IFRS standards use too many different terms to say how likely it is that an event will occur. And different people using IFRS Standards interpret those terms in different ways. Those clear messages appeared in a research report issued in 2016 by the Korea Accounting Standards Board (KASB) and Australian Accounting Standards Board (AASB). I summarised that report at https://accountingmiscellany.com/saying-how-likely-something-is
In this post, I discuss how I think the IASB could reduce the number of terms it uses and also make those terms more understandable.
Probability thresholds in IFRS Standards
IFRS Standards use terms of likelihood (such as ‘remote’, ‘probable and ‘virtually certain’) to set probability thresholds in some recognition, measurement and disclosure requirements. For example, a requirement might specify that a company must recognise an item if the probability that some specified event could occur (or that some specified state of affairs could arise) and the probability of that event (or state of affairs) exceeds some specified level. (In the rest of this post, I use probability threshold as shorthand for the probability level set in a probability threshold.)
The KASB/AASB report makes it clear that IFRS standards uses too many different terms to designate probability levels. In my view, there are several causes of this problem:
- IFRS Standards set too many different probability thresholds;
- IFRS Standards use too many different terms to label a single probability threshold; and
- some of the labels for probability thresholds are unclear, so people do not interpret and apply the thresholds consistently.
With those problems in mind, I discuss the following topics below:
- probability thresholds in recognition and measurement
- words for probability thresholds
- probability thresholds and disclosure
- updating probability thresholds in existing IFRS standards
Probability thresholds in recognition and measurement
The KASB/AASB report indicates that IFRS standards uses too many different terms to designate probability levels. I agree.
Once cause of this problem is that IFRS Standards set too many different probability thresholds. In my view, the IASB should create a structured menu containing a small number of different probability levels for use in probability thresholds. Whenever the IASB introduces a probability threshold in a new or amended Standard, it should pick a probability threshold from that menu.
In this section, I focus on probability thresholds in recognition and measurement (in practice, mainly recognition). I discuss probability thresholds in disclosure in a later section.
How many probability thresholds?
How many probability thresholds should the IASB designate in the future? I would answer that question using the following principles:
- Principle 1. One probability threshold should cover events that are so likely to happen that the chance of the event not happening can be disregarded. Those events can be treated in the same way as events that are certain.
- Principle 2. A 2nd probability threshold is needed at the mid-point of the probability scale (50%). IFRS Standards currently label this point as ‘more likely than not’.
- Principle 3. A 3rd probability threshold may be needed for events that are possible but have a probability below 50%. That 3rd threshold would serve little purpose if it were to be set close to 0% or 50%.
Deciding where to set that threshold—closer to 50%, closer to 0% or around 25%—may well be the most difficult issue in constructing a menu of probability thresholds.
- Principle 4. Symmetry: if the menu includes a probability threshold at one distance from one end of the probability range (0%-100%), it should include a probability threshold at the same distance from the other end of the range.
To illustrate, if the IASB designates a threshold for events that are almost certain to happen, it should also designate a threshold for events that are almost certain not to happen.
Symmetry can be achieved anyway by relabelling the event that X occurs as the event that not-X occurs (said differently, as the event that X does not occur). But the menu should show that symmetry explicitly, without relying on such relabelling.
Combining symmetry (principle 4) with principles 2 and 3 leads to 2 more probability thresholds:
- a 4th probability threshold covering events that are so likely not to happen that the chance of the event not happening can be disregarded. Those events could be treated in the same way as events that are certain not to occur.
- a 5th probability threshold at which it is possible that the event may not occur (and the probability of non-occurrence is at the 3rd probability threshold). Said differently, the 5th threshold probability would be set at the point where the probability of the event occurring is 100% minus the 3rd probability threshold.
So if the 3rd threshold is 30%, then the 5th threshold would be 70%. (I use percentages here for illustration only. I discuss below whether thresholds should be set as percentages.)
5 probability thresholds is enough
The above principles result in 5 thresholds. More than 5 probability thresholds would be too many. It is unrealistic to expect preparers, auditors, users of financial statements and regulators to work with more than 5 thresholds.
Words for probability thresholds
The KASB/AASB report highlights problems causes by words used to designate probability thresholds:
- IFRS Standards use too many different terms to label a single probability threshold; and
- some of the terms labelling probability thresholds are unclear
Too many different terms
The KASB / AASB report investigated 13 terms used in IFRS Standards to label probability thresholds. It concludes that 8 or 10 of them could be placed in 4 or perhaps 5 groups. Survey responses suggested that respondents viewed each term within the same group as having a similar meaning.
