Saying how likely something is: research methodology

In a separate post today, I summarised research by the Korea Accounting Standards Board (KASB) and Australian Accounting Standards Board (AASB). That research looked into how preparers of financial statements and auditors interpret various terms used in IFRS Standards to denote how likely an event is.

In this post, I comment on 2 aspects of the joint research by KASB and the AASB:

Did the 16 paragraphs studied all investigate ‘terms of likelihood?  

The questionnaire used in the research asked 2 main questions. One question focussed on 16 paragraphs presenting wording in existing IFRS Standards. This question asked respondents to think about the meaning of terms in a real context within the standard. Because it may be difficult to assess some of the responses without the context, I summarise in the following table each context paragraph given to respondents. In the summary of each paragraph, bold type highlights the ‘term of likelihood’ investigated. I also comment on some of the context paragraphs.

Summary of context paragraphMy comment
1. Definition of bearer plant: living plant that has remote likelihood of being sold as agricultural produce 
2. Definition of lease term: includes an extension option if, at inception of the lease, exercise by the lessee is reasonably certain 
3. Do not recognise government grants until there is reasonable assurance that the entity will both comply with the conditions and receive the grant 
4. Cease capitalising borrowing costs when substantially all preparation activities are complete.‘substantially all’ here does not denote the probability of a future event. So it is not a ‘term of likelihood’.
5. Carry out detailed impairment tests if increases in market interest rates are likely to have decreased an asset’s value in use materiallyThis is just one case (on a non-exhaustive list) when an entity must carry out an impairment test. ‘likely’ here refers to the probability that such a test (if carried out) would reveal that an impairment loss has already occurred.   ‘Likely’ is not used here directly as a probability threshold having direct consequences on recognition and measurement.

The overall principle in IAS 36 Impairment of Assets is that an entity must recognise an impairment loss whenever an asset is impaired. Arguably, the main function of the (non-exhaustive) list of indicators of possible impairment is to tell entities when it is safe not to carry out a detailed impairment test. That is safe when it is not likely that a detailed test would show an impairment.  
6. Definition of a contingent liability: a possible obligation (…) whose existence will be confirmed only by one or more future eventsIt is not clear whether the IASC (the IASB’s predecessor) intended ‘possible’ here to denote a precise probability with unequivocal effects. Also, the threshold having an unequivocal effect in this context is ‘remote’, tested by context 7. (The effect is on disclosure.)
7. Disclose a contingent liability, unless the possibility of an outflow of resources is remote 
8. When the recognition of income is virtually certain, then the related asset is an asset, not a contingent asset, and its recognition is appropriate. 
9. Reverse a provision if an outflow of resources is no longer probable. 
10. Recognise a provision when an outflow of resources is probableThe research used paragraphs 9 and 10 to investigate whether respondents apply the same level of probability to those 2 contexts.
11. It is highly unlikely that a change from the fair value model to the cost model will result in a more relevant presentation.‘highly unlikely’ is not a probability threshold. The standard (IAS 40 Investment Property) does not require an entity to assess the probability of some future event. This statement simply gives a warning that there is a very high bar for deciding to change from the fair value model to the cost model.
12. In assessing whether to recognise unused tax losses as a deferred tax asset, consider whether the losses result from identifiable causes which are unlikely to recur.‘unlikely to recur’ does not establish a precise threshold with unequivocal consequences. Assessing whether recurrence is likely is just one factor within an overall assessment. That overall assessment is of whether it is probable that sufficient taxable profit will be available to permit recovery of the deferred tax asset.
13. The definition of an insurance contract may be met even if the insured event is extremely unlikelyNo IFRS Standard uses ‘extremely unlikely’ as a threshold. This phrase appears only in IFRS 17 Insurance Contracts (and its predecessor IFRS 4), explaining that the definition of an insurance contract includes no probability threshold.
14. Disclose a sensitivity analysis showing effect of changes in a risk variable that were reasonably possible.This requirement is looking for disclosure of changes not already captured in a measurement. So this requirement would serve no purpose if ‘reasonably possible’ were viewed as a threshold high enough to cover only (or mostly) factors already reflected in the measurement.  
15. To apply hedge accounting to a hedge of a forecast transaction, that transaction must be highly probable.This requirement applies to future transactions that have not yet created recognisable assets or recognisable liabilities. Therefore, the requirement is intended to place strict discipline on the use of hedge accounting.
16. It is probable that the expected future economic benefits will flow to the entity.The report suggests that respondents were reading this sentence in the specific context of IAS 38 Intangible Assets.
Table, summary of 16 paragraphs studied in the KASB / AASB report

What did the research mean by ‘range’?

