Response from Dr. Ros Altmann, Director-General, Saga, to National Statistician’s Consultation on ‘improving the Retail Prices Index’
by Dr. Ros Altmann
(All material on this page is subject to copyright and must not be reproduced without the author’s permission.)
Purpose of inflation measures: As the consultation rightly points out, both RPI and CPI are measures of average inflation and do not measure the cost of living, but rather they measure ‘changes in the prices of goods and services consumed by households’. Therefore, I will approach my response to this consultation from the point of view of households buying goods and services.
Response to Consultation on the four options:
Option 1. No change
Response: For the moment, this might be better than the other options. A more thorough assessment of the accuracy of inflation calculation methodology is needed first, particularly as it reflects domestic purchasing, since RPI is important for many pensions uprating calculations and it could be that some elements of CPI should be changed to match RPI, now that CPI is to be used for uprating benefits and pensions in future. This Consultation assumes CPI is right and RPI is ‘wrong’ but in fact CPI may under-record inflation. Its geometric mean calculation does not cope properly when it is lowest price goods that rise most in price. It also does not cope with substitution, and will under-record the price rises faced by households who cannot or do not switch to lower priced goods.
Option 2. Change one approach to averaging price changes for clothing
Response: It is not clear that the CPI measure of clothing inflation is ‘right’ even if the RPI measure is not. Given the sharp rises in cotton prices, for example, it is entirely possible that CPI has under-recorded clothing price rises, especially if lowest price clothing has risen more in price than higher priced clothing. In this case, RPI may give a better reflection of household clothing price inflation than CPI.
Option 3. Change approach to averaging price changes for all categories using it
Response: Again, it will be important to consider how well Jevons and geometric means reflect price rises of goods whose price rises are correlated, or where the lowest price items rise in price more than highest value items. Also, it is important to ascertain which products can be realistically substituted for others in practice, not just in theory. If, in practice, many consumers (and especially the lowest income groups for whom inflation and benefit uprating is more important than others) cannot substitute goods easily to be able to buy those that have risen least in price or fallen most, then the geometric mean and CPI Jevons formula will under-record inflation. It would then be wrong to change RPI to the CPI averaging approach.
Option 4. Change RPI to align its formulae fully with CPI
Response: I do not believe this is appropriate – CPI may well under-record inflation and using Jevons for RPI will merely dumb down RPI, not make it more ‘accurate’. Given that changing the RPI formula so radically, and with such an obvious expected negative impact on the value of RPI inflation in future, I do not believe the ONS can seriously contemplate this option. The importance of this decision over time, its impact on the value of pensions, inflation linked national savings certificates and other RPI linked investments is such that any changes could have deleterious effects on future incomes and economic growth.
RPI and CPI have different purposes, so should not be the same: The purposes of the two inflation measures – RPI and CPI – are not the same. The retail prices index was originally designed to reflect the rise in prices of goods and services consumed by UK households in order to reflect the inflation rate they will face in their weekly expenditure. This measure was designed to reflect domestic inflation, which would then be used to adjust wages and benefits to ensure incomes would be preserved in real terms on average.
CPI reflects international comparisons, RPI reflects domestic considerations: The consumer prices index was introduced much later and the CPI measure was used in order to help with international comparisons of inflation, feed into macro-economic policy and National Accounts. Thus the CPI was designed more to fit in with other countries, while the RPI was designed to reflect domestic inflation for UK households. For example, the CPI includes expenditure by overseas visitors to the UK, whereas the RPI only looks at spending by domestic households in the UK.
Formula and composition both different, but one is not necessarily wrong: Because of cross-country differences in measurement, recording and expenditure patterns, it is inevitable that a measure of inflation that needs to harmonise with other countries, i.e. the CPI, will not necessarily best reflect price pressures for average households in the domestic country concerned. This is shown in the composition of the index, as well as the formula effect. For example, because it is difficult to compare housing costs across countries and because the UK has a much higher proportion of owner-occupation than other countries, the CPI omits housing costs in order to facilitate international comparison. But housing costs are an important part of rising inflation, so the RPI will better reflect national inflation than the CPI. [There are other differences in the baskets of goods and expenditure covered by each index – with CPI excluding council tax and underweighting insurance, while RPI excludes university accommodation and financial product charges and also leaves out the wealthiest and poorest households, but this is not really relevant to this consultation].
With all statistics, there is no perfect measure, all are estimates: As with all statistical measures, there is no perfect method of calculation, all are open to question and depend on assumptions or estimations to get as close as possible to reality, since perfect measurement is not achievable in a practical manner.
