British Telecommunications Plc v (1) BT Pension Scheme Trustees Ltd & Anor

“Becomes” and “inappropriate” are words of ordinary English meaning. It is common ground that inappropriate means something more than less appropriate. Accordingly, it is not enough that it would be better to use another index, or that another index has become more appropriate, or that RPI is merely undesirable. I agree with the second Defendant that the hurdle is therefore a high one. The determination of whether RPI has become inappropriate, however, cannot take place wholly without regard to possible alternatives. If, for example, either there was no other suitable index to replace RPI, or no other index had features of RPI that were particularly relevant to the protection of pensioners’ benefits under the Scheme, then it may be more difficult (although, even then, not impossible) to conclude that it was inappropriate to continue using it.

The concept of appropriateness does not exist in a vacuum, but relates to some other thing or purpose. In the case of the 2016 Rule, “becomes inappropriate” relates to the purpose for which the Rule exists, namely calculating an increase in pension payments so as to reflect increases in the cost of living of pensioners receiving benefits under the Scheme. I agree with the warning of Warren J in the Thales case (at [20]) against abstract appeals to the purpose of provisions being to “protect members from the effect of price inflation”, as these are empty without an attempt to see what provision is actually made as to how that should be done. (That does not preclude, however, reference to the purpose of the Rule in order to provide the parameters of the concept of “inappropriate”.)

This point is particularly apposite, given that much of the expert evidence (to which I refer in detail below) necessarily focused on the question of RPI’s appropriateness, generally speaking. To take an extreme example, even if RPI was universally regarded as inappropriate for all other uses, if it nevertheless remained appropriate for the purposes of measuring increases in private pensions, then the gateway in the rule would not have been passed. Conversely, if the population coverage of RPI was amended to exclude pensioners, and it could be shown that pensioners’ spending habits were substantially different to those remaining within coverage, then although RPI might remain appropriate for everyone else, it could well be said to have become inappropriate for the purpose of measuring increases in pensions payable under the Scheme.

I was referred by both the second Defendant and BT to the Oxford English Dictionary definition of “inappropriate”, namely “Not appropriate; unsuitable to the particular case; unfitting; improper.” When faced with a word in common usage such as “inappropriate”, there is in my judgment little to be gained from, and potential for confusion in, replacing it with another word of similar, but not necessarily identical, meaning found in a dictionary definition. The second Defendant seizes in particular upon the word “improper” in the dictionary definition, contending that the dictionary definitions point to a significant sense of impropriety in stigmatising something as inappropriate. This, in my judgment, demonstrates the risk of placing too much emphasis on a dictionary definition. While the word “improper” clearly imports a sense of impropriety, there is no such inherent sense in the concept of “inappropriate”.

The second Defendant submits that provided there is a respectable body of opinion that maintains that RPI is a suitable basis for indexing pensions, then it cannot be inappropriate to continue using it for that purpose. I do not think the position is as clear cut as this. On the assumption that my conclusion in respect of Issue 1 is correct, then the question is whether RPI has become, or has not become, inappropriate. The question is a binary one, and it is one for the Court. In the same way that, if a respectable body of opinion believes that RPI has become inappropriate, the Court is not bound to agree, the fact that a respectable body of opinion maintains that it is appropriate does not dictate the Court’s answer. A more nuanced point can, however, be made: the fact that a respectable body of opinion believes that it is appropriate to continue using RPI for at least some purposes is a factor to be weighed in the balance when deciding whether RPI has become inappropriate for the purpose of uprating pensions under the Scheme. I note that if my conclusion on Issue 1 is wrong, then the fact that a respectable body of opinion believes that RPI has become inappropriate would make a challenge to a decision by BT to that effect very difficult to sustain under the Wednesbury test.

The second Defendant also submits that the scope of “becomes inappropriate” is substantially narrowed by reason of the fact that the first limb of the gateway is that RPI ceases to be published. It is said that RPI becoming inappropriate should therefore be construed as the next worst thing to RPI ceasing to be published. Mr Furness submits that, together, the two limbs of the gateway indicate an intention that the use of RPI should only be discontinued if it becomes “effectively impossible”. In my judgment this goes too far. While I can readily see that the drafter might have intended that the next worst thing to RPI ceasing to be published should also trigger BT’s entitlement to adopt a different index, that in itself provides no clue as to the intended breadth of that next worst thing.

It is common ground that the requirement that RPI “becomes” inappropriate denotes a change in appropriateness. I have already dealt with the question whether the Rule is in this regard forward-looking, whether from 2002 (on its first implementation) or from 2016 (on replacement of the former version of the Rules). The dispute between the Claimant and the second Defendant on this issue is as to the nature and scope of the permissible changes.

The second Defendant contends that in order for RPI to have become inappropriate, it is necessary to identify either a direct change to the composition of RPI, or an indirect change to RPI such as a change in its status, for example its de-designation as a National Statistic. While the second Defendant accepts that the views expressed by expert commentators may be taken into account by the Court in determining whether the direct or indirect changes to RPI mean that it is now inappropriate, the fact that those views have been expressed, even if the commentators have an official capacity such as in the case of the National Statistician or the Executive Director of the Royal Statistical Society, is not in itself sufficient to constitute a change.

BT contends that it is unnecessary for there to have been any change, whether direct or indirect, to RPI itself in order for it to have “become” inappropriate. Instead, it would be enough if the consensus of opinion among commentators as to the appropriateness of RPI had changed. In that event, BT contends, RPI could be said to have become inappropriate within the meaning of the 2016 Rule.

Given my conclusion that the phrase “becomes inappropriate” in the 2016 Rules does not mean “becomes inappropriate since 5 April 2016”, this point is largely academic. That is because, although the only matters relied on by BT, since April 2016, as potentially causing RPI to have become inappropriate are views expressed by various experts, the matters relied on prior to April 2016 include the de-designation of RPI as a National Statistic (which occurred in March 2013).

Nevertheless, in my judgement, taking into account that the purpose of the 2016 rule is to ensure that at all times members benefits are to be increased by reference to an index which is appropriate, it is at least theoretically possible for RPI to have become inappropriate without there being any direct or indirect change according to the second Defendant’s understanding of those terms. To take an example, it is acknowledged by all parties that the fact that RPI is calculated using a method of aggregation of data called the “Carli” method is a serious flaw. It is also acknowledged by all parties that the Carli method has been integral to the calculation of RPI for many decades. According to the second defendant, even if all commentators now took the view that because RPI was calculated using the “Carli” method it was inappropriate as a tool for measuring inflation for any purpose, nevertheless it would not have “become inappropriate” within the meaning of the 2016 rule. According to BT, on the other hand, consequential matters involving changes in views expressed by the UKSA, ONS and other commentators and changes in use of RPI elsewhere, could be sufficient to cause the gateway in the 2016 Rule to have been passed through. I agree with BT’s interpretation of the 2016 Rule in this respect. It may be sufficient, to demonstrate that RPI had “become” inappropriate, that there had been a wholesale loss of confidence in it as an accurate measure of inflation and its widespread discontinuance. The point is, however, likely to remain purely theoretical, since it is unlikely that there would be such consequences without RPI also being de-designated as a National Statistic, which the second Defendant accepts is the sort of indirect change which could cause the gateway to be passed through.

Both parties referred me to the Thales decision (above). That case was also concerned with the interpretation of a gateway provision allowing RPI to be replaced for the purposes of indexing pensions. The question was whether the compilation of RPI had been “materially changed”. Warren J, at [43], concluded that this language precluded interpreting the gateway as permitting RPI to be replaced where, although the change in compilation was inconsequential, it had a material impact on the way in which RPI behaved. He went on to note that had the relevant wording been “a material change in the index” then that might have encompassed a change in the effect of the index. I do not regard that decision, focusing as it was on entirely different wording, as being of any assistance in the construction of the 2016 Rule. For completeness, I was also referred to a handful of other recent cases where a similar issue has arisen as to whether RPI should remain as the index for uprating pensions. The terms of the relevant rules, and the precise issues which arose for decision in those cases, are materially different, and there is no assistance to be gained from the decisions in them.

Drawing these threads together, and taking account of my conclusion on Issue 2, I conclude that in order for RPI to have “become inappropriate” two things must be satisfied. First, that RPI is now in fact inappropriate (and not merely less appropriate than any alternative index) for the purposes of calculating increases in pensions payable to members of the Scheme to reflect the inflation experienced by those members and, second, that this constitutes a change since 2002.

It is BT that contends that the default position (that RPI applies) is to be rejected, and it is BT that accordingly bears the burden of establishing that RPI has become inappropriate. As Mr Spink submitted, this is a multi-factorial question, requiring an examination of the benefits, disadvantages and changes in respect of RPI over two decades.

Each of BT and the Second Defendant called expert evidence. Mr Paul Johnson, called by BT, and Mr Simon Briscoe, called by the second Defendant, are undoubted experts in the fields of statistics and economics. Mr Johnson is director of the Institute for Fiscal Studies and was formerly director and chief micro-economist at HM Treasury and deputy head of the Government Economic Service. In May 2013, he was asked by the chairman of the United Kingdom Statistics Authority (“UKSA”) to carry out a review of the UK price indices. He published a far-reaching report in January 2015. Mr Briscoe sits on the Council of the Royal Statistical Society (“RSS”), has worked in the civil service, investment banking, and as statistics editor of the Financial Times over a three-decade career as a statistician and economist, and has written several books on economic statistics.

The experts have provided a wide-ranging review of the history, and respective merits, of RPI and its main rival, the consumer prices index (“CPI”). Subject to important qualifications which I return to below, this evidence is of considerable assistance in two respects.

First, in providing a comprehensive picture of how price indices are compiled, the various purposes for which they are produced, and of the advantages and disadvantages of RPI in that context. An understanding of these issues is critical, since the ultimate question to be decided is whether – having regard to the purpose for which reference to an inflation index is necessary under the 2016 Rule and the package of merits/disadvantages of RPI – RPI has become inappropriate for that purpose.

