Pension accounting under IAS 19

IASC completed a major revision of its standard on pension costs in 1998, when it issued IAS 19 Employee Benefits. But although that was a major update to the Standard, the revision did not involve a thorough overhaul of every important aspect of the Standard. Scope limitations adopted by the IASC at the start of that project left a lasting major effect on the Standard.

This post starts by examining the background. It reviews the 4 objectives of IASC’s revision of IAS 19 Employee Benefits between 1995 and 1998.

The post then looks briefly at initial inputs used in the project, as captured in an Issues Paper published in 1995. The post concludes by identifying the limitations of those inputs. I now realise that the most important limitation was a failure to consider how to account for pension benefits that are already vested.  


IASC completed its Comparability and Improvements project in late 1993 by issuing updated versions of 10 of its existing Standards. Although that package of 10 substantially improved Standards included IAS 19 Retirement Benefit Costs, IASC felt strongly that IAS 19 still needed much more work.

Soon after I joined the IASC staff in 1994, I was asked to write a project proposal for a project to improve IAS 19 further. I went on to manage that project throughout until IASC published IAS 19 Employee Benefits in 1998.

The aim of the new project was not to rethink every aspect of pension accounting. (Indeed, the Comparability and Improvements project had not rethought every aspect, either.) The project had 4 main objectives:

  • to consider concerns that the 1993 version of IAS 19 was too Anglo-centric.
  • to consider concerns that some features of IAS 19 appeared to delay recognition of some gains and losses. In the minds of some people, those features conflicted with IASC’s Conceptual Framework.
  • to review (and preferably reduce) the wide range of actuarial valuation techniques permitted by IAS 19.
  • to add requirements covering types of employee benefit other than pensions.

An Anglo-centric Standard?

Apparently, during the final stages of IASC’s work on the 1993 version of IAS 19, people complained that the Standard had been developed in a way that considered only the types of pension plan found in North America and the UK. People making these complaints felt that the Standard’s requirements were not appropriate for different types of pension plan found in, for example, the Netherlands, Japan or Germany.

To address those concerns, IASC considered it important for the Steering Committee working on the new project to include full representation from beyond the Anglosphere. IASC set up the Steering Committee with a chair from the Netherlands. Of the other 5 voting members, 2 were from Japan or Germany.

Conflict with the Conceptual Framework?

The decisions on pension costs made in IASC’s Comparability and Improvements project included some changes that would not result in immediate recognition of some actuarial gains and losses or of past service costs. Near the end of that project, some people expressed a view that lack of immediate recognition is inconsistent with IASC’s Conceptual Framework.

Actuarial gains and losses

For actuarial gains and losses, IASC decided in 1998 to permit a ‘corridor’ approach similar to one then in use in the US, though not identical.  Subsequently, the Board removed the ‘corridor’ from IAS 19, instead requiring actuarial gains and losses to be presented outside profit or loss in Other Comprehensive Income.

Past service cost

Past service cost arises when (a) an employer changes a pension plan and (b) the new terms include enhanced benefits for some past service (service already rendered by employees). When the Board of IASC first discussed past service cost, it was evenly divided on when past service cost should be recognised as an expense:

  • half of the Board thought it should be recognised immediately (because the cost relates to service already rendered).
  • the other half wanted to recognise past service cost over the remaining working life of current employees (on the view that benefit enhancements incentivise current employees).

In the exposure draft (E54, published in 1996), IASC explained that it was evenly divided. IASC asked respondents to say what they preferred, and why. Unsurprisingly, respondents’ opinion was just as divided as opinion among the Board itself. But in debating the alternatives, the Board came up with an intermediate approach. IASC decided that past service cost should be recognised over the period until the enhanced benefits become vested. (As a consequence, if enhanced benefits were already vested, their cost would be recognised immediately.)