Table 1 (reproduced from my earlier post) shows all 13 terms in (approximately) descending order of probability and labels the 5 groups A to E. The table shows 2 sets of results: for Australian respondents and for Korean respondents.
|Virtually certain||Group A||Group A|
|Substantially all||Group A||Group B|
|Highly probable||Group B||Group B|
|Reasonably certain||Group B||Group A|
|Unlikely||Group D||Group D|
|Highly unlikely||Group D||Group D|
|Extremely unlikely||Group E||Group E|
|Remote||Group E||Group E|
Notes on those groups A to E:
- ‘Reasonably certain’ is in group A for Korean respondents but in group B for Australians. Conversely, ‘substantially all’ is in group A for Australians, but in group B for Koreans.
- Korean responses did not lead to ‘probable’ and ‘likely’ being in any group. For Australians, ‘probable’ and ‘likely’ fell into a single group (group C).
- For both Australian and Korean responses, 3 of the terms (‘reasonably assured’, ‘reasonably possible’ and ‘possible’) did not fall into any group.
Reducing the number of terms
The KASB / AASB report suggests that within each group there might only be a need for one term. Following that suggestion would replace 10 existing terms by 5 terms (going by Australian responses) or 8 existing terms by 4 terms (going by Korean responses). That suggestion:
- does not consider whether the other 3 terms studied could (or should) be replaced.
- does not deal with other terms used in IFRS Standards but not covered in the KASB/AASB survey.
Is 5 the optimum number of groups?
The number of groups (5) identified in the KASB/AASB report is the same as the maximum number of probability thresholds that I recommend above for future standard-setting. Perhaps this provides the beginning of a basis for mapping existing terms to the 5 probability thresholds that I recommend above.
Only one term per threshold
I agree with the KASB/AASB report that some IFRS standards use different terms to label the same probability threshold. This causes confusion. In the future, for each probability threshold the IASB sets, it should aim to use only one term to label that threshold.
Unclear terms for probability thresholds
The KASB / AASB report shows that the terms used are not clear enough for people to interpret and apply them consistently. I comment below on selecting terms to label probability thresholds.
I have developed my comments assuming a future menu of only 5 probability thresholds, as I recommend above. I intend the comments to show a way of thinking about future use of terminology. I do not intend those comments as interpretations of existing uses of the terms in IFRS Standards.
- To my ear, the term ‘virtually certain’ conveys well the idea that an event is so close to being certain that it can be treated in the same way as an event that is certain. It seems to me that the terms ‘reasonably certain’ and ‘reasonably assured’ do not convey such a high level of near-certainty; I am not sure whether this is also the case for ‘highly probable’.
- At least sometimes, there may also be a conceptual difference between ‘virtually certain’ and ‘reasonably certain’ / ‘reasonably assured’ / ‘highly probable’. ‘Virtually certain’ is perhaps sometimes used to disregard procedural steps that lack substance: for example: (a) some options that have no genuine substance; or (b) legal steps that are still pending but are viewed as purely ceremonial.
- As explained in my earlier post, ‘substantially all’ doesn’t, in fact, designate a level of probability at all. Instead, it identifies whether an entity has a feature that gives it the possibility of rewards or exposes it to risks.
- The IASB used the term ‘likely’ rather than ‘probable’ for a few years in the early 2000s. This was mainly because of concerns that the term ‘probable’ put some readers off, by appearing to suggest excessively technical and rigorous statistical precision. To the best of my knowledge, the IASB never intended ‘likely’ to denote a threshold different from ‘probable’.
- As discussed below, IFRS Standards sometimes specify that ‘probable’ means ‘more likely than not’—immediately above 50%. When that specification is not present, many people may feel intuitively that ‘probable’ denotes a much higher threshold.
- Self-evidently, ‘highly probable’ denotes a higher probability than ‘probable’, and ‘highly unlikely’ denotes a lower probability than ‘unlikely’. I read ‘extremely unlikely’ as a synonym for ‘highly unlikely’.
- It may be unclear whether ‘it is unlikely that event X will occur’ is intended to mean the same as ‘it is likely that event X will not occur’. As confirmation that this lack of clarity exists, the KASB/AASB report says that past research has shown that probabilities assigned to mirror-image pairs such as ‘probable’ and improbable’ do not sum to 100%.
- The meaning of ‘reasonably possible’ may be unclear. (I comment below on the role of this phrase in disclosure requirements.)
- ‘Remote’ is well established accounting jargon (perhaps particularly in North America), but may be opaque to general readers. It isn’t clear to me whether it designates the same probability as ‘highly unlikely’ / ‘extremely unlikely’.