Second, in identifying the materials relied on by either side as relevant to the ultimate issue, including materials emanating from statutory bodies such as the UKSA and the Office of National Statistics (“ONS”) and papers from other respected commentators in the field.

In both respects, there was much common ground between the experts, particularly relating to the history of RPI and CPI, and the differences between them.

RPI is the UK’s oldest price index. According to a paper produced by the ONS in 2010, it began life as a compensation index, developed as an aid to protect ordinary workers from price increases associated with the First World War. RPI was made an official inflation measure in 1956.

CPI was introduced in 1996 as the harmonised index of consumer prices (HCIP) in response to EU regulation. It follows international legislation and guidelines.

The basic approach to the measurement of inflation adopted by both RPI and CPI is the same: they track the changing cost of a fixed basket of goods and services over time. There are, however, differences between them relating to population coverage, item coverage and, most importantly for present purposes, in the formulae used at the elementary aggregation of data stage.

So far as population coverage is concerned, RPI excludes the top 4% of households by income, and pensioner households dependent on state benefits. CPI includes the expenditure of all UK households.

So far as item coverage is concerned, RPI (but not CPI) includes expenditure of UK households abroad, mortgage interest payments, council tax, house depreciation, buildings insurance, house purchase costs, TV licences, road fund licences. On the other hand, CPI (but not RPI) includes spending by overseas visitors to the UK, stockbrokers fees, university accommodation fees, foreign student tuition fees and unit trust fees.

The use of formulae at the elementary aggregation of data stage is dictated by the lack of data of a sufficient level of particularity to be able to attribute weights to different individual items. There are a considerable number of formulae that could be used, but for present purposes two are the most important. The Carli formula and the Jevons formula. The essential difference between the two is that Carli uses an arithmetic mean whereas Jevons uses a geometric mean. Mathematically, for a given set of data, a geometric mean will always produce the same or a lower number than an arithmetic mean.

RPI uses, to a significant extent, the Carli formula. CPI uses almost exclusively the Jevons formula. The experts are agreed that, as a consequence of using Carli, RPI has an “upward bias” such that it is consistently higher than CPI, and other formulae based on a geometric mean. This is referred to as the “formula effect”.

In comparing formulae, three approaches are commonly used, the economic approach, the stochastic approach and the axiomatic approach. Whereas the economic approach had formerly been thought to support Jevons over Carli, in 2012 the ONS changed its view and concluded that no weight could be placed on the economic approach when choosing between the two indices. It is common ground that the stochastic approach provides no basis for preferring either Carli or Jevons.

The axiomatic approach posits a number of desirable mathematical properties for index numbers. These form tests (or axioms) against which alternative formulae can be ranked. It is common ground that the Carli formula fails three of the tests set out in the consumer price index manual for the International Labour Organisation, known as the transitivity test, the time reversal test and the price bounce test. The same features that cause Carli to fail one test cause it to fail all three. For this reason Mr Briscoe accepts that on the axiomatic approach RPI can be seen to be deficient when compared to CPI. He contends however that the real world impact of this deficiency is very slight. In particular, there is very little real-world evidence as to the impact which Carli’s failure to comply with these axioms has in practice. He also considers that there is no universal agreement as to whether the axiomatic approach is appropriate, or which axioms should be included or are the most important within it.

The formula effect is not new. In a paper published in 1999 by David Fenwick of the ONS, not only was the formula effect arising from the difference between a geometric mean and an arithmetic mean at the level of aggregating prices noted, but it was specifically noted that the formula effect in the UK was particularly great, and that this resulted at least in part from the relatively broad item descriptions used in the UK. Mr Fenwick noted the increase in the formula effect in recent years, and that a significant cause of this was likely to have been the introduction in 1996 of random sampling of outlets and locations. It was also noted that the formula effect was particularly pronounced for clothing.

By 2002, upon introduction of the 2002 Rules containing, for the first time, a rule in the same form as the 2016 Rule, the formula effect was well established, and was relatively stable, resulting in an upward bias in favour of RPI, as against CPI, of about 0.5%. In a document headed “the new inflation target” produced by HM Treasury in December 2003, it was noted that since 1997 when the ONS first published CPI, the formula effect had been stable and had contributed around 0.5% on average to the gap between CPI and RPIX (RPI excluding mortgage interest payments). The document recorded the Chancellor’s confirmation that the new inflation target would be based on CPI. The reason given was the importance of monetary policy decisions being based on the most relevant and accurate measure of inflation. RPI had long been designed to meet a variety of needs and this had necessitated some long-standing compromises in conceptual consistency. It was noted that in the last decade a global consensus had begun to emerge about the desirable form of consumer price indices appropriate for measuring inflation at the macroeconomic level. “This consensus has helped shape the CPI during its development, meaning that it has some distinct advantages over RPIX as a macroeconomic indicator of inflation, partly reflecting the fact that the latter was not developed specifically for this single purpose.”

The formula effect remained stable, contributing approximately 0.5% to the gap between RPI and CPI from 1997 until 2010, as confirmed by a study carried out by Ruth Miller for the Office of Budget Responsibility dated November 2011.

In 2010, as part of the numerous routine changes made to both CPI and RPI from year to year, the ONS implemented what were intended to be improvements in the collection and use of clothing prices. As explained by Mr Johnson (and agreed by Mr Briscoe), in collecting and comparing prices, a definition is needed for what counts as the same good one month to the next. If a price collector cannot find the same item in a shop this month as they found last month, then a new item is selected and the new index starts. Because of frequent changes in styles and fashions a large fraction of clothing items were classed as not comparable between one month and the next. The change implemented in 2010 consisted of guidance so as to allow more direct month-to-month comparisons to be made. In other words more items of clothing were henceforth to be considered comparable to one another.

What was intended as a routine improvement, however, had a dramatic effect. Between 2001 and 2009 clothing and footwear prices in the CPI fell by 5.3% on average, and by 2.5% on average in the RPI. In 2011 they rose by 2.3% in the CPI and by 11.5% in the RPI. More importantly, for present purposes, the clothing change resulted in an increase in the formula effect of between 0.4% and 0.5%. Since 2010 the formula effect has resulted in RPI being on average between 0.9% and 1% higher than CPI.

The impact of the clothing change on the formula effect is central to BT’s contention that RPI has become inappropriate.

There are important limitations on the usefulness of the expert evidence in this case.

First, much of the expert evidence was directed at the respective merits and disadvantages of RPI, as compared to other indices, in particular CPI. To the extent that the experts debated aspects of RPI or CPI which are not relied on by BT as demonstrating that RPI has become inappropriate, the evidence is largely irrelevant. As I have already noted, the question is not whether RPI is preferable to CPI, or vice versa, but whether RPI has become inappropriate for the purpose of calculating pension increases under the Scheme within the meaning of the 2016 Rule.

I include the following matters as falling within this category:

The second limitation on the expert evidence relates to the extent to which they engaged with the ultimate question, that is whether RPI has become inappropriate. I do not criticise either expert for this, as they were specifically requested to do so in the knowledge that the question was not one for them but for the court, but in circumstances where it might nevertheless provide some assistance to the court. In very short summary, it is Mr Johnson’s view that RPI is flawed for all purposes, including therefore for the purposes of the 2016 Rule, largely because of the ‘formula effect’, which has widened and become more apparent as a result of the clothing change in 2010. He considers that CPI, while not perfect, is the best measure available. Mr Briscoe, however, believes that RPI remains not only an appropriate, but currently the best, index for measuring increases in inflation for the purposes of uprating pensions.

In the end, I find that little assistance is to be gained from the respective experts’ views on the ultimate question for two reasons. First, because their views on the “appropriateness” or otherwise of RPI, generally or in respect of uprating pensions generally, is of only limited relevance to the question of its appropriateness specifically for the purpose of calculating pension increases under the Scheme. Second, because the question whether RPI has “become” inappropriate – again with specific reference to the meaning of that phrase in the 2016 Rule – is not one in respect of which either expert can claim any particular expertise.

BT challenged Mr Briscoe’s credibility as an expert in a number of respects. Mr Spink pointed to one aspect of Mr Briscoe’s written evidence in particular – namely his reliance on an article by the journalist Chris Giles as supporting his (Mr Briscoe’s) view that de-designation of RPI as a National Statistic had been wrong. It is plain, on a proper reading of Mr Giles’ article, that he provided no support for Mr Briscoe’s position, but was in fact criticising the UKSA for not having gone far enough, having recognised the flaw in the Carli formula, in ‘ditching’ its use in RPI. It was also said that Mr Briscoe showed an unwillingness to agree with reputable experts, and misinterpreted or refused to accept the obvious meaning of certain documents, for example when Mr Briscoe interpreted the National Statistician’s letter of 9 March 2016 (referring to RPI being maintained so as to be fit for purpose) as a reassuring sign that RPI is here to stay. While I accept that Mr Briscoe’s reliance on Mr Giles’ article was misplaced, and he was in some difficulty in recognising it, I do not accept that this affected the credibility of his evidence to the extent it was relevant to the issue I need to decide. The fact that he disagrees with other reputable experts emphasises that this is an area on which such experts disagree (as evidenced by the long-standing disagreement – which I explain further below – between the RSS and the UKSA on certain key issues). Insofar as Mr Briscoe exhibited a reluctance to agree with one or other interpretation of certain documents, the documents speak for themselves and the interpretation of them by the experts does not add materially to the relevant evidence.

The second Defendant criticised aspects of Mr Johnson’s evidence, contending that he tended to overstate his objections, and that he gave his evidence from the perspective of an economist, and was thus overly-focused on the macroeconomic needs of an index. I do not accept these criticisms. Nevertheless, for reasons I have just discussed, the fact that Mr Johnson credibly and forcefully opines on the inappropriateness of RPI in general is not determinative of the question that I have to decide based upon the interpretation of the specific provision in the 2016 Rules.

Before addressing directly whether each of the specific matters relied on by BT, whether individually or cumulative, has caused RPI to become inappropriate, I will summarise the most important materials identified by the experts as relating to this issue.

