« Out-sourcing your own job | Main | How the quest for ratings distorts research »


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

Nick, a couple things:

" The only real argument against lowering the inflation target is the fear that we would hit the zero lower bound on nominal interest rates more often."

Akerlof and Shiller, in their "Animal Spirits" book would beg to differ. And that's one Nobel between them.

Also, can the BoC really dictate mortgage downpayments?

Adam: fair point. The Bank of Canada's research suggests that absolute downward nominal wage rigidity is less of a problem at a 2% or 1% target, but we know the ZLB can be a problem even at the existing 2% target.

The BoC cannot dictate mortgage downpayments, but the government of Canada can. And there needs to be some sort of joint body with all the parties involved, including the BoC and CMHC, to take stock of what countercyclical capital/reserve/leverage/downpayment tools are available, who owns those tools, and who should decide when they need to be used.

Hi Nick. I know nothing about monetary economics, but I am a student of how politics interacts with policy.

I have serious concerns about the credibility of a commitment to price level targeting.

Imagine a positive one-year inflation shock happens. How politically realistic is it that any government would allow the BoC to push i-rates so high as to push down the price level to its planned path? Pushing back to an inflation-level target generates a lot of political heat. (Ask John Crow or read any newspaper clippings from the early 1990s.) I have doubts that a commitment to pushing back to a price-level path is politically sustainable.

Which I guess is just another way of saying that I think voters may have a hard time distinguishing these policies, which is the point you made.

It was a point many made. Although oddly enough, the credibility issue was almost always described in terms of the opposite scenario: people questioned whether or not the Bank would really set inflation targets above 2% if there was a negative shock. Back when I was in grad school, the story was always told the way you did: central banks couldn't credibly commit to disinflation.

Another thing that was interesting was that the proposals for change invariably came from within the Bank; the response of the academics was generally "yes, we see the theoretical benefits, but they look too small to justify dealing with the realities of implementing these ideas." Pretty much exactly the opposite to what you might expect in an exchange between central bankers and academics.

I might also add that it was great to finally meet Nick. But the experience was a bit jarring: now when I read his posts and comments, it's with his speaking voice.

What made it especially hard to justify major policy changes was the knowledge that IT had been a greater success than anyone had ever dreamed it would be twenty years ago. As someone put it, "we have something to lose now". Back in 1990, we didn't.

Kevin: *if* price level targeting gets built into expectations, then theory says that nominal interest rates *may* need to vary less than under inflation targeting. If inflation goes above target one period, then people expect lower than normal inflation the next year, which raises the real interest rate for a given nominal interest rate. But, if there is structural inflation inertia, then it will take a bigger move in real interest rates to get back to a price level target than an inflation target, just because you have to reverse inflation, not just stop it. So, in principle, the sum of the two effects could go either way.

But, if price level targeting does not get built into expectations (and my guess is that it will take time for this to happen), then what you say will be right.

However, price level targeting with a very slow return of price level to target could still deliver the long-run benefits of a more forecastable distant future price level (for pension planning etc.) by eliminating "base drift" at low cost.

On the one hand....

Yep. It was great to finally meet Stephen! After all these years! Like "pen pals", in the olden days. (And Stephen has been working in French so long he almost has a French accent, though not anywhere as strong as my English (UK) accent!)

That's why I refuse to meet any of you - I sound like Charles Nelson Reilly.

Thanks for the report from Ottawa!

As someone said at the workshop, inflation targeting was "policy ahead of theory". The Bank started targeting inflation, and the theory of why it would work came later. If we had to design a policy on a blank slate, we might not choose the current 2% inflation target. But we are no longer starting from scratch. Plus, a lot of people have made a lot of plans based on the 2% target. It's a quasi-constitution.

"Even if a recession in the rest of the world caused a fall in aggregate demand in Canada, why couldn't the Bank have loosened monetary policy enough to offset that fall in demand"

I thought we were considering the Canadian recession a terms of trade technology shock on AS.
Prof Gordon's posts.

There was potential for transmission of financial damage, the ABS paper and the Caisse i think? but it didn't happen.
Has our bank reserves increased to the same extent as America's? IF not does the IOR explain it.

