Which Measure of Inflation is Best for Monetary Policy?

Đọc lại tài liệu cũ:

Which Measure of Inflation is

Best for Monetary Policy?

Once again, here is an article complaining about using core inflation as the target for monetary policy. I've said this before, but it's worth repeating. There are both theoretical and empirical ways to define the price index to use in monetary policy. However, most of the discussion about policy involves only the empirical definition.

The empirical definition searches for the the price index that best predicts future inflation or best represents the inflation rate faced by a typical consumer. Almost always, the discussion revolves around whether core inflation properly measures the underlying inflation rate faced by consumers, and its usefulness for predicting future inflation rates, an essential component of policy.¹ But the reason why we look to inflation at all comes from theory, so it is useful to ask what theory says about which price index is the most useful to policymakers.

What is the theoretical definition? In models with price and wage sluggishness, it is the failure of prices to move quickly to clear markets that causes output to deviate from target. Thus, monetary policy makers need not be concerned with highly flexible prices, it is the sluggish prices that are the problem. The solution is to keep the problem (sluggish) prices as predictable as possible so that even if prices are set far in advance, they will remain optimal.

To do this, the sluggish prices must be stabilized - the flexible prices can take care of themselves. For policymakers, this implies that highly flexible prices such as food and energy can be removed from the index to isolate and highlight the problematic sluggish prices. The goal is not to find the index that best represents the cost of living, rather, the goal is to learn about current and expected future values of the index most useful for stabilization. All of the criticism about the index misrepresenting the actual cost of living misses this point entirely.

Here is an example of commentary that only talks about the empirical approach:

Strawberries out of the basket, by Wolfgang Munchau, Commentary, Financial Times: A former central banker once told the story of a French prime minister during the fourth republic who had asked why inflation always rose in late spring. His statisticians told him that this was due to the price of strawberries, a seasonal food. The prime minister then instructed his officials to take the strawberries out of the index and to include them in the winter months instead – when you could not buy strawberries anywhere in France.

I am not sure how true this story is, but it seems plausible. Taking inconvenient items out of the inflation index has a long tradition. Today, many central banks still take the strawberries out of the inflation index, along with all other food items and energy. The index you end up with is known as core inflation.

There is a prima facie case for a central bank to target core inflation. Core inflation is a far less noisy indicator than monthly headline inflation. ... There are several ways of constructing core indicators. The traditional way is to strip out the most volatile categories of the index permanently; for example, food and energy. There are alternative methods, such as the trimmed-mean method, used by the Federal Reserve Bank of Dallas among others, which strips only the most volatile items off the index each month.

In the US, core-inflation indicators have turned out to be reasonable measures of underlying inflation – but not always and probably not now. Core inflation was a particularly bad indicator in the 1970s when inflation rose sharply in response to the first and second oil crises...

Generally speaking, core inflation is a misleading indicator in times of protracted increases in energy prices. In such times, core inflation lags behind headline inflation. A monetary policy that follows a lagging indicator risks being too slow – “behind the curve”, as they say in the financial markets. ... One reason why core inflation may be a lagging indicator are the so-called second-round effects of energy prices – the pass-through to non-energy prices or to higher wages. ...

Another problem with core inflation is the impact of the unsuspected volatility of some of its components. In the US, the biggest single component of core inflation is owners’ equivalent rent (OER) – an imputed number that measures the cost of house ownership. ... As the housing boom has subsided, OER inflation has risen back to more normal levels. Last week’s sudden increase in US core inflation was due largely to the jump in the OER component in the core index...

Some private sector economists made the preposterous suggestion that housing should also be excluded from the core index. If you go down this route, you end up with a core inflation index that no longer bears any relationship to reality. ... A forward-looking central bank ignores headline inflation at its peril. ...

The test of whether it is permissible to use core inflation is not only whether the two indicators converge within a reasonable period of time, but whether they converge in the right direction. In both the US and the eurozone we are seeing core inflation catching up with headline inflation and not the other way round.

It seemed outrageous that the aforementioned French prime minister took the strawberries out of the inflation basket. But compared with what some central banks are doing today, those strawberries are peanuts.