In the Chancellor’s budget report of June 2010, the government announced that it would use CPI for the price indexation of benefits and tax credits from April 2011. The stated reasons were that CPI was believed to provide “a more appropriate measure of benefit and pension recipients’ inflation experiences than RPI, because it excluded the majority of housing costs faced by homeowners (low income households are subsidised separately through Housing Benefit, and the majority of pensioners own their home outright), and the differences in calculation mean it may be considered a better representation of the way consumers change their consumption patterns in response to price changes. This will also ensure consistency with the measure of inflation used by the Bank of England.” It was noted the change would also apply to public service pensions through the statutory link to the indexation of the Second State Pension. I note that the reasons given were not the same as those relied on by BT for showing how RPI has become inappropriate.

In October 2012, as a direct result of the impact which the clothing change had on the formula effect, the National Statistician consulted on options for improving RPI. The consultation was specifically concerned with the formula effect. Respondents were told that the consultation was on statistical methodology and best practice, although the consequences of any potential change were also noted. At paragraph 24 of the consultation paper, the main differences between CPI and RPI were said to be: (1) RPI’s exclusion of very high and low income households; (2) the exclusion by CPI of owner occupiers’ housing costs; and (3) the formula effect. In table 2 of the paper, the main strengths and weaknesses of Carli and Jevons were set out. Whereas Carli is described as passing most of the usual tests for a price index, Jevons was described as passing all of those tests. Carli’s weaknesses include, as well as failing some of the tests, being susceptible to price bouncing and having an upward bias. The only weakness identified in respect of Jevons was that if any one observation out of the set of observations is zero, their geometric mean is zero, whatever the value of the other observations. Based on this, the paper commented at paragraph 39 that “from an axiomatic point of view the Jevons index seems to be the index with the best properties and the Carli the weakest”.

In paragraph 32 of the paper it was noted that the primary cause of the formula effect is in relation to the collection of clothing data, referring to the widening of the gap from 0.5% to around 1% since 2010. It was noted in paragraph 33 that some of the differences between RPI and CPI counteract the impact of the formula effect. At paragraph 43, the ONS stated its conclusion that the choice of appropriate elementary aggregate formulae in CPI and RPI should be based on statistical considerations and empirical evidence. It suggested that the continued use of Carli in RPI should be reconsidered. Four options were presented: (1) no change, (2) changing one particular approach to averaging changes in price, for clothing; (3) changing one particular approach to averaging changes in prices, for all items; (4) changing RPI so that the formulae aligned fully with those used in CPI.

In advance of the consultation, the ONS had commissioned a detailed report by W.Erwin Diewert in 2012. This made a number of short and long-term recommendations on possible methodological improvements that could be made to the UK’s consumer price statistics. The most important recommendation for improvement was that RPI should drop its use of the Carli formula and replace it with either Jevons or another formula. The principal reason given for this recommendation was the upward bias in the Carli index. A number of other lesser recommendations were made including that the scope of RPI should be changed to cover the expenditures of all households in the UK on consumer goods and services.

A paper dated November 2012 by Peter Levell for the Institute for Fiscal Studies considered the case for replacing the Carli formula in RPI. It challenged certain of the conclusions reached by Diewert in his paper for the ONS earlier that year. Addressing each of the three approaches for comparison, namely the economic approach, the stochastic approach and the axiomatic approach, it agreed that the stochastic approach did not clearly favour one index over another, but considered that the economic approach did favour the use of the Jevons index. So far as the axiomatic approach was concerned, Mr Levell concluded that although a case remained against Carli, its failure to satisfy the time reversal test was perhaps not as serious as had been suggested. That was because although the failure to satisfy such a test was a problem, the index numbers into which the elementary aggregate eventually fed in both RPI and the CPI were themselves not time reversible. He also noted, however, that Carli failed a new, revised version of the price bouncing test.

The Pensions Management Institute provided a response to the consultation in October 2012, opposing the proposed changes in options 2 to 4, all of which would seem to reduce RPI and thus reduce the level of pensions payable in the future. The point was made that some pensioners had already chosen to purchase RPI increases at retirement in preference to either no increases or fixed rate increases, and that they would feel particularly penalised by changes known to reduce RPI.

The Association of British Insurers provided their response to the consultation in October 2012. It also opposed any change. It noted that it had been known since 1996 that RPI tended to produce higher answers than CPI as a result of the formula effect. Moreover the vast majority of RPI contracts, benefits and obligations had been instructed and built up, since 1996, in that very knowledge: “to change the basis of calculation is to change the value of the property rights between the parties that made the contracts and their associated benefits and obligations.” Whilst recognising that the increase in the ‘wedge’ between CPI and RPI that followed the changes in both indices was the legitimate focus of concern and the proper focus for action to remove unjustified differences, it noted that if technical problems had arisen in RPI, then an appropriately targeted technical solution was required.

The director-general of Saga, Dr Ros Altmann, provided a further response to the consultation, in which she made the following points. First, she noted the different purposes of RPI and CPI, the former having been designed to reflect the rising prices of goods and services consumed by UK households, the latter having been introduced in order to help with international comparisons of inflation and to feed into macroeconomic policy. Second, she noted that with all statistics there is no perfect measure, all are estimates, and whilst the ONS suggest that RPI is wrong, there was a real possibility that the CPI measure might be under-recording domestic inflation. In other words neither is right and maybe both needed changing. Third, and focusing particularly on the formula effect, she considered that the geometric mean (used in the CPI) is not always the best measure, and may be particularly unsuitable in the case of pensioners. That is because it is intended to operate as an approximation to substitution, i.e. when prices rise consumers move to an alternative that has not increased in price so much, whereas pensioners are exposed to proportionately higher price rises if they are less mobile between providers due to best and cheapest deals being only available online or in larger stores inaccessible to those who have no Internet or are less mobile. Moreover, while Carli may have an upward bias, Jevons may have a downward bias. That will be so if low-priced items rise most. Fourth, specifically in relation to the clothing change, she considered that while RPI may have overestimated inflation as a consequence, CPI may have underestimated inflation noting that, as the price of cotton (the raw material used in much clothing) rose sharply during the period 2010-2011, the price of cheaper clothing will have risen far more than the price of more expensive clothing, whereas the CPI suggested no price rise over that period. Again, this was a factor which was likely to have affected pensioners more than others.

The RSS’s response to the consultation, published in November 2012, also took a more conservative approach to potential change. It noted that the RSS had argued for some time that the formula effect should be eliminated or reduced to a minimal level, but the ultimate aim was to measure inflation as accurately as possible. It noted that there are a number of steps that could be taken to eliminate the formula effect, relating for example to price collection methods, choice of base month and sampling scheme. It recommended waiting a further year before making any substantial changes: “…It is clear to us from the work carried out so far that the accuracy of both CPI and RPI could be improved by implementing the results of the research programme and that changing the RPI now to bring it on a par with the CPI in terms of aggregation formula would be premature.” In particular, the RSS considered it unbalanced to consider changes to the RPI formulae without considering whether changes need to be made to the CPI formulae, as well as other issues. Addressing specifically the price bounce test, the RSS noted that while the problem of price bounce in Carli had been highly publicised, no assessment had been made of its impact, and that its belief was that while it was definitely a factor, and an undesirable factor, it may be less important than some commentary suggested.

The ONS provided its recommendations in response to the National Statistician’s consultation, so as to allow the CPAC to advise the National Statistician and the UKSA in January 2013. Table 1 of the document prepared by the ONS summarised responses received. Of a total of 406 responses, 82% were in favour of the first option set out in the National Statistician’s consultation paper, namely no change, leaving the formula effect as it was. At paragraph 16 however it was noted that the large majority of responses “…did not address methodological issues, but identified the impact that the changes implied by options 2-4 would have for them”. These were primarily in terms of the anticipated yield on investments and in the value of pension income. The ONS’s recommendations included: (1) that while Carli would not be chosen were ONS constructing a new price index, the formulae used at the elementary aggregate level in RPI should remain unchanged; (2) a new RPI based series should be developed, called RPIJ, using the Jevons formula in place of the Carli formula; (3) “That the basic formulation of RPI is accepted as currently defined and that any future changes should be limited to issues such as the annual update of the basket and weights, improvements to data validation and quality assurance etc.”

The first of these recommendations was expanded upon at paragraph 28 as follows: “ONS has concluded that there are sufficient arguments to support continuing production of the RPI on the current basis and recommends the formulae used at the elementary aggregate level in the RPI should remain unchanged. There are counterbalancing arguments for and against this outcome many of which favour option 3. However, the weight of users’ requirements and the desire for continuity in a long-run time series all militate against change.”

At paragraph 29, the reason given for the third of these recommendations was that the “ONS recognises that users of RPI will want to avoid uncertainty about potential changes in the future.” The third recommendation is what has since become referred to as the “freeze” in RPI.

The outcome of the National Statistician’s consultation was announced in a press release in January 2013. It stated that the National Statistician had concluded that the formula used to produce RPI “does not meet international standards and recommended that a new index be published”. It announced that RPIJ would be published from March 2013. It further announced that the “National Statistician also noted that there is significant value to users in maintaining the continuity of the existing RPI’s long time series without major change, so that it may continue to be used for long-term indexation and for index linked gilts and bonds in accordance with user expectations.” Accordingly, the national statistician recommended that the formulae at the elementary aggregate level in RPI should remain unchanged.

In a further document dated January 2013, related to the consultation, the ONS provided both plain English and statistical answers to a series of questions posed by respondents of the consultation.

In response to a question as to why the consultation had not been run 20 years ago, given that the issues with the index numbers had been known “for nearly a hundred years”, the ONS said that although the theoretical causes of the formula effect had been known for a number of years, the collective impact of these causes had been relatively stable (approximately 0.5%). Three factors had caused ONS to look again at the issue. First, the increase in the formula effect gap during 2010 as a result of the improvement in the measurement of clothing prices. Second, the changing user requirements for the CPI and RPI following the government’s decision to switch the indexation of many pensions and benefits to CPI. Third, the ONS had developed and uncovered new evidence, for example the analysis of consumer behaviour and its potential impact on the choice of formula plus a review of practices in other countries.

