1.1. This paper focuses on the above issues in relation to "defined benefit" arrangements. These arrangements, as their name implies, promise a defined benefit on retirement or death: the cost of such benefits can be known with certainty only at the time of payment. Funding plans are normally developed in a way to spread these costs in a reasonable manner and to provide funds from which benefits can be paid in the event of closure of the arrangements arising from, for example, the demise of the sponsoring employer.

The word "reasonable" is important here as costs of funding a pension benefit can be calculated according to a number of different actuarial methods which produce different spreading of costs. The method used the most in the U.K., the Projected Unit Method (PUM), produces costs to employers which grow with the age of the employee because the liabilities are based on accrued service rather than expected service. This has a number of important labour market implications, including potentially higher costs of employing older workers. Whether or not this is "reasonable" is a normative issue, but the particular funding method used is not without significant economic implications.

1.2. In a final salary defined benefits structure, a member who leaves service before retirement age receives less value for his service than a member who stays to retirement would receive for the same service. Transfer values, designed to provide a reasonable sum in exchange for an early leaver's benefits, neither exaggerate nor reduce this. To do so would introduce a degree of unfairness - between those who opt to transfer and those who opt to retain their pension in the scheme.

A worker who leaves a job has a choice between taking a transfer value or keeping a deferred pension. The transfer value is supposed to represent the expected value of the deferred pension. We argue in our report that deferred pensions are better for most workers than transfer values because of the constant growth rates built into the MFR norms in GN27; our report argued that actuarial discretion is very valuable for correcting these biases and GN27 is too limiting. The OFT report was primarily concerned with the unfairness of legislation concerning preserved benefits in which early leavers are penalised relative to those who stay.

1.3. Defined benefit structures do exist where early leavers receive identical benefits to those who stay. Under such benefit structures the amount of the benefit accrual in each year is independent of future movements in the individual's earnings (and therefore whether or not he remains in service) and instead are revalued up to retirement age in line with an index based on price inflation or general earnings inflation. SERPS is an example of such a structure. These benefits are often referred to as "Revalued Career Average" arrangements. In practice, these schemes have never really caught on in the U.K., probably because they are perceived to be more complicated to administer and explain to members than the more normal final salary arrangement. Nevertheless, such schemes play an important role in tackling the problem of preservation and greater encouragement could be given to the establishment of such structures. No mention of these benefit structures is made in the research paper.

We were asked by the OFT to examine the questions of portability and preservation of pension rights in the most common defined benefit schemes operating in the U.K. As the IFA response itself indicates, revalued career average schemes are not common, although we agree that they have some attractions.

However, revaluation by national average earnings instead of limited price indexation (LPI) would lead in general to a similar pension outcome. Our main concern, based on our knowledge of the diversity of lifetime earnings profiles of workers in different occupations, is that individual circumstances differ and indexing to national average earnings would pose undue burdens on some employers. We chose to make another policy recommendation which may well have its own problems, but it would work with most existing schemes.

1.4. On a point of detail, many schemes do provide a benefit in excess of the statutory minimum requirements for employees. Many schemes (and in particular most public sector schemes) go further than the statutory requirements by providing full inflation proofing of the preserved benefits during the period of deferment and when the pension benefits came into payment. Many other schemes go further than required for those who leave within the two year vesting period by allowing a transfer payment to be calculated as if the member had completed the vesting period. Vesting periods have historically been used to reduce disproportionate administrative burdens in dealing with small pensions: but with a modern administration system there is no particular reason why this period need be as long as two years and this is something which may be reviewed in future legislation but would need to be balanced against the extra costs that would result from dealing with a large number of small deferred pensions.

Schemes that pay beyond the statutory minimum: This was discussed in various places in our report, including the period in the 1970s when there was not full inflation proofing. We would be most interested in any surveys the Institute and Faculty of Actuaries has conducted on the question of the granting of benefits more generous than the statutory minimum since much of what the IFA says here is not readily available in the public domain.

Vesting periods: Pensions were traditionally provided as a reward for long, loyal service. Indeed, it was not until legislation was introduced that vesting periods were reduced and few employers provide a vesting period shorter than the statutory minimum.We refer to Ch.3 of E.M. Lee's An Introduction to Pension Schemes (published by the Institute and Faculty of Actuaries, 1986) for the story of `early leaver' Anthony Trollope who left the Post Office in 1867 after 33 years of service with no pension. Things have improved since then, of course. Nevertheless, we agree that a vesting period of less than two years might impose severe administrative burdens for deferred pensions. Transfer values could be offered on the basis of full service up to some specified minimum period (such as 5 years).