The deflation risk strikes me as huge.  Imagine that, like Kevin Milligan proposes, they suddenly have to produce 10% deflation following a 70s style supply shock.  Anybody want to commit right now that there is no chance that no matter how well people seemed to understand price targeting, that they won't suddenly get doubts and start anticipating a deflation spiral?

And those aren't the only risks.  You will drop long term rates by 2%, which will almost double land values/money supply.  And that's a lot more volatility with no apparent benefit.  The banks will love you, but do you really want that much leverage?

What are the benefits of (effectively) reducing the inflation target to 0%? The risks seem huge.

That's not to say, that falling expectations of NGDP or whatever. I know in my case, in the fall of 08, I didn't know any economics save for pop austrian. And I hit the blogs hard to figure out if this was going to be the next depression, or whatever and if I should withdraw from university in Jan. Plant a garden, survivalist ...... anyway. So what happened?

edeast: yes, if the rest of the world goes into recession, that will cause a decline in Canada's terms of trade. And that lowers our real national for a given GDP. But would that explain everything that has happened, like the increased unemployment, the disinflation? If the Bank had operated differently, and been quicker to loosen monetary policy, inflation could have stayed at target, and the real effects could have been smaller.

K: "You will drop long term rates by 2%, which will almost double land values/money supply."

Nominal interest rates would presumably fall by 2%, but real land values should depend on real interest rates.

What typically/theoretically happens with an adverse supply shock? I don't think that it explains everything. Our banks needed liquidity as well but I can't remember if they had the same mutual distrust/bank-run, as the states. Carney's recent speech. They certainly had less exposure to the shadow banking sector. Ok so what caused AD to go down in Canada, money demand? Excuse to link to a cool Bank of France paper, on the distribution of money demand. Fits in with the Gorton explanation, I think.
* ABS in previous comment should be Asset Backed Commercial Paper.

Does it matter whether everybody understands price level targeting? Or will there be transmission mechanisms whereby the knowledge of people who do understand it gets translated back to the realm that everybody else understands -- some kind of price or interest rate, or something? I'm going to start thinking about this.

Obviously there will be a time period over which people doubt the capability of the BOC to achieve its price level target. But if the BOC is capable, then the doubters will lose money and eventually come around. Once credibility is established it is not that complicated a subject: even if the average man on the street doesn't get it, it'll immediately show up in mortgage rates, union wage contracts, long-term commodity supply contracts, then residential gas contracts... it will trickle down.

Nick: What's "real land value?"

Why are the 2 choices "either stay as we are, with a 2% inflation target, or lower the target below 2% and move to price level path targeting at the same time."? Isn't there a third choice, "move to price level targeting, of a level that increases at 2%/year"?

Could you clarify what breaks if people don't understand the difference between inflation and price-level targeting? It seems like people should make roughly the same plans in either case, and that the errors from misunderstanding price level targeting won't cause any worse outcomes than plans under inflation targeting.


There are two main benefits of PLT over IT:

1. "Stabilising expectations": if AD starts to rise too high, the price level starts to rise above target, people expect lower than normal inflation in the immediate future, which raises real interest rates even for a given nominal interest rate, which reduces AD, which helps reduce the amount the price level rises today.

2. "Reduces long run price level uncertainty": under IT, temporary mistakes in inflation are never reversed. The price level follows a random walk, with the variance getting progressively larger as you move further into the future. With PLT, the price level is stationary (or trend-stationary. People will know in advance what their pensions will be worth. and can plan accordingly.

We lose those benefits if people don't know about PLT.

IT requires the Bank to stop temporary bursts of inflation. PLT requires the Bank to reverse temporary bursts of inflation. If there is inflation inertia, inflation really doesn't want to reverse direction, and it would take some real fluctuations in unemployment to make it reverse direction.

There is a third choice, PLT with a 2% trend. But in my opinion, I have real doubts people could understand it. It would be hard to explain.