Leaving an important issue unresolved in an exposure draft is not generally wise. The likely outcome (as in this case) is that respondents will be as divided as the standard setter was itself. So, respondents will not generate new ideas or arguments that will help the standard setter to reach a conclusion. But at least in this case, leaving the issue of past service cost unresolved did buy IASC some time. That extra time helped IASC move on and find a new solution. That new solution combines advantages of both options presented in the exposure draft.   

That new solution introduced a distinction between vested benefits and benefits that are not yet vested. As I discuss further below, thinking more about that distinction could pave the way for making pension accounting more consistent with accounting for provisions and for financial liabilities.

Actuarial valuation techniques

The 1993 version of IAS 19 permitted an unjustifiably wide range of actuarial methods for measuring pension liabilities. It included a list of 6 methods, grouped into 2 families (accrued benefit methods and prospective benefit methods).

The project came quickly to the conclusion that IASC should keep only one of the methods: the projected unit credit method. Confusingly, the projected unit credit method is not in the prospective benefit family. Instead, it is one of the 2 members of the accrued benefit family, the other being the unit credit method.

The unit credit method looks at each year of service by employees. It asks how much extra benefit employees gained by service in that year. The estimated present value of that extra unit of credit is the service cost for that year.

The distinction between unit credit and projected unit credit arises when pension benefits depend partly on future salaries. An example might be a pension that is 1% of final salary for each year of service. The unit credit method measures the current service cost for the current period as 1% of current salary. The projected unit credit method sees the service cost as 1% of estimated (projected) final salary.

The Steering Committee did not take long to select the projected unit credit method. It was already in use in the equivalent US standard (known then as SFAS 87). And no committee member saw any strong reason to go for a different method.

Even without the US precedent, I suspect IASC might well still have selected one of the accrued benefit methods (unit credit or projected unit credit). Both methods identifying the unit of benefit earned during the period and use the estimated cost of that unit cost as the service cost. That seems a natural approach for an accounting standard to take. (In contrast, the prospective benefit methods all involve averaging over employees’ working lives.)

Unit credit or projected unit credit?

What about the choice between the unit credit and projected unit credit methods? Without the US precedent, I suspect making that choice might have been harder. The ‘projected’ part of the projected unit credit method includes future salary increases that have not yet been made. Some people view companies as not yet having any liability to make future salary increases.

From first principles, I think the question of whether to include future salary increases is quite finely balanced. Overall, 2 arguments have always made me come down marginally in favour of including them:

  • a liability based on final salary is more onerous than an otherwise identical liability based only on current salary. The measurement should reflect that difference.
  • discounting removes inflation from future cash flows. Excluding future salary increases from the cash flows would remove inflation a second time.

Other employee benefits    

The 1993 version of IAS 19 dealt only with pension costs. IASC needed to show the International Organization of Securities Commissions (IOSCO) that its standards were comprehensive enough to be suitable for cross-border listings. As part of that exercise, IASC needed to develop reasonably quickly a standard covering other forms of employee benefit—particularly post-employment medical benefits and termination benefits.

The project built IAS 19 out from a standard covering just pensions into a standard covering all types of employee benefit.

But the section on the contentious and difficult topic of equity-based compensation was left as just a place-holder until IASC’s successor (the IASB) could complete IFRS 2.   

Initial collection of information

For the first meeting of the Steering Committee, I put together an issues paper listing the main accounting issues the committee would need to think about. I compiled it mainly by looking through everything I could find on the topic in IASC’s rather under-equipped library.   

After the meeting, the IASB’s Technical Director Liesel Knorr suggested we should publish the issues paper and seek public feedback. We did this, though in retrospect I doubt whether it was the right decision. The paper was not developed enough and was too open-ended to generate useful feedback. And I did not yet understand well enough how large a burden requests for feedback place on respondents.

Amusingly, a year or two later, the Society of Actuaries in the US ordered 500 copies of the Issues Paper from us as an educational resource. They said it was the most comprehensive guide they’d seen to pension accounting issues.

At the time, I didn’t realise that some important topics were missing from the Issues Paper. That was because the sources I had considered discussed only those accounting issues that standard setters had identified in past work on pension costs.       

