17 June 2010

Published in: Investment risk, ALM, Longevity - mortality

Inefficient inputs lead to outmoded outputs

Misleading assumptions often lead to inaccurate actuarial valuations of DB pension liabilities. Result: inefficient hedging strategies and an inappropriate contribution rate, says James Mushin

Modern risk management platforms, using analytics capabilities developed from the capital markets, are now enabling scheme sponsors and trustees to be more precise when measuring their defined benefit (DB) pension liabilities. The development of this technology has exposed the inaccuracies of formal actuarial valuations which, in some cases, can miscalculate by hundreds of millions of pounds.

The modelling process

Traditionally, DB pension scheme valuations are seen to be at their most accurate at the point of their formal actuarial valuation. And a key component of this valuation - and one that is critical in defining its precision - is the model the actuary uses to project the scheme's future cashflows.

In theory, at the time of valuation, actuaries will advise on reasonable assumptions to use, and therefore should be able to calculate a precise valuation that considers all available evidence. After this point, of course, the valuation may lose accuracy as unforeseen changes to economic conditions -- and to the situations of individual members within the scheme -- render the initial assumptions and calculations incorrect.

Inefficient models

However, in practice the valuation is often inaccurate at the very point in time that it is signed off by the actuary. This is partly because of the time taken to complete the valuation (in some cases up to 15 months) but it is also partly the result of the processes traditionally adopted by actuaries in the creation of their calculations.

This inaccurate starting point is then updated over time, usually in a very approximate way, to underpin investment strategy, accounting disclosures and other financially significant decisions, until the results of the next "accurate" valuation are finalized.

And, as any navigator will tell you, wrong assumptions made at the beginning of the journey become more and more exaggerated the further travelled. This means that while the original assumptions may have included conservative erring or approximations that created only a minor distortion, such errors may be compounded and magnified over time.

Most of the models rely on rounded ages...but, for the close management of longevity risk, such inbuilt inaccuracies can create major distortions  

Even in cases where the input data and assumptions may be reliable, the actuary's job is made more difficult by the models employed, which are often archaic and unable to cope with the significantly more intricate demands of modern pension scheme valuations. The rigid nature of these models restricts the effectiveness of actuarial calculations - forcing actuaries into a situation where they have to squeeze data into unsuitable models.

For example, most of the models rely on rounded ages. As such, a pensioner aged 75.4 years must be treated as a 75-year old for purposes of modelling. Logically, it could be argued that because, on average, half of scheme members' ages should be rounded up and half should be rounded down, the approximations will cancel out. And for younger people this may hold true.

For pensioners, however, this is not necessarily the case. The difference in life expectancy between a member aged 75.4 and one aged 75 is greater than the difference for a member aged 75 and one aged 74.6, so the impacts of rounding do not cancel out, and can actually introduce a bias into the calculations.

Actuaries state that this is a limitation of the model but a known limitation, and one that would be too laborious to remove. But, for the close management of longevity risk, such inbuilt inaccuracies can create major distortions.

Lack of flexibility

There are further examples of the lack of flexibility of actuarial models. Some older models, still in widespread use, are unable to handle curves (such as inflation curves) as input parameters. While actuaries may claim to deal with this by running flat-rate models and then making retrospective input adjustments in order to re-create a curve, the quality of results produced is severely diminished. In particular, when determining a hedging portfolio to match a scheme's liabilities, the cashflows produced from these models may not be reliable and could result in a sub-optimal hedge.

Some actuarial models also only accept information occurring at certain times of the year. Consequently, issues often arise when discounting the anticipated future cashflows of pension schemes in order to calculate the present value of the liabilities.

Most valuation reports do not provide all the assumptions the actuary has made, or they state assumptions in an ambiguous way 

For example, actuarial models may only recognize cashflows in the middle of the year, while the data on the discount rate may have spot values at the end of each year - making the effective matching of discount rates to cashflows, at best, a subjective evaluation.

And to top all this off, the inaccurate assumptions created by these processes tend to be used year-after-year with little regard for changes that may have occurred which may render the approximations invalid, for example, a redundancy exercise, or expected inflation becoming negative. These can compound the inaccuracies on "roll-forward" (requested by schemes between formal, triennial valuations) or future valuations.

An alternative solution

In an environment where pension scheme sponsors are keen to calculate and manage their risks more tightly, such approximations should no longer be viewed as acceptable. To counter this anachronistic approach to risk management, new solutions are being developed, including the advent of web-based risk management tools such as PensionsFirst Analytics' PFaroe.

These analytics platforms allow scheme sponsors and trustees to run cashflow models (which are flexible enough to reflect the exact details of each scheme's rules) under chosen inputted scenarios for valuation requirements with enhanced granularity and speed.

The first task when using these tools is to enter the old scheme data on to the platform in order to create a starting point for more accurate modelling. When the previous input assumptions and models are replaced with inputs and scenarios that more precisely reflect the scheme's idiosyncrasies significant discrepancies are identified.

Lack of transparency

This reconciliation process also throws up another shortcoming of some actuarial valuation reports - their lack of transparency. This makes it hard to understand even the assumptions that the actuary has made, let alone if they are reasonable.

In practice the valuation is often inaccurate at the very point in time that it is signed off by the actuary. This is partly because of the time taken to complete the valuation but also partly the result of the processes traditionally adopted by actuaries in the creation of their calculations.  

Guidance from the actuarial profession requires actuaries to disclose assumptions so that another actuary could reproduce their results with no material differences. Consequently, most valuation reports do not provide all the assumptions the actuary has made, or they state assumptions in an ambiguous way. A valuation report may state, for instance, that a pension scheme member is married with a "probability of 90%", yet fail to state whether this assumption holds at the date of valuation, at the date of retirement or at the date of death.

The increased transparency provided by new systems should advance the dialogue between sponsors, trustees and actuaries by providing more flexibility and visibility for the assumptions used - aiding the often protracted negotiation process of the modelling for formal actuarial valuations.

In conjunction with improvements to the quality of data held by pension scheme administrators - an issue which the Pensions Regulator is increasingly concerned with - these new techniques will help pension scheme sponsors to understand their pension liabilities better, and so help to secure the future financial viability of their schemes.

With de-risking becoming increasingly important, a more accurate valuation and cashflow projection will also be essential for pension scheme sponsors, trustees and their banks to design appropriate hedging strategies to match these liabilities.

James Mushin is a director at PensionsFirst Analytics, where he is responsible for bringing clients' pension schemes on to PFaroe, the firm's pension risk management platform. Prior to joining PensionsFirst, James worked in life insurance consulting where he focused on asset-liability modelling and economic capital techniques.

 

Back to top