K: "real land value" is the price of land adjusted for inflation. The real land value should equal real rents divided by the real interest rate.

jj: Yes, eventually people would learn, whether consciously or unconsciously, and we get to the rational expectations equilibrium. But it might take a generation. I'm not sure that my grandparents' generation ever understood inflation, for example. (Those born around 1900). There could be a long bad transition. If we were certain of the benefits of PLT, and knew we would never change our minds, it might be worth it. But we aren't that certain, and might change our minds again in 10 or 20 years, which means the investment in learning would be for naught.

edeast: my guess is that AD fell in Canada because: the wedge between the Bank's overnight rate and commercial interest rates increased; the stock market fell; world commodity prices fell; these caused expected future AD to fall which caused current AD to fall; and the BoC wasn't quick enough and aggressive enough to offset all these.

Nick: ok. But I dont know what "real rent" is. I understand real price changes and real rates. But I only understand your equation in nominal terms. How do I get a "real" rent (not a real change in rent) to divide by a real rate?

My practical point is that inflation may be stable at zero, but land prices will double if rates fall by half.

K: OK, lets do it in nominal terms. P is the price of land, and R is annual rent, and i is the nominal interest rate.

At zero inflation, the price of land is determined by P=R/i

Now assume 2% inflation. R is expected to grow at 2%. So the equation becomes P=R/(i-2%)

But i will also be higher by 2% if there is 2% inflation. So you get the same answer either way.

It's P=R/r where r is the real interest rate. P and R will both be growing at 2%, but in real terms (adjusted for inflation) they will stay the same over time.

Yup. I was trying to post a retraction but you got there first. Sorry and thanks, Nick.

First, I think I barely understand "price level path stability". I tried to find a nice explanation, but it appears that nobody explains it, they just discuss it!

But that lack of understanding prompts me to ask a question, that - as I think on it - matters a lot here. If my rough understanding is correct, as price levels move above a target - either static or moving - then the 'current inflation target' moves down. The idea is that expectations will build this change in - and I ask "what expected change, precisely, will be built in?". (The opposite direction can also be discussed, but I think my question can be fully addressed only looking at upward price level moves.)

If the price level moves 1% above the target, does this imply a reduction in the inflation target of 1%? Is it a particular Alice-in-Wonderland style "price level" - defined by the central bank, but not always corresponding to what we think we see elsewhere. And is the 1-for-1 up/down fixed and standard? Or might it just be "PL up -> target down", but the exact change depends on circumstances? And, after all this, we apparently still have to face the uncertainty we have today in thinking about how the reported inflation numbers will push large or small changes from the Bank.

Even if the way it works and the policy are understood, I find it easy to imagine that different observers of the central bank will build in differing expectations based on exactly the same observations. Right now, we can all agree that if the inflation target is 2%, we will know when we see inflation reports whether the result is above or below 2. But with price level path targetting, I can see that you and I might both see prices go up 4%, agree that the inflation target will rise - and then disagree on the expected magnitude of the target change, and also disagree on the expected magnitude of changes the Bank makes while pursuing the not-agreed upon target.

Chris S: I think you get it, and I think you also get one of the difficulties of communicating it.

Think of a trend path for the price level, where the trend path is rising at 2% per year. Suppose we start on the trend, but the Bank makes a small mistake, and the price level rises by 3% one year, so it's now 1% above the trend path. What does the Bank try to do next year?

Under inflation targeting, the answer is simple. The Bank aims for 2% next year. In effect, it shifts the trend path upwards by 1%, so the new path is parallel to the old.

Under price level path targeting, it leaves the original trend path unchanged, and tries to have lower than 2% inflation for a bit, to get back onto the trend path. But how quickly?

In communicating its policy, the Bank has to tell people the price level is 1% above trend, that the trend is upwards at 2% per year, but they will try to bring inflation down to trend over (say) the next 2 years, so it will be 1.5% next year, and 1.5% the year after, and 2% in the following year. Oh, and by the way, it's going to take 12 months for the squiggles in inflation over the last 11 months to disappear as well, so we expect it to be 2.2% next month, 1.7% the month after.....

Sure, gimme a nice clear graph, and a bunch of bright economics students who know this will be on the exam, and a good 30 minutes, and I reckon I could explain it all. But if I tried it out down the pub, I think they would be throwing things at me inside 30 seconds.

NR: I always like using historical data to put things into perspective.

