Eduzan

Learning Hub

Eduzan

Eduzan / 04 Financial Statement Analysis

FSA 5 – Evaluating Quality of Financial Reports

Worked examples are fully visible. Check-yourself items are study aids you can reveal one at a time.

Recognising weak financial reporting before the market does is one of the few genuinely repeatable edges available to an analyst. The payoff is asymmetric: spotting high-quality reporting merely raises confidence in work you were going to do anyway, while spotting poor reporting early can prevent a permanent loss of capital.

Two well-documented cases frame the reading. James Chanos took a short position in Enron in November 2000, more than a year before the company filed for bankruptcy protection in December 2001. His case rested on a mix of quantitative and qualitative evidence: a return on capital that was below that of comparable companies and below Enron’s own cost of capital, an aggressive revenue recognition policy, related-party disclosures that were complex and hard to follow, and gains that boosted earnings only once. Insider sales of stock and the departure of senior executives later confirmed the thesis. In a second case, analyst Enitan Adebonojo published reports in June 2001 questioning the accounting of the European food retailer Royal Ahold, including profits of acquired businesses presented as organic growth, capital gains on sale-and-leaseback deals booked as profit, and billions of debt held off the balance sheet. In 2003 the company conceded that it had substantially overstated the profits of the prior two years, the CEO and CFO resigned, regulators opened investigations and the market value fell sharply.

Reporting quality and results quality are different things

Reporting quality is a property of the information: high-quality reporting is decision-useful, meaning it is relevant and it faithfully represents what the business actually did during the period and what its financial position was at the end of it. Results quality, more commonly called earnings quality, is a property of the underlying business: it concerns the earnings and cash the company genuinely generated, and the financial position that followed, judged against expectations of current and future performance. Because the term earnings quality is the one used in practice, it is applied broadly here to cover the quality of earnings, of cash flow and of balance sheet items alike.

High-quality earnings do two things at once. They deliver a return at least equal to the cost of capital, and they come from activities the company is likely to be able to repeat. Earnings with both properties raise the value of a company by more than earnings with neither. Describing reported earnings as high quality therefore carries a double claim: that economic performance was value enhancing, and that the reporting itself was a good reflection of that reality.

Earnings can be low quality for two very different reasons. The reporting may be perfectly sound and simply be telling you that the business performed badly, or the reporting may misrepresent what happened. In principle the two attributes can separate. A company whose only earnings in a period come from a single lawsuit settlement, without which it would have posted a large loss, has low-quality earnings; if it measured that settlement properly and disclosed it clearly, its reporting quality is high. The practical caution is that weak performance is precisely the condition that tempts management to misreport, so the two attributes tend to deteriorate together.

How the two attributes interact

Earnings (results) quality	Low reporting quality	High reporting quality
High	Assessment of earnings quality is impeded, and so is valuation	Assessment is possible; high earnings quality raises company value
Low	Assessment of earnings quality is impeded, and so is valuation	Assessment is possible; low earnings quality reduces company value

High reporting quality is a necessary condition for judging results quality. It is not sufficient to guarantee good results, but without it the judgement cannot be made at all.

The spectrum

Combining the two attributes produces a continuum rather than a pass or fail test. At the top sit reports that comply with the accepted accounting standards of the reporting jurisdiction, are decision-useful, and describe earnings that are sustainable and earn an adequate return. GAAP here is generic: it means International Financial Reporting Standards, US GAAP or any other home-country framework. At the bottom sit reports built on transactions that never happened or that omit transactions that did.

Figure 1: The quality continuum of financial reports

Each step down is a specific failure: earnings that cannot be repeated, then bias inside the rules, then management of reported results, then departure from the rules altogether.

Every position on the continuum can be reached by asking two questions in order. Are the reports compliant with the applicable standards and decision-useful? Are the results of high quality, meaning do they provide an adequate return and can they be sustained? The order matters. A negative answer to the first question makes the second unanswerable.

Figure 2: The two questions, asked in order

A No at the first gate stops the analysis: without faithful reporting there is nothing reliable to evaluate. Only a Yes at both gates places a company at the top of the continuum.

Bias, and the vocabulary for describing it

Biased accounting choices produce reports that do not faithfully represent economic events. Bias operates on measured amounts and equally on presentation: a company can lay out its disclosures so that analysis is easy, or so that unfavourable facts are buried and favourable ones are foregrounded. Either way the damage is the same. Bias impedes the assessment of past performance, corrupts the forecast of future performance, and therefore distorts valuation.

Two labels are used throughout the reading. A choice is aggressive when it raises reported performance and financial position in the current period, at the likely cost of reduced reported performance later. A choice is conservative when it lowers reported performance and position now, possibly raising them later. Neither label is a synonym for wrong; both describe a direction of travel that the analyst must undo before comparing periods or companies.

Earnings management is a further category of bias. The classic form is smoothing, in which reported volatility is held below the volatility that faithful representation would produce. The mechanics are simple: understate earnings when operations are going well and overstate them when operations are struggling. The reported series looks more predictable than the business is.

Example 1 · Worked

Four companies report under IFRS. In each case the analyst has established the facts below.

Company A. Reports comply with IFRS and are clearly presented. Operating returns exceed the cost of capital and come from a long-standing product line.
Company B. Reports comply with IFRS and are clearly presented. The whole of this year’s profit comes from a single settlement of litigation; the trading operations lost money.
Company C. Reports comply with IFRS, but depreciable lives were extended and the allowance for doubtful accounts was cut, both without a change in circumstances.
Company D. A subsequent investigation established that a material part of reported revenue was invoiced to customers that did not exist.

1. Place each company on the quality continuum and state which of the two framework questions each one fails.

Solution. Company A sits at the top: it passes both questions, since the reporting is compliant and decision-useful and the results are sustainable and earn an adequate return. Company B passes the first question and fails the second, so it sits one step down; the reporting is fine, the earnings are not repeatable. Company C also passes the first question in a narrow legal sense, but the choices are biased, so the information is less decision-useful; it sits below Company B. Company D fails the first question outright and sits at the bottom of the continuum, on fictitious transactions.

2. Company C describes its choices as a change in estimate rather than as bias. Why does the distinction not help the analyst?

Solution. Whatever the label, both choices are aggressive: each raises current reported performance and financial position, and each is likely to depress reported performance in later periods when the true economics assert themselves. The analyst still has to reverse the effect before comparing Company C with prior periods or with peers, so the classification of the change is irrelevant to the adjustment that has to be made.

3. Company B is criticised for having low reporting quality. Is that criticism correct?

Solution. No. Reporting quality and results quality are separate attributes. Company B measured a genuine settlement correctly and disclosed it in a way that allows a user to strip it out. That is exactly what high reporting quality looks like. What is low is earnings quality, because the earnings are not derived from activities the company can be expected to repeat.

The choices that create quality problems fall into three families: the amount recognised, the period in which it is recognised, and the line in which it is classified. This section takes the first two; classification follows in the next section. Preparers also have a fourth option, which is to abandon the rules entirely and issue fraudulent reports.

A choice about one element rarely stays there. Financial statements are articulated, so a decision aimed at revenue or at an expense propagates through the system. The cleanest way to trace it is the accounting equation.

Rearranged, this is Assets minus Liabilities equals Equity, that is, Net assets equals Equity.

Any choice that touches the income statement touches equity, and once equity moves some other balance sheet item has to move with it or the statement will not balance. That single observation lets you predict the balance sheet footprint of almost any income statement manipulation.

Direction of the effects

Working through the common cases in the current period gives a checklist that is worth committing to memory.

Aggressive, premature or fictitious revenue recognition. Income is overstated, so equity is overstated. Assets, most often accounts receivable, are overstated as well.
Conservative revenue recognition, such as deferring revenue that has been earned. Net income, equity and assets are all understated.
Omitted or delayed expenses. Expenses are understated and income is overstated, so equity is overstated, together with overstated assets or understated liabilities. Understating bad debt expense leaves accounts receivable overstated. Understating depreciation or amortisation leaves the long-lived asset overstated. Understating interest, taxes or other expenses leaves accrued interest payable, taxes payable or another payable understated.
Understated contingent liabilities. Equity is overstated, because expenses are understated and income or other comprehensive income is overstated.
Overstated financial assets or understated financial liabilities carried at fair value. Equity is overstated through overstated unrealised gains or understated unrealised losses.
Operating cash flow. It can be lifted without any accounting entry at all, by deferring payment of payables, pulling collections forward from customers, deferring inventory purchases, and deferring operating expenditure such as maintenance and research.

The last item is worth pausing on because it is an operating choice rather than an accounting one. Nothing in the reports is wrong, yet the cash flow statement flatters the period. Poor-quality reports can understate as well as overstate results, depending on what management is trying to achieve; fraudulent reports, however, almost always overstate.

Example 2 · Worked

Satyam Computer Services Limited, an Indian information technology company founded in 1987, grew rapidly by supplying business process outsourcing worldwide. Its chief executive, Ramalinga Raju, was named Entrepreneur of the Year by Ernst and Young in 2007, and in 2008 the World Council for Corporate Governance cited the company for global excellence in corporate accountability. In 2009 the same chief executive resigned in a letter that set out a very large fraud. The collapse was rapid enough that the company became known as India’s Enron.

In late 2008 the World Bank ended its relationship with the company after establishing that Satyam had paid kickbacks to bank staff and had billed for services it never performed. That first disclosure drew wider scrutiny. The chief executive later admitted manufacturing bank statements in order to inflate cash and to show interest income that did not exist, and to creating fake salary accounts and taking the money paid to those employees. The head of internal audit created customer accounts and invoices that were equally fictitious in order to inflate revenue. The external auditors did not independently verify much of what the company gave them; even bank confirmations sent to them directly, which showed balances materially different from the reported figures, went unexamined.

