Recalculating inflation: billions in savings for governments and employers paid for by workers and pensioners

The top story in the Globe and Mail today reports on something I warned about a year ago: Statistics Canada is making changes to the way it calculates the Consumer Price Index.

At that time, I suspected changes to calculations of the CPI would be introduced as part of the renewal of the inflation target with the Bank of Canada, and they were clearly planning it at the time, but wanted to keep it under wraps.

While these changes may seem like obscure “refinements”, they could eventually lead to billions of dollars in lower payments from the federal and provincial government in Old Age Security Payments alone, billions in higher personal income tax revenues, and even higher savings from all employers from lower wage increases implicitly tied to CPI increases.

These savings for governments and employers of course won’t come out of thin air, but will be paid for by workers, pensioners, other income earners and income tax payers. For instance, a change that reduces the measured rate of CPI by 0.6% a year would mean a cumulative loss of $18,000 in income over ten years for someone with a starting salary of $50,000 with wage increases tied to inflation — not even including the impact of higher taxes and lower social transfers.

While there are some legitimate arguments for why the CPI overstates the real rate of inflation, there are also many reasons why the CPI understates the real changes in the cost of living — but these very well may be ignored by those making these changes, as they were largely ignored in the U.S. 1996 Boskin Commission.

For instance Canada’s CPI uses a new house price index for housing costs, which has increased at half the rate of resale homes. It also doesn’t account for faster depreciation and technological obsolescence of most goods, such as computers and cell phones. Nor does it account for increased commuting times, reduced public services, increased environmental costs or quality of life factors. All these factors affect the real cost of living for Canadians and they too should be accounted for in any revisions to any inflation or cost of living index as important as the CPI.

At the very least, the federal government should refer this issue to a public commission, as the United States Senate did in 1996 with the Boskin Commission, instead of unilaterally devising changes that will have such a large impact in secret behind closed doors.

Here’s the relevant text from the piece I wrote a year ago (pp. 9-10 in CUPE’s March 2011 Economic Climate for Bargaining)

Redefining inflation: who wins and who loses?

There’s another more obscure technical change that could lead to very significant impacts without even changing the inflation target.

This involves changing how inflation is measured. It may seem arcane, but it could ultimately lead to billions a year in losses for workers and pensioners, and billions in annual gains for governments and employers.

It has happened before. The 1996 Boskin Commission decided the US Consumer Price Index (CPI) overstated inflation by 1.1 percentage points a year mostly because it didn’t adequately account for the impacts of people shopping around for better deals. The changes to the U.S. CPI implemented since then saved the U.S. government hundreds of billions a year through lower social security payments and higher tax revenues.

By reducing the measured rate of inflation, they also reduced wage increases for most workers. Although the changes may seem small in any one year, they keep on growing cumulatively.

The Canadian government has also made more minor changes to reduce the way Canada’s CPI is measured, sometimes with little or no public notice. Now there is speculation the federal government may change the way the CPI is measured so that the official rate would be about 0.6 percentage points lower every year. It is important to understand this wouldn’t have any direct real impact on prices or inflation, but it would have very real consequences for wages, incomes, transfers and taxes that are linked to this measure of inflation.

A 0.6% annual reduction might not seem like a lot, but it really adds up and would mean major gains for federal and provincial governments paid for by workers, pensioners and other income earners. After ten years, a 0.6% annual decline would result in 6% lower wage, pension or transfer income. The cumulative loss over those ten years works out to over 30% of annual income, e.g: over $18,000 for a starting income of $50,000. The chart on the following page illustrates how these annual losses would grow over time.

Then on the flip side, workers would also end up paying more of their income in taxes because the tax brackets and credits would rise at a lower rate. The basic personal income tax credit would also end up being 6% lower in ten years time and taxes would be commensurately higher.

The major beneficiaries of this change would be governments through lower transfers and higher taxes and employers, through lower wages. The savings for them would be very significant and rapidly cumulate over time.

For instance, the annual savings to the federal government from lowering Old Age Security payments by 0.6% a year would rise from $210 million in the first year to over $1.2 billion in year five and almost $3 billion a year in ten years.

The increased revenues for the federal government just from a 0.6% lower annual increase in the basic personal income tax credit are of a similar magnitude: $180 million in the first year, rising to over $1 billion a year in year five and over $2.5 billion a year in ten years.

There are some legitimate arguments for why the Consumer Price Index may overstate the real rate of inflation and certainly better measures of inflation should be welcomed. However, there are also many reasons for why Canada’s CPI understates real changes in the cost of living but these appear to be ignored by those advocating for these changes.

