[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Pensions: CPI - a change of pace

The controversial recent decision of the United Kingdom’s coalition government to change the indexation of pensions from the Retail Price Index (RPI) to the Consumer Price Index (CPI) has some striking implications for pension providers and raises important questions for risk managers. This article examines the differences between the two respective inflation measures, with a particular focus on the magnitude and stability of such differences. Subsequently, we consider how this change in legislation could impact the risks faced by institutions that are obliged to manage inflation-linked cashflows.

The differences between RPI and CPI inflation
In most developed economies, central banks and statistical agencies collect price data for the construction of a single aggregate index by which to measure the rate at which prices are changing. Usually this is referred to as the Consumer Price Index (CPI). However, in the United Kingdom there are two headline price indices by which inflation rates are measured. As well as the CPI there is also the Retail Price Index (RPI). Historically, the RPI has been the more prominently reported of these indices. Key reasons for the differences between the two inflation measures are summarised here. Why are there differences between the CPI and the RPI?

>> Housing costs are excluded from the CPI: mortgage interest payments, council tax, housing depreciation, buildings insurance and ground rent, surveyors’ fees, estate agents fees and conveyancing costs are included within the RPI but not the CPI.
>> Other differences in coverage: different goods and services are included in the two indices. For example, unit trust and stockbroker charges are among some of the goods included in the CPI and excluded from the RPI.
>> Formula: the RPI uses an arithmetic mean to calculate the price of specific goods, whereas the CPI uses a geometric mean.
>> Weights and ‘other’ differences: items included in both indices’ baskets might be weighted differently due to inconsistencies in data sources and in the way that the goods are treated. ‘Other’ differences can exist because the above components of the gap between RPI and CPI are not strictly additive.

Index-linked government bonds, which were first traded in the UK in 1981, are linked to the RPI (and are not currently available for the CPI), which was originally designed to measure the true costs of living for the average household. The first step towards making the CPI the headline inflation measure in the UK was taken in 2003, when the then-Chancellor, Gordon Brown, changed the price index to be targeted by the Bank of England from the RPIX (RPI minus mortgage interest payments) to the CPI and adjusted the inflation target downwards from 2.5% to 2%. The main reason given for the 0.5% reduction in the inflation target was calculation differences, or the ‘formula effect’.

The spread between CPI and RPI rates of inflation can be broken down into the components mentioned in the bullet points previously; the time-varying impact of each of these is shown in Figure 1 (below). From observation of the chart it is evident that the formula effect has had a relatively stable impact upon the spread between CPI and RPI rates of inflation in the past. The next most stable component appears to be housing costs, although in 2008 these contributed to causing the CPI inflation to move above the RPI rate. Housing costs combined with mortgage interest payments pulled the RPI into negative territory during 2009. This can be attributed mainly to the rapid reduction of base interest rates pulling mortgage interest payments downwards.

For risk management purposes it is important to understand that there are material and unstable differences between RPI and CPI inflation rates. Going by historical data alone, we would expect CPI inflation rates to be around 66 basis points lower than RPI, although it is not guaranteed that this difference will persist in the future as the goods coverage of the CPI might be revised to include housing costs. Despite the likelihood being that, on average, CPI inflation will be lower than RPI, we can see from Figure 1 that this difference is not stable through time, with the CPI moving above RPI by more than 300 basis points in recent years. These factors have important implications for the management of CPI- and RPI-related risks.

The devil is in the detail
Following George Osborne’s June announcement that public sector pensions would be linked to the CPI as opposed to the RPI, the pensions minister, Steve Webb, announced in early July that private sector pension increases will also be applied at the CPI rate. It is clear that this could reduce the current liabilities of pension funds, however, because of uncertainty about the legalities of how the change can be practically implemented, and due to the instability of the difference between the two rates, it is not clear what the final impact of these new rules will be.

