There are two aspects of coverage that affect inflation measures:
- Item coverage
- Population coverage and weighting issues
This issue may come as a surprise. In theory, it would be expected that price indices cover everything that the population spends its money on. However in most cases they do not. Many indices leave out expenditure related to taxation and nearly all omit direct taxation costs such as income tax. They usually do not include elements related to savings and investments. The only exception sometimes is when that investment is housing; and most people would probably feel that is correct.
Indeed housing is one of the most important areas in which UK indices differ. Unlike RPI, CPI excludes mortgage interest payments, house price rises and rents. Also omitted are buildings insurance (but not contents insurance, bizarrely), ground rents and council tax. Housing accounts for over 25% of the RPI and so these are significant omissions from what most would regard as an ideal index.
What is fascinating is that their exclusion was not due to their lack of importance; it was merely that EU officials could not agree on a standardised way to measure them. Therefore, they left them out. This may have been a sensible decision for international comparisons, but it was not a good thing if you wanted to rely on it as a comprehensive cost of living measure, as we have now done in the UK.
These missing housing components do have an impact on CPI. The amount depends of course on what is happening to the cost of mortgages and house prices. Generally, it lowers the apparent CPI inflation figures when house prices and/or interest rates are increasing (which is most of the time). In August 2014, the exclusion of mortgages and houses from the CPI list reduced it by about 0.6 per cent, i.e. from 2.1 per cent to 1.5 per cent.
Given these problems with CPI, ONS decided in March 2013 to start producing a version of CPI that attempts to include the effects of housing called CPIH. Four years later in March 2017, CPIH became the main headline inflation measure in the UK.
CPIH is created by first estimating what are called “Owner Occupiers’ Housing” costs, or OOH, and then added them to CPI (along with Council Tax prices) – see method here. ONS are calculating OOH using something called “Rental Equivalence”, which in effect just involves looking at what happens to rent prices in the UK. They argue that changes in the prices of rents are a good proxy for housing costs and there is no need to measure things such as house prices, house insurance, water rates etc.
The problem is though that evidence suggests that rents are not that good an estimate of all housing costs and they can be affected by many other factors – not least supply and demand. Therefore in terms of a solution to solving the coverage issue, CPIH is not a good one – see full discussion here.
Population coverage and weighting issues
An ideal price index should cover the expenditure of the whole population and give each member of it an equal weight. In the UK, there are issues for CPI(H) and RPI in this respect but for different reasons.
The RPI index weights are based on the “Living Costs and Food Survey” that covers the complete household expenditure of over 6,000 households. However, two groups are excluded from these weights: (i) the 4 per cent of the population who are the wealthiest; and (ii) pensioners for whom three quarters of their income comes from state pensions and benefits. Between them these two groups account for 13 per cent of all spending.
The logic behind the above exclusions is that such households are likely to spend their money on atypical items and including them might distort the overall average. Some argue that RPI therefore reflects the “average household” better.
It is difficult to ascertain the precise effect of such exclusions. The rich undoubtedly have a different spending profile; pensioners who are mainly on benefits will spend a larger proportion of what little money they have on fuel and food. The impact of the latter can to some extent be measured by the pensioner versions of RPI that the ONS publishes monthly. These show that single person pensioners have experienced an average inflation rate 0.8 per cent above the rest of the population every year over the decade*. Given the proportion of pensioners in the population (and missing from RPI), this has lowered the overall RPI index by just under 0.1 per cent** e.g. from 2.5 per cent to 2.4 per cent in August 2014.
The issue for CPI(H) is different. Its weights are not survey based. Instead they are based largely on the “black box” (i.e. not published) of the government’s own calculations of GDP (and more specifically the Household Final Monetary Consumption Expenditure component of it). This means also that they have a different structure, which makes comparisons between the two difficult.
However one key issue with the GDP data is that it is value weighted. This means that the expenditure of the wealthy counts more. You can see the impact of this in the importance of items such as alcoholic drinks. They account for 2 per cent of the CPI index but over 6 per cent of RPI, as the poor spend relatively more of their income on drink. In contrast, 10 per cent of CPI is based on expenditure in restaurants and eating out, whilst less than 5 per cent of RPI is. Therefore RPI more accurately portrays the prices of the average person not only because it uses verifiable survey data but because it is not biased towards the expenditure of the more wealthy. The total impact of this on the CPI(H) index is not simple to calculate.
* Based on ONS price data for September (Q3) 2014. The figure was 0.5% for two person pensioner households excluded from RPI.
** Warwick University also has a personal inflation calculator, which allows simulation of both the missing groups. This again suggests that the omission of these two groups is reducing published RPI inflation estimates by around 0.1%. (Their calculator is predicting inflation of 0.5% higher for the 4% very rich and 0.6% higher for 9% pensioners on state benefits. Overall, this equates to a reduction of RPI of just over 0.1%.)