ONS suggests RPI is ‘wrong’ and CPI is ‘right’ – maybe both need changing: Clearly, given the different origins of the two inflation measures, they cannot be expected to be the same. The fact that there is a gap between measures of inflation on RPI and CPI is no surprise. However, for the purposes of this consultation, the ONS seems to believe that the CPI is the ‘better’ or ‘right’ measure and is seeking ways of aligning RPI and CPI more closely. An assumption that RPI ‘over-records’ inflation seems to lie behind the proposals put forward for change, but there is no consideration of the possibility that CPI ‘under-records’ inflation. All of the ‘alignment’ seems to revolve around lowering the RPI inflation figure, without considering whether in fact the CPI measure might be under-recording domestic inflation.
Why is ONS only consulting on accuracy of RPI, and not CPI?: It is indeed puzzling that the ONS is only consulting on the accuracy of the RPI, rather than also considering whether the CPI is adequately measuring inflation faced by households in the UK, especially bearing in mind that CPI was designed as a way of comparing international inflation, rather than measuring just domestic price pressures.
One-sided consultation does not ‘serve the public good’, especially as inflation is such an important number for the economy: The objective of the National Statistics Authority is to produce official statistics that ‘serve the public good’. This means that the inflation figures, or any other indices, must properly reflect the elements they are measuring. We submit that this consultation is therefore flawed, and fails to serve the public good, by not considering carefully the ways in which CPI may be under-recording inflation, as well as whether the RPI may be upwardly biased. This is especially so as the rate of inflation is such a crucial part of national life.
The inflation rate is reflected in so many people’s incomes and expenditure: Income, expenditure and future planning are all heavily influenced by inflation rates. As inflation rises, if incomes do not keep up with the increased living costs, the nation becomes poorer. If inflation uprating falls behind the real inflation rate faced by those whose incomes are being uprated, they will be unable to maintain their living standards. That will have a negative effect on national wellbeing. If wage rises consistently fall behind the actual inflation rates faced by most households, if pension rises fall behind the actual rises in living costs that most pensioner households face, if interest rates continue to lag behind rising prices, the nation will become poorer and growth will be lower as a result. Saga is concerned, therefore, to ensure that inflation is not ‘under-recorded’ and that this consultation does not result in a dumbing down of inflation figures to meet CPI levels, if CPI does not accurately reflect relevant inflation factors.
Formula effect is key to the response: This response focuses primarily on the formula effect, rather than the coverage issues, since that is the most relevant part of the consultation for these purposes. An important element of deciding on which calculation methodology is most statistically appropriate needs to consider how best to reflect consumer behaviour as closely as possible.
Geometric mean not always best measure: The use of a geometric mean in the CPI is said to be an approximation to the substitution that consumers carry out when one price rises and they move to an alternative that has not increased so much in price. However, it is well established that the geometric mean is only an approximation to substitution, it does not really measure that at all. Also, if substitution does not or cannot occur, CPI will again under-record the actual inflation rates faced by UK householders. A good example of this relates to pensioners, who will be exposed to proportionately higher price rises if they are less mobile between providers due to best and cheapest deals only being available on line, or in largest stores that are inaccessible to those who have no internet or are less mobile. This would also be relevant for gas, electricity and other commodities, where the best deals are often only on offer on-line or to those who are agile enough to shop around. The geometric mean will also not reflect domestic expenditure if price changes relate to changing consumer tastes, in other words, when price changes are driven by consumer demand, rather than by supply factors with suppliers offering goods more cheaply.
Carli may have upward bias, but Jevons also may have downward bias – appropriateness of use of geometric mean partly depends on ease of substitution: It is stated that the Carli index has an upward bias, particularly as it fails the reversibility test, however the Jevons index used for the CPI may under-record price pressures. Whether or not geometric means are more appropriate to use than arithmetic means depends on the extent to which products in the group are substitutes for each other and whether the consumers will happily and readily switch from one to another.
Jevons geometric mean will under-record inflation if low priced items rise most: Due to the fact that its calculation method will not fully reflect the rises in elementary groupings where the lowest price items have the highest price increases, the Jevons geometric mean will not fully reflect price rises which are concentrated more in low priced items than high priced ones. The Jevons calculation therefore depresses high price rises for lowest price goods and better reflects the situation where the highest price goods rise most in price.
This may be particularly relevant to clothing price inflation, which is the main example used in the consultation, to suggest that the RPI has over-recorded inflation and to highlight the supposed anomalies of the RPI.
Clothing price inflation differences between RPI and CPI may be result of CPI errors, not RPI: The Consultation criticism of RPI points out that clothing price rises seem to be far higher in 2010-2011 than recorded by CPI and that this has aggravated the formula effect in the measurement of clothing aggregates price rises. The differential between CPI and RPI clothing inflation is shown and the consultation seems to conclude that the RPI has over-recorded inflation, while CPI has measured clothing inflation correctly. However, as the price of cotton (the raw material used for much clothing) rose sharply during this period, the price of cheaper clothing (for which the raw material forms a larger proportion of the cost ) will have risen far more than the prices of more expensive clothing which reflect other factors not just the input costs.