1. Where do Satyam’s financial reports sit on the quality continuum?

Solution. At the very bottom. The reports were built on fictitious information, which is the lowest of the five positions on the continuum, below non-compliant accounting and far below biased choices made within the rules.

2. Express each of the four acts of misconduct in terms of the accounting equation: (A) the transactions with the World Bank, (B) the fictitious interest income, (C) the chief executive’s embezzlement, (D) the fictitious revenue.

Solution.
A. Billing for services never provided increases an asset, typically accounts receivable, and increases a revenue account such as service revenues. The kickbacks, if they were recorded at all, increase an expense such as commissions paid and either increase a liability such as commissions payable or reduce cash. Netting the two, income, net assets and equity are all overstated.
B. Interest income that does not exist overstates income; overstates assets such as cash and interest receivable; and overstates equity. The overstatements were concealed by falsifying the revenue and cash balances.
C. Paying wages to employees who do not exist increases an expense account such as wages and salaries and reduces the asset cash. Income and equity are therefore understated, and that understatement is matched by a real, fraudulent fall in cash, again hidden behind falsified revenue and cash balances.
D. Revenue that never happened overstates revenue and income; overstates assets such as cash and accounts receivable; and overstates equity.

3. Which documents had to be falsified to support the misconduct in question 2?

Solution. Invoices to the World Bank for services that were never delivered, bank statements, employee records, and customer accounts and invoices. Every one of them was designed to be shown to the external auditors.

4. What relationship in the published statements might have alerted a careful reader?

Solution. The growth in revenue set against the growth in balance sheet items that revenue creates: short-term and long-term trade receivables and unbilled revenue. The presence of long-term trade receivables and of an unbilled revenue account invites questions, as does a separate account for investments in bank deposits held apart from cash. None of this is proof. A well-constructed fraud is genuinely difficult to identify from the reports alone, which is why the qualitative evidence matters as much as the ratios.

Choices about amount and timing usually spread across several elements, several statements and several periods. Classification choices behave differently. They normally stay inside one statement and concern only where an item is shown, not how large it is or when it is recognised. That containment is exactly what makes them easy to overlook.

Classification within the balance sheet

On the balance sheet the motive is usually to improve a ratio or to conceal a developing problem. Accounts receivable is the favourite target because it carries so much diagnostic weight. A company that wants to hide a liquidity problem or a collection problem can sell the receivables to an outside party, transfer them to an entity it controls, convert them into notes receivable, or simply reclassify them within the balance sheet as long-term receivables. In three of those four routes the amounts are still receivables of one kind or another, but the reported accounts receivable balance falls. To an investor that looks like collection, and it flatters measures such as days of sales outstanding and receivables turnover.

Example 3 · Worked

Merck and Co., Inc. reported inventory of $3,411.8 million for 2002 in its 2002 annual report. In the 2003 annual report the same 2002 figure appeared as $2,964.3 million, with $447.5 million of inventory shown inside other assets, a long-term caption. The detail sat in Note 6.

Merck Note 6: inventories at 31 December ($ millions)

Component	2003	2002
Finished goods	$552.5	$1,262.3
Raw materials and work in process	2,309.8	2,073.8
Supplies	90.5	75.7
Total at approximate current cost	$2,952.8	$3,411.8
Reduction to LIFO cost	Nil	Nil
Total	$2,952.8	$3,411.8
Shown as: inventories	$2,554.7	$2,964.3
Shown as: other assets	398.1	447.5

Inventory measured on a LIFO basis was approximately 51 percent of the total at 31 December 2003 and 39 percent at 31 December 2002. The amounts moved to other assets were stocks held for product launches and not expected to sell within one year. The fall in finished goods is mainly the spin-off of Medco Health in 2003.

1. Moving part of inventory into other assets will most likely cause days of inventory on hand to: (A) fall, (B) be unaltered, (C) rise.

Solution. A is correct. Days of inventory on hand computed from the reported inventory figure will most likely fall, because inventory measured against cost of goods sold has been reduced while cost of goods sold is unchanged.

2. Following the same reclassification, the current ratio will most likely: (A) fall, (B) be unaltered, (C) rise.

Solution. A is correct. Current assets fall by the amount transferred out of inventory, $447.5 million in 2002, while current liabilities are untouched, so the ratio decreases.

3. The explanation in the note is entirely reasonable. Why is the reclassification still an analytical problem?

Solution. The logic is sound: current assets are those expected to be consumed or converted to cash within the operating cycle, usually a year, so stocks earmarked for launches beyond that horizon belong outside current assets. The problem is comparability, not principle. Inventory turnover, cost of sales divided by average inventory, can be computed for 2003 and cannot be computed on the same basis for 2002 or any earlier year, because the amount that would have been reclassified in those years was never disclosed. A time series of Merck inventory turnover therefore mixes two definitions, and the analyst has to say so rather than read a trend into it.

Operating against non-operating, and other comprehensive income

Splitting revenues between operating and non-operating helps a user judge how much of the total is likely to persist. The same split can be turned around. Presenting revenue as arising from core, continuing operations, or presenting an ordinary expense as non-operating, invites the user to treat an inflated amount of income as sustainable. The same trick appears outside the statements, in non-GAAP metrics where income-reducing items are described as non-recurring.

Classification between net income and other comprehensive income has a similar effect on comparability. Two otherwise identical companies that classify their investments differently can report different net income, because the change in value flows through profit for one and through other comprehensive income for the other. Neither company is wrong; a comparison that ignores the difference is.

Classification within the cash flow statement

Management has a clear incentive to maximise the amount of cash shown as operating. Operating activities are the day-to-day business, selling inventory or delivering services. For most companies the disposal of property or other long-lived assets is not part of that, and pushing such proceeds into the operating section overstates the ability of the business to generate cash from what it actually does. A second route is capitalising expenditure that should have been expensed, which moves the outflow from operating into investing without touching the total.

A working list of warning signs

The table below groups the recurring issues, the actions a company can take, and the signals those actions tend to leave behind. The signals may appear in the statements, in the notes, or only in ratios the analyst computes and then tracks over time or against peers. Bias most often pushes net income upward, but a newly appointed management team, or a management team running a company in difficulty, may prefer to depress current income so that later periods look better.

Issues, actions and the traces they leave

Issue	Actions available	Signals to watch
Operating income or net income overstated, or not sustainable, through accelerated revenue, understated expenses or misclassified gains and losses	Contingent sales carrying a right of return; channel stuffing, which induces customers to order goods they would not otherwise order, or to order them earlier, through generous terms; bill and hold sales, where customers order goods that stay on the seller premises; outright fictitious revenue; capitalising expenditure as an asset; showing non-operating income or gains inside operations; describing ordinary expenses as non-recurring or non-operating; routing gains through net income while routing losses through other comprehensive income	Revenue growing faster than the industry or the peer group; rising discounts and rising customer returns; receivables growing faster than revenue; a large share of annual revenue falling in the final quarter of a business that is not seasonal; operating cash flow far below operating income; the contents of operating revenue and operating expense changing from year to year; operating margin rising; aggressive assumptions such as long depreciable lives; losses appearing in non-operating income or other comprehensive income while gains appear in operating income or net income; compensation tied largely to reported financial results
Balance sheet items misstated, which may in turn affect the income statement: assets or liabilities over- or understated, or misclassified	Choosing the fair value model and the inputs that feed it; moving items from current to non-current; over- or understating reserves and allowances; understating identifiable assets so that more of an acquisition price falls into goodwill	Models and inputs that bias the fair value measurement; inputs applied inconsistently to assets and to liabilities; assets that are normally current, such as receivables and inventory, sitting in non-current captions; allowances and reserves that move about over time or that do not resemble those of peers; goodwill large relative to total assets; special purpose vehicles in use; large movements in deferred tax assets and liabilities; significant obligations kept off the balance sheet
Operating cash flow overstated	Managing the timing of operating flows; classifying flows so that the operating total benefits	Accounts payable rising while receivables and inventory fall; capitalised expenditure appearing in investing activities; sale and leaseback transactions; growing bank overdrafts

Business combinations concentrate several quality problems in one transaction. The acquisition method requires one party to be identified as the acquirer and the results to be presented on a consolidated basis, and both of those features can be put to work.

Acquisitions as a cash flow disguise

A company whose own ability to generate cash is fading has an obvious reason to buy one that can. If the purchase is paid in cash it appears in investing activities; if it is paid in equity it may not appear on the cash flow statement at all. Either way, consolidated operating cash flow from that point includes the cash generated by the acquired business, which masks the deterioration at the acquirer. The boost is one-off and may not be repeatable. No standard requires a company to disclose what its cash flow would have been with and without acquisitions, so an investor cannot reliably establish whether the underlying problem is getting worse.

The incentive runs in both directions and in both directions of time. Managers at an acquirer paying in stock have a reason to inflate reported earnings before the deal, because that raises the value of the shares being handed over, and there is evidence that this happens. Managers at a target have the mirror incentive, to lift earnings and secure a better price. After a deal closes, acquiring managers may push earnings upward again to make the acquisition look like a success.

Misreporting can also be the reason for a deal rather than a by-product of it. Acquisitions complicate a set of financial statements and can bury earlier misstatements inside the complication. Companies later accused of accounting fraud by the US Securities and Exchange Commission have been found more likely than others to make acquisitions, and more likely to buy targets that reduce the comparability and consistency of the statements, for example targets with little public information and operations unlike their own.

