Should calculation of inflation be the outcome of best scientific method?
CALCULATING the rate of inflation is not without controversy. The method used to measure changes in price of a basket of goods and services can make a difference to the result. So, should the calculation of inflation be the outcome of the best scientific method as determined by experts, or should members of the public get to choose the method that yields the most desirable outcome for themselves?
Back in the days when SA had both the consumer price index (CPI) and the CPIX (CPI excluding mortgage costs), our office would receive calls from divorcées needing to rerate maintenance agreements.
The party needing to pay maintenance would want to use the measure showing the lowest rate of inflation, and the one to receive payment would enquire which measure showed the highest rate of change.
This is a simplified example of what is playing out on the political stages of two large western economies.
One aspect of the present fiscal tug of war in the US involves the method used to calculate inflation, with a diverse range of stakeholders arguing for retaining a method that yields higher rates of inflation rather than one that is statistically superior.
Similarly, the government of the UK opted earlier this year to retain an outdated inflation formula for the indexation of public pensions and government bonds — because the arrangement results in a higher rate of inflation and therefore higher payouts to these interest groups.
In the US, President Barack Obama has proposed using an inflation measure known as the chained CPI to adjust public grants and benefits and to adjust tax bands.
The Bureau for Labour Statistics, tasked with compiling the CPI and other economic statistics, calls it “a better approximation of changes in the cost of living”.
One US senator described it as “a devious and underhanded way to wage class warfare against working families”.
So which is it?
All CPIs use the proportions of consumer expenditure for a particular time period as the weights for the different products and services that make up the CPI basket.
The weights are the relative effect of price changes of the different basket items on the aggregate index.
In most countries, the spending proportions are changed only periodically, giving rise to a fixed-weight or Laspeyres-type index (after the 18th-century French economist who first designed the basic index formula).
The chained CPI is what is known as a superlative price index, where the weights and prices are from the same period.
This aims to overcome the inherent upward bias of fixed-weight indices, which occurs because consumers adjust their spending habits over time in response to price changes. Economic and index theory supposes that consumers reduce their purchases of items that have become relatively more expensive, and increase those that have become relatively cheaper. This is known as substitution. A fixed-weight index only accounts for this when the weights are changed after a number of years.
The chained CPI, however, continuously updates the expenditure weights — meaning it tracks the actual changes in the cost of living of consumers in real time.
A focus of the fiscal debate in the US concerns the long-term costs of social security expenditure.
These are increased each year at the official rate of inflation, which is assumed to be an indicator of changes in the cost of living — defined as the change in the minimum amount of money required by a household to obtain a constant level of satisfaction between two periods.
In practice, this means being able to buy the same set of goods and services in the two periods. Where the argument comes to the nub, is that the current (fixed-weight) CPI escalator shows a higher rate of inflation than the chained CPI — as expected.
Lobby groups representing the elderly, and others dependent on government grants, have argued very strongly that the chained CPI should not be adopted — usually for this simple reason.
Some advocates in this consortium have suggested that an experimental “pensioners” CPI should rather be used to index pension benefits. Indeed, we have compiled such an index in SA for many years.
There is merit in this argument as the basket of goods underlying that index is more closely matched to the purchasing patterns of the elderly (for example, less on durables and more on medical care).
However, this yields an inflation rate high- er than both the official and chained CPIs and would therefore be less appealing to those trying to balance the budget.
In the UK, the government decided to retain a statistically flawed index formula in the form of the retail price index (RPI) for indexing government bonds and public pensions because it resulted in a higher rate of inflation than the scientifically superior consumer price index (CPI).
This decision followed a recommendation of the UK statistics office following a threemonth consultation exercise during which numerous stakeholder groups argued for the retention of the RPI.
There are a number of coverage differences between the RPI and CPI in the UK — most important, the present exclusion of housing costs from the CPI.
However, the central point of the discussion is the seemingly arcane matter of the formula used to calculate the price indices at the elementary level.
The RPI uses a “Carli” index, which is the simple arithmetic average (sum the numbers and divide by n) of price changes.
The alternative CPI uses the international norm of a “Jevons” index, which uses a geometric average (multiply the numbers and obtain the nth root). The Jevons index is less swayed by a few big price changes (outliers) and is therefore proven to deliver a more accurate picture of price changes for a particular product type. SA’s CPI uses a Jevons index formula.
The difference in formula is worth about 1.2 percentage points a year to the British inflation rate, which compounds over the life of a pensioner or the term of a government bond. This is because the Jevons formula allows for substitution behaviour by consumers, which will result in a lower rate of change in the cost of living.
In the face of an outcry by those receiving pensions, and also those holding inflationlinked government bonds, the UK statistics office decided to retain the RPI.
It will, as a partial compromise, publish an additional, adjusted RPI measure, which uses the Jevons index (RPIJ) and will give this more prominence in public announcements each month.
A few weeks ago on these pages, Gavin Keeton of Rhodes University ably demonstrated why most people perceive inflation to be higher than it actually is.
In the cases discussed here, there is no dispute that the proposed inflation measures provide an objectively more accurate reflection of the rate of consumer inflation.
But, vocal lobby groups have shouted down what is scientifically superior on the basis of their (biased) perceptions and parochial interests.
Could this be a dangerous portent of a Luddite-like future in which populism triumphs over sound statistical method?