I discuss below the things missing from the 1995 Issues Paper under 2 headings:

  • vested benefits
  • other

Vested benefits

One important early decision was that the measurement of pension liabilities should include not only vested benefits (ones not conditional on future service), but also unvested benefits.

That was an important and indeed foundational decision. I still believe it is also the only decision producing an answer that is both useful and workable. To take an extreme example, in some pension plans in France in the 1990s, employees would receive no pension at all if they left the company one day before their normal retirement date. Restricting measurement of such liabilities to only vested benefits would mean recognising no liability at all until the retirement date—a ridiculous outcome.   

But we failed to give any attention at all to the corollary of accounting for unvested liabilities—how should you measure pension liabilities that are already vested?  Once a pension liability is vested, it is really just a special type of financial liability, or perhaps a provision or even an insurance contract. So, logically:

  • accounting for a vested pension liability ought to be similar to the accounting for a financial liability (or provision, or even an insurance contract);
  • as a pension liability moves closer to vesting, measurement of the unvested liability should converge towards the future measurement of the liability once it has become vested.

Now, in fairness to us at the time, IASC then had no standard—or even well-advanced project—on financial liabilities (future IAS 39 and even more future IFRS 9), provisions (future IAS 37) or insurance contracts (future IFRS 17). So, we had no pre-existing benchmarks we could refer to in setting the accounting for the vested part of a pension liability. We had only the existing pension standards (particularly the most up-to-date one, the US’s SFAS 87).

If I had to write IAS 19 from scratch today, I would start by defining the accounting for a liability that has just become vested:

  • I would measure vested benefits of uncertain amount or timing (such as annuities or post-employment medical benefits) using IAS 37 Provisions, Contingent Liabilities and Contingent Assets. (I might use IFRS 17 Insurance Contracts instead, but this might not be viable without a contractual service margin.)
  • for vested benefits of certain amount and timing, I might also use IAS 37 or I might use IFRS 9 Financial Instruments.    

Other topics

I will discuss discount rates under IAS 19 in a future post.

IAS 19 did not cover the following topics in 1998, though since 2011 it has covered them:

  • how to measure pension benefits containing embedded derivatives. An example is benefits providing the higher of 2 amounts calculated on different bases.
  • how best to reflect risks borne by employees rather than by the employer.
  • whether contributions by employees are always best treated as benefits that happen to be negative.

IAS 19 still does not address some other topics:

  • whether the Standard should state explicitly what measurement basis it applies to pension liabilities, rather than just providing an organised catalogue of individual measurement decisions. (I suspect someone picking a measurement basis today might call it something like ‘fulfilment value’—though it is worth remembering that labels of that kind were not yet in use in the 1990s.)
  • whether estimates of cash flows should sometimes consider a range of possible outcomes, not just the single outcome estimated to be the most likely outcome.
  • benefits varying with the level of plan assets. The IASB conducted some limited research on these plans between 2018 and 2021, but decided not to amend IAS 19 in this respect.  

Links to other posts

I have mentioned some other aspects of pension accounting in earlier posts:


I doubt whether there will be any appetite to change IAS 19 much in the next few decades. But, if the IASB does ever decide to change IAS 19, I believe there is a viable way forward.

IASC’s project between 1995 and 1998 made some important improvements to pension accounting. But, in retrospect it failed to ask what is—I now think—the most important question of all: how should companies account for pension benefits that are already vested?

The most natural answer to that question would be: in the same way as for provisions (applying IAS 37) or, conceivably, in the same way as for financial liabilities (IFRS 9) or insurance contracts (IFRS 17). That answer seems natural today, but it was not yet available in 1998 because none of those Standards existed then.

Once that question is answered, the accounting for unvested benefits would need to be designed so that their measurement converges towards the measurement of vested benefits on the day when vesting occurs.     

Leave a comment

Your email address will not be published. Required fields are marked *