Goodwill and the incentive it creates

At the acquisition date the acquirer must measure and recognise the identifiable assets acquired and liabilities assumed at fair value. That can bring on to the balance sheet items the target never recognised, such as intangibles it developed internally and certain contingent liabilities. Whatever the purchase price exceeds the recognised value of those identified net assets becomes goodwill.

Goodwill is not amortised. It is tested for impairment, and in the absence of a charge it sits on the balance sheet indefinitely. That default treatment, which carries no future amortisation expense, gives an acquirer a reason to understate the value of amortisable intangibles. Goodwill is a residual, so anything not assigned to an identifiable intangible lands there and escapes the income statement. Understating the fair value of acquired assets has the same double effect: it avoids future charges to expense and it raises goodwill.

The cost of the bias is that an uneconomic acquisition is not recognised as such until an impairment charge arrives, which may be years later. Managements may be willing to accept that risk because they expect to persuade analysts and investors that an impairment is a non-recurring, non-cash charge that can be looked through, and many will indeed look through it. The presence of significant goodwill should therefore make an investor more curious, not less: curious about the company record on recognising impairments, and about the process it uses to test goodwill. Outside assets and liabilities with quoted prices in active markets for identical items, fair value measurement leaves management with judgement, and judgement moves reported values. It is reasonable to ask whether the goodwill on a balance sheet reflects economic reality at all.

Example 4 · Worked

An SEC enforcement action concerned the financial statements of Digilog, Inc. To develop and launch a new product, Digilog set up a separate entity, DBS, capitalised with $10 million of convertible debt issued to Digilog. On conversion Digilog would own close to 100 percent of DBS. The initial owners equity of DBS consisted of a few thousand dollars of common stock issued to the manager of DBS.

For the first two years Digilog did not consolidate DBS, arguing that the manager controlled it by virtue of owning all the outstanding common shares. DBS ran substantial losses in those two years, while Digilog recorded interest income on its investment in the convertible debt. Once DBS turned profitable, Digilog exercised the conversion option and consolidated from that point on. The SEC took the view that the contractual and operating relationships made the two a single enterprise for reporting purposes, and Digilog auditors settled. The settlement recorded the view that consolidation would have produced the most meaningful presentation. Years later, and after Enron, the concept of a variable interest entity emerged: consolidation becomes necessary even where voting control is absent, provided the investor is able to influence how the entity is financed and operated and either bears, or is entitled to, returns from it that vary. IFRS does not use the term but its provisions are similar.

1. Under the variable interest entity rules, Digilog would most likely argue against consolidating DBS on the ground that: (A) Digilog does not have voting control, (B) the interest income Digilog receives from DBS is not variable, (C) the manager of DBS holds operational and financial control.

Solution. C is correct. Digilog would assert that operational and financial control rests with the manager of DBS. The assertion is unlikely to survive: the manager invested a few thousand dollars against $10 million from Digilog. The absence of voting control, answer A, is not by itself enough to avoid consolidation. Answer B fails because Digilog is exposed to variable returns through the possible losses and through the conversion option.

Impairments and restructuring charges

Two items almost always need separate thought. An asset impairment is a write-down required when circumstances show that the carrying amount exceeds the benefits the asset is expected to deliver. A restructuring charge, in IFRS terminology, covers the sale or termination of a line of business, closure of locations, changes to the management structure, or a fundamental reorganisation. Each of those events can also create a liability, such as a commitment to pay severance or to settle a lease.

Example 5 · Worked

On 25 April 2013 Fuji Electric Co., Ltd, a Japanese company reporting under its home-country standards, announced an impairment loss of ¥6.5 billion on land, buildings, structures and leased assets used in its solar cell and module business. The entire loss was recorded in the 2012 fiscal year, which ended on 31 March. Assets and net income were each reduced by ¥6.5 billion.

Elan Corporation, plc, a biotechnology company based in Ireland, reported US$42.4 million of restructuring and other costs in fiscal 2012 in connection with its decision to close a research facility in San Francisco, with the loss of about 200 jobs, and to move much of its activity back to Ireland as business conditions changed. Part of the cost related to current and deferred severance obligations.

1. Recognising an impairment loss and restructuring charges in a single period, although consistent with most accounting frameworks, is most likely to overstate: (A) net income of prior periods, (B) net income of the current period, (C) net income of future periods.

Solution. A is correct. Both the impairment and the restructuring almost certainly arose from activities of earlier years, so those earlier years carried expenses that were never charged and their reported net income was overstated. Current net income, unless the charge is expected to repeat, is understated rather than overstated. Future net income could be overstated if reversals occur, but reversals are not likely. Taking the whole charge now is an example of conservative accounting.

2. How should the analyst treat the two charges in a forecast?

Solution. It depends on whether events of this kind happen often enough to be part of permanent earnings. The decline in the solar business at Fuji Electric evidently developed over more than the single year in which the loss was booked, and the Elan restructuring plausibly relates to earlier periods too. Where such events recur, the analyst should normalise by spreading the charge across past periods as well as the current one. Where an item is genuinely one-off, the financial effect of a natural disaster being the clearest case, normalising means excluding it. The choice requires judgement and is only as good as the analyst knowledge of the underlying facts.

Three related items should trigger the same question, which is whether an earlier period was overstated. A revision to an ongoing estimate, such as the remaining economic life of an asset, invites the question whether the change should have been made sooner. A sudden increase in an allowance or a reserve invites the question whether the earlier estimates were flattering rather than unbiased. A large accrual for losses, environmental or litigation-related, suggests that earlier periods were overstated because the loss was not accrued when it arose. Reserves and allowances are also the standard instruments for smoothing.

What the rules leave out

Some divergence from economic reality is not the fault of the preparer at all. Accounting standards do not permit research and development expenditure to be capitalised, yet research and development creates assets that produce future benefits. The prohibition exists because it is genuinely hard to say in advance which spending will produce a benefit and which will produce nothing, but the consequence is that a real economic asset never appears. Standards also route some value changes away from net income: marketable securities classified as available for sale have their fair value changes reported in other comprehensive income, while the same changes on securities classified as trading run through net income.

No basis of accounting recognises every economic asset and liability, so part of the work is deciding what is missing. A frequent example is the sales order backlog. Under most frameworks revenue is not recognised, and no asset is created, until performance has occurred and the other criteria are met. In large-scale manufacturing, aircraft assembly being the standard case, the backlog can be a substantial unrecognised asset. When it is significant it is normally discussed in the management commentary, which gives the analyst something to work with for adjustments and forecasts.

The mirror problem is deciding whether an item shown in other comprehensive income belongs in the analyst view of net income. The usual candidates are unrealised holding gains and losses on certain equity investments; unrealised holding gains, and subsequent losses, on property and equipment where the revaluation option is elected, which is available under IFRS only; the owners equity effects of translating the financial statements of a foreign operation into the reporting currency of the group; certain changes in the net pension liability or asset; and gains and losses on derivatives, and on certain non-derivative instruments denominated in a foreign currency, that are accounted for as hedges of future cash flows. Where the analyst concludes that a significant item belongs in net income, both the reported and the forecast amounts should be adjusted.

Why an impairment is not simply a non-cash item to be ignored

The claim that an impairment is non-cash is true and beside the point. Cash left the business when the asset was bought; the impairment is the admission that the cash did not buy what was expected. Treating the charge as noise therefore does two things at once: it accepts the earlier capital spending as sound, and it accepts the earlier years of earnings as correct even though those years were relieved of a cost that was accruing. If write-downs recur at a company, the right response is not to exclude them but to spread them, because the pattern is telling you that the depreciation and amortisation charged in normal years is too low.

Before any analysis begins, the purpose has to be settled. What question is the work supposed to answer? How much detail does that question require? What data exist? Which factors and relationships will drive the answer? What are the analytical limitations, and do they threaten the conclusion? Those five questions frame every financial analysis, and in the specific context of report quality they narrow to the two framework questions set out at the start of this lesson: whether the reports are compliant and decision-useful, and whether the results provide an adequate return and can be sustained.

What follows is a general guide rather than a procedure to be executed in order. Companies differ, and a particular project may need steps added, reordered, emphasised or dropped.

The seven steps

Understand the company and its industry. Knowing what the business actually does explains why particular accounting principles are appropriate and why particular metrics matter. Knowing the principles used by the company and by its competitors establishes what the norm is, which is the only way to judge whether a treatment is out of line.
Learn about management. Establish whether there are specific incentives to misreport. Read the disclosures on compensation and on insider transactions, paying particular attention to insider sales of the shares, and read the related-party disclosures.
Identify the significant accounting areas, especially those where management judgement, or an unusual rule, drives a large part of reported performance.
Make comparisons, in three directions. Compare this year statements and significant disclosures with last year: are there major differences in line items or in key disclosures such as risk, segments, or the classification of specific expense and revenue items, and are the reasons for any change apparent? Compare the accounting policies with those of the closest competitors, and where they differ, work out the direction of the effect. Then compare performance with the closest competitors using ratio analysis.
Check the standing warning signs. Falling receivables turnover can mean revenue is fictitious or premature, or that the allowance for doubtful accounts is too small. Falling inventory turnover can mean obsolescence that has not been recognised. Net income above cash provided by operations can mean that aggressive accrual policies have pushed current expenses into later periods.
Look at the segments. For companies operating across several geographies or product lines, and particularly for multinationals, consider whether inventory, sales and expenses have been shifted so that the company appears well positioned in a region or a segment the investment community regards as attractive. A segment performing strongly while the consolidated result stands still or deteriorates is the signature to look for.
Apply quantitative tools to assess how likely misreporting is.

The first six steps are qualitative. They rely on reading, comparing and asking, and they are where most of the diagnostic value lies. The seventh is different in kind, and it is the subject of the next section.