If I was to look at historical CPI, from StatsCan, Sept 1990 CPI= 78.8, Sept 2010 = 116.9

So, 116.9/78.8 = 1.484 using IT (not sure when 2% target precisely started)

Now, if I was to go with a 2% PLT, I would expect (1+.02)**20 = 1.486

Now, assuming I'm doing the math correctly (could be a big assumption), I don't see a significant difference. So, what can I conclude? Maybe the BofC is already doing some form of PLT?

JVFM: That precisely was one of the puzzles, mentioned at the workshop. It depends a little on when you start, and which price index you use, (core or total CPI), but what we have actually had over the last 20 years of IT is almost indistinguishable from PLT. Why?

1. Positive shocks tend to get followed by negative shocks. It's just the way shocks are.

2. The BoC was secretly doing PLT, despite saying it was doing IT.

3. Sheer fluke.

4. Someone (closely connected to the BoC) told me another theory yesterday, about how interest rate smoothing would lead to this, but I didn't understand it.

It doesn't work in Sweden, by the way. Which supports the sheer fluke theory.

Maybe the answer lies in the briefcase.

Whatever the reason, the evidence adds weight to the "if it ain't broke, don't fix it" crowd.

JVFM: Yep. But on the other hand (there we go again), it says if we are doing PLT anyhow, either by accident or design, we might as well say we are doing it, so people know what to expect, and can plan better.

Well, that argues to some extent the point that Richard Fisher, President of the Federal Reserve Bank of Dallas made about regulatory certainty that Simon van Norden blogged about earlier: http://tinyurl.com/22qozxl

In terms of investment decisions, I just don't think that firms are that sensitive to whether the BofC is using IT or PLT (in the short term where discounted cash flow has its greatest effect).

Maybe a bigger concern in other parts of the economy that I am not considering.


This may be a dumb question, but since absorbing Scott Sumner's views on MP almost wholly, I don't understand this credibility problem much. Afterall aren't the markets smarter than the experts or the people? All the BoC would arguably have to do is gain credibility with the markets, and the rest will, well, be transmitted. What am I missing?

Contemplationist: you are missing the difference between rational expectations (your perspective) and adaptive expectations. If the market had rational expectations there would be no such thing as demand shocks. The only explanation of the current state of reduced output would be a negative supply shock. But then prices would be rising faster, not slower. So markets aren't *that* rational.

"If the market had rational expectations there would be no such thing as demand shocks. "

A patently false statement.

Arrrggh. Yeah, youre right. That was nonsense. Having coffee now. Sorry, Contemplationist.

Shocks are by definition exogenous.

Contemplationist: I had to think about your comment. If you are missing something, it's not anything obvious (to me). Maybe it's this: the "markets" we are talking about aren't necessarily just the markets for bonds and financial assets. We are talking about markets for regular goods and services like restaurant meals, and real assets. Markets where it's not so easy for informed traders to quickly bring prices into line.

Also, I expect I am basically a bit more sceptical or nuanced in my acceptance of EMH than Scott.

"Also, I expect I am basically a bit more sceptical or nuanced in my acceptance of EMH than Scott."

Somewhere, hidden behind the non-sensical blather, I think it was something like this that I had in mind.  The (New Keynesian?) distinctions, which Adam P has down cold, between rational actors and efficient markets do cause me significant trouble.  I'm simply too trained in efficient equilibrium.  As I attempted a coherent retelling of my story (after Adam's smack down) I realized that it was principally about market clearing failures and not expectations, and that it therefore didn't address Contemplationist's point which was about markets being smarter (i.e. rational expectations).  I.e. apart from being nonsense, the explanation was irrelevant.  

In the end, I should have just made a trivial point, which is not a distinction between models with or without market failures.  It's simply the point that maybe expectations are rather more adaptive than rational, especially when it comes to workers and consumers rather than investors.

I have a question for the macro modelers.  Are rational expectation models way harder to solve than adaptive ones, because of the problem of determining a solution that is coherent with its own expectation?  And the solutions less intuitive?