I discuss below some other aspects of terminology for probability thresholds:
- numerical guidance
- probable as more likely than not
- more likely than not: a numerical percentage
- expected value
- US GAAP: similarities and differences
Some respondents to the KASB / AASB survey suggested that the IASB should specify probability levels using percentages and not only using words. Superficially, it might seem that specifying percentages would lead to more consistent judgements by preparers of financial statements.
Nevertheless, I do not see any benefit in specifying percentages. When significant uncertainty exists, I doubt whether preparers can estimate the level of uncertainty with great precision. For instance, suppose that a preparer estimates that an event has a probability in a range of about 65% to about 85%. In my view, setting a probability threshold of, say, 75% would not lead to the preparer making more consistent judgments or to more useful accounting than setting a threshold that relies on a narrative description rather than on a percentage.
Probable as more likely than not
The IASB’s predecessor (IASC) introduced the term ‘more likely than not’ into its standards in 1998 when it issued IAS 37 Provisions, Contingent Liabilities and Contingent Assets. IAS 37 uses the term ‘probable’ as a threshold in one of its recognition criteria: paragraph 14(b) of IAS 37 sets a criterion that ‘it is probable that an outflow of resources embodying economic benefits will be required to settle the obligation’. Paragraph 23 goes on to say: ‘For the purpose of this Standard, an outflow of resources or other event is regarded as probable if the event is more likely than not to occur […]’.
IASC used the term ‘probable’ because it was (at that time) in the IASB’s Conceptual Framework and also appeared in recognition criteria in other IASC standards. The IASC wanted to give guidance on what ‘probable’ meant, but felt that the project to develop IAS 37 was not the right place to revisit what ‘probable’ means in other standards that already used the term. So, IASC included a footnote to paragraph 23 stating “The interpretation of ‘probable’ in this Standard as ‘more likely than not’ does not necessarily apply in other standards.”
I have often seen blanket statements that ‘probable’ means ‘more likely than not’ in IFRS Standards. Indeed, the IASB itself made such statements in paragraphs BC81 of the Basis for Conclusions on IFRS 5 Non-current Assets Held for Sale and Discontinued Operations and BC211 of the Basis for Conclusions on IFRS 15 Revenue from Contracts with Customers. That is an understandable position to take, and not surprising, but it goes further than IASC felt able to go in 1998 when it issued IAS 37.
More likely than not: a numerical percentage
One term in IFRS Standards does specify a precise level of probability denoting a percentage. That term is ‘more likely than not’. It refers to an event that has a probability of more than 50%. Unfortunately, it seems that most people do not understand that term until someone explains it too them.
I think the reason why the term is difficult to understand is that it uses highly condensed wording. Here is fuller wording: an event is more likely than not if it is more likely that the specified event will occur than that the event will not occur. Thus, an event is more likely than not if its probability is, for example, 50.0001%. An event is not ‘more likely than not’ if the probability of the event is only 50%.
Paragraph 23 of IAS 37 does include the full explanation of the phrase ‘more likely than not’: ‘the probability that the event will occur is greater than the probability that it will not occur’. Although that explanation is there, I suspect it is not prominent enough, particularly for use outside IAS 37.
Also, although the phrase ‘more likely than not’ refers to a precise percentage (immediately above 50%), the phrase and supporting guidance spell that out only in words, without using the numeric symbol ‘50%’. As a result, many people do not understand that definition immediately. Some people might understand the phrase ‘more likely than not’ more easily if the supporting guidance included that numeric symbol. That description already contains that threshold of (immediately above) 50%, but explains it only in words, not using the numeric symbol having the same meaning.
Sometimes IFRS Standards uses the term ‘expected’ to denote the expected value in the statistical sense of the probability-weighted average of all possible outcomes. ‘Expected’ in the phrase ‘expected value’ does not designate a probability level and so cannot be used as a probability threshold.1
1. For some types of probability distribution, the expected value is an amount that equals the most likely amount. Nevertheless, that fact does not turn the word ‘expected’ in the phrase ‘expected value’ into a probability level.
At other times, the Standards use the term expected’ more vaguely, perhaps as a synonym for the most likely outcome or perhaps as a synonym for a term denoting some particular probability threshold (such as ‘probable’).
To avoid confusion, when I was writing standards for the IASB:
- I used ‘expected’ only in the first sense—in the phrase ‘expected value’.
- to refer to the most likely outcome or a specified probability threshold, I used wording referring explicitly to that outcome or threshold.
- to refer in a general sense to an entity’s estimates of possible outcomes and the probabilities of those outcomes but without specifying a probability level, I used more generic terms, such as ‘estimated’.