Example 6 · Worked

An analyst is starting work on a manufacturer that has reported five consecutive years of rising operating margin in an industry where margins have been flat. The analyst has the last two annual reports, the peer group accounts, and the compensation disclosures.

1. Which of the seven steps addresses the margin observation most directly, and what should the analyst do with it?

Solution. Step 5. Rising operating margin is itself one of the standing warning signs, and it is more compelling because it stands against the industry pattern rather than with it. The step tells the analyst to treat it as a question, not a conclusion. The specific follow-ups are the ones that would produce a rising margin without a real improvement: expenses capitalised rather than charged, ordinary expenses relabelled as non-operating or non-recurring, and revenue recognised earlier than it should be. Step 4 then provides the test, by comparing the accounting policies and the ratios directly with the closest competitors.

2. The compensation note shows that most of the incentive pay of senior management is based on reported operating income. How does that change the work?

Solution. It does not prove anything, but it belongs in step 2 and it raises the prior probability that the margin trend is reported rather than real. Compensation tied largely to financial results is one of the listed warning signs. The practical effect is to raise the priority of the step 3 review of the significant accounting areas, particularly those where judgement decides how much cost reaches the income statement, and to make the insider transaction disclosures worth reading closely.

Step 7 asks how likely it is that a company is misreporting. Where the answer is that it is quite likely, the analyst has to take special care with everything that follows, including the qualitative work.

Messod Beneish and co-authors built a probit model to identify quantitative indicators of earnings manipulation and to convert them into a single probability. The output is the M-score.

The M-score gives the probability of earnings manipulation.

Each of the eight variables is an index, generally this year divided by last year, and each carries an intuition about why manipulation would move it.

The eight Beneish variables

Variable	Definition	Why it belongs
DSR, days sales receivable index	(Receivables in t divided by Sales in t) divided by (Receivables in t minus 1 divided by Sales in t minus 1)	A shift in the relationship between receivables and sales can point to revenue recognised inappropriately
GMI, gross margin index	Gross margin in t minus 1 divided by gross margin in t	Margins that are deteriorating can predispose a company to manipulate
AQI, asset quality index	The ratio, this year over last year, of one minus (PPE plus current assets) divided by total assets	A change in the share of assets that is neither PPE nor current can point to expenditure being capitalised too freely
SGI, sales growth index	Sales in t divided by sales in t minus 1	Managing the appearance of continued growth, and the capital needs that real growth creates, can predispose a company to manipulate sales and earnings
DEPI, depreciation index	Depreciation rate in t minus 1 divided by depreciation rate in t, where the rate is depreciation divided by (depreciation plus PPE)	A falling depreciation rate can point to understated depreciation used to lift earnings
SGAI, selling, general and administrative expense index	(SGA in t divided by sales in t) divided by (SGA in t minus 1 divided by sales in t minus 1)	Rising fixed SGA implies falling administrative and marketing efficiency, which can predispose a company to manipulate
Accruals	(Income before extraordinary items minus cash from operations) divided by total assets	Higher accruals can indicate manipulation
LEVI, leverage index	Leverage in t divided by leverage in t minus 1, leverage being debt to assets	Rising leverage can predispose a company to manipulate

In the empirical results of Beneish (1999) the statistically significant variables were the days sales receivable index, the gross margin index, the asset quality index, the sales growth index and accruals.

Turning the score into a probability

The M-score is a normally distributed variable with a mean of zero and a standard deviation of 1.0, so the probability of manipulation is read from the cumulative standard normal distribution, taking the area in the left tail. An M-score of −1.49 corresponds to a probability of 6.8 percent and an M-score of −1.78 to a probability of 3.8 percent. Higher scores, meaning less negative ones, mean a higher probability.

Where the cut-off is placed depends on the relative cost of the two errors: a Type I error classifies a manipulator as a non-manipulator, and a Type II error classifies a non-manipulator as a manipulator. The cut-off is chosen to minimise the cost of misclassification. Beneish took the relevant cut-off for investors to be a probability of manipulation of 3.8 percent, which is an M-score above −1.78.

Example 7 · Worked

The table sets out the Beneish variables and the resulting M-score for XYZ Corporation, a hypothetical company.

XYZ Corporation M-score

Variable	Value	Beneish coefficient	Product
DSR	1.300	0.920	1.196
GMI	1.100	0.528	0.581
AQI	0.800	0.404	0.323
SGI	1.100	0.892	0.981
DEPI	1.100	0.115	0.127
SGAI	0.600	−0.172	−0.103
Accruals	0.150	4.679	0.702
LEVI	0.600	−0.327	−0.196
Intercept			−4.840
M-score			−1.231
Probability of manipulation			10.93%

1. Using −1.78 as the cut-off, would the model lead an analyst to flag XYZ as a likely manipulator?

Solution. Yes. Each product is the variable value multiplied by its coefficient: 1.300 × 0.920 = 1.196, 1.100 × 0.528 = 0.581, 0.800 × 0.404 = 0.323, 1.100 × 0.892 = 0.981, 1.100 × 0.115 = 0.127, 0.600 × (−0.172) = −0.103, 0.150 × 4.679 = 0.702 and 0.600 × (−0.327) = −0.196. Adding the eight products to the intercept of −4.840 gives approximately −1.23; the exhibit reports an M-score of −1.231 and a probability of manipulation of 10.93 percent. The score exceeds the cut-off of −1.78, so the probability is higher than acceptable. Classification does depend on the relative cost of Type I and Type II errors, but 10.93 percent is far above the 3.8 percent that Beneish identified as the relevant cut-off.

2. DSR, GMI, SGI and DEPI are all above one. What does each of those tell the analyst?

Solution.
DSR above one means receivables have risen as a percentage of sales. That can indicate revenue recognised inappropriately: XYZ may have shipped goods early and taken revenue that belongs to later periods. The alternative reading is that customers are becoming less able to pay, which is still a problem for the analyst.
GMI above one means gross margins were higher last year than this year. Deteriorating margins can predispose a company to manipulate earnings.
SGI above one means sales grew relative to the prior year. Growth creates both the wish to sustain the perception of growth and the need for capital to fund it, and either can predispose a company to manipulate.
DEPI above one means the depreciation rate was higher in the prior year, so it has fallen. A falling depreciation rate can indicate manipulated earnings.

Other models, and what none of them can do

Researchers have tested many other inputs for detecting misstatement. Those found useful include accruals quality, deferred taxes, a change of auditor, market-to-book value, whether the company is publicly listed and traded, differences between the growth rates of financial and non-financial variables such as the number of patents, employees and products, and features of corporate governance and incentive compensation.

Two limitations apply to all of them. The first is structural: accounting is a partial representation of economic reality, so a model built on accounting numbers can only establish association. Cause and effect require something the model cannot supply, whether interviews, surveys or an investigation by a regulator with enforcement powers. The second is adaptive. Manipulators know as much about the models as analysts do. The 2013 study by Beneish and co-authors found that the predictive power of the model has been declining over time, which is what would be expected if managers were using the model itself to test how detectable a given tactic would be. The screens remain useful, and the search for better ones continues, but the qualitative work cannot be delegated to them.

High earnings quality is normally evidenced by earnings that are sustainable and that represent a return equal to or above the cost of capital. Low earnings quality is the opposite: earnings insufficient to cover the cost of capital, or earnings derived from one-off activity, or reported information that simply does not tell the user anything useful about performance. Note that the phrase high-quality earnings already assumes high reporting quality; the two cannot be separated in a valuation.

Four families of indicator are used in practice: recurring earnings, earnings persistence and the related measures of accruals, the beating of benchmarks, and after-the-fact confirmations such as enforcement actions and restatements. This section takes the first; the next takes the rest.

Recurring earnings

When current and prior earnings are used to forecast future earnings, as they are in any earnings-based valuation, the analyst wants the part that will recur. Earnings from subsidiaries already selected for disposal, which have to be shown separately as discontinued operations, are normally excluded outright. Many other items can be non-recurring: a one-off asset sale, a one-off settlement of litigation, a one-off tax settlement. The higher the proportion of such items inside reported earnings, the less sustainable those earnings are and the lower their quality.

Example 8 · Worked

Enron Corp. was an energy distribution company. The excerpt below comes from its consolidated income statement for the years ended 31 December.

Enron and subsidiaries, income statement excerpt ($ millions, except per share amounts)

Line	2000	1999	1998
Revenues in total	$100,789	$40,112	$31,260
Costs and expenses in total	98,836	39,310	29,882
Operating income	$1,953	$802	$1,378
Equity affiliates outside the consolidation, share of their earnings	$87	$309	$97
Disposals of non-merchant assets, gains arising	146	541	56
TNPC, Inc., gain arising when it issued stock	121	0	0
Income from interest	212	162	88
Net figure for other income	−37	181	−37
Income before interest, minority interests and income taxes	$2,482	$1,995	$1,582

1. Compare the trend in operating income with the trend in income before interest, minority interests and income taxes.

Solution. Operating income moves violently. It falls from $1,378 million in 1998 to $802 million in 1999, then more than doubles to $1,953 million in 2000. Income before interest, minority interests and income taxes does the opposite: it rises smoothly and substantially in each year, by 24 percent in 2000 relative to 1999 and by 26 percent in 1999 relative to 1998.

2. Which items look non-recurring rather than the product of routine operations, and how large are they?

Solution. Two items stand out: the gains arising on disposals of non-merchant assets, and the gain recorded when TNPC issued stock. Gains from sales of non-merchant assets do appear in every year, but the activity is not part of energy distribution. Two further items outside operations swing about violently as well: the share of earnings taken from affiliates outside the consolidation, and the net figure for other income. Their significance is twofold. The smooth upward path of Enron income is produced directly by these items, and they are large: in 1999 they represent 52 percent of income before interest, minority interests and income taxes, since ($309 + $541 + $181) divided by $1,995 is $1,031 divided by $1,995.

