Auditing Estimates: Point Estimates, Ranges and Detecting Management Bias

An estimate can pass every individual test and still be biased. If management's allowance, warranty accrual, inventory reserve and deferred tax valuation allowance each land at the favorable edge of what is reasonable, no single entry is wrong, and the financial statements are. The audit question is not only "is this number reasonable?" but "which way do the numbers lean, and why?"

The mechanics of the three audit approaches, risk factors and documentation are in auditing accounting estimates under AU-C 540 (SAS 143). This post is about the end of the process: comparing management's number to the auditor's, using ranges and detecting bias.

What the standards say about bias

SAS 143, codified in AU-C 540, is described by the AICPA as improving, expanding and modernizing existing requirements for auditing estimates, not as creating wholly new ones. The Journal of Accountancy's summary says it enhances auditors' focus on the factors driving estimation uncertainty and on potential management bias, and that it applies to audits of financial statements for periods ending on or after December 15, 2023.

The PCAOB's standards for issuer audits spell out the comparison step in detail, and they are the clearest published text for the ideas below. They are cited here by paragraph:

• AS 2501.16: the auditor evaluates whether the company has a reasonable basis for its significant assumptions and, where it selected assumptions from a range of potential assumptions, for that selection. It also evaluates whether the significant assumptions are consistent with other relevant evidence where applicable.
• AS 2501.25: if the auditor's independent expectation is a range rather than a point estimate, the auditor determines that the range encompasses only reasonable outcomes under the framework and is supported by sufficient appropriate evidence.
• AS 2501.26: the auditor compares the independent expectation to the company's estimate and evaluates the difference.
• AS 2810.27: the auditor evaluates whether the difference between estimates best supported by the audit evidence and the estimates in the financial statements, which are individually reasonable, indicates possible bias.
• AS 2401.64-.65: the auditor performs a retrospective review of prior-year estimates against actual results and, if bias is identified, evaluates whether it represents a fraud risk.

Point estimate or range

A point estimate is one number. It is the natural choice when the auditor can recompute with data and a model that has little judgment in it, such as depreciation on a defined schedule.

A range is the natural choice when reasonable people disagree about assumptions. The discipline in AS 2501.25 is that the range holds only reasonable outcomes. A range wide enough to contain everything is not evidence. If management's number sits inside a narrow range with evidence behind each end, the estimate is supported. If it sits outside, the difference is a misstatement to be evaluated.

Three tests keep a range honest:

1. Each end has evidence. Write the assumption set that produces the low end and the high end.
2. The range is narrower than materiality, or nearly so. If the range is several times materiality, the estimate is not auditable to a conclusion without more evidence, and the team should say so.
3. The range was set before management's number was seen, or the team documents why not. Anchoring on management's figure is the most common way ranges become convenient. The Journal of Accountancy notes that auditors are particularly vulnerable to anchoring because they begin from management-provided statements.

Bias is directional

AS 2810.27 gives the sharpest test. If every estimate is individually reasonable but the effect of each difference from the best-supported estimate is to increase earnings, the auditor evaluates whether that indicates bias. It also says bias can result from the cumulative effect of changes in several estimates. If estimates were grouped at one end of the range of reasonable estimates last year and at the other end this year, the auditor evaluates whether management is using swings in estimates to achieve a desired outcome, for example to offset higher or lower than expected earnings.

The Journal of Accountancy adds that bias is not necessarily intended. It lists anchoring, overconfidence, and interpreting evidence to fit existing beliefs, among others, and suggests seeking evidence that could disprove management's assumptions and asking why an initial assessment could be wrong.

Worked example

All figures are illustrative. Pre-tax income is $1,900,000. Materiality is $190,000 (10 percent of pre-tax income) and the clearly trivial threshold is $9,500.

The team compares each estimate to the point best supported by its evidence:

Each difference is below materiality and four of five are on the same side. The net is $105,000, which is 55 percent of materiality and 5.5 percent of pre-tax income. That is a misstatement to add to the schedule of uncorrected items, and the one-directional pattern is the more important finding.

The retrospective review adds history. Management's allowance was $390,000 last year; actual write-offs against that population were $445,000. Last year's inventory reserve was $200,000 and actual disposals were $228,000. In both years the estimate came in below the outcome. The team concludes there is possible bias, evaluates whether it is a fraud risk under the standard, and discusses it with those charged with governance.

A client that says "each of these is within a range" has not answered the question. The aggregate is the finding.

The retrospective review: how to do it

AS 2401.64 directs the review at significant accounting estimates reflected in the prior year's financial statements. A workable routine:

1. List last year's significant estimates and the amounts recorded.
2. Obtain the actual outcome (write-offs, claims paid, disposals, tax assessed).
3. Compute the variance by estimate and the sign of each.
4. Look for the same sign across estimates and across years.
5. Ask management to explain the variances, and test the explanations.

The standard adds that the retrospective review does not call into question the auditor's prior-year judgments made on the information available then.

Responding to bias

If the team finds bias, AS 2810.26 requires evaluation of whether its effect, together with uncorrected misstatements, results in material misstatement, and whether risk assessments, in particular fraud risk, and the related responses remain appropriate. Practical steps:

• Propose adjustment to the best-supported amount, not only to the edge of the range.
• Tell management what the pattern looks like and ask why.
• Extend testing on other estimates where the same person judged.
• Consider the communication to those charged with governance.
• Document the reasoning, not only the arithmetic.

See materiality in auditing and SAB 99 on qualitative factors for why a smooth earnings pattern changes the evaluation of small amounts.

Documentation checklist

• The independent expectation or range, with the evidence behind it.
• The comparison to management's number and the difference.
• The retrospective review and its results.
• The aggregate schedule showing the direction of each difference.
• The conclusion on bias and the effect on risk assessment.

FAQ

Is a difference inside the auditor's range a misstatement?

Not by itself. The auditor still evaluates whether the pattern of differences across estimates indicates bias.

Does management bias always mean fraud?

No. Bias can be unintentional. The auditor evaluates whether the circumstances represent a risk of fraud and responds accordingly.

How many years should a retrospective review cover?

At minimum the prior year's estimates against actual results. A longer history is more informative where the data is available.

Sources

• AICPA, Implementing SAS 143: Re-evaluate and improve auditing estimates
• Journal of Accountancy, Accounting estimates and disclosures addressed in new auditing standard (July 2020)
• Journal of Accountancy, I'm not biased, am I? (February 2015)
• PCAOB, AS 2501, Auditing Accounting Estimates, Including Fair Value Measurements (paragraphs .16, .25 and .26); AS 2810, Evaluating Audit Results (paragraphs .24-.27); and AS 2401 (paragraphs .63-.65), in the PCAOB auditing standards for audits after December 15, 2024

CORAA prepares ready-to-review audit files for US CPA firms from QuickBooks books; you review and sign. See /us.

AU-C references are to AICPA Professional Standards, paraphrased here; read the standard itself before relying on a requirement. This page is guidance, not professional advice. Report an error