The reason I am asking is that the problem of intuitively understanding the (recursive) effects of our expectations (including the future dynamics of expectations) are so daunting that I can't begin to conceive of myself as forming rational expectations in that context.  When it comes to forecasting macro variables, I can do little better than, e.g. extrapolating a downward line on inflation, applying Taylor like rules of thumb, etc. I am certain that my own human/moral/observational biases weigh vastly greater than objective computation.  I'm also pretty sure these biases are not independent of the biases of all the other humans; i.e. they don't cancel out.  It truly strains my credulity that it is possible for us to form self-consistent expectations of a system that is that complex, non-linear, innovative, *and* recursive.  

Is the only way that this can work if the Fed tells us what the true model says and we all believe it unequivocally?  If so, the market can't be smarter than the fed. Also, the Fed needs more than credibility.  They also need to know the true model.

K: " It's simply the point that maybe expectations are rather more adaptive than rational, especially when it comes to workers and consumers rather than investors."

I would lean towards that. Except, I would say that it takes a long time for most people to learn the new form of adaptive expectations (rules of thumb) that work best when the policy regime changed. I did a post on this some time back, arguing that rational vs adaptive expectations was a false dichotomy: http://worthwhile.typepad.com/worthwhile_canadian_initi/2010/07/rational-vs-adaptive-expectations-a-false-dichotomy.html

"I have a question for the macro modelers. Are rational expectation models way harder to solve than adaptive ones, because of the problem of determining a solution that is coherent with its own expectation? And the solutions less intuitive?"

They used to be harder to solve in the olden days, when we were first learning how to do it. I don't think they are necessarily harder to solve now. We learned some tricks. But sometimes the solutions are a bit less intuitive. Or, it takes a different intuition. Because adaptive expectations depend only on what happened in the past. So it was easier to tell a story like: "A causes B which causes C which causes D, etc.". There's more simultaneity in rational expectations. To explain why the RE solution is the solution, we sort of conjecture that it is, and then explain why no individual would want to move away from it.

What I think you are getting at is that, even if an RE equilibrium exists, and is unique, it is not at all obvious if the economy would ever get there. Would people ever *learn* the RE equilibrium? Because the very fact that they are learning changes how the economy operates. The RE equilibrium is not obviously a stable equilibrium, in other words.

Some economists worry about this. Others don't. I think we ought to worry about it.

Thanks, Nick.  

"Would people ever *learn* the RE equilibrium?"

Here's my (train of) thought on this matter...

Imagine that there is a true model of the world.  Lets imagine that this model has a very large number of stochastic variables.  Lets also imagine that we don't know the model but we can make perfect observations of the variables.  If we assume that variable are continuous then we can work out the exact dynamics to within an arbitrary drift associated with each underlying Brownian motion.  But we can't observe those drifts, even in principle.  But, if markets are complete, we can hedge all risk away, and we   only need to care about risk neutral expectations.  The drifts don't matter.  Completeness is a suspect assumption, but we can make it in principle.

But if we allow for jumps things become much worse.  If we don't know the model then we don't know the jump sizes before they occur.  And we can't know the intensities of the jumps until we have observed a very large number of them.  And at any time scale, no matter how long, there may be jumps that you have never observed but which are proportionately larger and therefore still relevant for expectations.  And it's not right to imagine a complete market in jump risk that you don't even know exists.  And some jumps may have a continuum of sizes, so even if you knew the distribution (which you can't)  you would need an infinite number of assets to hedge any possible exposure, so a finite market wouldn't be complete even if you knew the model.  

A sophisticated modeler would at this point object that not knowing the dynamic is nonsense.  From a Bayesian perspective there is a distribution of probabilities of having any one of all the possible models each with millions of stochastic variables each with its own complex properties.  And this Bayesian distribution is simply a part of the *known* model of the world.  So now our model is the model of all possible models of the world.  And given that we don't have the number of assets required to complete the dimensionality of the model of all possible models we are left with the task of computing expectations from that model by inferring its dynamics by direct measurement of the observable variables.  But more seriously, we are left with the task of setting forth the Bayesian prior of the model of all possible models. We must do this because the prior *is* the model.  But you can't *learn* the prior. There's an infinite choice of them and no one of them is any more rational than any other. They are *all* rational. And every one of us has a different prior.  That is why we disagree with each other.