3. James Chanos told the US Congress that a number of one-time gains that boosted the earnings of Enron were among the items that strengthened his conviction that the market was mispricing the stock. What does that tell you about how the reported figures were being used?

Solution. It suggests that at least part of the market was feeding reported income straight into an earnings-based valuation without stripping out the non-recurring items first. Doing so capitalises gains that will not repeat, and the resulting value is too high by a multiple of the error.

Classification shifting

Deciding that an item is non-recurring is a judgement, and judgements can be steered. That creates a second, subtler problem: the amount a reader of the income statement identifies as repeatable can be inflated without touching net income at all. Investors label the earnings expected from the ordinary business as recurring or core, and in the absence of special items such as restructuring charges, employee separation costs, goodwill impairments or gains on disposals, operating income is a fair proxy for them. Classification shifting exploits that. Reclassifying a normal expense into a special item, or moving operating costs into income-decreasing discontinued operations, leaves net income unchanged and raises reported core earnings.

The evidence is anecdotal and, worse for the analyst, it arrives late.

Cases in which expenses were shifted out of operations

Company	What was found
Borden, food and chemicals	The SEC concluded that $146 million of operating expenses had been shown as part of a special item, restructuring charges, when they belonged in selling, general and administrative expenses
AmeriServe Food Distribution Inc.	The company went bankrupt four months after completing a $200 million junk bond issue. An examiner appointed by the bankruptcy court found that substantial operating expenses had been presented as restructuring charges, which concealed serious underperformance and delayed the point at which every party recognised how severe the problems were
Waste Management	In 1998 the company issued what was then the largest restatement in SEC history. The enforcement documentation shows that operating income had been improperly inflated by netting non-operating gains, from the sale of investments and from discontinued operations, against unrelated operating expenses
IBM	Revised disclosures, prompted by SEC scrutiny and by requests from analysts, showed that income from intellectual property had been shown as an offset to selling, general and administrative expenses. Operating expenses were therefore understated and core earnings overstated, by $1.5 billion in 2001 and $1.7 billion in 2000

Because the evidence emerges only after the fact, it is of limited use for anticipating problems. What the analyst can do is treat income-decreasing special items with particular suspicion, and treat them with more suspicion still when the company is simultaneously reporting unusually strong operating earnings for the period, or when the classification of the item is what allowed the company to meet or beat a forecast of operating earnings.

Pro forma and non-GAAP measures

Companies know that investors separate recurring from non-recurring items, so many of them volunteer help. Alongside the components of income on the face of the statement, a company may disclose a pro forma or adjusted figure, described as a non-GAAP measure or a non-IFRS measure, from which non-recurring items have been removed. Such disclosures must be accompanied by a reconciliation to the reported figure.

The reconciliation requirement does not settle the question, because the classification behind it is still a judgement and the person making it has an interest in the answer. Groupon, an online discount provider, included in its original initial public offering filing a pro forma measure of operating income that excluded online marketing costs. The SEC found the measure misleading and required the company to remove it. Voluntary adjustments can be genuinely informative, but the analyst has to check, item by item, that what has been excluded is really non-recurring.

Persistence is the second property of high earnings quality: the sustainability of earnings once the obviously non-recurring items are removed, together with the persistence of growth in them. The assumption behind every earnings-based valuation model is that more persistent earnings are better inputs. Persistence can be measured as the coefficient on current earnings in a simple regression.

A higher coefficient on current earnings means more persistent earnings.

Splitting earnings into cash and accruals

Earnings can be decomposed into a cash component and an accruals component. The accruals component exists because the rules put revenue in the period it is earned and expenses in the period they are incurred, rather than in the period cash moves. A sale on account produces accounting income when the sale is made; if the cash arrives later, the gap between reported net income and cash collected is an accrual.

Research has established that the cash component is the more persistent of the two. Running the regression with both components separately, the coefficient on cash flow is higher than the coefficient on accruals.

The coefficient on cash flow exceeds the coefficient on accruals.

That single result is why so many quality indicators reduce to measuring the size of the accruals component. Earnings carrying a larger accruals component are less persistent, and therefore of lower quality.

A further distinction matters more than the split itself. Accruals arising from the normal transactions of the period are non-discretionary. Accruals arising from transactions or accounting choices outside the normal, possibly made in order to distort the reported result, are discretionary. Outlier discretionary accruals are the indicator of possibly manipulated, and therefore low-quality, earnings.

Identifying them takes two stages. First model what normal accruals should be, as a function of the economic factors that generate them: growth in credit sales, which should produce growth in accounts receivable, and the amount of depreciable assets, which should produce depreciation. Then regress total accruals on those factors and treat the residual as the proxy for abnormal accruals. The approach began in academic work and was adopted in practice. The SEC has described its own Accounting Quality Model as an extension of the traditional approach, which was usually based on the Jones Model or the Modified Jones Model, in that discretionary accrual factors form part of the estimation itself: total accruals across all registrants are estimated as a function of a large set of proxies for the discretionary and non-discretionary components, discretionary accruals are computed from the estimates, and the result is used to screen the firms that appear to be managing earnings most aggressively.

A cruder version is available to anyone. Compare the magnitude of total accruals across companies, scaling them so the comparison means something, for example by average assets or by average net operating income. High accruals on that basis are an indicator of possibly manipulated and thus low-quality earnings. The most dramatic version of the signal is a company reporting positive net income alongside negative operating cash flow.

Example 9 · Worked

Allou Health and Beauty Care, Inc. manufactured and distributed hair and skin care products. After the periods below, its warehouses were destroyed by a fire for which management was found responsible, and the company was later shown to have fraudulently inflated its sales and inventories in those years.

Allou Health and Beauty Care, annual 10-K data, years ended 31 March

Line	2002	2001	2000
Revenues, net	$564,151,260	$548,146,953	$421,046,773
Costs of revenue	500,890,588	482,590,356	367,963,675
Gross profit	$63,260,672	$65,556,597	$53,083,098
Income from operations	27,276,779	28,490,063	22,256,558
Income from continuing operations	$6,589,658	$2,458,367	$7,043,548
Cash flow: net income from continuing operations	$6,589,658	$2,458,367	$7,043,548
Cash flow: change in accounts receivable	(24,076,150)	(9,725,776)	(25,691,508)
Cash flow: change in inventories	(9,074,118)	(12,644,519)	(40,834,355)
Net cash used in operating activities	$(17,397,230)	$(34,195,838)	$(27,137,652)

The gap between income from operations and income from continuing operations reflects interest expense and the income tax provision in each year, and in 2001 also a $5,642,678 loss on the impairment of investments.

1. Judged on the income statement alone, how did Allou perform?

Solution. The picture is unremarkable and mildly reassuring. Revenue grew in each of the three years, though more slowly in the latest. Gross margin drifted down but stayed fairly stable: $63,260,672 divided by $564,151,260 is 11.2 percent in 2002, against 12.0 percent in 2001 and 12.6 percent in 2000. Operating margin behaved the same way, holding at around 5 percent: $27,276,779 divided by $564,151,260 is 4.8 percent in 2002, against 5.2 percent in 2001 and 5.3 percent in 2000. Income from continuing operations was sharply lower in 2001 because of the impairment loss, but the company reported positive net income in every year. On the income statement alone, performance looks reasonably stable.

2. Now compare income from continuing operations with cash from operating activities.

Solution. The comparison destroys the impression. Allou reported positive income from continuing operations and negative operating cash flow in all three years, of $(17,397,230) in 2002, $(34,195,838) in 2001 and $(27,137,652) in 2000. Persistently negative operating cash flow cannot be sustained by a going concern.

3. What do the reconciling items tell you?

Solution. Accounts receivable and inventories rose in every year, and between them those two increases account for most of the gap between income from continuing operations and net cash used in operations. In 2000 alone the increases were $25,691,508 in receivables and $40,834,355 in inventories against income from continuing operations of $7,043,548. The company appears to be accumulating stock it cannot sell and failing to collect what it has billed. The mechanics are worth stating: under the indirect method an increase in a current asset is deducted from net income in arriving at operating cash flow.

Why the accrual signal cannot be applied mechanically

Enron shows the same profile as Allou at the quarterly level and the opposite profile annually.

Enron: reported income against operating cash flow ($ millions)

Period	Net income	Operating cash flow
Three months to 31 March 2001	425	(464)
Three months to 31 March 2000	338	(457)
Year to 31 December 2000	979	4,779
Year to 31 December 1999	893	1,228
Year to 31 December 1998	703	1,640

Quarterly figures are from the 10-Q filings and annual figures from the 10-K filings.

The quarterly data show positive net income with negative operating cash flow in quarters later shown to have been misreported. The annual data show the reverse: operating cash flow exceeded net income in all three years of the fraud. In 1998 it was more than double net income, at $1,640 million against $703 million; in 1999 it was about 38 percent higher, at $1,228 million against $893 million; and in 2000 it was almost five times net income, at $4,779 million against $979 million. Some of the fraudulent transactions were designed precisely to generate operating cash flow. The right response to any of these patterns is to ask why the difference exists, in either direction, because the capacity to generate cash from operations ultimately governs what the company can invest and how it can finance itself.

