen to the deficit if that number grows to 5% in the absence of a strong recovery in RGDP?

Second, correcting a target overshoot in the presence of a persistent output gap is easier said then done for obvious political reasons. It is much easier to simply change the target then it is to jeopardize the banking system and risk deflation.

Third, uncertainty about the Fed's actions and inflation-fighting credibility depresses expected real returns to real assets, or at least increases their expected standard error. This, in turn, dampens investments and exacerbates the persistence of the output gap. Thus, if inflation overshoots, the more the Fed fudges the target, the more it has to fudge the target.

Again, none of the above is relevant as long as:
-trend RGDP growth is inherently, organically robust
-inflation expectations remain anchored in the presence of structural fiscal deficits

[Andrew F]

Is there such thing as a credible inflation target whenever there is a large, sustained cyclical fiscal deficit? I don't think it matters whether you're looking at inflation or PLP targeting. In either case, there will be incentive to fire up the presses and cause inflation.

[scott sumner]

Nice post. Here's my take. If everything you know about macro tells you that a higher level of AD would have been desirable in late 2008 and all through 2009, and if an inflation or price level target is telling you that you are too expansionary, and need to reduce inflation during that time period, then maybe inflation and/or the price level are the wrong target. Maybe you need a target variable that is consistent with what we believe needs to happen to AD. I have argued for NGDP. NGDP in 2009 was far below any reasonable target path in the US (and I believe Canada as well.) And NGDP was a bit too high in 1999-2000 and 2005-06. So it tends to give monetary policymakers the right signals, whereas inflation and the price level do not.

Any effective inflation or price level target will need to include the output gap. But if we are going in that direction, why not adopt a much simpler policy that won't confuse markets--an NGDP target path. One other advantage; you mentioned the bad luck to enter a recession when inflation is above target. One problem was that we had an adverse supply shock in early 2008, as oil prices spiked. This drove inflation above target. On the other hand supply shocks don't raise NGDP. In the first half of 2008 NGDP growth was less than 3%, so the Fed would have felt much more comfortable cutting rates in September 2008 if they had been focusing on NGDP. Instead they focused on inflation, and decided to stand pat. I'm sure even Bernanke would admit that a rate cut in September 2008 would have been better in retrospect, indeed a deep cut.

[Andrew Usher]

The Globe and Mail completely butchered its description of price level targeting:

The central bank is putting considerable effort into studying an
intriguing, yet untested, technique called price-level targeting.
Instead of targeting the rate of inflation, the central bank would
attempt achieve a specified increase in inflation over a period of
time.

An increase in inflation as the target? So the Bank wants to make the price level I(2) and inflation I(1) with positive drift? That sounds like a great policy... from Germany in the early 1920s and Zimbabwe in recent years, targeting an acceleration rate of inflation is very different from price level targeting which is essentially a move to make the price level trend stationary...

If the author intended to mean a one time increase in the target that is still very different from a move to price level targeting.

In the next graph:

While the method would theoretically give policy makers more flexibility, a major drawback is that even practitioners struggle to fully understand the concept, let alone explain it.

So the author does not understand.

[Nick Rowe]

I think Scott's got a good point above. It wouldn't completely eliminate the problem (if P were too far above target, even if real GDP were falling, it would still imply a tighter monetary policy). But switching to NGDP level targeting would reduce the likelihood of this happening.

[westslope]

Great post, discussion and yes, there probably is an article there.

Careful with the wording guys. Wouldn't want to inadvertently start a sect of Canadian central bank conspiracy theorists.

In addition to the issues of dynamic or time consistency, credibility of commitment, I would be worried about increased pressures to improve the data. I would also guess that shocks to price levels will tend to be asymmetric, causing price levels to rise in booms and occasionally sharply decrease during recession events. In terms of distribution effects, financially-incompetent people holding large cash savings will tend to do well in the event the real value of the currency increases.