Jevons index does not pick up price rises properly if lowest price goods rise most in price: If this is the case, the Carli geometric mean will not pick up the price rises properly. Indeed, the CPI figures suggest that clothes prices rose by less than average inflation, which, in light of the sharp rise in raw material costs, does suggest price rises were under-recorded by CPI and that RPI figures for clothing may be more fairly reflective of actual price changes faced by those buying ordinary, rather than expensive designer clothing.
Disadvantages of Jevons geometric mean which have not been explained in consultation: Use of a geometric mean is therefore not always appropriate:
- If consumers cannot or do not substitute cheaper goods for more expensive ones as their relative price falls
- if the cheaper goods are of poorer quality than those with higher prices
- if prices of low-cost goods in the aggregates rise faster than the prices of higher priced goods in those items
- if relative price changes are positively correlated
Part of the problem with current CPI methodology may be the use of the ‘average of relatives’ geometric mean, rather than the ‘ratio of averages’ and there may also be an impact from using January as the base month each year, when there are so many January sales discounts at that time.
Pensioner inflation measures: RPI and CPI inflation figures are really important in the context of the economy and particularly so for the older generations. As Director-General of Saga, I have been contacted by many of our subscribers who are most anxious about the proposed changes to RPI, because they fear that it will reduce their incomes even further. In fact, it would seem that CPI geometric mean calculations are likely to underestimate pensioner inflation, because pensioners are likely to spend more of their money on lower priced goods which will have been affected by rising commodity prices and they are also least likely to be able to shop around either on line, or by getting to the shops where prices are cheapest, as they are less mobile and tech-savvy.
In fact, the ONS does calculate its own pensioner inflation indices, which highlight that pensioner inflation is much higher than the national average inflation figures over the past few years. These figures are shown in the Table below and relate only to pensioner households with more than three quarters of their income from state benefits:
Comparison of ONS inflation data for pensioners and all ages, since 2007
|Date||ONS Pensioner RPI (single)||ONS Pensioner RPI (couple)||RPI – national||CPI – national|
|CUMULATIVE change since
Taking the start of the financial crisis as the month when Northern Rock failed, the cumulative pensioner inflation increase, as calculated by the ONS on its own pensioner inflation indices, has been over 27%, while national inflation (both RPI and CPI) has seen a cumulative increase of around 17%. Thus the poorest pensioners have become even poorer over the period, even if their incomes have risen in line with national inflation. This is of concern to us and has also been a factor weakening private consumption, since the impact of inflation on older generations has caused them to cut their spending. The Table below shows the relative price movements since Q4 2007 as measured by the ONS, for pensioner inflation, national all-age RPI and national all-age CPI. [Interestingly, the CPI measure has actually been higher over the period, due to the omission of housing costs which fell sharply in the RPI figures as interest rates tumbled during the early part of the period. As it happens, this did not benefit pensioners at all, because they are almost all mortgage-free, so the national inflation measure seriously understated the price rises faced by the oldest members of society. This does call into question the validity of the Government’s arguments for changing from RPI to CPI as a measure to be used for uprating pensions and other benefits.]
Impact of under-recording inflation would be serious for pensions and other incomes: If the result of this consultation were to be that the National Statistician decides to dumb down the RPI inflation measure, by aligning it to CPI, and if CPI does indeed under-record inflation, then the impact on the incomes of those who receive inflation uprating in line with a lower figure will be significantly negative over time. If the purpose of inflation uprating is to help keep incomes stable in real terms, so that households can continue to buy the goods and services they need, then a shortfall in the uprating will leave them consistently poorer over time and the power of compounding is significant in the long-term. For example, an annual pension of £10,000 a year, uprated for 30 years in line with inflation will move as follows:
|Annual inflation increase||3.00%|
|Annual inflation increase||2.50%|
|Years on lower uprating||30|
|Using inflation measure of 3% Final value of annual pension||£24,273|
|Using inflation measure of 2.5% Final value of annual pension||£20,976|
|Cumulative loss of pension||£40,024 after 30 years of lower increases|
So, if the inflation experienced by pensioners in terms of the change in prices of goods and services consumed by average households is actually 3%, but the uprating of their pensions under-records the price rises overall and measures them as 2.5%, these pensioners will end up being able to afford to buy far less than before, with their standard of living continually declining through their retirement. It is therefore essential that the ONS ensures that its measures of inflation do accurately reflect price rises for the goods and services which households buy, with no artificial lowering of the inflation measures.
Dr. Ros Altmann