Slide 1

Professor Michael Mainelli, Executive Chairman, The Z/Yen Group

[An edited version of this article appeared as "Accounting for Confidence" Accountancy Futures (March 2011), pages 56-57.]

Beginning a decade ago with embarrassing failures of large firms boasting successful-looking financial statements, followed by current punctured balance sheet and going concern statements of global financial institutions, surely, this is a good time to rethink auditing. And in analysing what might be done, a little science couldn’t hurt.

When people move from science to accounting, they are stunned to find that auditors do not practice measurement science. Scientific measurement specifies accuracy and precision. Accuracy specifies how closely a stated value is to the actual value. Precision specifies how likely repeated measurements under unchanged conditions will produce the same results. Take your bathroom scale. You’d like accuracy and precision, but your weight fluctuates. With an expensive, more accurate scale that weighs you down to the gram, you have more accuracy. For greater precision you need consistent conditions, not jumping, controlling for temperature and calibrating regularly. A measurement system can be accurate but not precise, precise but not accurate, neither, or both. If your bathroom scale contains a systematic error, then increasing sample size by weighing yourself more often increases precision but not accuracy.

Scientists view measurement as a process that produces a range, and they express those ranges using intervals. There is a big difference between point estimation and interval estimation. The former is about a single value, the latter about a range. Auditors provide a point estimate, scientists an interval. Without digressing into statistical discussions of credible intervals versus confidence intervals, scientists express most measurements as X, with an interval. In a simple example assuming a normal distribution, you might be said to weigh 78 kilos ± 0.65 kilos. If +.65 or – 0.65 are one standard deviation from the mean, then 68% of the times you measure your weight the scale will show a value between 77.35 kilos and 78.65 kilos. There are ways of expressing more complex distributions found in science and finance. Statistical terms, such as mean, mode, median, deviation, or skew, are common terms used to describe some of a distribution’s ‘look and feel’. The key point is that scientists are trying to express characteristics of a distribution, not a single point.

Social scientists, not just physical scientists, use confidence levels to report interval estimates. In a poll of voting-intentions, pollsters might express the result as 40% of respondents intend to vote for a certain party. A 90% confidence interval for the proportion in the whole population having the same intention on the survey date might be 37% to 43%. From the same data the pollster might also provide a 95% confidence interval, which might be 34% to 46%.

What might a world of auditors using interval estimation with confidence levels look like? Well, the end result would be the presentation of major entries for the profit & loss, balance sheet and cashflow statements as distributions. For example, a profit might be stated as £83,120,000 ± £1,500,000. On a balance sheet, the value of freehold land might be stated as £150,000,000 ± 45,000,000, perhaps recognising a wide range of interesting properties and the illiquidity of property holdings. Next to each value would be confirmation of the confidence level, e.g. 95% confidence that another audit would have produced a value within that range. Finally, there would be a picture, a histogram of the range, so people can see the shape of things. For want of a term that distinguishes the use of intervals and confidence levels from the use of points or discrete values, let’s use “Confidence Accounting”, a term used by several proponents of the shift to interval estimates such as the Long Finance initiative.

What might the use of intervals and confidence levels achieve? Greater simplicity, greater impact and competition on quality. Confidence Accounting simplifies many discussions of which number to pick, e.g. the lower of cost and value, because the range itself is expressed clearly. There are numerous examples of difficult single numbers in audits, think of assets in exploration, or environmental liabilities, just to get going. For users, presentation would be easier to understand. Footnotes would be simplified or made redundant in many cases. Confidence Accounting goes to the heart of the mark-to-market debate. Different instruments have different values for different entities. A hedge fund that has been caught out on a long-term instrument in the short-term is different from a pension fund that can hold the instrument to maturity. Surely presenting a range of potential future valuations, informed by historic prices, is better than just marked-to-the-market price at a particular valuation date. There is greater value in an audit that confirms a range of market values applicable to a specific firm.

The impact of the audit process would be enhanced by a richer dialogue with managers, shareholders and other investors. Which is worse, forcing directors to specify a single number, such as a guess-timated mean, or asking them to specify their views of the likely range of outcomes? Some organisations will want to provide extremely wide ranges in their distributions. Where this reflects reality, so be it. In other cases managers will hope that a wide range removes some responsibility of meeting target. If they are consistently providing silly future estimates, and these are recorded in the financial accounts, the silly estimates are there for investors to judge. Markets will price the value of tighter distribution ranges, and auditors will be able to sell appropriately the value of greater work to provide better disclosure.

Audit buyers have limited information on which to base choices, so they choose what other people have chosen, an algorithm that over time spirals in on a few brands not competing on measured quality. Major firms assert quality, bragging that compliance has grown in burden and cost. Dispiritingly, it is impossible for an outsider to evaluate quality through analysis of published audits. Under Confidence Accounting, external assessors could evaluate performance. Any major firm will have a number of client failures over a period, say the past decade, but are these within the bounds of their audited financial statements? If so, perhaps a good, or even too prudent, auditor. If not, perhaps a sloppy, or statistically unusual, auditor.

Common counter-charges to Confidence Accounting are complexity and ‘gaming’. To a traditional point estimator, Confidence Accounting looks more complicated. Many people will claim that the mythical “Aunt Agatha” cannot understand all this. Yes, the profession will need to work on specifying standard measures and representations, but the profession should worry more about the scientific measurement ignorance of its members before worrying about Aunt Agatha. Others will point out that these changes may lead to managers gaming a new system. Perhaps, but managers are gaming a system that provides too many ‘get-outs’ based on the unfairness of reporting on single numbers. Others will invoke the ultimate clincher for the status quo, Confidence Accounting is against vested interests, doesn’t solve everything and might reduce the unmeasured quality of present practice. But reform will come, and better that it is based more on science than on more compliance.

Work needs to be done, largely in three areas – commitment by the auditing establishment to reform, restructuring audit skills, and better communication to users of financial information. Auditors are taught that good financial information is accurate, complete, relevant, reliable and timely. Confidence Accounting more accurately reflects the situation, provides more complete information, is more relevant to many users, allows reliability to be checked and takes better account of timing issues. Surely it is timely to start discussing Confidence Accounting now.

Established in 02007 by Z/Yen Group in conjunction with Gresham College, the Long Finance initiative began with the conundrum - “when would we know our financial system is working?” Long Finance aims to “improve society’s understanding and use of finance over the long term”, in contrast to the short-termism that defines today’s financial and economic views. Long Finance is a community which can be explored and joined at – there you can also find "In Search of the Eternal Coin: A Long Finance View Of History".