By the theoretical argument, it is not "prepostorous" to take asset prices out if the index or "outrageous" to remove food prices as these are highly flexible, and within this theoretical structure, non-problematic prices. That the "core inflation index ... no longer bears any relationship to reality" is beside the point. The point of targeting an index is stabilization, it has little to do with accurately measuring the price of some market basket of goods. It would be nice, agree or not, if people writing on this topic would at least recognize the theoretical side of the problem. For anyone who wants to follow up, there is nice NBER research summary of Woodford's work on these issues.
_________________
¹In general, the empirical approach would search for the price index most highly correlated with future economic activity (or someother variable of interest) so I'm not sure the term "empirical approach" is the best label to use here, but it will do for now.

Posted by Mark Thoma on June 18, 2006 at 12:33 PM in Economics, Inflation, Monetary Policy | Permalink

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/t/trackback/5137797

Listed below are links to weblogs that reference Which Measure of Inflation is Best for Monetary Policy?:

Comments

I don’t think the article even characterizes the empirical facts accurately: “In both the US and the eurozone we are seeing core inflation catching up with headline inflation...” Only if by “catching up” he means that the one has finally gotten out of the starting gate while the other continues to speed away. That image might lend itself to a nice cartoon caption: “Hey, guys, 2.1%! I’m catching up! I’m catching up!” (Of course, if you exclude rent – which is reasonable from a practical point of view today for reasons I discuss here – then the core isn’t even in position yet, much less out of the starting gate. And the same is probably true if you include rent but exclude indirect energy costs.)

Posted by: knzn | Jun 18, 2006 3:11:39 PM

OK, but now looking at the NBER paper, I think I see what the problem is. According to the bio of Wolfgang Munchau, he holds “an M.A. in International Journalism” and his experience, while quite impressive, is all in journalism and not at all in economics per se. I have a PhD in economics (with macro as one of my specialties), and even I would have to read the paper closely several times and think about it for several hours before I had any sense of how its arguments might apply to the real world (which I probably wouldn’t do anyhow, since I’m too busy with other things). And this is just a summary paper.

Of course, you make it sound pretty simple at the top of this post, and maybe it is that simple, if you believe that stabilizing output is the main objective. But somebody has to start making a more aggressive case for that objective. I doubt that most financial journalists have even considered the possibility. Somebody needs to prod them to get serious about the question, “Why is inflation bad?” But I’m afraid that getting serious about that question ends up requiring more time and energy than most non-specialists have available for the topic.

Posted by: knzn | Jun 18, 2006 3:57:05 PM

Nice analysis, post and comments :) From my perspective, let the bond market explain how to interpret price data, and I find a complacent bond market as I have for since the Federal
reserve tightening sequence began and that tells me be confident.

Posted by: anne | Jun 18, 2006 4:05:07 PM

«The empirical definition searches for the the price index that best predicts future inflation»

This seems to be to be strikingly disingenuous, because it is does not say what prices are relevant for «future inflation».

It is a thoroughly circular argument, which begs the question of what ''inflation'' is; while in effect ''inflation'' is whatever is defined by a particular index and not viceversa.

Challenges of the day:

* Define ''inflation'' uniquely.

* Define ''inflation'' in any way without using an index.

« or best represents the inflation rate faced by a typical consumer.»

Why only by a «consumer?» (never mind defining «typical») What makes «consumer» prices special?

They are just prices after all, and the notion that only «consumer» prices matter to whatever ''inflation'' may be is totally contrary to your ostensible rationale for controlling ''inflation'' (whatever that is):

«Thus, monetary policy makers need not be concerned with highly flexible prices, it is the sluggish prices that are the problem.

The solution is to keep the problem (sluggish) prices as predictable»

If the goal is to achieve better price efficiency and price signals by stabilizing «sluggish prices» (I am assuming this for the sake of argument only), than all such inflexible prices matter, not just those for consumer goods, thus prices for services (from tuition to insurance) or any form of assets also matter.

Otherwise stabilizing just the inflexible prices of consumer goods may well cause enormous distortions, as the inflexible prices that are not stabilized take the strain. For example, some service and asset prices are sluggish, and not stabilizing them, but letting them rip ahead for example in spiraling asset price bubbles most likely leads to large resource misallocations.

«as possible so that even if prices are set far in advance, they will remain optimal.»

This sounds rather like the whoosh of hands waving to me, because «remain optimal» begs the question, as there is not reason to assume that such prices are optimal at any point in time.