In my view, the IASB should, in the future, use ‘expected’ only when that term is part of the phrase ‘expected value’.
US GAAP: similarities and differences
It might sometimes be unclear whether terminology in IFRS Standards has the same meaning as terminology in US Generally Accepted Accounting Principles (US GAAP). The IASB decided to use the term ‘highly probable’ in IFRS 5 Non-current Assets Held for Sale and Discontinued Operations and IFRS 15 Revenue from Contracts with Customers—both Standards developed working together with the US Financial Accounting Standards Board (FASB).
According to paragraphs BC81 of the Basis for Conclusions on IFRS 5 and BC211 of the Basis for Conclusions on IFRS 15, the IASB made this decision with the aim of achieving the same meaning in IFRS 5 and IFRS 15 as the US GAAP term ‘probable’, defined in US GAAP as ‘likely to occur’. Paragraph BC81 states that the IASB regarded ‘highly probable’ as implying a “significantly higher probability” than ‘more likely than not’. Another place where IFRS Standards use ‘highly probable’ is in requirements on hedge accounting.
When the IASB looks at establishing which terms to use to label probability thresholds, it will be important to consider implications for translation. Some terms may be easier to translate clearly than other terms.
The KASB/AASB report makes interesting comments on some translation issues and on seeking explicit seek input on translation and interpretation issues during standard-setting outreach and consultation.
Probability thresholds and disclosure
The discussion above is mainly about probability thresholds used in decisions about whether to recognise something. What appear to be probability thresholds also appear in some disclosure requirements. For example, some IFRS Standards require disclosure of a sensitivity analysis showing effects of changes in a risk variable that were ‘reasonably possible’.
This requirement is looking for disclosure of changes not already captured in the measurement of a recognised item. In this context, ‘reasonably possible’ serves a purpose as a probability threshold only if ‘reasonably possible’ is lower than the probability threshold (if any) applied in deciding whether to recognise the related item.
Furthermore, ‘reasonably’ seems to indicate that the objective of the requirement is not to disclose all possible changes in a risk variable, but to disclose all changes whose probability or magnitude may be high enough to interest users of financial statements. To meet that objective, it may be best to view ‘reasonably possible’ in that context as a threshold reflecting both probability and magnitude—not as a threshold of a fixed probability only:
- if the effects are very large, disclosure might be needed at a lower probability level.
- if the effects are small, disclosure might not be needed even at a high probability level.
As a further example, IAS 37 requires companies to disclose contingent liabilities, unless the possibility of an outflow of resources embodying economic benefits is ‘remote’. It might be best to view ‘remote’ in that requirement as referring to a threshold reflecting both probability and magnitude—in the same manner as I have suggested for ‘reasonably possible’. (I am not taking a position on whether ‘remote’ and ‘not reasonably possible’ are the same threshold or different thresholds).
Updating probability thresholds in existing IFRS standards
I have suggested above that the IASB should begin using:
- no more than 5 different probability thresholds in future standard-setting; and
- for each probability threshold, only a single label for that threshold.
Ideally, at the same time as beginning to make those changes, the IASB would also apply those changes to all existing IFRS standards that set probability thresholds.
Unfortunately, I doubt whether such an exercise would pass any reasonable cost-benefit test. Probability thresholds are tightly linked to recognition and derecognition criteria and measurement decisions. Those issues are usually fundamental to most Standards. So, conforming the terminology would not just be a narrow-scope housekeeping exercise to update terminology. Any such exercise would need to consider whether conforming the terminology would change any different accounting outcomes in practice, and whether any such change would past a cost-benefit test.
And even if there were ultimately no changes in outcomes, preparers of financial statements (and auditors and regulators) would still incur costs in:
- assessing, and commenting on, an exposure draft; and
- once the change has been finalised, assessing whether the change does, in fact, cause any changes for a particular company.
Updating existing standards over time
Although a blanket exercise to update probability thresholds (and their labels) in all existing IFRS Standards would not pass a cost-benefit test, the IASB should look to make targeted updates over time as it amends existing standards containing probability thresholds.
In this post, I suggest that the IASB should:
- use no more than 5 different probability levels (probability thresholds) when it sets new requirements;
- for each of the 5 probability thresholds, always use the same label for that threshold when it sets new requirements; and
- remain alert for case-by-case opportunities to update existing probability thresholds to conform to the new guidance (no more than 5 thresholds, 1 term per threshold). Updating all probability thresholds in existing IFRS Standards would not pass a cost-benefit test.
The KASB/AASB report provided useful insights into how IFRS standards use language and how IFRS standards are translated. There is scope for a lot more research on these topics and I encourage researchers to investigate them.