Ronan Lyons | Personal Website
Ronan Lyons | Personal Website

Who cares about measuring inflation correctly? Taxpayers should

Technical issues with the measurement of inflation… surely the dryest of dry topics, right? Well, I guess everything’s dry until it becomes a multi-billion euro issue, at which point it becomes very interesting. I’d be surprised if too many non-economists got excited over a recent article by Stephen Cecchetti of the BIS in Switzerland on the merits and in particular demerits of “core inflation” measures, in which he reviews two key problems with consumer price indices. Nonetheless, it’s worth recapping both those problems, because their relevance doesn’t diminish. The first is a short-term issue and quite well known:

“Over short intervals …[consumer price indices] contain temporary noise. For example, floods or droughts might drive prices of raw food temporarily higher, only to have those increases quickly reversed when the weather improves. Since this volatility is transitory, it is important that policymakers not respond to it. If interest rates were adjusted in response to the short-term inflation moves, it would amplify the noise, destabilizing the economy – the opposite of what we want them to do.”

In Ireland, we have seen this recently with people disregarding some of the large fall in prices in the last 12-18 months because of a one-off extreme comparison between variable mortgage interest rates in the summer of 2008 and again in 2009.

What is perhaps more worrisome, however, is the long-run issue with consumer price indices. Ever wonder, for example, why the ECB targets 2% inflation, rather than 0%? It’s not because they’ve given up on ever achieving the price stability of 0% inflation and 2% seems like a realistic enough goal. It’s because, to the best of our knowledge, somewhere close to 2% is price stability, at least when some margin for error is factored in when avoiding deflation. This is because there are two main methodological biases in measuring inflation and both of them are in the same direction – upwards.

The first is known as the substitution bias. In Irish terms, one could think of this as the Aldi effect. The Central Statistics Office surveys people every five years for the Irish CPI and then follow the goods and services they buy for the next five years, to get a consistent price series over time. Of course, people may switch to cheaper goods over the mean time. With the recession kicking in, many consumers may have switched from doing 80% of their shopping in the local convenience store and 20% in the nearest German discounter to perhaps a 50-50 mix. This change won’t be captured in measures of inflation because, by the time the CSO notices it, it’s time to be consistent from one month to the next with the new basket.

The greater the number of people inclined to switch to cheaper alternatives (or the more the typical person switches), the more incorrect the price index will be. For example, if you think it is plausible that unemployment is more likely to shake people into new purchasing habits, you might suggest that the last two years have seen a significant growth in the risk of the substitution bias.

The second type of bias, the quality bias, is even more subversive. You could think of it as the iPod effect. Quality bias relates to the product itself, rather than where it’s bought or sold. Either the product has got better over time in some hard to measure way, or else it is a new product that is not being captured.

An interesting way of thinking about this is the cost of computing power over time. The price per transistor is now about 1/600th of what it was thirty years ago. This means that computers have suffered a staggering 20% deflation on average every year for three decades (and yet no-one’s the least bit worried). If you think computers are alone in this, you might be interested to know that clothes now in Ireland are as cheap in nominal terms as they were in 1980. But I digress slightly. The take-away point is that when it comes to prices, every single innovation ‘under the hood’ that makes the products we buy better is overstating inflation.

Is this such a big worry? If anything, surely we all win when products get better. Obviously, product improvement is a Good Thing. Where would be all we if we were still stuck with our 8Gb iPhones, rather than our shiny new 32Gb 3Gs model, I’m sure you’re all thinking. The reason it’s a problem is that people think, and pay out, in terms of the measured cost of living. And by people, I mean in particular the government.

A US report from 1996, the Boskin Report, found that the CPI overstates inflation by somewhere between 0.8 and 1.6 percentage points each year, and most likely in the region of 1.1% a year. Because the federal government relies on the CPI, this bias makes the poor measurement of inflation the fourth biggest government expenditure item per year, after social security, healthcare and defence. Their best estimates of what the poor measurement of inflation would cost the taxpayer between 1996 and 2008 was over $1,000bn.