For instance, Canada’s CPI uses a new house price index for housing costs, which has increased at half the rate of resale homes. It also doesn’t account for faster depreciation and technological obsolescence of most goods, such as computers and cell phones. Nor does it account for increased commuting times, reduced public services, increased environmental costs or quality of life factors. All these factors affect the real cost of living for Canadians and they too should be accounted for in any revisions to any inflation or cost of living index as important as the CPI.

Update (14/2/12): The Globe published a piece I wrote on this in their Economy Lab section today.

The Star also published an article on this issue, interviewing Susan Eng of the Canadian Association of Retired Persons (CARP) and Erin Weir, a frequent contributor to this blog. (However, I don’t think the TStar article is accurate in saying that in the U.S. inflation is thought to be overstated by as much as 1 percentage point.)


  • Can I get further explanation about the impact on tax brackets and credits?

  • Hi Paul:

    Personal income tax brackets and many tax deductions and credits, including the Basic Personal Amount, the child tax credit, the GST credit and Old Age Security benefits are indexed to CPI inflation.

    Any reductions in the measured rate of CPI inflation will mean that the value of these thresholds, deductions and benefits will be reduced every year. It may be by less than a percent a year, but over the years it really adds up. In effect, it is equivalent to partial de-indexation of these deductions and benefits — and hurts those dependent on pensions, indexed benefits and lower incomes the most.

    See the following for a list of some of the items from the federal government that are indexed by inflation and would be affected.

  • Hi Toby, nice article and some very good points that you raise.

    Well I am going to do something here that I have yet to do. I am actually going to post some praise for Stephen Gordon. Gordo actually makes a healthy contribution from the tech lab on a further complication to the points you raise Toby for the upcoming CPI changes.

    And for me, a half breed methods kind of person, I actually think what Gordo points out here could be potentially more damaging than the actual problem that Toby makes note of.

    Essentially what Stephan correctly states without a proper census to benchmark the SHS survey, a lot of control in the sampling plan is lost and hence sampling errors could very well overwhelm the accuracy and reliability of the CPI estimates given the disruption and challenges that the voluntary NHS has brought to the ecosystem of national statistical programs. Good work Gordo.


  • The flip side is that downward estimates on the CPI should have the positive effect of changing the BoCs interest rate decisions.

  • The language in the Globe piece was quite disconcerting because it clearly positioned a CPI revision as a tool for cost containment. Is Stats Can being asked to *lower* the CPI rather than improve its precision?

    The agency is more than 18 months into the CPI Enhancement Initiative, which attempts to improve the precision of its work. Statscan received extra funding to tackle the “measurement bias” – cash that rises to as much as $15-million in the fifth year of the effort – and make the index more attuned to Canadians’ spending habits.

    The effort is under way as the Harper Conservatives try to squeeze savings out of government and trim expenses in order to balance the budget by 2015-16.

    I still vividly remember my index numbers course with Erwin Diewert at UBC and I know there’s no perfect index. I can appreciate as well as any economist that the current CPI calculation methodology leaves a lot to be desired and I am all for making refinements to our measurements. But the sole consideration of this process must be to improve accuracy, not to cut costs.

  • You’re absolutely right, Travis.

    Lower measured rates of inflation will enable the Bank of Canada to keep interest rates lower than they might otherwise. That’s exactly what happened in the United States. Alan Greenspan was instrumental in getting the Boskin Commission underway.

    The impact of the Boskin Commission recommendations allowed Greenspan to keep interest rates low for so long enabling the credit-fuelled boom and resulting bust.

    Barry Ritzholtz pulls no punches:

    “Boskin is the economist/weasel/fraud who helped to officially distort the CPI, making it more or less worthless as a measure of inflation. The Boskin Commission was an act of fraud, a backdoor method to suppress Social Security cost of living adjustments (COLAs).

    The Boskin Commission’s massive government falsehood allowed former Fed Chair Alan Greenspan to take rates to absurdly low levels, as the official CPI data showed no inflation, despite double digit price increases.

    As such, he is one of the contributors to the financial collapse.”

    Greg Mankiw is also brutally frank about it:

    “The debate about the CPI was really a political debate about how, and by how much, to cut real entitlements.”

  • OK, but all I was pointing out is that having higher estimates of inflation are not inconsequential either. The real problem is the fetishization of price stability and inflation. My second best option would be:

    RGDP indexing under a regime that admits sovereign governments do not have to tax or borrow to spend.
    Under such a regime the definition of full employment could be much lower with downward biased estimates of inflation. So it is really the regime that is the problem not any particular estimate of the price level.

