CPI or RPI? Fairness or sleight of hand?
The following article was just published by Citywire. In it Stuart considers the merit of the Government’s change in the indexation of certain state and private benefits, from RPI to CPI, which has been greeted with suspicion.
The Budget announced a change from RPI to CPI as the basis of future inflation adjustment to certain benefits, including the state pension. Subsequently the DWP announced that it would use CPI, not RPI as currently, as the basis of Limited Price Indexation for private pensions and of indexation of public-sector pension benefits.
The Government’s message is consistent, based on the context of the index use and therefore the components of the index. If benefit claimants do not own their own home, mortgage costs and housing depreciation are not relevant but rent is. If recipients of public sector pension do not have a mortgage, mortgage costs are not relevant. By emphasising the home ownership components, it has (possibly unintentionally) led people into thinking the inclusion of home ownership costs explains the past higher trend increase in the RPI than the CPI and is the basis of a systematic and sustainable bias.
Media reporting has picked up on this implication and reported it without question. But it misses an important point. The reason why, other things being equal, we should expect RPI to lead to higher levels (and so increase the cost to the taxpayer of providing indexed benefits) is that it uses arithmetic means whereas the CPI uses geometric means. The NSO has quantified the effect since the CPI was introduced in 1997 as an average downward bias of 0.5% pa. This is almost exactly the actual difference in the indices that has been widely attributed in the media to the housing component.
House prices affect both the mortgage interest payments and the housing depreciation components of the RPI. They affect mortgage costs because the NSO attempts to capture the changing national average mortgage level as the base for multiplying by a current interest rate, as distinct from the unchanged mortgage of the same typical household. Similar tricky concepts influence the differences in competing house price indices. House prices affect the depreciation component (whose presence I will not try to explain) similarly although the effect ought to be less because it separates the building element from the plot. Both mortgage interest payments and depreciation have a weight of about 5% in the RPI. Rents have a similar weight but are common to both indices.
Over the long term, it is reasonable to expect house prices to rise faster than general costs, because of the links to earnings, via credit. A realistic long-term trend in relative prices is about 2% pa. I suspect it would be less (it was zero in the US until the late 1990s) if development land were more freely available. Building costs ought not to have a trend different from general prices. With a mean interest rate change of zero, this translates into a bias between 0.1 and 2% pa, mainly depending on the dodgy depreciation component. Clearly, the component difference between the two indices is smaller than the computational difference and also overwhelms the debate about which index is right according to the context.
For pensioners in particular, who either rent or own homes but do not typically have mortgages, including or excluding a 5% weight for mortgage interest payments, however logical, is trivial relative to the computational bias.
There is a debate to be had about the most appropriate index based on its computation as well as its components, so that we can be clear that the change is fair to both beneficiaries and taxpayers.
Generally, we might expect statisticians to prefer to compute a series with a mean change different from zero using geometric means. In the specific case of the CPI, it also applies a concept that consumption patterns change with shifts in relative prices, which seems intuitively more accurate and would not arise if aggregating arithmetic means. I can see why a fair-minded, disinterested decision maker (a politician?) might opt for the CPI.
Understanding the concepts behind the Government’s change is important if reacting to clients’ questions and suspicions. Clients are not helped by comments such from Buck Consultants in the Sunday Times last weekend that ‘the change had effectively wiped £67,000 off the total value of the pension pot’ for someone earning £100,000 with 30 years service. For a start they assumed 0.75% pa difference in trend (where does that come from?) and in any case why assume this is their right?
Understanding the concepts is also important for assessing the impact on the ILG market. Here we need to separate the impact on investors betting against the implied inflation rate or breakeven rate of inflation from the impact on investors hedging the inflation risk implicit in their liabilities or goal outcomes.
The first ought to affect yields almost immediately, as a once off change. It would reflect the expected difference in RPI and CPI not as a long-term trend but over the term of their bet. This impact may now be complete.
The second poses a technical problem, until the market creates CPI as well as RPI swaps. But the impact on risk preferences is neutral. ILGs, whatever the index used, still represent a tight hedge for inflation compared with the very loose match provided by real assets like equities and property. The separation of risky bets and risk free hedges will still look to many clients like a better way to manage risks than relying on diversification effects from using more and more building blocks with less and less certainty about the correlations.