In Ireland, the Government faces a similar problem. This year, it will pay out over €40bn in pay, pensions and benefits to public servants, pensioners and social welfare recipients, about twice as much as it paid in 2002. In the same period, the increase in the CPI has been about 30%. Therein lies the rub, as it is this measure that determines how much the government changes things like pay and benefits each year.

Breaking down Exchequer pay and pensions into CPI changes (the running to stand still argument often presented by unions) and other changes (one would like to think productivity changes), and adjusting the CPI figure each year downward by 1.1%, as per the Boskin report, the figures from 2002 on can be recalculated (taking 2001 as a somewhat arbitrary neutral starting point). This highlights how serious a measurement error this is, because it compounds each year – as is shown in the graph below.

Index-linked public expenditure: what it was and what it could have been
Index-linked public expenditure: what it was and what it could have been

The per annum costs of not measuring inflation correctly when paying current and former public servants quickly rose from €117m in 2002 to almost €1.6bn this year. Accurately measuring the change in prices from 2001, the Government would have saved about €6.5bn in its pay bill by now. The same exercise can be done for social welfare payments, whose gross costs have grown from less than €8bn in 2001 to an estimated €22bn this year. The per annum saving this year is estimated to be as large as €1.8bn.

Adding up the cumulative savings from public sector pay and pensions and social welfare, if the Government had more accurately measured changes in the cost of living since 2001, it could have saved a total of almost €13bn. For no particular reason*, here are the Government’s estimates for 2009 net expenditure, €47.4bn, and tax revenue, €34.4bn (*hint: look at the gap between the two).

In its ‘Introduction to the CPI‘, the CSO notes that the CPI “a pure price index, not a cost of living index. To produce a cost of living index is a major and complex task.” It certainly is, although one suspects if the Government knew the compounding opportunity cost of not having a true cost of living index, they might take a different slant. This problem is also not just going to go away. The longer it’s ignored, the bigger it gets.