This highlights some of the economic arguments which could be used to justify portability losses. In practice, we suspect that few could use the "training cost" argument to penalise early leavers. Final salary schemes date from the time of very low inflation, so that most of an individual's earnings progression arose from seniority or promotional awards and employers were happy to translate these increases into higher pensions.

We accept that pension schemes were set up to reward long, loyal service. Nevertheless, the vesting period can be used to `recover' the training costs of workers who leave very early on, even if this is not explicitly stated. Surely, this must be conceded if the IFA can argue that vesting periods are used to reduce the administrative burdens of dealing with small pensions!

3.2. There is a minor error on page 22. The main factor that causes a transfer value to result in a larger pension at retirement than the deferred benefit is if the member's career progression with his new employer exceeds the assumptions made by the new employer's scheme when calculating the benefits in return for the transfer payment - it has nothing to do with relative career progressions in the individual's former and new jobs.

This point is correct. However, the reader is unlikely to have been misled because this particular sentence followed another sentence which made the correct argument about the worker having private information about his career prospects.

3.3. The reference to Minimum Cash Equivalents in the context of the contracting-out requirements (see footnote 9 on page 26) is unclear. Perhaps this reference arises from a misunderstanding which features later in Chapter 6 (and see comment 6.2 below) that the contracting-out rules require the calculation the "Minimum Cash Equivalent"

We also accept this point which is indeed related to an incorrect sentence in Chapter 6 - see 6.2 below.

4.1. This chapter attempts to quantify portability losses and, in our view, contains a number of worrying misconceptions. We explain these misconceptions in paragraphs 4.2 and 4.3 below. However, we must first correct what seems to be a fundamental misunderstanding as to why the PUM (projected unit method) valuation technique which is adopted by actuaries produces an increased cost for older members compared with younger members. Section 4.2 of the report incorrectly states that this arises because "...pay is almost invariably backloaded ..." and also "...greater service has already accrued for an older worker than for a younger member." (see first paragraph on page 44). These are not the reasons - rather the PUM produces a higher cost for the older member simply because of the reduced term to retirement age and, therefore, of the period for accruing investment returns on a matching asset.

In this chapter, we calculated, using actuarial methods, the precise size of the portability losses experienced by early leavers with respect to two benchmarks: (1) a worker with the same salary experience as the early leaver who remains in a single pension scheme all his career and who therefore receives a full service pension, and (2) a worker paying the same total contributions into his pension scheme as the early leaver but who belongs to a scheme that does not backload contributions since it has a constant, age-independent marginal contribution rate. In the first case, we derive the cash equivalent loss (or transfer value loss). In the second case, we derive the backloading loss. These are precisely calculated losses (using actuarial methods) with respect to these two benchmarks and the benchmarks can be used as an objective basis for comparing the position of the early leaver. There can be no question about that. Where differences can occur, of course, is over interpretation.

We strongly disagree with the IFA's comments on Chapter 4. The particular point made in 4.1 is thoroughly muddled. Our statement looks at the situation from the liability side (and correctly states that "an additional year of service from an older worker buys a larger pension entitlement than an additional year of service from a younger member" because "pay is almost invariably backloaded... and ... because greater service has already accrued for an older worker than a younger worker"). The IFA merely looks at the same issue from the asset side - the higher cost arises "because of the reduced term to retirement age and, therefore, of the period for accruing investment returns on a matching asset". Both statements are perfectly consistent!

4.2. As mentioned above, we are concerned about other misconceptions. In particular, great emphasis is made of the "cash equivalent loss" and "backloading loss" neither of which is in any way relevant to calculation of the cash equivalent. As stated earlier in this document, the cash equivalent aims to place a fair value on the preserved benefit which the leaver would otherwise leave behind in his former employer's scheme. It is completely independent of the funding plan (ie projected unit method, current unit method, or indeed, any other funding method which might be used). To calculate cash equivalents in any other way would be bound to introduce some degree of unfairness between those who opt to transfer out and those who do not.