Unless of course one's propaganda includes Say's Law, and thinks that monopolies or oligopolies are good stuff, or never heard of the ''second best'' theorem, or thinks that ''Sraffa'' is a brand of wines.

«The goal is not to find the index that best represents the cost of living, rather, the goal is to learn about current and expected future values of the index most useful for stabilization.»

But this sounds like a complete fantasy to me: because it is most obvious that indices like the CPI are used for indexation, in particular COLAs for salaries and pensions, and even for contracts or bonds or whatever.

The historical rationale for COLA indices is to provide automatic indexation so that economic agents don't renegotiate constantly their ''sell'' prices, with the risk of supercompensating for increases in their ''buy'' prices, leading to a race to hyperinflation.

But then indices have been ruthlessly ''adjusted'' to minimize their reported growth, to lower the cost of indexation, but such is the power of mumbo jumbo that not many have realized it.

«For policymakers, this implies that highly flexible prices such as food and energy can be removed from the index to isolate and highlight the problematic sluggish prices.»

But the rationale usually mentioned for taking out food and energy of indices used for indexation is completely different: that since these often move down as well as up, this would mean that COLA adjusted nominal salaries and pensions could decrease as well as increase, and this would be politically unacceptable, and just ignoring downward movements would be practically unacceptable too.

The rationales for trying to stabilize ''inflation'' (whatever that is) at all are:

* Prevent feedback loops.

* Rapidly varying prices causes rational resource decisions to be more difficult.

Indeed the last point is completely contrary to the idea that one should try to stabilize only the already inflexible prices; quite the opposite, adding a measure of smoothness to the movement of ''twitchy'' prices improves resource allocation in the long term (just consider the consequences of the hog cycle).

(this assumes that in the real world, even if not in the delirium of traditional economics, adjustments involve both prices and quantities and intertemporal stocks, so that smoothing is possible).

Now, I don't dispute that an index of the more sticky prices could/should be built and used for monetary policy, but it seems to me a grave ''mistake'' to assume that:

* Of all inflexible prices, this index should only measure consumer goods prices.

* This index should be the one used for COLAs or other forms of indexing.

* The purpose of monetary policy should be to make inflexible prices more inflexible still while letting inflexible prices twitch all over the spectrum.

Posted by: Blissex | Jun 18, 2006 4:53:27 PM

mark great post the ground rent component of the core is a real fine take out candidate

flexible or not its not a product of labor  or of any price makers calc its an asset balloon that might require sectoral credit flow adjustment and it surely needs tracking but it only can effect wage push indirectly my optimal core index would be blatant wage control the change in nominal wage rates less the change in labor productivity calling a rose by any other name ....
just creates a foxy confusion

Posted by: js paine/slink | Jun 18, 2006 6:20:50 PM

Mark Thoma refers to: "highly flexible prices such as food and energy." I presume he means that the prices of food and energy respond quickly to changes in supply and demand. I don't know about food in the US, but the assumption that the price of oil (crude oil) is the immediate result of the real supply and demand situation seems pretty heroic to me.

Posted by: gordon | Jun 18, 2006 6:24:25 PM

Might it be better to mask out the volatility of commodities such as food and energy by using larger units of time for their sampling than is used for the "sluggish" items. Surely any long and sustained changes in the prices of food and energy do matter even if their short term fluctuations are a distractions.

Posted by: Ed | Jun 18, 2006 6:30:24 PM

"price of oil (crude oil) is the immediate result of the real supply and demand"

Well, under the (extremely bizarre) idea that the only prices that matter for stabilization are flexible consumer prices, it is not the price that "crude oil" that matters, but the retail price of refined products. Well, even if most BTUs of energy are traded wholesale under long term fixed price contracts, the price of retail energy products is indeed quite flexible and sensitive to demand and supply.

All this again under the assumptions that monetary policy should target the same indices used for COLAs, that targeting implies the long term stabilization of already stable prices, and that the only stable prices that should be further stabilized are consumer prices.

