In its recently published exposure draft on pensions accounting, the International Accounting Standards Board (IASB) proposed that entities should present a sensitivity analysis to show how changes in key actuarial assumptions might reasonably be expected to affect:

• The defined benefit obligation (DBO) at the end of a reporting period, and;
• The service cost component of the pensions calculation at the start of the reporting period [paragraphs 125I, BC62, BC63-66].

The feedback from those constituents was that the approach - culled from International Financial Reporting Standard 7, Financial Instruments - was a non-starter in terms of its complexity and because it fails to deliver decision-useful information.

This leaves the IASB with a problem: How do you design a disclosure requirement that will require entities to communicate the extent to which a DBO is about more than a single time-point estimate on a balance sheet? The answer, it seems, is to explore the possibility of applying the notion of a measurement uncertainty analysis to pensions accounting. But as anyone who attended the recent IASB employee benefits working group on 27 September will know, no-one really understands what a measurement uncertainty analysis is, let alone how it might apply to a DBO.

To get a grip on the concept of a measurement uncertainty analysis requires a quick reprise of the IASB's fair value measurement literature. The IASB's fair value measurement guidance is set out in what at the moment is a draft IFRS - effectively an import of a US accounting standard FAS157, Fair Value Measurements. Fair value measurement itself is based around a three-level hierarchy made up of:

• Level one: real-world quoted or observable prices;
• Level two: modelled prices based on observable inputs, and;
• Level three: unobservable inputs.

Put simply, if a level-one price is a quoted equity, a level-three price is the kind of fantasy number that wreaked havoc during the credit crisis when the models used to create values for securitised sub-prime mortgages turned out to be less than reliable. In a bid to address this measurement uncertainty - to be polite about it - the IASB issued an exposure draft in June.

The purpose of this document was to propose a requirement for entities to disclose just how soft those level-three prices might be. Summing up the proposed disclosure methodology to the working group, IASB fair value measurement project supremo Hilary Eastman said it requires preparers to change "one or more input to another one, such as changing a growth rate from one to another, where it would have been reasonable [to do so]".

And put like that it all sounds simple enough. The trouble is that no-one in the room seemed to agree. US pensions consultant Jim Verlautz had no idea where to start: "I don't understand how to do this, how to apply this for a defined-benefit obligation." Diana Scott, also a consultant in the US, but with Towers Watson, broadly endorsed his view. "I think that looking at sensitivity analysis and saying I'm looking at one input in isolation is the wrong answer because we know that there are going to be certain relationships. But putting everything together, I just don't know how you do it," she said.

And in the context of a DBO, she was not alone in struggling to see the distinction between a sensitivity analysis and measurement uncertainty analysis: "I see a sort of a blend between the sensitivity and the measurement uncertainty. There are a number of assumptions ... [and the] financial assumptions I think are relatively easy. You can say there is an inter-relationship to a certain extent between them, although I struggle when you get to something like the healthcar cost-trend rate... So I'm kind of struggling with ‘how do I do this?'."

What the staff are trying to do is to establish whether a measurement uncertainty analysis methodology that was originally intended for level three fair values, can in some way be applied to the defined-benefit obligation - even though the DBO is not a fair value.

Andrea Pryde explained that: "The proposal in the exposure draft was based on the disclosure in IFRS7. When we're doing the IFRS7 disclosure here we thought that not having the inter-relationships was useful because it was too hard to do. But then the comment letters come back and say it is not very useful because it is less meaningful.

"I agree that the objectives in the disclosures [are] probably a little bit muddy so we put them in partly because of comments from this working group that it is difficult to think of pensions as one number and it is difficult to understand how the variability around that. So we would just like to explore that a little bit more and see what the real purpose is and whether something like the measurement uncertainty analysis would be a better approach."

HSBC, which figured among the respondents to the June 2010 exposure draft that proposed the level-three fair-value disclosures, pointed out that the need to consider inter-relationships between different components of the measurement uncertainty analysis might lead to "excessive additional operational complexity".

"As an example of this operational difficulty," the London-listed bank argues in its comment leter, "there is a lack of guidance as to the level in the analysis at which the criteria should apply. For example, while it may be relatively practical to apply the criteria to a given instrument type traded within a given location, it becomes impractical to apply the criteria across a large portfolio of different instruments traded across different geographies, where the set of inputs being adjusted becomes very large."

The letter continues: "Therefore, whilst we understand the logic of including the effect of inter-dependencies between unobservable [level-three] inputs, we have concerns that the additional operational complexity involved may not result in an improvement to the information given to the users of financial statements under the current requirements."
If the banks can't do it, the IASB really has to ask itself who can.