    Prudential loan practices (verifying income, significant down payments, low multiples of income etc.) would have prevented the boom in consumer credit. But that was not the regime. The neoliberal model was and still remains predicated on credit financed consumption.

    We need a different regime not simply a different measure of inflation.

  • My personal favourite reason to believe that inflation is currently under-estimated by StatsCan is their “hedonic” adjustments. I took at look at how they do it for “computer and peripherals” since that is an area I know something about, and to me it very much looks like they fail to properly account for the fact that early adopters are always more willing to pay for new features than is the general population. Some new thing comes along, they estimate how much it’s worth to the people crazy enough to buy it immediately; it then quickly becomes a standard feature you can’t buy a computer without getting, making it impossible to revise the estimate of its hedonic worth. I imagine they probably do the same thing for cars, houses, and microwave ovens. The result is understated inflation.

  • Congratulations, Toby, on having been ahead of the curve. I made similar comments to The Toronto Star.

    I agree with Travis that the silver lining of lower inflation estimates should be more accommodative monetary policy. (However, last time I suggested lower interest rates, Travis seemed pretty skeptical.)

  • Erin,

    At these rates I am. If rates were 6% and inflation was being underestimated it might make a difference. Anyway you can either contest the validity of my original argument about interest rates (near zero, marginal changes are near irrelevant and dangerous outside of a regime change) or you can say I was saying something I was not.

    Oh and as far as past positions: how is that story you told about low interest rates causing inflation coming along?


  • The story I have consistently told is that, with the economy weak and inflation low, the Bank of Canada clearly has room to keep interest rates low and perhaps reduce them further. That story has held up very well.

  • Statcan has stated that the sole consideration is to improve accuracy.

    There always seems to be a lot of noise about hedonic quality adjustments but the fact is, they are virtually nonexistent in the Canadian CPI. As far as I know, the only index where hedonic quality adjustments are made is in the computer index.

    Keep in mind that quality adjustments, hedonic or otherwise, can work both ways.

    There also seems to be a common misconception that reducing substitution bias means that statisticians assume that people stop buying steak and buy hamburger. Not true. They use current expenditure information that accurately reflects the proportional shift away from goods whose prices are increasing, to other goods.

  • Ron Murphy : “There always seems to be a lot of noise about hedonic quality adjustments but the fact is, they are virtually nonexistent in the Canadian CPI.”

    Yeah, I guess it was that noise about the US inflation measures that provoked my curiosity as to how StatCan does things. It isn’t “hedonic” as they call it, it seems I was mistaken to use that term or take the specific approach it describes as representative, but I had noticed that there was a whole lot of “quality” adjustment going on in cars, televisions, and so on. For the most part it appears that the conventional way of doing things just pretty much goes unquestioned.

    On looking at some actual price changes from 2002 to 2012: The results are not as bad as I would’ve guessed for televisions to the extent they’re representative of “video equipment”. The 10-year-old version of a similar-size television would have about twice the “quality” which isn’t so bad considering that the new ones are all LCD and HD. For cars, it seems more obviously wrong; prices are generally higher today for several models I don’t believe have changed much at all, but the price index is substantially lower. Makers of automobiles are always out to convince everyone that whatever new stuff they’ve added each year is worth quite a lot, and whatever method they currently use for the CPI accepts their story a little too easily methinks. Then there’s clothing which IMO ought to receive zero adjustment of any kind for quality except when some entirely new sort of fabric is invented. Fashions have changed, the function that clothing serves has not. My experience of buying men’s clothing this year and twenty years ago does not agree with the premise that prices have gone down and quality improved.

    Most of the quality changes are through “splices”, in which it is just assumed that any market price increase in the new item to be measured over the old one reflects exactly a quality improvement. To me that seems like a bizarre assumption. Occasionally, direct measurement is impossible since the old product is no longer on the market by the time they go to sample it. They apply best judgement when necessary, which is not usually, and amazingly enough it’s usually decided when human judgement is required that quality didn’t actually improve as much as the price did. The increase in the price index in these cases is on average less than the price increase in a surprisingly linear way, or it was in the period considered here.

    Anyway, it does get complicated, and after browsing a small selection of research on the subject, the impression I get is that there was some debate about this twenty years ago, but nothing came of it because although the current methods of quality adjustment are probably flawed and probably do tend to under-estimate inflation particularly when it’s generally at a high level, it’s unclear to what extent other measurement errors offset this and nobody really came up with anything better.

Leave a Reply

Your email address will not be published. Required fields are marked *