WorldCom Inc.: reported income against operating cash flow ($ millions)

Year to 31 December	1999	2000	2001
Net income (loss)	$4,013	$4,153	$1,501
Net cash provided by operating activities	11,005	7,666	7,994

WorldCom makes the point even more sharply. Cash from operating activities exceeded net income in each of the three years, which on the usual reading suggests high-quality earnings, and yet the reports were fraudulent. The central misstatement was the improper capitalisation of costs that should have been expensed. Because capital expenditure appears as an investing outflow rather than an operating one, the fraud inflated operating cash flow rather than depressing it. An analyst who compared cash flow with net income and did not look at investing activities would have drawn precisely the wrong conclusion. Accrual measures are indicators, not tests, and net income itself rests on subjective estimates such as the expected life of long-term assets, which are easy to move.

Mean reversion

Forecasting earnings is the analyst core task, and forecasts are more accurate and more credible when the earnings stream persists. Persistence comes from the cash flow element. The accruals element adds information about performance, but it also detracts from stability, because of the estimation involved in producing it.

Empirical work confirms what intuition suggests: nothing lasts. Extreme levels of earnings, high and low alike, revert to normal over time. This is a natural property of competitive markets rather than an accounting artefact. A company earning poorly will close or shrink the losing operations and replace weak managers with better ones, and earnings improve. A company earning abnormally well attracts competitors, unless the barriers to entry are insurmountable, and new entrants cut prices to establish a position, which erodes the incumbent profits. In either direction the expectation should be a return to the mean.

Figure 3: Extreme earnings revert toward the mean

Portfolios sorted on an extreme value of a return measure move toward the centre. Neither the strong nor the weak group holds its starting position, so neither can be extrapolated.

Nissim and Penman (2001) demonstrated that the principle holds across a wide range of accounting measures. Their sample covered listings on the New York Stock Exchange and on the American Stock Exchange across the years 1963 to 1999, and the measures they followed included income after a charge for capital, the operating version of that same residual measure, the return earned on common equity, the return earned on net operating assets, the rate at which common equity grew, and profit margins on core sales. Starting from 1964 data they sorted companies into 10 equal portfolios on their ranking for a given measure and followed the median value in each portfolio through each of the next five-year periods, re-sorting the portfolios at the end of every fifth year. Extending the process through 1994 produced means of portfolio medians across seven rankings, and the findings were consistent across the metrics.

Net operating assets are the assets used in operations, net of the liabilities that operations generate.

Return on net operating assets makes the point concretely. The observed range of RNOA ran from 35 percent down to −5 percent at the start of the observations and had compressed to a range of 22 percent down to 7 percent by the end of the study. The portfolios that were not outliers in either direction in Year 1 behaved differently: they stayed constant, or nearly so, over the whole observation period. Reversion is a property of the extremes.

How accruals affect the speed of reversion

The lesson for forecasting is that neither very high nor very low earnings can simply be extrapolated. Useful forecasts have to work from normalised earnings over the relevant valuation horizon. Since earnings are the sum of cash flows and accruals, and since the cash component is the persistent one, earnings dominated by cash revert more slowly. A significant accruals component hastens the reversion, and it hastens it further when the accrual elements are outliers relative to the normal amount of accruals in that company earnings. Building a forecast therefore requires a realistic cash flow model and realistic estimates of accruals, not one or the other.

Figure 4: Accruals accelerate the return to the mean

Both series start at the same extreme. The accrual-heavy series falls back toward the mean much sooner, so a forecast that extrapolates it is wrong faster and by more.

Beating benchmarks, and after-the-fact indicators

Announcing earnings that meet or beat a benchmark such as the consensus of analyst forecasts usually lifts the share price. That does not make it a mark of quality. Exactly meeting a benchmark, or beating it by a hair, has been proposed as an indicator of manipulation and therefore of low-quality earnings, and academic work has documented a statistically large cluster of outcomes just above zero when actual results are compared with benchmarks, which some read as evidence of earnings management. Others dispute that reading. What is not disputed is that a company which consistently lands exactly on, or barely above, its benchmark invites questions about its earnings quality.

The last family of indicators is external and confirmatory: enforcement actions by regulators and restatements of previously issued statements. Both are unambiguous, and both arrive too late to be worth much. The value of recognising poor earnings quality lies in doing it before the deficiency is widely known, so external indicators rank low in usefulness. They still deserve attention, because when one appears the correct response is to reopen the decisions that were based on the earlier reports rather than to defend them.

The purpose of analysing earnings is to understand how persistent and how sustainable they are. If the reported figures do not describe the reality the company faces, any forecast built on them inherits the flaw. Reporting is full of choices and estimates, and where there are choices and estimates there is temptation. Companies that look like extraordinary performers frequently turn out to be ordinary or worse once a regulator establishes what the accounting choices were.

Studying past cases is the cheapest way to learn the patterns. Reviewing 227 of its own enforcement cases from the years 1997 to 2002, the SEC reported that revenue recognition was the area in which accounting misrepresentation arose most often. That is not surprising. Revenue is the largest single number on the income statement and arguably the most important; its size, its leverage on earnings and the discretion available in the recognition policy make it the account most likely to be deliberately misstated. Analysts tend to concentrate on the quantitative side of revenue, asking how fast it grew and whether the growth was organic or acquired, and much less often on how the revenue was generated. That second question, whether revenue came from discounting or from bill-and-hold arrangements, is where the quality lies.

What Sunbeam did

Sunbeam Corporation made household appliances and outdoor products. In the middle and late 1990s its new chief executive, Albert Dunlap, appeared to have engineered a turnaround by cutting costs and raising revenues. The reality was different, and rigorous analysis of the statements in the early phase of the misreporting would have supported far more scepticism than the market showed. The revenue transactions included the following.

One-time disposals of product lines were included in sales for the first quarter of 1997, with no indication that non-recurring items sat inside revenue.
At the end of that same quarter, in March, revenue and income were booked on a sale of barbecue grills to a wholesaler. The wholesaler held the goods over the quarter end without taking on the risks of ownership, could return them at will, and would have its shipping paid in both directions by Sunbeam. Every grill came back in the third quarter of 1997.
Customers were induced to order more than they otherwise would, through discounts and other incentives, often while retaining the right to return what they bought. Pulling future sales into the present in this way is channel stuffing. The policy was not disclosed, and it was used routinely at the end of 1997 and the start of 1998.
Sunbeam used bill-and-hold arrangements, in which revenue is recognised when the invoice is issued while the goods stay on the premises of the seller. These are unusual transactions and the requirements are strict: the buyer must request the treatment, must have a genuine business reason for the request, and must accept the risks of ownership. Supporting criteria include the past experience of the seller with such transactions, specifically that buyers did take possession and that the transactions were not reversed.

Neither the channel stuffing nor the bill-and-hold sales had a business purpose beyond letting the seller accelerate revenue and letting buyers take advantage of that eagerness at no risk. In the description used by the SEC, they amounted to little more than projected orders dressed as sales. Sunbeam did not make any of this clear, and its disclosures from the fourth quarter of 1996 to the middle of 1998 were inadequate. The methods nevertheless left traces.

The traces in the receivables

Inducing customers to buy goods they do not yet need, through favourable payment terms or wide latitude to return them, tends to lengthen days of sales outstanding and to raise returns. Revenue growth also tends to exceed both the company own history and the industry. Collection behaviour, expressed through the receivables measures, is therefore where the aggressiveness shows up.

Sunbeam sales and receivables, with 1997 restated for a receivables sale ($ millions)

Measure	1995	1996	1997	1997 pro forma
Total revenue	$1,016.9	$984.2	$1,168.2	$1,168.2
Change on prior year	Not shown	−3.2%	18.7%	18.7%
Gross accounts receivable	$216.2	$213.4	$295.6	$354.6
Change on prior year	Not shown	−1.3%	38.5%	66.1%
Receivables as a share of revenue	21.3%	21.7%	25.3%	30.4%
Change in that share	0.7%	0.4%	3.6%	8.7%
Days of sales outstanding	77.7	79.2	92.3	110.8
Accounts receivable turnover	4.7	4.6	4.0	3.2

Based on the original 10-K filings. Computing days of sales outstanding directly from the reported balances gives 77.6, 79.1 and 92.4 for the three years, a first-decimal rounding difference from the figures carried in the pro forma exhibit; it does not affect any conclusion below.

Turnover is the number of times receivables converted into cash during the year.

Four readings follow from the table. First, revenue fell 3.2 percent in 1996, the year the misreporting started, then rose 18.7 percent in 1997 as the revenue enhancement programmes ran. What should have caught the eye is not that increase but the far larger simultaneous increase in receivables. Receivables rising faster than revenue is the signature of a company pulling future sales forward with discounts or generous return rights, which is exactly what Sunbeam was doing.

Second, receivables as a share of revenue is another way of expressing the same thing. A rising share means a smaller proportion of sales has been collected. That may mean customers have become less able to pay. It may also mean the seller manufactured period-end sales by shipping goods nobody wanted, since the shipment generates documentation that looks like evidence of a sale; receivables and revenue then rise by the same absolute amount, which raises the ratio, and the goods come back in the next period. Entirely fictitious revenue produces the identical pattern, because revenue from a customer that does not exist raises receivables by the same amount. Either way, a rising relationship between receivables and revenue tells the analyst that collection has slowed or that something is wrong with recognition.

Third, days of sales outstanding rose every year, from 77.7 to 79.2 to 92.3. Receivables were not being paid on time, or the revenue was never genuine. Receivables turnover, which is the same information inverted, fell from 4.7 to 4.6 to 4.0. A trend of slower collection is at best evidence of worsening efficiency and at worst a signal to question the sales themselves.

Fourth, the receivables were of poor quality on the simplest possible comparison. In 1997 they grew 38.5 percent while revenue grew 18.7 percent. An analyst who also read the notes would have found more. A note headed Accounts Receivable Securitization Facility recorded that an arrangement for selling receivables had been put in place during December 1997, and that proceeds of roughly $59 million had reached the company by 28 December 1997 under it. The receivables behind those proceeds were left out of the year-end balance. Adding them back gives the pro forma column: receivables would have risen 66.1 percent rather than 38.5 percent, the share of revenue would have reached 30.4 percent, and days of sales outstanding would have been 110.8. Had the sale not happened, the deterioration would have been very hard to overlook.

Comparison with the peer group

Sunbeam looks bad against its own history. It looks far worse against its industry. The medians below come from a group of other consumer products companies: Harman International, Jarden, Leggett and Platt, Mohawk Industries, Newell Rubbermaid and Tupperware Brands.

Sunbeam against the industry median

Measure	1995	1996	1997
Sunbeam days of sales outstanding	77.7	79.2	92.3
Sunbeam receivables turnover	4.7	4.6	4.0
Industry median days of sales outstanding	44.6	46.7	50.4
Industry median receivables turnover	8.2	7.8	7.3
Sunbeam shortfall, days of sales outstanding	33.0	32.5	41.9
Sunbeam shortfall, receivables turnover	(3.5)	(3.2)	(3.3)

Based on information in company 10-K filings.

Sunbeam collected its receivables in roughly twice the time of the median peer in 1995 and 1996, and the gap widened to 41.9 days by 1997. A company whose collection cycle is that far outside the norm of its own industry, and moving away from it, is not merely inefficient.

Example 10 · Worked

In the annual report for December 1997 the revenue recognition note had been expanded from the prior year. It stated that revenue from product sales was recognised principally on shipment to customers; that in limited circumstances, at the request of the customer, seasonal product might be sold on a bill and hold basis provided the goods were complete, packaged, ready for shipment and segregated, and provided the risks of ownership and legal title had passed; and that such bill and hold sales at 29 December 1997 amounted to approximately 3 percent of consolidated revenues.

Working the disclosure through with reasonable assumptions about the gross profit on those sales, 28.3 percent, and the applicable tax rate, 35 percent, produces the table below.

Effect of the bill-and-hold sales on net income ($ millions)

Step	Amount
1997 revenue	$1,168.18
Bill-and-hold sales per the note	3.0%
Bill-and-hold sales in 1997	$35.05
Assumed gross profit margin	28.3%
Gross profit contribution	$9.92
After-tax earnings contribution	$6.45
Total earnings from continuing operations	$109.42
Share of earnings attributable to bill-and-hold sales	5.9%

1. Reproduce the arithmetic and state what it shows.

Solution. Bill-and-hold sales are $1,168.18 × 3.0% = $35.05 million. At a gross margin of 28.3 percent the gross profit contribution is $35.05 × 0.283 = $9.92 million. Taxing that at 35 percent leaves $9.92 × (1 − 0.35) = $6.45 million. Against total earnings from continuing operations of $109.42 million, that is $6.45 divided by $109.42, or 5.9 percent. A disclosure of 3 percent of revenue therefore turns into nearly 6 percent of net income, because the sales carry a margin and the base being compared against is much smaller than revenue.

2. Why should a 3 percent disclosure have been treated as important rather than immaterial?

Solution. Two reasons. The first is the arithmetic above: the percentage looks small only because it is quoted against revenue. The second is qualitative and matters more. The company had voluntarily expanded its revenue recognition note to describe a method of questionable quality, and had quantified it. That is a company telling the reader where to look. Establishing that nearly 6 percent of net income rested on such transactions should have been enough to discourage a favourable view of the shares.

MicroStrategy, Inc. sold software and information services, grew quickly, and came to the public market in 1998. After the flotation its transactions became more complex: fewer outright sales of software, more arrangements bundling several deliverables together, including obligations to provide services.

The distinction matters because the two revenue streams are recognised at different times. Product revenue is normally recognised at once, subject to the delivery terms and to customer acceptance, whereas service revenue is recognised as the service is performed. The standards then governing bundled arrangements allowed revenue to be taken on a software delivery on two conditions, both of which had to hold: the sale of the software had to be separable from the services element, and the service revenues had to be accounted for on their own.

The revenue recognition policy in the accounting policies note of the 1998 Form 10-K said that the company followed exactly that: revenue from product licensing arrangements recognised generally after a licensing agreement was executed and the product shipped, provided no significant obligations remained and the receivable was considered collectible; services revenue, covering training and consulting, recognised as the service was performed; and maintenance revenue deferred and recognised evenly over contract terms running from 12 to 36 months.

What happened in practice was different. The ambiguity in bundled arrangements was used to characterise service revenues as part of the software sale and recognise them early. In the fourth quarter of 1998 the company entered a $4.5 million transaction covering software licences and a wide range of consulting services. Most of the licences the customer acquired were intended for applications MicroStrategy had yet to develop, and yet the entire $4.5 million was recognised as software revenue. In the fourth quarter of 1999 the pattern repeated on a larger scale: the elements of a multiple-deliverable arrangement were allocated improperly toward the software element, which brought $14.1 million of product revenue into the quarter, an amount that was material.

What an outsider could see

Without the contracts themselves, nobody could approve or reject the allocation with certainty. The company still left a trail visible to anyone who had read its stated policy.

MicroStrategy revenue by type, 1996 to 1998 ($ thousands)

Type	1996	1997	1998
Licences	$15,873	$36,601	$72,721
Support	6,730	16,956	33,709
Total	$22,603	$53,557	$106,430
Licences, share of total	70.2%	68.3%	68.3%
Support, share of total	29.8%	31.7%	31.7%
Total share	100.0%	100.0%	100.0%

Based on the income statement in the 1998 Form 10-K.

The support share rose slightly between 1996 and 1997 and then flattened in 1998, the first year in which the two categories are known to have been mischaracterised. With hindsight, had the $4.5 million of consulting services not been recognised at all, total revenue would have been $101.930 million and support revenue would have represented 33.1 percent of the total rather than 31.7 percent. That is the size of the distortion the annual figures conceal.

MicroStrategy revenue mix by quarter, first quarter 1998 to fourth quarter 1999

Quarter	Licences	Support
1Q98	71.8%	28.2%
2Q98	68.3%	31.7%
3Q98	62.7%	37.3%
4Q98	70.7%	29.3%
1Q99	64.6%	35.4%
2Q99	68.1%	31.9%
3Q99	70.1%	29.9%
4Q99	73.2%	26.8%

Example 11 · Worked

1. Using the quarterly table, identify the two quarters an analyst should have questioned, and explain why.

Solution. The support share climbs through the first three quarters of 1998, from 28.2 percent to 31.7 percent to 37.3 percent, and then drops sharply to 29.3 percent in the fourth quarter, which is the quarter in which $4.5 million that should have been deferred was characterised as software licence revenue. The share then recovers to 35.4 percent in the first quarter of 1999 and declines through the year to 26.8 percent in the fourth quarter, which is the quarter in which a further $14.1 million was mischaracterised. In both cases the licence share reaches its local peak in the quarter the misallocation occurred. There is no business reason why the split between licensing and support should swing this much from quarter to quarter, and the swings are the questions to put to management. The answers, and how well they hold together, are the information.

2. What general habit does this case argue for?

Solution. Reading the accounting policy note first, and then watching for deviations from the trends that policy implies. If the note establishes that the company recognises revenue on contracts with multiple deliverables, the analyst knows in advance that misallocation between elements is a live risk and knows which series will reveal it. Trend analysis will rarely pinpoint a single manipulated transaction, but it is enough to raise a legitimate doubt about the accounting, which is all an outsider needs in order to act.

A checklist for revenue quality

Start with the policy. Read the revenue recognition policies in the most recent annual report before anything else. Establish the shipping terms; whether customer rights of return are limited or extensive; whether rebates affect revenue and, if so, how they are accounted for and what estimates they involve; and whether one arrangement carries multiple deliverables, in which case ask whether revenue is deferred until the later elements are delivered and whether deferred revenue appears on the balance sheet.
Age matters. Receivables do not improve with time. Track days of sales outstanding and receivables turnover over a meaningful period, and compare the days of sales outstanding of the company with those of similar competitors over the same periods.
Cash or accrual. A high ratio of receivables to revenue may mean nothing, or it may mean channel stuffing, which implies future returns of inventory or weaker demand ahead. Compare the ratio over time, and compare it with competitors or an industry measure.
Anchor to the physical world. Where a company routinely reports non-financial data, relate revenue to it. Airlines disclose miles flown and capacity, retailers disclose square footage and store counts, and companies in every industry disclose headcount. Revenue per unit can then be compared with competitors or an industry measure.
Trends and composition. Examine how much revenue comes from product sales or licences and how much from services, and whether that relationship has changed and why. Examine whether the movement in total revenue makes sense against the movement in accounts receivable.
Relationships. Establish whether the company transacts with entities owned by its senior officers or shareholders. Where such an entity is private and a public company recognises revenue from it, the arrangement can become a dumping ground for obsolete or damaged inventory while revenue is inflated.

Overstating revenue is not the only route to higher earnings. The same review of SEC enforcement cases from 1997 to 2002 put improper expense recognition second on the list. It usually means understating expenses, and the effect on earnings is identical to overstating revenue. It also leaves traces.

WorldCom was a global communications company supplying phone and internet services to business and consumer markets, and it became a major player during the 1990s largely by acquisition. To keep delivering the earnings analysts expected, it improperly capitalised operating expenses known as line costs. These were fees paid to third-party network providers for the right to use their networks, and the correct treatment is to charge them as an operating expense. The practice began in 1999 and continued into the first quarter of 2002. The company filed for bankruptcy in July 2002, and restatements followed.