In response to questions about the upward bias of RPI as a result of the use of the Carli formula, the ONS noted that it is not possible to state categorically that either RPI or CPI is biased as a measure of the true rate of inflation, given that inflation is a latent variable. The problem with the Carli formula was that it produces higher results than one would intuitively feel reasonable. The example given was that if a set of prices might increase and then fall back to their original level then, intuitively, an index calculated using these prices should return to its original value. However, the Carli index suggested there had been a price increase.

In March 2013 the UKSA published its assessment of compliance with the Code of Practice for Official Statistics. This was a report prepared under the provisions of the Statistics and Registration Service Act 2007, which allows the UKSA to reassess whether the code of practice continues to be complied with in relation to official statistics already designated as National Statistics. The report covered RPI. In its summary of conclusions it stated as follows:

“1.2.1.

The statistics authority judges that the RPI … does not comply with Principle 4 and specifically with Principle 4, Practice 5 of the Code. This view is based primarily on: i) the finding that the methods used to produce the RPI are not consistent with internationally recognised best practices; and ii) the decision to freeze the methods used to produce the RPI and only to contemplate “routine” changes.

1.2.2

The statistics authority notes and supports the decision by the National Statistician that, to meet the needs of existing users of the RPI in its current form, ONS will not amend its basic formulation. This has the effect that the RPI is inconsistent with the code of practice.”

At paragraph 3.4 of the document, it was noted that Principle 4 of the code requires that statistical methods be consistent with internationally recognised best practices, that research of other national statistical institutes had shown that only the UK was found to use the Carli formula, that the Carli formula was the primary source of the formula effect, and that “this formulation does not meet current international standards”. Accordingly, RPI did not comply with the code.

At paragraph 3.5, the ONS’s recommendation to “effectively freeze” the formula used in RPI, by limiting changes in the future to issues such as the annual update, and improvements to data validation and quality assurance, was said to be inconsistent with the requirement in paragraph 5 of Principle 5 of the code (“Seek to achieve continuous improvement in statistical processes by, for example, undertaking regular reviews of releasing statistical work in progress such as experimental statistics”).

As a consequence of failing to meet the requirements of the code, paragraph 1.2.3 of the document stated that the UKSA “has cancelled the designation of RPI, including the sub- indices and variants listed in section 1.1.2, as National Statistics”.

On 15 December 2014 the ONS released a paper presenting analysis of the inflation rates experienced by different types of household in the UK. One of the tasks it had carried out was to compare the rate of inflation according to CPI (which it described as a ‘plutocratic’ measure) with a rate of inflation weighting each household equally (which it described as a ‘democratic’ measure). It concluded that CPI is broadly representative of the price experience of households around two-thirds of the way up the expenditure distribution, and that an equivalent democratic price index is on average 0.3% higher than the plutocratic measure over the period. I note in passing that neither RPI nor CPI is a democratic measure of inflation.

The UKSA had invited Mr Johnson in 2013 to conduct a review of UK price indices. His report was published in January 2015. It commented that inflation can mean a range of different things and that the consumer price statistics are used for a variety of purposes.

The report’s conclusion was that CPIH - once issues with the measurement of owner-occupier housing costs were resolved - could provide a good estimate of price changes across the economy. It recommended that the ONS adopt this as its main measure of inflation.

So far as RPI was concerned, the report recommended that the ONS and the UKSA should restate their position that RPI is a flawed statistical measure which should not be used for new purposes and whose use should be discontinued for all purposes unless there are contractual commitments at stake. Moreover, government and regulators should work towards ending the use of RPI as soon as practicable. The first priority should be producing CPI and CPIH to the best possible statistical standard.

The report noted the National Statistician’s decision to freeze RPI and commented that the logic of the current position was that RPI is purely a legacy measure, produced because of the needs of users who have long-term contracts. It meant that the improvements to the methodology of CPI and CPIH would not be carried over to RPI.

The report also considered whether ONS should publish an alternative measure to capture inflation as experienced by households, sometimes referred to as a household index or an operating index. It did not however recommend the development of a single household index, arguing that to do so would potentially create confusion and encourage inflation rate shopping which would be inappropriate for the operating of pensions and benefits and would be no better than CPIH for understanding price changes faced by groups of households. Instead it recommended that the ONS publish on an annual basis a set of inflation indices as experienced by a range of household types.

The report called on the UKSA to monitor use of consumer price statistics and speak out against inappropriate use, particularly of RPI. It recommended that the UKSA should encourage users to move to CPIH when it was reinstated as a National Statistic.

In March 2015 research undertaken by the Office for Budget Responsibility identified the various sources of the formula effect. This confirmed the conclusion reached by the ONS in its paper of 20 March 2012, that on average since 2010 the formula effect had accounted for 0.9% of the difference between RPI and CPI. Other components, listed as housing, coverage and weights tended to balance each other out to a significant degree.

In June 2015 in a working paper for the Centre for Research in Social Policy, entitled “Inflation and the Minimum Income Standard – Past and Future Measure”, Donald Hirsch of Loughborough University considered whether RPI remained a suitable index for the purposes of identifying the Minimum Income Standard. For present purposes, the importance of the paper lies in its conclusions as to the respective merits of RPI and CPI for measuring inflation among particular groups of users, specifically those on low incomes. Referring to the work undertaken by Dr Altmann in 2012 (referred to above) the paper pointed out that there are some circumstances (for example in relation to clothing, following the clothing change in 2010) in which CPI appears to be under-estimating inflation to a greater extent than RPI over-estimates it.

The UKSA issued a consultation paper entitled “measuring consumer prices: the options for change” in June 2015. It summarised the recommendations of the Johnson review.

In section 3 of the consultation paper, under the heading “RPI”, the UKSA stated that it held a clear position on RPI, set out in a range of statements since 2013, namely that it considers that the methods used to produce RPI are not consistent with international best practice and consequently that it is not deserving of National Statistics accreditation. It also noted that in the light of the continuing use of RPI in commercial contracts and index-linked gilts, the ONS could not simply discontinue RPI and they did not propose to do so.

In September 2015 the RSS provided a response to this consultation. It disagreed with the suggestion that the ONS should identify a main measure of price change across the economy. Instead it recommended there should be two: one of which would measure inflation from the perspective of the individual household (the uses of which would include allowing government to monitor the impact of inflation on individuals and households and for operating in respect of wages, pensions etc.); the second would satisfy macroeconomic needs such as inflation targeting.

The RSS identified the following differences between the macroeconomic index and a household index. “The former should be expenditure (plutocratic) weighted, the latter household (democratic) weighted or as near to this as possible. The former includes spending by overseas visitors, the latter should not. The former should not include interest payments; the latter is likely to include at least mortgage interest. The former considers the household sector as a whole; the latter individual households – this affects the treatment of owner occupier costs and insurance premiums in particular. The former, if it follows EU rules, should not include taxes other than those related to consumption. The latter should include council tax which is directly related to the consumption of household services. Coverage of owner occupied housing costs is a difficult area but a household index would almost certainly include some items considered inappropriate for a macroeconomic index. A macroeconomic index must accord with economic theory and its focus must be on consumers as a whole while a ‘household’ index should be compiled from the perspective of an individual household.”

It was noted that the CPI was compiled solely for comparison between EU countries, and was later used as the target for interest rate setting by the European Central Bank. It was not designed to measure inflation from a household perspective. Nearly all EU countries, and all the major ones, used their own national indices as their main operating index. The UK would therefore be out of line with most international practice if it adopted the CPI or a close derivative such as CPIH as its main operating index.

The paper referred to the findings of the ONS in December 2014 (above) as to CPI not being a suitable measure of inflation experienced by the median household, and went on to state: “While the RPI is expenditure-based, the exclusion of the top 4% of households by income, and of pensioners mainly dependent on state benefits, means that it is noticeably closer in practice to a household weighted index than the CPI.”

The RSS therefore recommended two indices: the CPI; and a household/household inflation index.

Commenting on RPI specifically, the RSS said: “The RPI was originally intended to measure inflation from the household perspective. To a large extent it could still fulfil the role of a household index but its use of Carli has made it unacceptable to the UKSA. It still enjoys far wider public confidence than the CPI, as demonstrated, for example, by the results of the 2012 consultation and the number of signatures to the e-petition protesting about the change from RPI to CPI for uprating public sector pensions. Some of this confidence may stem from the higher estimates of inflation it generally gives, or from its time-honoured status, but other factors, such as its inclusion of mortgage interest, also count. Whatever the reason, it can only be supplanted in public confidence by an index which clearly reflects actual household expenditures. This is an important point. It will be difficult, and take much time, to phase the RPI out not just because it is widely embedded in contracts but due to the extent of public confidence in it. If the UKSA wishes its use to decline as quickly as possible, an alternative the people can recognise as reflecting their own experience is vital. Neither the CPI, nor any index closely derived from it or established on similar principles, will do.”

Addressing the problem of Carli, the RSS said: “the perceived problems with the use of the Carli formula (and we note that there are credible challenges to the current expert consensus) could be addressed.”

Finally the RSS recommended against freezing RPI, commenting: “it will necessarily continue in use, and in wide use, for decades to come. It should therefore be improved and updated along with all other consumer price indices calculated in order, first, to follow changes in household purchasing behaviour and, second, to take advantage of technological improvements that are applied to other consumer price indices … The only constraint on changes to it should be those which would change its character in the ways proposed, and rejected, by the 2012 consultation, i.e. those which would change its character fundamentally, for example the wholesale replacement of the Carli formula by Jevons.”

The ONS’s technical advisory panel met in January 2016 to consider a paper presented by a Mr Mark Courtney in which he sought to justify the use of the Carli formula.

In his paper, Mr Courtney had stated the following conclusions. First, so far as coverage was concerned it was generally agreed that RPI coverage, including housing costs and excluding expenditure by the wealthy or the poorest pensioner households, was the most appropriate for an uprating index. Second, there were economic arguments that favoured Carli over Jevons. Third, the fact that Carli failed the transitive test meant that there was potential for an upward bias or downward bias to an index that used it, however both theoretical arguments and empirical evidence from the UK indicated that such bias was tiny overall. He therefore regarded RPI as rehabilitated: “it is the best consumer price index for uprating purposes that we have. The differences from the CPI can be ascribed almost entirely to underestimation by the CPI. Of course, the RPI is by no means perfect. There is a lot that can be done to make both the RPI and the CPI better. But, for the moment, the RPI is the best index that we have.” In a somewhat scathing attack on the UKSA, Mr Courtney found it quite extraordinary that RPI should have been downgraded from being a National Statistic. He recorded the fact that in December 2010 the UKSA had reviewed it and found it quite satisfactory subject to one requirement, that the ONS publish information and the reasons for the differences in scope and methods between the CPI and RPI and explain the implications that those differences had for the uses to which the statistics were put. Following publication by the ONS of such an explanation in October 2011, the UKSA had duly confirmed RPI as a National Statistic in January 2012. He questioned what had happened during the following year leading to the downgrading of RPI in March 2013, noting that the ONS had consulted on a proposal to remove the Carli formula from RPI but that in light of overwhelming responses in favour of retaining the existing formulae, had dropped that proposal.

The minutes of the meeting record that while panel members agreed that the relative merits of the mathematical properties of elementary aggregate formulae were finely balanced when considered from a theoretical standpoint, the performance of each formula in practice was an important consideration. In particular, the changes made to the collection of clothing prices in 2010 and the resulting large impact on RPI suggested that the Carli index was less suitable, and ONS research had shown the Carli index was usually higher than either the Jevons or Dutot formulae. This reflects Mr Johnson’s view: while the direct correlation between the failure of Carli to satisfy statistical tests and the uplift in RPI has not been established, he concludes that there is an upward bias in RPI “relative to most other plausible ways of doing it … relative to the best method we have of calculating inflation.”

The Claimant places particular reliance on a letter from the national statistician, Mr John Pullinger, to the chair of the UKSA dated 9 March 2016. The claimant’s focus is on two paragraphs on the second page of the letter specifically dealing with RPI. It is important, however, to read those paragraphs in context. The purpose of the letter, as stated in the first paragraph, was to update the chair of the UKSA on Mr Pullinger’s thinking for the future of consumer price statistics. He began by referring to the UKSA’s report of 3 March 2016 which included a number of actions that were needed in order for CPIH to regain its status as a national statistic. He then set out the approach he intended to take, subject to that work, under four points.

First, he was inclined to consider that CPIH should become the ONS’s preferred measure of consumer inflation and the focal point of ONS commentary in due course. He cited two reasons for this: that it is important that a measure of owner-occupiers’ housing costs is included in the measure which they made the focal point of their commentary, and it is important to focus on a measure that can continue to develop to meet the needs of UK users without being constrained by international regulations (as would be the case with CPI). He noted that this change to their commentary could not take place until CPIH was re-accredited as a national statistic.

Second, he said that he had listened to calls for a measure showing the effect of changes in payments for goods and services, referred to as a “household inflation index”. He considered that such an index would involve fundamental differences in a number of important aspects to the traditional measurement of consumer inflation, including the potential inclusion of asset prices and interest payments, plus giving each household expenditure equal weight. He noted the ONS would publish proposals for the timetable on the development of such an index in July 2016.

Third he noted that users had sought clarification on the future of RPI. He said: “put simply, I believe that the RPI is not a good measure of inflation and does not realistically have the potential to become one. I strongly discourage the use of RPI as a measure of inflation as there are far superior alternatives. Nonetheless, RPI is still used for a number of legacy purposes and its production is mandated by legislation. My intention is that from the start of 2017, ONS would publish the minimum of RPI-related data necessary to ensure the critical and essential needs of existing users are met.” Having said that for the avoidance of doubt RPIJ would no longer be published, he went on: “The RPI would continue to be maintained through routine changes. This covers all changes required to continue production of a consistent, fit for purpose RPI (for example the annual update of the basket and weights, computer systems upgrades and improvements to data validation and quality assurance methods). With due consideration to the requirements of the Statistics and Registration Services Act 2007, ONS would only consider making methodological changes to RPI if not to do so would inhibit the improvement of CPIH and the consumer prices index.”

Fourth, he said that ensuring CPIH was reaccredited as a National Statistic was the top priority in the short term, but that they would be preparing for other changes such as including council tax in the CPIH calculation which he thought was sensible.

On 28 April 2016 the Department for Work and Pensions issued a statistical notice. Having noted the Johnson review and the UKSA consultation in 2015, it concluded, at paragraph 8, that “the current position as set out above, however, makes an ongoing use of RPI in our publication untenable.”

On 25 May 2016 OFWAT announced its decision to move from RPI to CPI (or CPIH) with effect from the 2019 price review in relation to revenue indexation, and in phases from 2020 to 2025 in respect of its regulatory capital value indexation. This followed on from, and largely adopted the views expressed in, a report prepared by Oxera dated 31 March 2016. At section 3 of its decision document it explained the reasons for the change. These included, as summarised at paragraph 3.1.1, that “RPI is losing its legitimacy: it is statistically flawed, is discredited and has been discontinued as a national statistic.” As noted in paragraph 3.1, “for price controls to be legitimate, the measure of inflation used must also be legitimate. This means the measure of inflation used needs to be recognised and accepted by customers and their representatives. It means that the measure of inflation used must be statistically robust to maintain the credibility of the regulatory regime.” It was also noted that RPI is upwardly biased due to the calculation method used, and was increasingly falling out of use. Reference was made both to the fact that the basic state pension was operated by reference to CPI, and that the revaluation of pension protection fund compensation started to be calculated in line with CPI as of March 2011.

In October 2016 the ONS produced an article containing information about the users and uses of consumer price inflation statistics.

It identified the following users and uses of the CPI alone:

It identified the following users and uses of RPI alone

It identified the following users that used both CPI and RPI:

In November 2016 the ONS published an article detailing which RPI and related indices would be continued or discontinued in 2017. It stated that RPI was to be continued as it was used in a large number of commercial contracts including index-linked gilts. RPIJ, however was to be discontinued. The arguments against continuing to publish it included that it shared the weaknesses of RPI that went beyond the Carli formula, such as population coverage and use of a direct measure of house prices to estimate owner occupiers housing costs. Both RPI and RPIJ would continue to be maintained through routine changes.

The National Statistician, in a letter dated 9 December 2016, responded to comments received on his November statement from the RPI/CPI user group. Much of the letter concerns the relative merits of CPIH, and the extent to which users including economic regulators and pension industry experts had indicated a willingness to consider adopting it once re-designation had been achieved. Explaining its benefits, he commented as follows: “While the CPI may have had its origins in a desire to better measure inflation on a macroeconomic level it does not logically follow that it – or any variant – is only suitable for that purpose. This is particularly the case with CPIH, which is not bound by the same regulations as CPI and which can therefore be modified to better suit UK purposes, for example by the inclusion of council tax.” He went on to note that the ONS’s view was that neither mortgage payments nor house prices are a good measure of housing costs. Mortgage payments mix up how houses are financed with the price of housing, and house prices partly reflected that houses are an asset as well as a provider of housing services. That was a significant factor in deciding to cease publication of RPIJ. So far as RPI was concerned, the letter stated: “it is currently used in a number of long-term contracts so it would be impractical to cease its publication and will therefore continue for these legacy purposes.”

In March 2017 the ONS published a guide to changes in consumer price inflation statistics. It noted that certain RPI-related data will be discontinued although RPI would continue to be published. Paragraph 7 explained this resulted from the National Statistician’s letter of March 2016, stating that the ONS would publish only the minimum of RPI related data necessary to ensure that the critical essential needs of existing users are met.

On 31 July 2017 the director-general for regulation at the Office for Statistics Regulation wrote to John Pullinger, the National Statistician, to summarise the Office for Statistics Regulation’s review of whether CPIH had met the standards to have its National Statistics status restored. He concluded that it had, and he was pleased to confirm the re-designation of CPIH as a National Statistic.

On the same date a message was posted on the ONS website, commenting that CPIH was now “our lead measure of inflation”. It continued: “We will continue to produce RPI for legacy uses. However, the RPI is a flawed measure of inflation with serious shortcomings and we do not recommend its use.”

In July 2017, the Joseph Rowntree Foundation published its annual report on a Minimum Income Standard for the UK in 2017. Building on the contents of its 2016 Report, which itself referenced the work undertaken by (among others) Donald Hirsch (see above), it noted that RPI produced a better estimate of inflation for the purposes of a minimum income standard, but that in the case of clothing, comparisons of RPI/CPI with a re-costed minimum price basket resulted in a rate of inflation that was somewhere between the two. Accordingly, for the purposes of the minimum income standard, uprating in respect of clothing (between bi-annual re-pricing) would be undertaken on the basis of an assumption that prices had risen by the average of RPI and CPI.

In a letter dated 15 September 2017, the chair of the UKSA, Sir David Norgrove, wrote to Chris Giles, a journalist who had previously published articles critical of the UKSA’s stance in relation to RPI, and who had emailed the UKSA with specific concerns in July 2017. Mr Giles’ view was that the ONS was in breach of the Statistics and Registration of Services Act 2007 in accepting that RPI was not a good measure of inflation, but proposing to do nothing about it. He also considered that in failing to correct known errors in RPI, the UKSA was imposing large financial burdens on (among others) tax-payers, students with loans and rail users. Sir David reiterated that the UKSA had been consistent in its view that RPI is not an appropriate method of calculating inflation, and had acted accordingly to safeguard the quality of official statistics. He also said that he shared Mr Giles’ regret that RPI is still used more widely than for index-linked gilts, including for student loan repayments and for rail fares. He defended the continuing publication of RPI on the grounds, as previously stated by the UKSA and as confirmed by the Johnson review, that despite its statistical inadequacy, RPI could not simply be discontinued given its significant value to users in maintaining the continuity of the existing RPI’s long-time series without major change.