To gain some insight into how this will affect pension providers, we consider three different eventualities:

1 A legal requirement to link pensions specifically to the RPI: no switch to CPI.
2 No legal complications: all pension inflation links switched from the RPI to the CPI.
3 CPI rates become the minimum guaranteed increase: pensions linked to the maximum of CPI and RPI.

We proceed by taking a set of real (yet to be inflated) pensioner cashflows. These ’dummy‘ cashflows are designed to approximate the liability profile of a typical pension fund and are displayed in Figure 2 (below). To isolate the core issues, we consider only the pensioners and ignore longevity risk. The chart shows a monotonically declining liability profile that corresponds to an unknown nominal (i.e. money) payout in any future year. For example, in year 20 the fund has promised to pay around £2m real pounds, which corresponds to some nominal amount rolled up by an inflation index.

To simulate the realised nominal value of these cashflows we used an economic scenario generator. Projecting CPI and RPI inflation rates simultaneously is not a trivial task: models must be configured to capture the structural linkage and corresponding high correlation between the two measures, while also respecting the long-term differences in their level and volatility. Moreover, it is important to consider the impact of market-implied expectations of interest rates and inflation as measured by the forward curve and RPI-linked bonds. The simulation data used in the analysis below meets these criteria.

Consider the first of the three potential situations: cashflows are RPI-linked and let us assume that the pension fund has hedged them exactly by buying RPI-linked bonds. In this case there is no uncertainty about the capability of the fund to pay pensioners what it owes them. However, there is uncertainty about what the realised nominal liability will be. This is shown in Figure 3 (below).

The same set of cashflows rolled up at projected CPI inflation is shown in Figure 4 (below). A comparison of Figures 3 and 4 shows that there is more uncertainty around what nominal payouts will be in RPI terms than CPI. The charts also show the upside risk associated with the realised nominal liabilities is greater than the downside risk; the funnel of doubt is positively skewed. This is consistent with the real-world risk of very high inflation rates: hyperinflation has occurred in many of the world’s economies in the past, but ‘hyperdeflation’ has not — the models are able to capture this dynamic.

The factors driving the shape of the charts come into play when we consider different conditions under which a mismatch of inflation-linked assets and liabilities might occur with respect to a change in indexation.

Matching real liabilities: a best and worst case scenario
We now turn to the latter two of the legislative possibilities specified previously. Consider the situation in which real liabilities are inflated by the CPI but expected real cashflows are exactly matched by RPI-linked bonds. The future distribution of the cashflow differences is shown in Figure 5 (below).

The chart shows that, under these circumstances, the pension fund can expect to be in surplus. There is one important caveat to this: the expected surplus is uncertain. There is a probability that the same assets that would have, without risk, covered RPI-linked payouts will not be sufficient to cover the same cashflows inflated by the CPI rate.

Now consider the situation in which the CPI rate is the minimum guaranteed increase. In this case, by matching cashflows with RPIlinked assets the fund expects to be in deficit. The uncertainty around the evolution of this position is shown in Figure 6 (below).

On average, the pension fund can now expect to be unable to match nominal cashflows. Importantly, under these circumstances the uncertainty faced about the magnitude of the potential mismatch is several times what it would be if the switch were to be to straightforward CPI indexation.

It is evident that, with regards to how this legislation will affect the UK pensions industry, the devil is in the detail. Changing the inflation rate to which pensions are linked from the RPI to the CPI raises important issues for risk managers. Due to there being material differences in the way that the two indices are constructed, questions need to be asked (and answered) with regards to the management of an inflation-linked liability that cannot currently be directly hedged with liquid market instruments. Furthermore, the impact of the proposed change will be highly dependent upon the detail of how it is implemented in practice. Until these issues are resolved it will remain difficult to gauge precisely what the outcomes of this new legislation will be. Pension funds currently face elevated uncertainty about the nature of the inflation risks that they bear.


Harry Hibbert is an analyst and Rudolf Puchy a senior consultant at Barrie & Hibbert