This comment is not correct. The "cash equivalent loss" is the difference between the cash equivalent under current legislation and that which would have been obtained from calculating a preserved benefit using a projected final salary. We refer the reader to our Executive Summary:

"Cash Equivalent Losses arise because the early leaver's leaving salary is revalued to retirement age at a less favourable rate than used to determine the projected final salary..."

The IFA response is focusing on the difference between transfer values and deferred pensions. Of course, the "cash equivalent loss" is not relevant for computing the cash equivalent as in current actuarial practice, but that does not mean that the "cash equivalent loss" is not an important concept. To say that the cash equivalent aims to place a fair value on the preserved benefit is completely missing the point, since the aim of the study (as noted even in the title!!) was to examine whether preserved benefits were "fair" relative to full service benefits.

4.3. The "backloading loss" is, in our view, a misconceived idea and is based on the thought that an employer "ought" to pay the same amount of contribution irrespective of an employee's age. Possibly, this misconception is based on the thought that this is the way defined contribution schemes operate - but in fact, many such schemes do link contributions to the member's age.

Again, this comment is not correct. We were not making any normative judgements about what employers "ought" to do, but were rather reflecting what they do as well as identifying the nature of the higher pension costs that employers incur for older workers. The one operational comment that could have been made here by the IFA is that when accounting for age-related contributions, a different measure of the size of the backloading loss is produced.

5.1. This contains many figures and graphs to illustrate how early leaver benefits compare to full service benefits for different individuals - the main factor being the number of job changes and future earnings growth. As indicated earlier, these are a feature of final salary schemes and other defined benefit arrangements (ie career average arrangements) show no differences between benefits for early leavers and those who remain in employment.

This chapter applies actuarial formulae to estimate the differences between full service and short service benefits for some typical UK workers and in particular provides estimates of the portability losses experienced by early leavers who do not enjoy full service credits on transfer (as happens in the public service, for example). Clearly, these losses are a feature of final salary schemes, but these are the types of scheme that most members of occupational schemes have. We are pleased that the IFA does not dispute the size of the losses that we have identified. Indeed, the losses we have identified are of the same order of magnitude as those found by the actuary Bryn Davies in his report "What's wrong with transfer values?", published by Pensions Investment Research Consultants in 1990.

6.1 This discusses actuarial discretion in connection with transfer values but unfortunately contains a number of errors and misconceptions. Firstly, there is an assertion that the wording of GN11 implies that actuaries can modify the MFR assumptions to take account of individual circumstances in such a way as to give rise to the systematic calculation of transfer values for individuals which are materially lower than the MFR assumptions. This is simply not the case. The wording to GN11 to which the author refers is to provide some degree of flexibility on factors such as, for example, whether to calculate the transfer value using the member's age in years and months, or years and days; or whether market value should be that applicable to the date of calculation or, say, the start of the month when the calculation was made. These flexibilities are of a minor administrative nature and the actuary must be satisfied that they do not produce any significant bias which is likely to result in transfer values materially smaller than MFR.

We think the IFA has overreacted to this chapter. As economists, we believe that individual circumstances do matter and economists vary assumptions (such as the discount rate when performing cost- benefit analyses) depending on the specific circumstances. Similarly, in finance, risk-adjusted rates of return depend on individual attitudes to risk. We accept that professional judgement is important. Indeed, our executive summary specifically stressed the value of actuarial discretion.

We stated in the report that in practice "most transfer values are not calculated on an individual basis by an actuary. The most common procedure is for actuaries to lay down a set of factors or formulae (often presented in the form of easy to read tables) which scheme administrators use to calculate transfer values" and acknowledged correctly that the MFR rules tightened the range of transfer values.

Nevertheless, the Goode Report was concerned about the extent of actuarial discretion and recommended that it should be tightened. This was achieved by the 1995 Pensions Act in the form of the MFR rules.

Actuaries still have enormous discretion (as evidenced by the most recent Coopers & Lybrand survey of actuarial assumptions). If actuarial discretion were not large, how otherwise would it be possible for a scheme to have a surplus of 20% or more using one actuarial valuation method, yet still satisfy the 1986 Finance Act upper limit of 5% on surpluses on the basis of the actuarial valuation method stipulated under this act?