So that's my real point about learning.  Without market completeness it's required.  But learning (modeling) is projection onto a framework (or prior).  And we each have our own framework.  So we will never learn the same thing.  And therefore, in an unknown world, rational expectations aren't unique.

Sounds like you are against the ergodic hypothesis. Paul Davidson, introduced me to this critique of Samuelson, as I started to read post-Keynesian work. But I can't explain it. Reading about the transmission of Keynesianism to NA was interesting though.

Or maybe not, since you argue that there is a universal model. But rather that the agents are limited in discovering the infinite priors. here's a second from Davidson, that explains the difference."Knight's epistemology". The first link is pretty long, and not sure if it's the best explanation.

Edeast: Thanks for making the link to statistical physics. Though I think the question is irrelevant in practice, perhaps it is a relevant theoretical question to ask whether we can say that *all* models have equal a priori probability (and that's our prior). So perhaps the question is whether or not the set of all models is measurable? If it is, then maybe there is a unique prior and my claim above might be wrong (at least in principle). I haven't read your second paper yet. The link didn't work but I will try to find it tomorrow.

Price level targetting isn't any harder to understand than today's unpredictable inflation, with the right explanation.

Inflation targeting: you nod off while driving, and wake up driving beside the road, in the ditch. Oh well, you might as well drive the rest of the way in the ditch now that you're here.

Price level targeting: if you ever end up in the ditch, you will try to get back on the road.

I dunno... It makes sense to me.

Sorry for the broken link, it's the first paper at his home page.

Explaining how Taleb's interpretation of black swans or your 'jumps' is, of lack of a priori.
"In an ergodic universe, any single event will appear to be unique to the observer only if she does not have a sufficient a priori or statistical knowledge of reality to properly classify this event with a group of similar conditional events. "

The non ergodic claim is more fundamental. Still trying to wrap my mind around it. Because you can use non-parametric methods to solve for variables in an ergodic system. I mean the biggest problem with black-scholes, is the normal distribution, adding the fat-tails or the Mandelbrot critique, is a hatchet job around it. So it's not just about all models having equal a priori, ...

or wait a second. If all models have equal a priori probability, maybe you need to have a locally (space/time) optimized prior. I'm thinking evolution here. or geniuses before their time or... be in a localized universe, infinite expansion-contraction etc.

My brain can't deal with whether the set of all models is measurable, Maybe Cantor's work with set theory.

Concerning your fourth question Nick, you started by saying that "The whole point of having flexible exchange rates and an independent monetary policy is supposed to be that it allows the Bank of Canada to set aggregate demand for domestic objectives regardless of what happens in the rest of the world." I am not sure this is in fact the reason we have either flexible exchange rates or monetary independence (such as it is in Canada). But in any event, if there is a need to offset real shocks hitting us from the rest of the world, and these happen quite often, can the Bank contribute to this stabilization while pursuing an inflation target defined as that target is now?


Hi Stan! Welcome!

What other reason would there be for having a Canadian central bank with an independent monetary policy and flexible exchange rates?

If a real shock reduces Aggregate Supply in Canada, then there is little the Bank of Canada can or should do. But the Canadian recession looked more like an AD shock than an AS shock. Excess supply, unemployment, disinflation.

The transmission mechanism by which monetary policy changes the economy is the banking system: provision of favorable monetary conditions for banks (cheap borrowing costs) makes lending by banks profitable at lower interest rates.

Bank lending in Canada has increased over the past 3 years across all measured categories except business lending (i.e. consumer and real estate loans are up even as consumer and real estate lending have contracted in the US). The demand for business loans in Canada is of course highly contingent upon global economic growth (Canuck businesses sell to a global market).

In short, Canada has done the maximum possible to 'stimulate' domestic aggregate demand in the face of falling global aggregate demand. Canada has had a business recession, but not a consumer or household recession.

Frankly, it's hard to imagine that a monetary policy looser than the current one could have much of an effect because Canadian monetary policy can't make up for lower global aggregate demand.

Also, remember that, by definition, today's monetary policy must have effects down the road...

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad