Measures of inflation under the microscpe
Colin Wilson and Chris Bull explain why actuaries should be aware of the differences between the retail price index and the consumer price index
01 JULY 2012 | COLIN WILSON AND CHRIS BULL
Recent changes to UK price indices mean that actuaries need to take a long hard look at the assumptions they make regarding future inflation and real returns. The consumer price index (CPI) has increased in importance for pension indexation as opposed to the retail price index (RPI), and this has already meant that actuaries have had to become more specific about what they mean by inflation and inflation-adjusted investment returns.
The effect of changes to price collection methodology means that it is also necessary to consider how the actual level of inflation measured may have altered, and the impact this has on expected real returns. Further changes are expected in the future, so this topic will not go away.
There is more than one way to measure inflation. Several conceptual methods exist, with the optimal choice depending on the use to which the index will be put. Furthermore, changes that are made through time mean that a single index may have different characteristics at different times.
In 2010, a significant change was made to how the prices for clothing and footwear were collected in the UK. It is widely accepted that this has increased the difference between CPI and RPI, and it has also been suggested that the actual rate of price inflation measured using both indices will now be higher than it would have been using the previous method. This may mean that achieving returns in excess of these inflation measures could be harder in future than it was in the past.
In the UK, two headline price indices are widely used as measures of inflation: CPI and RPI. Both track the cost of a fixed basket of goods and services. But they differ in their coverage and population base, and the way in which individual price quotes are combined. At the first stage of price aggregation, CPI mainly uses geometric means but also some arithmetic means, whereas RPI uses two forms of arithmetic means.
A geometric mean is always lower than either of these arithmetic means, and this ‘formula effect’ results in a lower CPI increase. Historically, annual CPI inflation has averaged around 0.7% less than RPI inflation; the main contributors to this are the formula effect discussed above and the inclusion of housing costs and mortgage interest payments in RPI but not in CPI – although the effect of this can fluctuate.
At the start of 2010, changes were made to how the prices of clothing and footwear were collected. This aimed to increase the sample size each month, better reflect consumer spending patterns and increase the number of price quotes in the calculation of the base period index. These changes appear to have increased the dispersion in price changes collected, increasing the difference between the arithmetic and geometric means and increasing the formula effect significantly.
Before 2010, the formula effect made a relatively stable contribution to the difference between RPI and CPI of about 0.5% a year; however, since 2010 it has averaged 0.9% a year to the end of 2011 (see chart). The increase in the formula effect was widely noted. Less attention appears to have been given to whether the changes in price collection methods altered the actual level of inflation measured.
The Bank of England has suggested that the way clothing price data was gathered in the past probably made inflation appear lower than it really was. This is because it is likely that previous practices led to seasonal falls in prices – such as in sales – being captured but not the subsequent recovery in prices when sales ended.
By assuming that UK CPI clothing prices were broadly in line with imported prices and euro-area prices, the Bank of England estimated that annual CPI inflation may have been underestimated by up to 0.3% a year between 1997 and 2009. Part of the fall in UK CPI clothing prices may have been the result of retailers substituting cheaper imported goods for more expensive domestically produced goods. An assumption that annual CPI inflation was underestimated by 0.2% a year may therefore be reasonable.
If the increase in the formula effect is entirely because of the change in price collection methods, then annual RPI inflation will have been underestimated by a further 0.4% a year, so in total it could have been underestimated by around 0.6% a year in the past, relative to what it would have been if the new methodology had been in place.
It is clear that the change in price collection methods in 2010 has substantially changed the measurement of inflation. Actuaries often use gilt or swap implied inflation as a starting point for making inflation assumptions. In this case, a change in how inflation is measured may not directly affect how assumptions are set, as it will be assumed that changes in expectations will already be priced in by the market. Also, the Bank of England’s target of annual CPI inflation of 2% remains an anchor for inflation expectations.
Even if assumptions about the level of future inflation remain unchanged, thought needs to be given to other assumptions, where these are expressed as an amount in excess of an index.
With, as yet, no liquid market in CPI instruments, many institutions continue to hedge CPI liabilities using RPI-linked gilts or swaps. For those doing so, it is important to understand the differences between the measures. An increase in the expected difference between the two indices may mean that liabilities become over-hedged.
As salary inflation is often assumed to be linked to price inflation, it is common to use a salary inflation assumption that is expressed relative to price inflation. But then one must consider whether price inflation measured in the past is consistent with how it will be measured in the future. If historic price inflation has been underestimated, then this will lead to an overestimation of the amount by which average earnings have exceeded prices by an equivalent amount. One therefore needs to take care in how historic data is used when setting assumptions about the future. If real asset return assumptions are also derived in this way, then one will encounter the same problem.
Problems in measuring inflation are not confined to the UK. When using international data, for example to produce a global equity real return, you need to consider whether indices are comparable. The introduction of harmonised indices of consumer prices (HICPs), of which the UK CPI is one, into the EU has dramatically increased the degree of comparability. However, differences remain. For example, there is some scope to use methods other than geometric means for price aggregation, and there are differences in the latitude given in terms of the products for which prices can be collected.
The development of indices does not stand still. Starting with clothing prices, the Office for National Statistics is working to examine the causes of the formula effect and determine how unjustifiable causes of this effect can be removed. It is also looking at the inclusion of owner-occupiers’ housing costs into the CPI. Any changes are likely to be implemented from early 2013. The expected impact will depend on the approaches adopted.
There will be other changes to the indices before this date, such as how new car prices are measured and the inclusion of vehicle excise duty, television licences and trade union subscriptions in the CPI. Not all these can be expected to have as significant an effect as the changes discussed here. Nevertheless, the risks of further changes to index characteristics should not be ignored.