Thursday 10 February 2022

What Does the CPI Measure? The Case of Housing and Cars

Ask anyone if there is an inflation problem and they will likely say, "Well Canada's CPI is up 4.8% in December 2021 and that's the highest since 1991." Editorialists will say such things as "the cost of living is out of control" and support their argument by pointing to the rate of increase of the Consumer Price Index.

When I was downloading some CPI data from Statistics Canada I noticed this footnote:

The Consumer Price Index (CPI) is not a cost-of-living index. The objective behind a cost-of-living index is to measure changes in expenditures necessary for consumers to maintain a constant standard of living. The idea is that consumers would normally switch between products as the price relationship of goods changes. If, for example, consumers get the same satisfaction from drinking tea as they do from coffee, then it is possible to substitute tea for coffee if the price of tea falls relative to the price of coffee. The cheaper of the interchangeable products may be chosen. We could compute a cost-of-living index for an individual if we had complete information about that person's taste and spending habits. To do this for a large number of people, let alone the total population of Canada, is impossible. For this reason, regularly published price indexes are based on the fixed-basket concept rather than the cost-of-living concept.

So, it's not a cost of living index. I thought that I'd better go to the Bank of Canada website to see what the Bank says because, after all, they target keeping CPI inflation at 2%. So this is from the BoC's website: 

The CPI is a simple and familiar measure of price changes, or inflation. Employers use it to make cost-of-living adjustments in wages and salaries. Governments use it to adjust income taxes and social benefits such as the Canada Pension Plan and Old Age Security.

OK, so now I'm confused. It's not a cost of living index, but the Bank of Canada says employers and governments use it as a cost of living index.

I know some of the history. When inflation was out of control in the 1970s, there was lots of talk about a wage-price spiral. Prices were rising fast, so workers often went on strike and bargained for higher wages. When higher wages were granted to cover the rising cost of living, producers of goods and services raised their prices and a wage price spiral ensued. During a period of high inflation, employers and governments come under pressure to increase wages and government benefits. This was especially true for programs and wages that were indexed to the CPI. Some economists then argued that the CPI was overstating inflation and the rise in the cost of living. In the US, the Boskin Commission recommended a variety of changes to way inflation was measured to reduce what was viewed as an "upward bias" in the CPI. One of the changes was to introduce hedonic pricing for some items in the CPI to try to capture the effects of quality change. Anyway, the recommendations were adopted by many countries, including Canada, and this had the effect of dampening inflation as measured by the CPI.

So today, even with these changes made to the CPI, Statistics Canada says it is not a cost of living index, but it is widely used as a cost of living index to adjust some wages as well as many public and private pension payments.

These days, the complaint most often heard, at least form the general public, is that the CPI understates inflation and the rising cost of living that they are experiencing. Most of these complaints come from looking at the way housing and automobiles are treated in the CPI. We hear from the real estate industry that house prices were up 26.6% from a year earlier in December 2021, but the CPI measure of housing (what they call "owned accommodation) is only up 5.8%. What gives? We hear from the auto industry that because of a global shortage of microchips, new cars can't be completed and their prices are rising sharply. We hear from the used car industry that the shortage of new cars and trucks has pushed consumers into buying used vehicles and their prices are soaring by as much as 30-40%. But Canada's CPI for motor vehicle purchase was up just 7.2% in December, while the US CPI for new and used vehicles was up 21%. What gives?   

Housing Prices in the CPI

Statistics Canada has heard these complaints before. In 2017, during an earlier spike in house prices when the CPI was not tracking what people were seeing happening to house prices, StatCan published a note on the measurement of housing prices in the CPI. Here are some excerpts from the note:

There are many approaches to processing owned accommodation: payment, net acquisition, rental equivalence, user cost, and exclusion altogether. Depending on the main purpose of its CPI (indexation, deflation of expenditures or incomes, monitoring monetary policies), each statistical agency adopts a version of one of these approaches.....

The “net acquisition” approach consists of treating owner-occupied dwellings like all other durable goods in the CPI, which means attributing all dwelling purchase costs to the acquisition period, even if the useful life of the dwelling spans well beyond this period. For this option, net dwelling purchases in the reference year would be used as the expenditure weight for owned accommodation. Since it provides the possibility of instantly showing the impact of dwelling price changes on the CPI, the “net acquisition” approach has interesting characteristics for measuring price inflation and monitoring monetary policy. However, since it does not take into account the stream of services generated by the owned accommodation across time, it is less favoured for a CPI that is used primarily for indexation...

... there is a general consensus that the acquisition approach is the most suitable for handling owned accommodation, when the primary use is tracking inflation in the housing market, since this approach makes it possible to automatically show the impact of changes in dwelling prices... 

On the other hand, an owned accommodation index like the one adopted by the Canadian CPI, based on the user cost approach does not seek to measure changes in dwelling prices. Its purpose is to determine changes in the cost of using a stock of dwellings. This makes it sensitive to past movements in housing prices, as well as imputed costs such as depreciation. Those wishing to use the CPI as an indicator of inflation in the housing market must be aware of the inherent limitations of such an approach, and should not expect it to produce an index that automatically tracks movements in dwelling prices. 

Now, I'm getting more confused. StatCan says the CPI is not a cost of living index. But when it comes to choosing a method of measuring housing prices, they choose a method which they judge is more appropriate for (cost-of-living) indexation. They reject a method which is "most suitable... for tracking inflation in the housing market" and "monitoring monetary policy".