On 8 November 2017 the ONS published an article on the development of the household costs indices. Under the heading “Why are we developing the household costs indices?”, it noted that the original idea for such an index was developed by Astin and Leyland in 2015, supported by the RSS, who had considered that CPI, which had been developed as a technical index for comparing inflation between EU countries and as a target measure for interest rate setting by the European Central Bank, was not, for various reasons, fully suitable for measuring inflation as experienced by households.

On 15 November 2017, the Executive Director of the RSS wrote to the Chancellor of the Exchequer expressing the RSS members’ increasing concern at the continuing use of two different inflation indices for what is essentially the same purpose i.e. compensation for changing prices. The letter complained in particular that all too often government formulae which affected people’s incomes (such as pensions and benefit increases) used CPI, which normally provided a lower estimate of inflation, whereas many key formulae which affected people’s outgoings (such as student loan repayments and rail fares) were related to RPI, which generally gives a higher estimate. This was described as not only unfair but unjustifiable. In particular it was said that it was indefensible that millions of people’s outgoings are still being linked to RPI nearly 5 years after its technical shortcomings led to it losing National Statistics status. Particularly as it seemed unlikely RPI would ever regain that status after the ONS had made it clear it had no intention of “in effect, repairing this index”. The last paragraph read as follows: “accordingly we urge you to announce, in your budget, a timetable for ending long-standing but increasingly untenable uses of the retail prices index so that, in future, people’s incomings and outgoings are both increased in ways which – unlike RPI – command widespread confidence and are, statistically, fit for purpose.”

I record that the second Defendant objected to the late introduction of this document into evidence, the timing of which precluded Mr Briscoe from undertaking any research into the view of the RSS, and in particular whether the RSS had changed the views it had previously expressed in September 2015 when responding to the UKSA’s consultation paper dated June 2015. In my judgment, and paying due regard to the Second Defendant’s and Mr Briscoe’s lack of opportunity to carry out investigations into the RSS’s current views, this letter does not demonstrate that the RSS has changed its views as to the appropriateness of RPI, per se, as expressed in September 2015. I note, in particular, that the thrust of this document is an objection to the mis-match between RPI being used for the purposes of increasing millions of people’s outgoings, but CPI was used for the purposes of increasing their income. Beyond this, I consider that the letter demonstrates no more than an acceptance by the RSS that its recommendations made in responding to the UKSA’s 2015 consultation (including its recommendation that RPI ought not to be frozen) had not been taken on board.

In its closing submissions, BT relied on the following matters (occurring in and since 2010, and reflecting the matters referred to in paragraph 9.2(a)-(f) of the Re-amended Claim Form) as having caused RPI to become inappropriate:

I will address first certain of these matters in isolation, before addressing the question whether, by reference to all of these matters cumulatively, it is now to be said that RPI has become inappropriate within the meaning of the 2016 Rule.

The clothing change undoubtedly had a significant impact on the formula effect, and is an important factor to take into account when considering the cumulative effect of the seven matters relied on by BT. In itself, however, I do not regard it as having caused RPI to become inappropriate, for two main reasons. The first relates to the broader points I make below in connection with the cumulative effect of the matters relied on, namely that the formula effect, and the ‘upward bias’ had been known factors in RPI in 2002. RPI was not inappropriate for the purposes of the 2016 Rule before the clothing change, and the only impact of the clothing change was to increase, in percentage terms, the formula effect. This is an instance where, in my judgment, it is relevant to consider the impact on the protection for pensioners if RPI was replaced by any other index. In particular, given that any other index would remove not only that part of the upward bias introduced by the clothing change, but the whole of the upward bias (already running at an average of 0.5%) which had always been present, that would deprive the pensioners of an important element of protection which opting for the index with in-built upward bias gave them. The second reason is that RPI remained a National Statistic, and continued to have widespread support (as shown by the responses to the consultation in 2012) for some time after 2010.

The initial ‘freeze’ in RPI occurred with the announcement of the National Statistician in January 2013. It was far from a complete freeze, however, since it envisaged that changes would continue to be made to RPI relating “…to issues such as the annual update of the basket and weights, improvements to data validation and quality assurance etc.” What was not to change was “the basic formulation of the RPI”, which was accepted as currently defined. This meant, in particular, that the RPI would continue to be based, to the same extent as it currently was, on the Carli formula.

The ‘freeze’ was reiterated in the statement from Mr Pullinger, the National Statistician, in March 2016. This confirmed that routine changes would continue to be made, including so as to update the basket and weights, computer systems upgrades and improvements to data validation and quality assurance methods. These were described as examples of changes that would continue to be made as “required to continue production of a consistent, fit for purpose RPI”. Whereas in 2013 it had been stated that there would be no change to the basic formulation of RPI, it was now stated that “methodological changes” to RPI would be made if those changes were necessary in order to make improvements to CPI and CPIH.

In my judgment, the imposition of the freeze in 2013 did not cause RPI to become inappropriate for the purposes of calculating pension increases under the Scheme. As was made clear in the March 2016 announcement, the freeze did not prevent RPI from remaining “fit for purpose”. That announcement also recognised that RPI continued to be widely used for legacy purposes. The legacy uses of RPI included its widespread use in private pension schemes – as the responses to the UKSA’s 2012 consultation made clear (in particular, those from the RSS and the Pensions Management Institute). These legacy purposes were those for which RPI’s fitness was to be maintained. As Mr Johnson accepts, the imposition of the freeze has not so far caused any change. His concerns as to possible future uncertainty are met, in my judgment, by the fact that the UKSA is committed to ensuring that RPI remains fit for purpose, as well as by the fact that, if by reason of its inability to change, RPI in the future becomes inappropriate, then any decision made now does not inhibit a different conclusion being reached then.

The UKSA’s and ONS’s reasons for the imposition of the freeze do not detract from this conclusion. The ONS’s paper of January 2013 notes – as the negative qualities of RPI – that it is its use of Carli which is the primary source of the formula effect, and that Carli does not meet current international standards. These were matters that were known to be attributes of RPI in 2002, and thus not matters which can be said to have caused RPI to become inappropriate since 2002. In any event, the essential conclusion of the ONS was to reject the proposals for change, in large part because “…Carli’s properties are well understood and there is significant value to users in continuity of the RPI’s long time series.” This was expanded upon (at p.11 of the ONS paper), as follows: “…the respondents’ views, that the properties of the Carli are known and have been used in the construction of the RPI since its creation are also pertinent, especially in the context of the Code of Practice for Official Statistics.” At paragraph 28, the “weight of users’ requirements and the desire for continuity” were cited as reasons militating against change. While the ONS (1) stated that it would not advocate the use of Carli in the construction of a price increase index if it were starting from scratch, (2) recognised there were counterbalancing arguments for change, and (3) noted that “[t]he RPI’s continued use should be reassessed by users, and other indices that are closer to international best standards should be considered as alternatives”, I do not consider that the views of the ONS, derived from the paper as a whole, constituted a conclusion that it was inappropriate for RPI to continue to be used in the legacy contexts where it was already in use.

The reiteration of the ‘freeze’ in 2016 cannot – if the original imposition had not rendered RPI inappropriate as from 2013 – render it inappropriate from 2016. There was no substantive worsening of the ‘freeze’. If anything, there was a minor relaxation to the extent that it made clear that methodological changes would continue to be made, if necessary so as to improve CPI/CPIH.

The decision to introduce RPIJ as a new variant of RPI was made at the same time as the initial ‘freeze’ of RPI. It was the counterbalancing arguments for change (as noted above, in connection with the decision to ‘freeze’ RPI) that led to its introduction. In particular, the ONS paper of January 2013 noted that the new series would aid analysis of the impact which Carli had on the RPI. Given my conclusion that neither the fact of, nor the reasons for, the ‘freeze’ caused RPI to become inappropriate for the purpose of calculating pension increases under the Scheme, the mere fact that RPIJ was introduced as an alternative cannot, in my judgment, cause RPI to have become inappropriate. The most that could be said is that RPIJ was a more appropriate index than RPI. That, however, fails to meet the threshold set out in the gateway in the 2016 Rule.

Similarly, the subsequent abandonment of RPIJ did not cause RPI to become inappropriate. The reasons for its abandonment included that it was little used (most users having shifted to CPI rather than RPIJ), and that it appeared to lead to confusion (as found by Mr Johnson in his 2015 review). In its paper of November 2016 identifying the indices that would be maintained in 2017, the ONS noted that “unless there is a strong justification, having multiple measures does not encourage a clear understanding of inflation in the UK”.

While it was also noted that RPIJ shared the weaknesses of RPI that went beyond use of the Carli formula, such as population coverage and use of a direct measure of house prices to estimate owner occupiers’ housing costs, these are not factors relied on by BT as causing RPI to have become inappropriate for the purposes of calculating increases in pensions.

The fact, taken alone, that RPI was de-designated as a National Statistic is not sufficient, in my judgment, to cause RPI to become inappropriate for the purposes of calculating pension increases under the Scheme, albeit such a decision by the UK authority charged with oversight of statistics is a very important factor in considering the appropriateness, generally speaking, of RPI as a measure of inflation. There is some textual support for this conclusion in the wording of the 2016 Rule, in that the first gateway comprises two elements: that RPI is no longer published, or that it has become inappropriate. The fact that the first element is cessation of publication altogether, and not merely that it ceases to be published as a National Statistic suggests that while loss of National Statistic status may well be a factor when considering whether it has become inappropriate, loss of such status is not in itself determinative of that question. Moreover, for reasons similar to those I have set out above in relation to the ‘freeze’ on RPI, its continued publication by the UKSA and its maintenance for the express purpose of ensuring it remains “fit for purpose” preclude the conclusion that it has become – by reason of its de-designation as a National Statistic alone – inappropriate (i.e. unfit for purpose) for the legacy purposes for which it is being maintained.

In light of the above points on certain of the individual matters relied on by BT, the critical question is whether the cumulative effect of them has led, now, to the conclusion that RPI has become inappropriate.

There is considerable force in the proposition that, having regard to all seven of the matters relied on by BT, RPI is now to be considered inappropriate as a measure of inflation given in particular that: (1) as a result of the clothing change, the formula effect has nearly doubled; (2) it fails internationally recognised tests; (3) it has, as a result, been de-designated as a National Statistic; (4) the UK authorities charged with oversight of statistics have publicly described it as a flawed measure of inflation and strongly discouraged its use; and (5) it is in the process of being replaced by CPI in a number of contexts.

The question I have to determine, however, is not whether RPI is inappropriate, generally speaking, as a measure of inflation, but whether it is inappropriate specifically for the purpose of uprating pensions within the meaning of the 2016 Rule. Similar to the analysis undertaken by the ONS when determining to retain the Carli formula in RPI in January 2013, while it may well be true that it would be inappropriate to use RPI if starting from scratch, for example in a pension scheme created today, it does not necessarily follow that RPI has “become inappropriate” for the purposes of uprating pensions in the Scheme. In this context, I consider that two factors (developed in the following paragraphs) are particularly important: first, the flaws which underlie all of the matters relied on by BT were present, and known to be present, in RPI in 2002, albeit that the formula effect has worsened, and the perception of those flaws has hardened, in the intervening years; and second, the purpose of the 2016 Rule is to provide protection for pensioners against increases in the real cost of living to which they are likely to be subjected.

The flaws in RPI that are relied on by the UKSA, the ONS and most of the commentators, including Mr Johnson, who consider it to be inappropriate are (1) the fact that it uses the Carli formula which does not comply with internationally recognised standards; and (2) the fact that it contains an upward bias, as a result of using the Carli formula, compared to CPI or other indices using the Jevons formula. It is these two flaws that underlie each of the seven matters listed above that are relied on by BT. For example, the decision to ‘freeze’ RPI was taken following the consultation in 2012 which was specifically concerned with the formula effect; the de-designation of RPI occurred for two reasons, first because of the ‘freeze’ (itself a consequence of the formula effect) and the failure of Carli to meet international best standards; and the various statements from the UKSA, the ONS and the National Statistician rely on the same flaws in the RPI.

These are flaws which pre-date the first introduction of the relevant wording, in the 2002 Rules. As noted above, in the HM Treasury document published in December 2003, the formula effect was identified as having contributed approximately 0.5% to the gap between CPI and RPIX since 1997. Similarly, David Fenwick’s 1999 paper for the ONS referred to the formula effect as a known quantity, commented on the fact that a formula based on a geometric mean would always produce a lower number than that based on an arithmetic mean, noted that the formula effect was particularly pronounced for clothing, and noted that the introduction of broader item descriptions in 1996 contributed to the formula effect. Mr Johnson, in cross-examination, acknowledged that the criticisms which he makes of RPI, including that it ‘overstated’ inflation by a percentage point a year could have been made equally in 2002, albeit that the extent of the ‘overstatement’ was increased as a result of the clothing changes in 2010. He also acknowledged that the fact that RPI failed the axiomatic test had been known for a long time, and certainly in 2002 (Q: If you were doing the research from scratch in 2002, let us say, you, like Professor Diewert, would have identified these axioms and would have noted that going back for a long time before then academic statisticians had identified the problem with the arithmetic mean and the failure of the time reversibility: it was well-known? A: Yes, that has been known for a long period.”)

Two consequences follow, in my judgment, from this. First, the starting point for considering whether RPI has become inappropriate within the meaning of the 2016 Rule is that RPI must be taken to have been appropriate, for the purposes of calculating increases in pensions payable under the Scheme, notwithstanding the fact that RPI is based on a formula that failed parts of the axiomatic tests and that it produced a headline number for inflation that was 0.5% higher than CPI. Second, the drafter of the 2016 Rule has opted, in the knowledge that CPI produced a lower rate of inflation, to identify as the default index for increasing pension payments, an index which (by reason of the formula on which it is based) produced a higher rate of inflation than an index based on a different formula, in particular CPI. Members therefore had the protection of an index known to have built into it the same upward bias which is the main focus of the current objection to RPI.

While I acknowledge (as noted above) the limitations in an appeal in the abstract to the purpose of an uprating provision being to protect pensioners, nevertheless in considering the concept of “appropriateness” within the rule it is in my judgment legitimate to have regard to that purpose. In particular, it would be an important factor against concluding that RPI has become inappropriate if jettisoning RPI would introduce a material risk that increases in pensions would not keep rate with increases in the cost of living likely to be experienced by the relevant pensioners. This is particularly so taking into account that the pension is likely to constitute the principal, if not the sole, source of income for the relevant pensioners.

On the basis of the evidence I have summarised above, I consider that there are reasonable grounds to conclude that jettisoning RPI would lead to such a material risk for the pensioners under the Scheme.

First, as is common ground, inflation is a latent variable, and any index can do no more than provide an estimate of the increase in cost of living as experienced by any given household, or even type of household. Thus, it is impossible to say that RPI is wrong and CPI is right, or even that RPI is more wrong (or right) than CPI, as an estimate of the likely increase in cost of living for pensioners under the Scheme.

Mr Johnson, while accepting that this is so, says nevertheless that “if you know there is a problem in the construction of an index in the way that the Carli creates biases, then whilst you might not know the right answer, you do know that you are trying to get there in the wrong way.” This is, however, not a complete answer to the latent variable point, since it is common ground that there is no established correlation between the fact that Carli fails certain of the axioms, and the extent of the formula effect.

Second, there is convincing evidence (see, for example, the papers cited above from Dr Altmann, the Rowntree Foundation, Donald Hirsch and the RSS) to support Mr Briscoe’s conclusion that there are certain respects in which CPI might be said to underestimate inflation (albeit probably to a lesser extent) as well as respects in which RPI might be said to overestimate it, and that in some instances pensioners are likely to be particularly affected. While the most significant cause of the difference between RPI and CPI is the formula effect, there are other material causes, including property costs, which accounted on average for 0.5% of the increase. Although another factor – termed “weights” in the Office of Budgetary Responsibility’s revised assumptions for the long-run wedge between CPI and RPI – accounts for a negative difference of -0.4%, that does not detract from the point that there are reasons why RPI is greater than CPI which are not in themselves sufficient to render RPI inappropriate (and in any event have not caused RPI to become inappropriate since 2002). As I have noted in connection with the clothing change, a consequence of replacing RPI by an index based on Jevons would wholly remove the upward bias, a substantial part of which was inherent in RPI on its adoption as the default index in 2002.

Third, it is significant in my judgment that RPI continues to be used in a number of contexts related to outgoings of large sections of the population, including in many contexts dictated by government, or national regulator policy. I accept that the weight of opinion, based on the materials I have been shown, supports the view both that (a) the lack of consistency between the use of RPI for outgoings, as compared to income, is regrettable and (b) the single measure of inflation that should be used for both sides of the balance sheet is something other than RPI (probably CPI, or a variant of it). I also accept that to the extent that items which have increased by reference to RPI over the previous twelve months are included in the basket of goods reviewed for the purposes of CPI, then this will compensate to some extent for the mis-match between incomes and outgoings. Nevertheless, it is this mis-match that is the target of the RSS’s ire in the letter from its Executive Director of 15 November 2017 – describing it as “unfair” and “unjustifiable”, and it is the very existence of such unfairness which contributes to the risk that uprating the Scheme pensions, for the forthcoming year, by reference to CPI might lead to the pensions failing to keep pace with the real increases in outgoings as experienced over that year. This is an example of the fact that a conclusion as to ‘appropriateness’ in a general sense (or for other purposes) is not determinative of the question as to appropriateness for the specific purpose of the 2016 Rule: a conclusion by the UKSA, ONS and National Statistician that RPI ought not to be used at all is not determinative of the question whether it is now inappropriate to use it for uprating pensions under the Scheme, in light of the reality of its continued use in a number of other contexts.

I add that this is not to suggest that it is safe to rely on the continued use of RPI by other users as an objective endorsement of RPI’s appropriateness, any more than its discontinuance by other users can in itself be relied on as an objective endorsement of its inappropriateness. That is particularly so when the user in question is the state itself, where the mismatch between income and outgoings is particularly acute (taxes and other revenue being based on RPI, while benefits and state pensions are based on CPI). The constraints of policy and the national budget make it difficult to assume that such decisions are founded on an objective, reliable assessment of RPI.

The landscape, so far as the number and importance of users continuing to rely on RPI, is not constant, but is in a state of flux. This highlights one of the particularly acute difficulties in answering the question raised by the gateway in the 2016 Rule, when the continuing appropriateness of RPI is linked to (1) changes in perception as to flaws that have always been there, and (2) changes in its use in other contexts. That is, accepting that at some point in time the state of the general consensus as to RPI and the extent to which it is no longer used elsewhere will render it inappropriate for use in the 2016 Scheme, the identification of the tipping point from appropriate to inappropriate, along a continuum of change, is a highly fact-sensitive judgment call.

The recent decision by OFWAT demonstrates both aspects. Its decision was based among other things on its conclusion that RPI “is losing its legitimacy”, but it will nevertheless continue to make use of RPI until 2019 (for some purposes) and until between 2020 and 2025 (for other purposes). OFWAT’s decision is particularly relevant to the last of the seven matters relied on by BT. However, in that context insofar as the question is whether RPI continues to be appropriate for uprating pensions under the Scheme in January 2018, OFWAT’s decision is for this reason less relevant.