The 2nd main question in the questionnaire asked about 13 terms of likelihood used in IFRS Standards. Before looking at that question, it is worth noting how IFRS Standards use terms that describe levels of probability. The standards use those terms in setting criteria for recognising assets or liabilities or for disclosing information. For example, some IFRS Standards require recognition of an asset or liability if it is probable that some specified event will occur. Some IFRS Standards (but not all) go further: they state that an event is probable if it is ‘more likely than not’ that the event will occur.

That ‘more likely than not’ condition is met if the probability is greater than 50%—regardless of whether the probability is, for example, 50.001%, 70%, 99% or even 100%. In understanding the phrase ‘more likely than not’, what matters is only the lower threshold (in this case, immediately above 50%). There is no need to specify that the outcome of applying that condition is meet if the probability is within a range from 50.001% to 100%, only that the probability is above 50%.

Thus, for most terms denoting a level of likelihood, what matters in answering accounting questions is the lower threshold identified by that term—the lowest probability at which some specified thing should be done. On the other hand, for a few terms with a problem near 0%, what matters may be an upper threshold—the highest probability at which some specified thing should be done.

In summary, in considering accounting questions, the only thing that matters is the threshold itself—not the entire range lying beyond the threshold.

What ranges did the questionnaire ask for?

With that background in mind, let’s turn to the questionnaire’s question about ranges. That question said:

‘Listed below are the terms of likelihood that are contained in IFRS which relate to a level of probability of a transaction or event occurring. Please indicate the range of probability that best corresponds, in your professional opinion, to each term of likelihood in percentage (%) terms on a scale of 0% to 100%.’

Unfortunately it is unclear, in my view, what that question means by the ‘range of probability’ ‘corresponding to’ a term of likelihood. For example, does it mean:

  • the top and bottom of a range when the condition using that term is met? For the above ‘more likely than not’ example, that range would be 50.001% to 100%.
  • different probabilities that might be appropriate in different circumstances? For example, in thinking about whether an event is ‘probable’, respondents gave a point estimate around 20 percentage points higher than they did in thinking about an event that is ‘no longer probable’.  Would those people give a range of 20%? Furthermore, some people might view a term as denoting one probability for assets and a different probability for liabilities. Would those people give the range lying between those two points?
  • some lack of precision about exactly where a threshold lies? For example, some people might view some term as denoting a ‘fuzzy’ probability of X% plus or minus Y%, rather than a precise probability of exactly X%. Would those people give a range from X%-Y% to X%+Y%?
  • an estimate by one person of the range of point estimates that would be provided by different people in the population?

Because of this lack of clarity, it is not obvious how to interpret the ranges reported by respondents. Nevertheless, the results of the survey do give some indication of differences in interpretation.

What ranges did the report provide?

The report provided some information on ranges. Table 5 in the report ‘presents range of numerical probability for each term of likelihood used in the analysis which is interpreted by Australian and Korean accounting professionals.’ That table presents, for each term studied, the minimum and maximum, both for Australia and for Korea. Those minima and maxima appear graphically in Figure 17 (reproduced in my other post of today).

A note to table 5 says ‘Minimum and maximum of terms of likelihood presented above are mean value.’ Therefore, I assume that the minima and maxima denote the means of the minima and maxima for the ranges reported by respondents in their answers to the question on ranges.

Because the question on ranges was so unclear, I wonder whether more useful information might have come from looking at the range and distribution of point estimates. The report does contain figures showing some of the distributions of point estimates, but does not devote much attention to those and other distributions of the point estimates.

Point estimates

The other main question in the survey (on the 16 paragraphs) asked respondents for a point estimate rather than a range.  It asked respondents to ‘indicate the numerical probability that best corresponds, in your professional opinion, to each term of likelihood in percentage (%) terms on a scale of 0% to 100%.’ supporting examples made it clear that the survey was looking for answers that identified the threshold at which a condition would be met. So, although the wording of the question wasn’t totally clear on this point, I think respondents would have realised that this question was asking about thresholds—and not about, for example, a typical probability lying somewhere within the range lying beyond the threshold.

Leave a comment

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