Stephen Cecchetti, Head of the Monetary and Economic Department at the Bank for International Settlements in Switzerland, in a discussion on the merits and in particular demerits of “core inflation” measures, reviews two key problems with consumer price indices. The first is a short-term issue and quite well known. As he succintly says:
“Over short intervals …[consumer price indices] contain temporary noise. For example, floods or droughts might drive prices of raw food temporarily higher, only to have those increases quickly reversed when the weather improves. Since this volatility is transitory, it is important that policymakers not respond to it. If interest rates were adjusted in response to the short-term inflation moves, it would amplify the noise, destabilizing the economy – the opposite of what we want them to do.”
In Ireland, we have seen this recently with people disregarding some of the large fall in prices in the last 12-18 months because of a one-off extreme comparison between variable mortgage interest rates in the summer of 2008 and again in 2009.
What is perhaps more worrisome, however, is the long-run issue with consumer price indices. Ever wonder, for example, why the ECB targets 2% inflation, rather than 0%? It’s not just because they’ve given up on ever achieving the price stability of 0% inflation and 2% seems like a realistic enough goal. It’s because, to the best of our knowledge, somewhere close to 2% *is* price stability, at least when some margin for error is factored in when avoiding deflation. This is because there are two main methodological biases in measuring inflation and both of them are in the same direction – upwards.
The first is known as the substitution bias. In Irish terms, one could think of this as the Aldi effect. When the Central Statistics Office survey people every five years, they then follow those goods and services for the next five years, to get a consistent price series over time. Of course, people may switch to cheaper goods over the mean time. With the recession kicking in, many consumers may have switched from doing 80% of their shopping in the local convenience store and 20% in the nearest German discounter to perhaps a 50-50 mix. This change won’t be captured in measures of inflation because, by the time the CSO notices it, it’s time to be consistent from one month to the next with the new basket.
The greater the number of people inclined to switch to cheaper alternatives (or the more the typical person switches), the more incorrect the price index will be. For example, if you think it is plausible that unemployment is more likely to shake people into new purchasing habits, you might suggest that the last two years have seen a significant growth in the risk of the substitution bias.
The second type of bias is even more subversive (after all, in theory, these days a government could probably use debit card or loyalty card data to figure out who’s switching to what where). This bias is the quality bias and relates to the product itself, rather than where it’s bought or sold. Either the product has got better over time in some hard to measure way, or else it is a new product that is not being captured.
An interesting way of thinking about this is the cost of computing power over time (the one thing to note is that this is not so hard to measure). The number of transistors in top-end computers has gone from about 2,000 in the early 1970s to 2 billion less than thirty years later. The price per transistor is now about 1/600th of what it was then. This is an amazing amount of deflation, an average of more than 20% deflation every year, (and yet no-one’s the least bit worried). If you think computers are alone in this, you might be interested to know that clothes now in Ireland are as cheap in nominal terms as they were in 1980. But I digress slightly. The take-away point is that when it comes to prices, every single innovation ‘under the hood’ that makes the products we buy better is overstating inflation.
Is this such a big worry? If anything, surely we all win when products get better. Obviously, product improvement is a Good Thing. Where would be all we if we were still stuck with our 8Gb iPhones, rather than our shiny new 32Gb 3Gs model, I’m sure you’re all thinking. The reason it’s a problem is that people think, and pay out, in terms of the measured cost of living. And by people, I mean in particular the government.
A US report from 1996, the Boskin Report, found that the CPI overstates inflation by somewhere between 0.8 and 1.6 percentage points each year, and most likely in the region of 1.1% a year. The sheer proportion of the federal government that is index-linked means that this bias makes the poor measurement of inflation the fourth biggest government program per year, after social security, healthcare and defence. The best estimates are that poor measurement of inflation since then has cost the US government over $1,000bn.
In Ireland, the Government faces a similar problem. This year, it will pay out over €40bn in pay, pensions and benefits to public servants, pensioners and social welfare recipients. This is about twice as much as it paid in 2002. In the same period, the increase in prices as per the CPI is about 30%. Therein lies the rub, as it is this measure that determines how much the government changes things like pay and benefits each year.
It is possible to break down the change each year in per person cost of Exchequer pay and pensions into CPI changes (the running to stand still argument often presented by unions) and other changes (one would like to think productivity changes). Adjusting the CPI figure each year downward by 1.1%, the Boskin report’s central estimate, the figures from 2002 on can be recalculated.
Recalculating the figures based on a more accurate figure for inflation highlights how serious a measurement error this is, because it compounds each year. Taking 2001 as an arbitrary starting point, the per annum costs of not measuring inflation correctly quickly rise from €117m in 2002 to almost €1.6bn this year. Accurately measuring the change in prices from 2001 one would have saved the Government somewhere in the region of €6.5bn by now. Likewise with social welfare, whose gross costs have grown from less than €8bn in 2001 to an estimated €22bn this year. The per annum saving this year is estimated to be as large as €1.8bn.
Adding up the cumulative savings from public sector pay and pensions and social welfare, and we’re talking a total saving over the past eight years of almost €13bn. For no particular reason*, here are the Government’s estimates for 2009 net expenditure, €47.4bn, and tax revenue, €34.4bn (*hint: look at the gap between the two).
In its ‘Introduction to the CPI’, the CSO notes that the CPI “a pure price index, not a cost of living index. To produce a cost of living index is a major and complex task.” It certainly is, although one suspects if the Government knew the compounding opportunity cost of not having a cost of living index, they might take a different slant.
  • Kevin Denny ,

    I think there is a strong argument here for increasing the frequency of the Household Budget Survey: too much can happen in 5 years. The equivalent in the UK, the FES, is annual.

    There is an older argument for not targetting 0% inflation which has some merit: essentially its about real & relative wage flexibility. It is very difficult to reduce nominal wages so this means that it can be very difficult to change relative wages. If you have a bit of inflation in the system then nominal wages need to be constantly adjusted. This gives you the opportunity to make changes in real wages & hence to also change relativities: modest inflation helps oil the wheels of the labour market.

    Leave a comment