Posted by: Blissex | Jun 19, 2006 6:52:15 AM

"my optimal core index would be blatant wage control"

Well, I should perhaps voice at this point my suspicion that the (extremely bizarre) arguments in this post may seem to me prevarication, code for "the only inflation that matters is wages, as long as they stay low and grow slower than productivity, as measured by us, everybody else is really happy", as stabilizing already stable prices, excluding massively internationally traded commodities like food and energy, probably correlates quite a bit with stabilizing the puchasing power of wage earners.

The possible reason why the post does not come out clear on this is that it would be somewhat polically daft to explicitly state, as you do, that monetary policy is an instrument of industrial policy, and industrial policy is about ensuring a distribution of income as favourable to capital as possible, under the guise of anti-''inflationary'' policy; after all those who get income from capital usually have incomes high enough that they spend only a small part of them, and there are few enough of them that what they spend does not add a lot to inflationary pressures (except for Rolexes and Ferraris).

When in a few years the top 1% of earners triple their share of income from 8% to 16% (excluding capital gains!):

http://www.economist.com/images/20060617/CSF838.gif
http://www.economist.com/world/displaystory.cfm?story_id=7055911

this sequesters and neutralizes a lot of spending power that would have driven inflation higher: the lower 30% of income earners have a share of around 8%, and if the 8% that was added to the top 1% had gone to the bottom 30% it would have almost doubled their spending power and raised price pressures considerably.

Posted by: Blissex | Jun 19, 2006 7:10:48 AM

I think there are 3 questions:

1) What is inflation? 2) How do we measure it? 3) How do we predict it?

1) Inflation is a general increase in all prices. 2) Since there is more than one price in the economy and relative prices of goods can change too inflation is not easy to measure. We observe only some prices and we see them change at different rates ... how do we distinguish the relative price changes from the general price changes? There is also another problem when we add fixed prices, some prices cannot adjust all the time and when they do they might be adjusting to past, present and future inflation. Some goods carry more information about the general price changes than others. We have a problem of signal extraction.

3) Suppose that we solved the problem of measuring inflation we can then ask if there is a way to predict future inflation. Inflation may have some AR component in the short run but for long run inflation there is no better indicator than money growth.

Posted by: Alejandro | Jun 19, 2006 8:09:16 AM

"calling a rose by any other name .... just creates a foxy confusion"

It creates very politically useful dissembling. I found recently on DeLong's blog this very enjoyable and agreeable post on political realities and the joys of prevarication:

http://delong.typepad.com/sdj/2006/06/what_should_the.html#comment-18345435

Posted by: Blissex | Jun 19, 2006 8:25:17 AM

"Suppose that we solved the problem of measuring inflation"Uhm, reminds me of the defeatist attitude of that guy whose motto was "hypotheses non fingo". ''Optimism'' is indeed a more positive attitude...

Inflation is predicted and readily so and accurately so by the bond market, but seemingly few analysts attend to or understand how to read the bond market.

I applaud this post for its suggestion to theory, and how inflation should be measured, and how price stability should be achieved, but let's not lose sight of what measurements like the CPI are actually used for.Inflation protected treasuries have been developed so that investors (primarily old people looking for steady, reliable income) have a hedge against inflation. Likewise, Social Security increases are pegged to CPI as a sort of cost of living increase. Indirectly, many employers use the CPI as a measure of what to give their employees as a cost of living increase.

Removing vital measures of the "consumer" (aka wage earner/retiree) price index, such as food, energy, medical costs, home prices, and now, as some of you propose, rental increases, does nothing more than steal money from the poor and middle class. The current mind-bending miscalculation of the CPI is regressive enough, let's not make it more regressive.

That said, it wouldn't be a bad idea to build a different index used to predict future inflation, based on more theoretical measures. This could help those in the know better address price stability concerns.

Posted by: plymster | Jun 19, 2006 9:04:03 AM

"Inflation is predicted and readily so and accurately so by the bond market,"

While I agree with the sentiment that one can look at the bond markets to have a sense of how ''inflation'' (whatever that is) is expected to go, "readily so and accurately so" sounds to me like a wild exageration.Small example of a peculiar price signal: recently some government have started selling again 30 year fixed rate nonindexed bonds, after having refinanced national debt at ever lower rates as frequently as possible by using short term paper.

The government thinks that these long term bonds are going to be a bargain for the government, and the markets think they are going to a bargain for them instead...

They can't be both right, or can they? :-)Also, is insider trading *by the government* illegal? :-) but seemingly few analysts attend to or understand how to read the bond market.