Well, what difference does it make? Suppose, that we equally-weighted Statistics Canada's New House Price Index (up 11.7%) and the Teranet House Price Index for existing homes (up 15.5%)  and used the result (up 13.2%) in the CPI instead of StatCan's user cost approach to measuring owned accommodation prices (up 5.8%). Owned accommodation has a weight of 19.2% in Canada's CPI, so using StatCan's method, it contributes (.192 x 5.8%) or 1.1 percentage points to the December inflation rate of 4.8%. If we used the equally weighted increase in new and existing house prices (13.2%), the contribution of owned accommodation would be (.192 x 13.2%) or 2.6 percentage points to the CPI inflation rate.  

Motor Vehicles in the CPI

Prior to the pandemic, most people, including most economists probably thought that the price of motor vehicles was treated the same in measuring both the Canadian and US CPI. But they are not. This became evident when US inflation started spiking in the spring of 2021 and one of the leading causes was the sharp rise in prices of used vehicles. When that happened, I for one, did take the trouble to look at Canada's CPI breakdown to see what was happening to used vehicle prices in Canada. But there was nothing there. A phone call and an email to StatCan returned the information that, "No, Canada's CPI does not include used cars, only new cars." 

So then one has to look for industry information to see what is happening with used vehicle prices. The US has the Manheim index which provides a clear picture that used car prices have surged in the US and most of this surge is being picked up in the US CPI. Manheim does not have an index for Canada, but some time spent searching the web shows that CarGurus has an estimate that in Canada, used vehicle prices are up 35-40%. Well, how important could that be, you might ask?

Google is a wonderful tool for getting answers. I'm not a statistician and the following comments might be questioned by an expert, but here goes.

First, how important are used vehicle sales? Well, in the US new vehicle sales in 2021 were about 15 million units. Used vehicle sales were 40.1 million. Almost three times as many used vehicles are sold to consumers compared to new vehicles. 

The average price of new vehicles in the US rose to a record US$47,000 in December 2021 according to Kelly Blue Book and increase of 15.2%. The US CPI measured US new vehicle prices to be up 11.8% in December. The difference is probably mostly accounted for by the "hedonic quality adjustment" done for the CPI. But if you are a consumer who bought a car, the average price was up 15.2% as the Blue Book says.

The average price of used vehicles in the US rose to a record US$28,200 according to Cox Automotive in December 2021, up by a stunning 46.6% from year earlier. The US CPI measured used vehicle prices to be up 37.3% in December. 

Based on the US CPI measures, new and used vehicles contributed 1.5 percentage points to US CPI inflation of 7.1% in December. If the industry measures of price increases were used the contribution would have been 1.9% and total inflation would have been 7.5%.

In Canada, new vehicle sales were 1.66 million in 2021. There is no industry data that I could find on Canadian used car sales, but since the auto markets in Canada and the US are very similar, one could estimate that used vehicle sales in Canada are probably in the 4 to 4.5 million range. The best guide the used vehicle prices is the CarGurus price index which recently showed a 34% increase over a year ago, lower than the US industry estimate, but quite close to the US CPI estimate.

Based on Canada's CPI measures, the increase in new vehicle prices added 0.45 percentage points to Canada's CPI inflation of 4.8% in December 2021. Of course, used vehicle prices contributed zero percentage points because they are not included. So, even though the auto markets in Canada and the US experienced very similar conditions in 2021, vehicle prices contributed just .45% to Canada's inflation rate compared with 1.5% in the US. Thus, it would seem that Canada's CPI is probably understated by at least 1 percentage point due to the exclusion of used car prices. 

If used vehicle prices had been included, using the CarGurus estimate, and if Blue Book prices had been used for new cars (rather than the hedonically quality-adjusted estimates used by Statistics Canada) then the contribution of new and used vehicle prices to Canada's CPI could have been as much as 1.7 percentage points (instead of 0.45 pct pts).

Concluding Comments

A lot of people are experiencing inflation differently from the official statistics. In the 1970s, the argument was that the CPI was overstating inflation. Today, it feels like the CPI is significantly understating inflation. Statistics Canada admits that the method it uses to measure owned accommodation prices is not the best measure for gauging inflation and for monetary policy purposes. The Bank of Canada does not acknowledge this. The exclusion of used vehicle prices from calculation of Canada's CPI is clearly understating inflation as it is experienced by consumers. If we used easily understandable alternate measures of prices for houses, CPI inflation would be about 1.5 percentage points higher. If we included used vehicles and used industry estimates for vehicle price increases would be as much as 1.7 percentage points higher. Just adjusting the two components, which make up about 25% of the CPI, would boost our estimate of actual inflation being experienced on average by actual people by something between 2.5 and 3.2 percentage points. That would put December's CPI inflation not at 4.8%, but somewhere between 7.3% and 8.0%. That is probably closer to how consumers are experiencing inflation today than the official estimate.  

Despite a warning from Statistics Canada that the CPI is not a cost of living index, it is widely used for cost of living adjustments. The Bank of Canada continues to express optimism that once supply chain pressures ease, inflation should drift back down toward the 2% target, but consumers can't be blamed for being skeptical. Despite commentators suggesting that Canada's inflation is not as bad as in the US, it probably is as bad or worse.