The CPI release: Ignore the headline, and look at the core

A recurring source of frustration for me is that when Statistics Canada publishes its monthly numbers on CPI inflation, the data that really matter don't appear in the press release* so I have to wait until its Cansim data base is updated before I can comment. One consequence of this omission is that first reactions to the most recent release include stories like this one pointing to headline CPI inflation rates that are declining and are likely to go negative, and we are told of the increasing danger of deflation.

But that is simply not the case. The Bank of Canada doesn't use the headline numbers as a guide for its policy decisions. It uses core inflation, which strips out the effect of swings in prices of gasoline and food. And when you look at core inflation, you see that the threat of deflation is in fact receding.

Part of the problem is that attention is almost exclusively focused on the year-over-year (y/y) rates of inflation. But in times like these, it's not clear why the reference point should be a period when gasoline prices were at $1.30/l. Indeed, as the reference point moves to where gasoline prices hit their peak, we should expect the y/y number to drift even lower. But that's simply an artifact of the convention of reporting only the average rate of inflation over the past 12 months: why not the past 3 years, or the past 3 months?

Here's a graph of annualised inflation rates of headline CPI over other horizons.

Last summer, gasoline prices were rising and taking the headline CPI with it. If the Bank of Canada used headline CPI as a guide to inflation, it would have seen inflation rates of almost 8%, and would have started to raise interest rates.

But it doesn't, and it didn't. Here is the graph for core inflation, using the same scale on the vertical axis:

When you take out short-term swings in the most volatile elements of the CPI, you get a series that is much more stable, and which - in the Bank's view - is a much better measure of trends in inflation.

Now, none of this is to say that deflation has never been a serious concern. If we change the scale of the vertical axis in that last graph, we see that core inflation dropped sharply in January and February, while the real economy was in free fall. It was during that period that the Bank hastened to prepare the ground for quantitative easing, in case the trend continued.

Happily, it didn't. The last couple of releases have core CPI growing at around 2% at annual rates, and we're also seeing signs (the changing shape of the yield curve, for instance) of a softening in financial markets' perceptions of the risk of deflation.

So deflation is better thought of a danger that is receding (for now, at least), not one whose arrival is imminent.

*Yes, StatsCan does publish a core CPI number in its press release, but it's unadjusted for seasonality and can only be used to calculate y/y core inflation.

Comments

[David]

“But in times like these, it's not clear why the reference point should be a period when gasoline prices were at $1.30/l. Indeed, as the reference point moves to where gasoline prices hit their peak, we should expect the y/y number to drift even lower. But that's simply an artifact of the convention of reporting only the average rate of inflation over the past 12 months: why not the past 3 years, or the past 3 months?”

Thanks for saying this. The focus on the year over year numbers in both Canada and the US struck me as odd given that the degree of monetary accommodation changed substantially in the fall and then again this spring (with QE in the US).

I also have a question (with a longish intro).

My assumption has been that one of the dangers of using various inflation measures as a guide to monetary policy is the possibility that the figure could be distorted or biased by relative price changes and therefore be a poor indicator of changes in the overall price level. Presumably, this could arise if the index included price changes for certain goods(which might reflect relative price increases, for example) but excluded those for others (which might reflect compensating relative price decreases). Setting aside the issue of existing asset prices, the equation of exchange implies that the ideal index would reflect all GNP-component transactions (doesn’t it?). Given that, I don’t understand why such emphasis is placed on the prices of certain baskets of consumer goods. The use of a less than completely broad-based index then necessitates the exclusion of the relatively more volatile prices (i.e., gasoline and food) which presumably are more likely to represent relative price changes.

Why isn’t a more broadly-based inflation measure/target used?

[Nick Rowe]

David:

"Setting aside the issue of existing asset prices, the equation of exchange implies that the ideal index would reflect all GNP-component transactions (doesn’t it?)."

I don't see why that follows. If the definition of Pb that the Bank of Canada targets is different from the Pa in MV=PY, then the Bank will need to allow M to vary to offset changes in Pb/Pa. But the Bank would not necessarily see that as a problem, since it is letting M vary anyway.

The main argument for targeting CPI is that it is a definition of the price index that the public is familiar with. It also roughly represents the price index needed to deflate nominal income to get real purchasing power.

But the argument against targeting CPI is this: some prices are sticky, and some are flexible. The sticky prices don't want to change, and if we try to make them change, we get problems, like excess demand or supply. So the point of monetary policy is to try to make sure that the prices that don't want to change don't need to change.