Sunday, July 7, 2019

Thinking about the CPI 1


Statistics Canada recently released  the May 2019 figures for the Consumer Price Index.  The 12 month inflation rate (meaning the percentage increase in the CPI from May 2018 to May 2019) was 2.4%, up from 2% in April 2019 (i.e. the increase in April 2019 relative to April 2018).  Putting it that way makes it sound as if all prices rose at the same general rate.  In fact there was considerable variation among commodities.

For example, fresh vegetables rose 16.7% over the year, contributing to a rise in food prices overall of 3.4% while gasoline fell 3.7%.  The StatCan report gives a list of the main upward and downward contributors to the 12 month increase. 


The inflation rate figure which makes the headlines is actually a weighted average of the 12 month rates of increase in the prices of the component commodities, with weights equal to the share of the typical consumer’s budget which is spent on each commodity.  The broad category of “Food”, for example, has a weight of 16.48% while gasoline has a weight of 3.13%. 

The StatCan release also reports on the May values of the inflation measures which the Bank of Canada pays closest attention to.  The Bank’s CPI-Trim, which omits the items whose year-on-year percentage changes are in the upper and lower tails of the month’s distribution of 12 month percentage changes rose 2.3% from May 2018 while CPI-Median, which reports the price change at the 50th percentile of the distribution of increases rose 2.1%.  Interestingly, CPI-Common, which uses a statistical approach to trying to estimate what we might term general price pressures showed the smallest increase, at 1.8%.  The Bank gives its definitions of each of the measures here, where they also graph the modified measures along with the Total CPI back to 1996.  Interestingly enough, something which you won’t see from the StatCan release is that not only has the CPI-Common remained constant through the past several months, it has actually dropped since the beginning of 2019, even though the Total CPI has been trending up since January.

The various modified CPI figures – CPI-trim, CPI-Median and CPI-Common represent the Bank’s attempt to get at what it refers to as Core inflation.  The general idea is that the rates of inflation for any particular commodity can be thought of as the general rate of inflation plus or minus a commodity-specific element.  The general rate of inflation is, roughly speaking, the rate at which all prices and incomes are tending to increase, while the commodity specific factors reflect conditions in the markets for the individual commodities.  If there’s excess demand building up for a particular commodity, we’d expect its piece to rise faster than the general rate of price increase while if there’s excess supply we’d expect the price of that commodity to rise more slowly than most, if not actually fall.  The calculation of Total CPI inflation is complicated by the fact that the Total CPI level includes indirect taxes, so changes in those taxes will show up in the changes in the levels of commodity CPIs.  Obviously, if tax rates remain constant over time they’ll stay in the level but not have any effect on the percentage change (inflation) figures.

What the Bank wants to know is that general rate of inflation – it doesn’t want to be adjusting policy interest rates on the basis of short term fluctuations in demand and supply conditions for individual commodities, it wants to stay focused on overall price pressures, whether upward or downward. 

That’s not as easy as it sounds.

Originally the Bank did this by producing what it called CPIX, meaning a modified CPI figure which it calculated by omitting the commodities which were typically the most volatile over time, and redistributing their weight across the other commodities.  This was less than entirely satisfactory, as the Bank knew, because it meant dropping the same commodities each month, regardless of how volatile they happened to be at that time.  CPI-Trim and CPI-Median are extensions of the CPIX approach, but instead of always dropping the same commodities, they involve dropping whichever ones happen to be in the appropriate tails of the inflation distribution that month.

CPI-Common is a different beast.  In calculating it the Bank uses all of the individual inflation rates observed over time, and uses statistical methods to try and extract a common trend from them, with the difference of any individual commodity’s inflation rate from the overall trend being due to commodity-specific demand or supply factors in that month.  This is an approach which seems to have been developed by the New Zealand central bank (the Reserve Bank of New Zealand) and which has been adopted by a number of other central banks in recent years.  So by this calculation, even though Canada’s overall inflation rate has been picking up through 2019, the common element, which the Bank takes to reflect broad inflationary pressure, has remained pretty steady.

One interesting aspect of all of this is that the Bank’s inflation target, as agreed on with the Department of Finance, is in terms of the Total CPI, not CPI-Common.  So at present, core inflation (which is really what we think of as the rate of inflation generated by monetary expansion, and therefore is the part of overall inflation which we assume the Bank has the most influence over) is steady while the rate the Bank is supposed to target has been rising.  

That’s not necessarily a problem (from a policy point of view – explaining it to the press might be a different matter).  So long as the statistical procedure used to separate out the common trend element is working ok, deviations from it, even when they appear to be persistent, should average out.  If the Common element is really the anchor inflation rate – the rate which everyone builds into their own pricing plans – then if the Bank can control it, the inflation rates on individual commodities should fluctuate around it in a pretty predictable manner. 

Remember that the CPI-Common inflation rate is based on a statistical decomposition of the pattern of all of the individual commodities inflation rates, so what we’re looking for is a way for the Bank to control that central trend (which, never forget, means that we’re assuming that the central trend which our statistical procedure gives us is in fact what we want it to be).  That was one thing the Bank was looking at when it was developing the CPI-Common.  One thing it found was that there was a consistent relation between the output gap and CPI-Common, with a lag of about 5 quarters, meaning that a change in the output gap today will have its clearest effect on CPI-Common a bit over a year from today.  That’s something to keep in mind when you’re doing policy modelling: simply relating changes in explanatory variables today to changes in dependent variables today will give you a misleading story.

Fortunately, statistical methods allowing for lagged responses are pretty well developed.  We'll look at some of those in another post.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.