From the TUC

The Great Brain Robbery: How statisticians’ counting tricks hit your pay packet

19 Feb 2015, by Guest in Economics

At the elementary level of price data aggregation a consumer price index can utilise the ratio of averages or the average of relatives  … blah, blah, blah … everyone’s stopped listening.

There’s no getting away from the fact that debates on inflation measurement can be a fiendishly complex subject, conducted in impenetrable essays among great statistical brains, using language that seems a million miles from the concerns of Fair Pay Fortnight and ordinary workers’ experiences of rising living costs.

But a threat has been growing that will have a powerful impact on pay bargaining and the value of wages — unless we come to grips with the details of calculating inflation.

Standing uncontested for decades as the prime measure of inflation in the UK, the Retail Price Index (RPI) has traditionally been seen as the key reference point for almost all pay bargaining. However, the governing bodies for statistics in the UK have launched a series of reviews over recent years that have diminished the importance given to RPI in official publications and raised the prominence of the alternative Consumer Price Index (CPI).

This culminated in a 2015 report commissioned by the UK Statistics Authority  which recommended that RPI should eventually be discontinued and an adaptation of CPI should be adopted as the UK’s principal measure of inflation.

In recent years, the government’s massive budget cuts have been forming the dominant backdrop to pay bargaining in public services, but the knock-on effects of these developments in inflation figures are also emerging across the economy. Examples are growing of companies that have traditionally agreed RPI-linked pay deals seeking to switch to deals based on CPI.

That’s a problem because the rate of CPI averages almost one per cent lower than RPI and so reduces pay rises by hundreds of pounds every year.

To investigate the reasons behind the gap between RPI and CPI, former Treasury economics adviser Dr Mark Courtney conducted a rigorous investigation at the end of last year. His study found that about a third of the difference is down to variation in the types of expenditure covered.

RPI is most relevant to pay bargaining as it is more tightly drawn around the working population by excluding the spending of the ultra-rich, tourists and a large chunk of pensioners.

RPI also fulfils the basic requirement of measuring inflation of housing costs fully, while CPI omits owner-occupier housing from its calculation – an incredibly major fault for any reasonable estimate of changes in the cost of living.

However, two-thirds of the difference between CPI and RPI was found to be down to the statistical methods used in their calculation.

If you asked someone to calculate the average number of people attending a meeting if, on one day 10 people attend, the next 20 and the next 30, most would add up the total number attending and divide by the number of meetings to give the answer 20. That’s called the arithmetic mean and that’s what the RPI uses in calculating average price changes.

The geometric mean used in the CPI would tackle this example by multiplying the number of attendees together and finding the cube root, giving an average attendance of just over 18!

In statistical terms this is a very simplistic example, but it does show how the techniques used to put together CPI have an inherent tendency to understate the real level of inflation.

The reasons put forward by the Office for National Statistics (ONS) to justify using this technique have varied over time. Initially it was justified on the basis of “consumer substitution” – which meant that it accounted for people switching to cheaper products when prices of alternatives went up. However, this position was abandoned when one of the foremost statistical experts on the subject changed his view and stated that the position was no longer supported by the evidence.

At other times, great emphasis has been given to “price bounce,” which supposedly showed RPI tends to overstate inflation. Yet the best estimates of “price bounce” now suggest that that it adds much less than 0.1 per cent a year.

With those arguments on the wane, the justification normally reverts to fear that the UK is out of line with standard international practice. Yet the reason countries such as the US and Australia switched to methods more in line with the CPI was due to the now discredited “consumer substitution” argument.

However, whatever the justification put forward, ONS has been relentless in ploughing on with downgrading RPI in the face of overwhelming opposition. When it conducted a consultation on RPI in 2012 it received 352responses, of which 332 sought to maintain RPI on its current basis. Among responses from statistical organisations and statistical experts, the weight of feeling was equally as strong with 44 out of 51 in favour.

The UK Statistics Authority is now set to begin a new consultation over the summer on the report that threatens the future of RPI and it is critical that we mount a vigorous campaign in defence of RPI. If we fail, we will have allowed an act of statistical trickery to change the landscape of pay bargaining in the UK for the worse.

This blog was written as part of Fair Pay Fortnight, which runs from 16 February to 1 March. To read the main findings of research by Dr Mark Courtney on Consumer Price Indices in the UK click here. To see the full report click here.

One Response to The Great Brain Robbery: How statisticians’ counting tricks hit your pay packet

  1. John
    Feb 20th 2015, 3:21 am

    Thankyou for this guest article Karen Jennings.

    After first reading I cannot help but think that the Government, the ONS would just like to use the statistical calculations that give THEM their best results rather than true objective results. I realise that the components used these calculations can be confusing, debateable, etc but if most people, companies, organisations prefer the RPI, then please let us just stick with that.

    Good luck with this.