6.2. The second major misconception relates to the benefits subject to an MFR minimum when calculating a transfer value. The research paper asserts that it is only the contracting-out part which was subject to an MFR minimum! This is incorrect and the whole benefit must be subject to the MFR transfer value minimum. This misconception significantly affects the development of Chapter 6 and invalidates much of what is written.

We agree that the sentence in question is incorrect. Actuaries can still calculate a transfer value on the standard basis (provided it does not fall below the MFR minimum); as stressed in the report, calculations on the standard basis are subject to more discretion than Minimum Cash Equivalent calculations.

We would welcome a new survey by the IFA on the range of realised assumptions affecting the size of transfer values.

6.3.With regard to discretionary benefits, GN11, in line with legislation, is clear that the actuary must allow for discretionary benefits in a way which reflects established practice unless he is directed to do otherwise by trustees. Once the extent of the established practice has been quantified, the allowance in transfer values is automatic unless the trustees direct the actuary to ignore them. If the trustees do make such a direction, then GN11 requires the actuary to prepare a written report which will, in essence, give the actuary's opinion as to whether or not the resources of the scheme would have permitted such an allowance for discretionary increases. This report must be disclosed to members on request.

We cannot see how this is inconsistent with what we wrote in our report.

6.4. In section 6.1 of the report there is commentary on the discount rate to be used by actuaries in accordance with GN11 (see, in particular, page 81). While the authors would be correct to state that actuaries have a certain degree of discretion as to the choice of discount rate because GN11 is not precisely prescriptive on this point, we take issue with the statement that the "actuary has enormous discretion over the discount rate". In practice, the actuary must exercise professional judgement as to where the rate is to be set and must be satisfied that the rate of return corresponds to an appropriate choice of notional matching assets (see paragraph 3.1 of GN11). As with all matters involving the exercise of professional judgement, the actuary must be prepared, if challenged, to justify the decisions made.

The fact that actuaries can set a discount rate that corresponds to an appropriate choice of unspecified `notional' matching assets gives actuaries, we would maintain, enormous discretion over the discount rate. And as Tables 6.1 and 6.2 of our report show, of all the parameters subject to actuarial discretion, it is the discount factor that has the greatest effect in influencing the size of both outgoing transfer values and incoming service credits.

We would like to see a study undertaken of the effects of actuarial discretion as operating in the U.K. In particular, on the basis of survey evidence on the distribution of actuarial assumptions, it would be possible to generate a distribution of outcomes for both outgoing transfer values and incoming service credits. Only then could the true impact of actuarial discretion be appropriately assessed.

7.1 . The suggestion of adopting an amalgam of contribution roll up and "value of benefit" when calculating transfer payments has a number of potential problems, as follows -

a) There is no guarantee that the resultant figure will be "fair" compared with the alternative deferred benefit.

b) There are the practical problems of determining the appropriate investment roll-up! This could be negative in periods of falling stock markets.

c) What contributions should be used? If the employer is paying nil contributions because of a surplus position arguably a nil value should be used: alternatively, if the employer is paying extra contributions to correct a deficiency, this could unfairly benefit the transferring member. If the "normal" contribution rate is used (ie the rate the employer would pay if there were no deficiency or surplus) then this is superficially attractive, but brings its own problems:

In short, the resultant transfer payment is likely to subject to an even greater degree of subjectivity and variation than the current methods which the author criticises.

Our policy proposal as outlined in Chapter 8 was very much a compromise and is clearly stated as such. So we accept most of the criticisms of it outlined on the left. However, it was an attempt to improve the position of early leavers and it would still do so despite the fact that it would be subject to a similar and possibly even greater degree of subjectivity and variation than existing arrangements.

The really disappointing feature of the IFA response is that they put forward no proposals to reduce the portability losses suffered by early leavers. Indeed they appear to regard what they currently do as "fair" and that any other procedure would introduce a "degree of unfairness" between early leavers and long stayers, despite the fact that they admit that "in a final salary defined benefits structure, a member who leaves service before retirement age receives less value for his service than a member who stays to retirement would receive for the same service" (Para 1.2).

In our view, it is attitudes such as these (together with both their intrinsic complexity and inherent unsuitability to a flexible dynamic, labour market) that discredits defined benefit schemes and is likely to leave them with a constantly reducing share of the pensions market. Without substantial changes (perhaps with the help of the IFA), the defined contribution system is destined to be the "preferred pension option" of an increasing number of UK individuals and firms.