I do not suggest that it is wholly irrelevant, however, because its expression of view as to the appropriateness of RPI is to be taken into account as part of the general consensus as to RPI. Nevertheless, this again highlights an important distinction between the appropriateness, generally, of RPI and its appropriateness in the context of the Scheme. OFWAT was concerned that any index used for increasing prices had the confidence of the public and could therefore be said to have legitimacy. The same considerations of public confidence and legitimacy pull in the opposite direction where the question is whether pension payments should now be uprated by reference to CPI, and RPI continues to be used to increase a large variety of outgoings. Of more relevance, in that context, is the fact that the responses to the consultation in 2012 demonstrated widespread continuing confidence in RPI, particularly in the context of uprating pensions. While recognising that this may well have been to a large extent driven by self-interest, that is not a reason for dismissing it altogether. It was sufficient to dissuade the ONS from opting to replace the Carli formula in RPI, and it remains a relevant consideration in the context of the determination to be made in this case, given the importance of the 2016 Rule as a measure of protection for pensioners as described above.

Having regard to the totality of the matters relied on by BT and the second Defendant as demonstrating, respectively, the disadvantages and merits of RPI, notwithstanding the powerful statements from the UKSA, ONS and others to the effect that RPI is flawed and that it ought not to be used as a measure of inflation, I have reached the conclusion for the above reasons that RPI has not at this time “become inappropriate” for the purposes of uprating pensions, within the meaning of that phrase in the 2016 Rule so as to meet the gateway threshold.

I will address the issues of construction that arise in relation to the 1993 Rule before considering the application of the gateway test.

BT submits that although the wording of the 1993 Rule is not identical to the 2016 Rule, the same broad range of facts and matters can be taken into account by the decision-maker in determining whether the gateway has been passed through. Accordingly, says BT, the decision-maker is not confined to considering merely amendments to the compilation of RPI, but can and should look at the question more broadly. Thus, if RPI can now be said to be invalid as a continuous basis for the purposes of calculating increases, then the gateway is passed through whether that is as a result of (1) operational changes, (2) direct effects of those changes, such as de-designation, or (3) other consequences flowing from those changes and their direct effects.

The second Defendant submits that the scope of the 1993 Rule is significantly narrower than the scope of the 2016 Rule. In particular, the 1993 Rule is not engaged at all unless there has been an amendment to RPI, that is, a direct change in some way to RPI, to its compilation or to the way it is calculated. Unlike the 2016 Rule, therefore, the act of de-designating it as a National Statistic would not be sufficient to allow the gateway to be passed. Moreover, it is not any amendment that is required, but an amendment which invalidates RPI as a continuous basis for the purposes of calculating increases. The second Defendant places particular emphasis on the use of the word “continuous”, and contends that the question is: has the continuity of RPI been invalidated by the amendment?

So far as the concept of amendment is concerned, I accept the submissions of the second Defendant. Mr Spink again sought to derive support from the obiter comment of Warren J in the Thales case, to the effect that if the rule in that case had referred to a “material change in the index” then it might have been arguable that a change in the impact on RPI would have been sufficient. As already noted, I gain no support from a decision on materially different wording. To my mind the word amendment, which is a more specific word than “change”, is intended to connote an alteration to the formulation, or method of calculation, of RPI.

Accordingly, I reject the submission that the 1993 Rule can be engaged by a change in the status of RPI, let alone a change in the views of commentators as to the validity of RPI. There was some debate as to whether any amendment to RPI would count, or only an amendment of a sufficiently material nature. This was in the context that operational adjustments are regularly being made to the list of items included within RPI to reflect changing spending habits. In my judgment, there is no such limit on the type of amendment to RPI. That is because the drafter has built in to the provision a bespoke control mechanism, namely that the amendment must be one which “invalidates… etc.” There is therefore no need to imply any other control mechanism (such as an abstract notion of materiality) on the type of amendment that will trigger the operation of the gateway.

I also agree with the second Defendant that the requirement that the amendment “invalidates” RPI “as a continuous basis for the purposes of calculating increases” has a significantly narrowing effect on the gateway. In this regard it is instructive to note that the mechanism for calculating an increase in pensions set out in sub-paragraph (1)(a) of the 1993 Rule requires a comparison to be made between the index figure in RPI for the January just gone, with the index figure in RPI for the previous January. Payments are to be increased by the percentage ratio by which the former exceeds the latter (or 5% if that is less). I understood it to be common ground that if there was a change in the formulation of RPI, for example if RPI began using the Jevons formula instead of the Carli formula, then that would be the type of change which would invalidate RPI as a continuous basis for the purposes of calculating increases. That would be because comparing the index figure for the January of the prior year (calculated using Carli) and the index figure for the January just gone (calculated using Jevons) would be like comparing apples with pears. I note that in the event that the 1993 Rule gateway is passed, BT’s obligation is to substitute another index “or appropriate basis of comparison”. This lends support, in my judgement, to the second Defendant’s submission, since it emphasises that the essential function of the index is to provide a comparison as between the current and prior year, and it is only if an amendment to RPI causes it no longer to be valid for the purpose of making such a comparison that the gateway is passed through.

BT submits that the word “continuous” is intended to refer to the need for the relevant index to provide consistency into the future. The difficulty with this submission, in my judgement, is that it is unlikely to be possible to determine whether a change that is now made to the formulation of RPI will mean that in future years it ceases to be a valid basis of comparison, from one year to the next. Mr Spink relies, in particular, on the decision to “freeze” RPI. This is a good example of the difficulties to which his interpretation of the word “continuous” gives rise, because it is common ground among the experts that the effect of the “freeze” on RPI in the future is difficult to predict. The fact that the 1993 Rule requires a repeat performance each year of the calculation required by sub-paragraph (1)(a) lends further support to the view that it is concerned with continuity between the two particular years in issue, rather than seeking to predict whether the index will remain consistent in future years.

The second Defendant advances similar arguments to those it made in relation to the equivalent issues (Issues 5.1 and 5.2) under the 2016 Rule. My conclusions on the inapplicability of the principle derived from family trust cases to the 2016 Rule apply equally to the 1993 Rule. Accordingly, I reject the submission that the rule derived from trusts law, that a failure to exercise a mere power within a reasonable time causes the power to lapse, is to be applied by analogy in the case of the 1993 Rule.

There is, however, a further point that arises on the 1993 Rule to be considered. This was not specifically advanced by the second Defendant, but it follows logically from my conclusion as to the meaning of the rule. Since I have held that (a) the gateway is triggered only by an amendment to RPI, and (b) ‘invalid’ within the meaning of the rule relates to the continuity of RPI from year to year, it follows that it is at least possible that an amendment that occurs in a particular year may have the effect of so changing RPI that it ceases to be valid for the purposes of making a comparison with the previous year but that (if RPI was not then replaced), in respect of the comparison to be made in future years, the invalidity does not persist. For example, if RPI had been changed in 2010 so that it was based on the Jevons formula, then the experts were agreed that this would constitute an amendment that fell within the gateway in the 1993 Rule. That is because it would involve a comparison between apples (an index based on Jevons) and pears (an index based on Carli). But, if no action was taken, such that RPI remained the applicable index in the following year, then in 2011 the comparison would be between two similar indices. In this way, a failure to reach a determination that RPI had been so amended as to render it invalid, before the next date for applying the relevant index to update pensions in payment, would have the consequences that (a) the pensions paid in 2010 in reliance on RPI could probably not be disturbed (although since no-one seeks to upset pensions payments made in 2010, and I therefore heard no argument on this specific point, I do not reach a concluded view on it) and (b) the basis for reaching a determination in future years, in reliance on the clothing change that occurred in 2010, was removed.

I refer to paragraphs 50 - 55, for the reasons for rejecting the second Defendant’s contention that the adoption of the 2016 Rules necessarily constituted an implied exercise of the power to determine that RPI remained appropriate.

In relation to the 1993 Rule, the position is even further removed from the Davis v Richards & Wallington principle. The second Defendant’s argument proceeds on the basis that re-basing, as a matter of construction, the appropriateness of RPI in April 2016 (assuming in her favour the answer to Issue 2) cannot have occurred unless BT had also concluded that RPI had not been so amended as to invalidate it as a continuous basis for the purposes of calculating increases under the 1993 Rule. This proposition is a greater non-sequitur than the argument in relation to the 2016 Rule, and I reject it.

It is common ground that the relevant decision-maker under the 1993 Rule is BT, and that the Court’s role is limited to assessing whether any decision taken by BT was made rationally and in good faith. It is also common ground that BT has not yet taken the relevant decision. I am nevertheless asked to conclude whether, if BT were now to determine that RPI had been so amended as to invalidate it, that would be a determination that was made irrationally. In so doing, it is necessary to make an important assumption: that the first limb of the Wednesbury test (that the decision-maker did not exclude from consideration relevant factors, and did not take into account irrelevant factors) is either inapplicable, or would be satisfied in respect of BT’s determination. On that assumption, I am in effect being asked to determine whether no reasonable decision maker in the position of BT could determine that RPI had been so amended as to invalidate it within the meaning of the 1993 Rule.

Given my conclusion as to the meaning of “amendment” within the 1993 Rule, there is only one matter, of the many relied upon by BT, that passes the threshold requirement of being an amendment to RPI, and that is the clothing change. All of the rest (including the de-designation and the ‘freeze’ of RPI) do not constitute an amendment in the requisite sense.

Moreover, in light of my conclusion as to the meaning of “so… as to invalidate it as a continuous basis for the purposes of calculating increases”, the clothing change could only fall within the definition of the 1993 Rule if it invalidated the continuity of RPI, so as to prevent a like-for-like comparison between the current and prior year. In my judgment, given the routine nature of the clothing change, and the fact that its impact was limited to contributing to a widening of the already existing formula effect, it does not qualify as such an amendment. More importantly, I do not consider that a rational decision maker in the position of BT could determine otherwise, because to do so would constitute a mis-interpretation of the rule.

Finally, the widening of the formula effect occurred in 2010. Even if that had the effect of precluding a valid comparison as between 2009 and 2010, then for the reasons given in paragraph 228 above, the formula effect has remained constant ever since so that the clothing change, in 2010, cannot have constituted an amendment that rendered RPI invalid for the purposes of comparison as between any of the consecutive years since 2011.

Accordingly, my conclusions on the eight issues raised for determination are as follows: