Why year-ended inflation data?

Can anyone explain why inflation rates are conventionally reported on a ‘year-ended’ basis, despite the fact that we have quarterly price indices?

The Reserve Bank Governer’s press release of 5 February said that

CPI inflation on a year ended basis picked up to 3 per cent in the December Quarter…

There was a pretty detailed analysis of inflation rates and forecasts in the Financial Review last week too, all using the year-ended data.

But if we want to compare the last quarter’s inflation rate with that of the previous quarter, then why would we use a measure that contains seventy five percent of the same information? The CPI increased by 0.7% in the September Quarter and 0.9% in the December Quarter. This is all you needs to know, really, although if you prefer to think in terms of annual percentage changes, you can convert those numbers to annualised rates of 2.8% and 3.8%: thus, ‘the CPI inflation rate’ was one percent higher in the December Quarter than in the September Quarter.

The corresponding year-ended increases are 1.9% and 3.0%. At first glance this looks like a similar result, but it’s easy to show that that’s a chance coincidence. The first number is, loosely speaking, the sum of -0.1, 0.1, 1.2 and 0.7, while the second is the sum of 0.1, 1.2, 0.7 and 0.9. The net effect is that the December 2006 observation has been replaced by the December 2007 one in the calculation. And because that December ’06 figure was an exceptionally low one, indeed a negative one, dropping it makes the contrast between those year-ended figures look more dramatic. (To see this, suppose that the CPI had increased in two even jumps of 0.4% in September and December 2006, instead of a 0.9% rise followed by a 0.1% fall. It turns out that the year-end result for September 2007 would have been 2.4% instead of 1.9%.)

So, when the Governer says of the 3 percent figure, ‘This was a little higher than was expected a few months ago’, he may really mean that. But if the expectation was a rational one, he must have formed it on the basis of some other information than the previous year-ended figure, which is a captive of volatility in an irrelevant, long-gone epoch.

As far as I know, there is no convention of comparing four-period moving averages of any other quarterly data. Sure, we calculate annual averages for the unemployment rate, and annual aggregates for the trade balance and so on, but we don’t compare these with their equivalents for the previous quarter. Moving averages are good for discerning trends, but for something like monetary policy, which operates in monthly decision cycles, supposedly responding to the freshest data, it doesn’t make much sense.

In the case of forecasts, it’s even weirder. To forecast the moving average, you would need to start with a raw estimate for each quarter’s CPI, then calculate implicit quarterly CPIs from that, and then use those calculations to find the year-ended versions. How could such numbers possibly be more useful for policy analysis than the raw inflation forecasts themselves?

This entry was posted in Economics and public policy. Bookmark the permalink.
Subscribe
Notify of

Inline Feedbacks
Mark U
16 years ago

The year ended calculation is the percentage difference between the price level in the current quarter and the price level four quarters ago. That is, it is the inflation rate over the last year (or over four quarters). I don’t see anything particularly odd about using this figure. In fact, the Commonwealth Treasury provides forecasts of a whole range of economic variables on this basis in its Budget papers.

16 years ago

Well, thanks for making a comment, Mark. But I did lay out some quite specific objections, so unfortunately it will require something more than ‘I don’t see anything odd’ to overcome them.

sdfc
16 years ago

Why is forecasting an annual rate weird? Are you suggesting they forecast quarterly inflation out to mid-2010?

16 years ago

sdfc

Ricky Ponting’s last four ODI scores were 25, 10, 11 and 124. Making a forecast for the next match, I could either phrase it: (1) His four-game average will increase from 43 to 61; or (2) He’ll score 100.

What advantage do you see in the former method?