Posted by: Blissex | Jun 19, 2006 9:09:50 AM

Markets are funny. Often times when you want to buy something the seller sees the buyer and changes his mind and demands a higher price for no other reason than he can. Sometimes the seller sees a queue of buyers and doesn't ration or sell scale up but lets prices gap to what they think the market will bear. IN other situations (like a panic) where buyers are queued up in a fashion representing an elastic demand curve, one may observe that the buyers simply evaporate into thin-air. They run-away run away, run away.

I make no pejorative judgements upon either actions. But what is clear is this: looking ONLY at either the historical transacted price (recent prices in comparison earlier prices), or the typical amount that might be expected to trade (aggregate demand & money supply), separately is wholly and unsatisfactorily insufficient for estimating anything other historical change in price. And at certain times it is hugely bogus possibly mis-estimating dramatically in one direction or the other depending upon the conditions.

Alejandro (and the monetarists generally) are wrong to grant complete primacy to money growth since money growth may indeed be accompanied by and justified by output and underlying growth or other plausible mitigating factors, or at the other extreme, "pushing on a string". But importantly, (and this is my biggest beef with all the inflation-related threads these past few days) the inflation apologists must also concede and recognize that "excess liquidity" represents "potential inflation" and that metaphorically 'waiting to see the whites of its eyes' before acting is fraught with risks - not least that once the genie is out, it will be dislocatingly large to put it back again, in addition to the moral hazard that accompanies the fact that IF my friends know you're going to wait, they will game it for keeps with net loss to society.

Posted by: Robert | Jun 19, 2006 9:12:43 AM

Alejandro (and the monetarists generally) are wrong to grant complete primacy to money growth since money growth may indeed be accompanied by and justified by output and underlying growth or other plausible mitigating factors, or at the other extreme, "pushing on a string". Robert, I'm ommiting the long run growth from the equation and you are right to point it out. But I (and monetarist in general all know it is there) as well as possible changes in velocity.

MV=PY inflation = money growth - output growth + velocity growth

Posted by: Alejandro | Jun 19, 2006 10:24:58 AM

Nice criticism :) Then consider bond market projections of inflation and whether accuracy is a proper description of bond market decisions. After all, the bond market gives only a day to day projection and conditions change and possibly change in dramatic terms. Again however, the bond market reflects the purchase analysis of large numbers of professional investors and they need to be conservative and accurate. Suppose however we change or modify "accuracy." The bond market reflects the projections of large numbers of professional traders, with a significant stake in being right. Fine.

I'm not a rigid monetarist - don't even know enough about it to be called a monetarist... but my gut says they are right in that if you want to control 'inflation' then measure & control money supply.

Mind you, I am sympathetic to the importance of monetarist concerns. But surely in an open international fiat system it becomes difficult to know whether the liquidity you're creating is building a golf course on the Pampas, or yet another mall in Short Hills. This is no small complication, and should if nothing else, keep the monetarists from getting dogmatic.

As for the bond as indicator of inflation, I would point out that both the JGB market and market for US Treasury Bonds are bent. Not UFO, Area 17 or the alledged PPT bent, but sufficient interference to question how good a measure of estimating inflation it really is. By how much its off is another very virtuous question.

The JGB market is twisted because the MoF is the great milk-teat from which those who must (city banks, trust banks, coop banks, regional banks, mutual & life insurers & prop/cat insurers) suckle. And then there is Jabba-The-Hut, the Postal Savings Bank, who makes Mammon look a goat-herder. How bent? Surely the Japanese can't be any worse than the Swiss (same demography, fertility, growth & Xenophobia), and the Swiss 10yr Swaps yields are something over 3% (and short end is 1.50% vs. ZIRP).

As for the US Bond, we will never know what the equilibrium rate would have been these past few years without Asian CB intervention and, without float of the RMB. And without ZIRP. I think most market folk reckon +100 to +150bps is a reasonable place to start.

Posted by: Robert | Jun 19, 2006 11:40:01 AM

I think that a carefully constructed core inflation metric is the way to go. It lets economists and bankers ignore the folks wheeling carts full of trillion dollar bills to buy their groceries. Then again, I'm bucking for a post at the Weimar Central Bank.