The auditor was Arthur Andersen, which had access to the records. The special committee that investigated the failure concluded that the auditor had decided, year after year and mistakenly, that the risk of fraud was minimal, and had therefore never designed procedures capable of addressing it. Although the audit was controls-based and relied on the internal controls of the company, it failed to appreciate the nature and extent of top-side adjustments made by senior management through reserve reversals with little or no support, through highly questionable revenue items, and through the entries capitalising line costs. Corroborating tests were not performed in many areas, and the absence of variances in the statements and schedules, inside a highly volatile business, was read as a reason for less scrutiny rather than more.

Where the money had to go

If the auditors missed it, the question is whether an outsider could have done better. Probably not to the level of the line costs themselves. But an outsider looking for relationships between accounts that do not make sense, which is precisely what an auditor is expected to do, had a clear opening. An understated operating expense has to be matched by an increase somewhere else, and here the increase was in gross property, plant and equipment.

WorldCom, common-size asset side of the balance sheet

Asset	1997	1998	1999	2000	2001
Cash and equivalents	0%	2%	1%	1%	1%
Receivables, net	5%	6%	6%	7%	5%
Inventories	0%	0%	0%	0%	0%
Other current assets	2%	4%	4%	2%	2%
Current assets in total	7%	12%	11%	10%	8%
Property, plant and equipment, gross	30%	31%	37%	45%	47%
Accumulated depreciation	3%	2%	5%	7%	9%
Property, plant and equipment, net	27%	29%	32%	38%	38%
Equity investments	Not available	Not available	Not available	Not available	1%
Other investments	0%	0%	0%	2%	1%
Intangibles	61%	54%	52%	47%	49%
Other assets	5%	5%	5%	3%	3%
Total assets	100%	100%	100%	100%	100%

Based on information from the Standard and Poor Research Insight database.

The fraud began in 1999. Gross property, plant and equipment had been 30 percent of total assets in 1997 and 31 percent in 1998. It became 37 percent in 1999, then 45 percent in 2000 and 47 percent in 2001. Nothing in the strategy of the company changed to justify a shift of that size. An analyst in 1999 could not have named the line costs, but the buildup of cost inside property, plant and equipment was reason enough to suspect that some expense in the income statement was being under-reported.

Example 12 · Worked

1. Using only the common-size table, quantify the shift in gross property, plant and equipment from 1998 to 2001 and explain why the accumulated depreciation line does not contradict the capitalisation story.

Solution. Gross property, plant and equipment rose from 31 percent of total assets in 1998 to 47 percent in 2001, a shift of 16 percentage points, with the whole of the move occurring after the fraud began. Accumulated depreciation rose over the same period as well, from 2 percent to 9 percent of total assets, which is what would be expected once a growing asset base begins to be depreciated. That is the point: capitalising an operating cost does not remove the cost, it converts it into depreciation spread over the life of the supposed asset. Net property, plant and equipment therefore rose from 29 percent to 38 percent, and current-period earnings were relieved of the difference between the cost capitalised and the depreciation charged on it.

2. WorldCom reported operating cash flow above net income in every year of the fraud, as the earlier comparison of the two series showed. Reconcile that with the capitalisation.

Solution. The two facts are the same fact. Capitalised expenditure is presented as an investing outflow, not an operating one, so every dollar of line cost moved out of the income statement also moved out of the operating section of the cash flow statement. Operating cash flow was therefore inflated by exactly the mechanism that inflated earnings. This is why the comparison of operating cash flow with net income must be read alongside investing activities rather than on its own.

A checklist for expense quality

Start with the policy. Understand the cost capitalisation policies in the most recent annual report. Which costs are capitalised into inventory, how is obsolescence dealt with, and are there reserves for obsolescence that could be raised or lowered at will? What are the depreciation policies and the depreciable lives, how do they compare with those of competitors, and have they changed?
Examine the non-current accounts every quarter for movements quarter on quarter and year on year. An unusual increase in cost may point to improper capitalisation.
Tie the margins to the balance sheet. Gross and operating margins are watched closely at each results announcement but are rarely related to the balance sheet, and they should be. If non-current assets are building unusually while margins improve or hold flat, improper capitalisation is one explanation. The industry backdrop matters too: stable margins and growing balance sheet accounts in a slumping industry is the combination to worry about.
Compute turnover ratios for total assets, for property, plant and equipment, and for other assets, dividing revenue by the asset classification in each case. Slowing turnover with falling revenue may mean the assets produce something demand is leaving, which points to future write-downs. Slowing turnover with steady or rising revenue points to improper capitalisation.
Compare depreciation or amortisation with the relevant asset base. Is it drifting up or down without a good reason, and how does it compare with competitors?
Compare capital expenditure with gross property, plant and equipment over time. A rising proportion may mean the company is capitalising more aggressively to keep costs out of current expense.
Relationships again. Where the company deals with entities owned by senior officers or shareholders, and particularly where those entities are private, transactions may be priced to move wealth out of the public company. The same transfer can run through excessive compensation, direct loans or guarantees, and the practice is known as tunneling. The reverse also occurs: the manager-owned entity transfers resources into the public company to keep it viable and preserve the option of extracting value later, which is known as propping. Sham dealings in either direction can be reported so as to improve the profits of the public company and thereby the performance-based pay of its managers.

The general point is that earnings quality should be established rather than assumed. Waiting until an accounting problem surfaces means waiting until it is too late, and high reported growth in earnings that later proved fraudulent has preceded a great many bankruptcies.

Bankruptcy prediction reaches beyond earnings into cash flow and the balance sheet, which is consistent with the broad use of the term earnings quality throughout this reading. A range of approaches exists for quantifying how likely a company is to default on its debt or to enter bankruptcy.

The Altman model

The earliest well-known model is Altman (1968), built on research that used ratio analysis to identify likely failures. Its important contribution was to combine several ratios into one measure. Viewing ratios one at a time creates a specific error: a company with weak profitability or a weak solvency position looks like a bankruptcy candidate even when its liquidity position is strong. Using discriminant analysis, Altman produced a function that separates bankrupt from non-bankrupt companies.

A higher Z-score is better.

The five Altman ratios

Term	Ratio	What it captures
X1	Net working capital divided by total assets	Short-term liquidity risk
X2	Retained earnings divided by total assets	Accumulated profitability, and indirectly relative age, since retained earnings build up over time
X3	EBIT divided by total assets	Profitability, as a variant of return on assets
X4	Market value of equity divided by book value of liabilities	Leverage, expressed as equity over debt, so a higher number means greater solvency
X5	Sales divided by total assets	Activity, the ability to generate sales from the asset base

The discriminant function as it appeared in the original 1968 article looked different.

The same five ratios, scaled differently.

Altman (2000) explained the difference. Because of the original computer format, the variables X1 through X4 have to be entered as absolute percentage values: a company whose net working capital to total assets is 10 percent is entered as 10.0, not as 0.10. Only X5 is entered differently, so that a sales to total assets ratio of 200 percent is entered as 2.0. Restating the coefficients to accept decimal inputs produces the first version, which is why that is the form normally quoted.

Interpretation rests on two thresholds. In Altman application of the model to a sample of manufacturing companies that had made losses, a score below 1.81 indicated a high probability of bankruptcy and a score above 3.00 indicated a low probability. Scores between 1.81 and 3.00 were not clear indicators either way.

What came afterwards

Later work addressed two identified weaknesses. The first is that the Altman model is single-period and static: it uses one set of financial measures observed at one date. Shumway (2001) answered this with a hazard model, which uses all available years of data to compute the bankruptcy risk of each company at each point in time.

The second weakness is shared by all accounting-based models. Financial statements measure past performance and are prepared on the going-concern assumption, so the reported values on the balance sheet already assume that the company is not failing, which is precisely the assumption under test. Market-based models avoid the circularity. Models building on the concept, associated with Merton, of equity as a call option on the assets of the company infer the default probability from the value of equity, the amount of debt, equity returns and equity volatility. Credit default swap data and corporate bond data can be used to derive default probabilities as well. The evidence suggests that the most effective models use both kinds of input: Bharath and Shumway (2008), for instance, build a default probability from five inputs: what the equity is worth in the market, the face amount of the debt, how volatile the equity is, how the stock performed against the market across the preceding year, and net income measured against total assets.

Why the models get less useful as they get better known

Every published screen is available to the people it is designed to catch. Beneish and co-authors documented declining predictive power for their own model over time, and the most plausible explanation is that managers run the model on their own numbers before deciding how to report. The same logic applies to the accrual screens and, to a lesser extent, to bankruptcy scores. Two consequences follow. First, a clean score is much weaker evidence than a dirty one: failing a screen is informative, passing one may only mean the company knew where the line was. Second, the qualitative steps of the general procedure gain value precisely as the quantitative ones lose it, because reading disclosures, comparing policies with competitors and asking management direct questions cannot be reverse-engineered in the same way.

The cash flow statement escapes some of the discretion built into accrual accounting, which is why analysts lean on it heavily. It does not escape all of it. Management retains ways to influence what the statement shows.

Operating cash flow is the component that matters most for judging performance and for valuing a company or its securities, so discussions of cash flow quality concentrate there. The vocabulary works exactly as it does for earnings. High-quality cash flow means both that economic performance was good, in the sense of value enhancing, and that reporting quality was high, so the disclosed figures reasonably reflect economic reality. Low-quality cash flow means either that the reporting correctly describes bad performance or that it misrepresents what happened.

What good looks like

The life cycle of the company and the profile of the industry come first. A start-up might be expected to run negative operating and investing cash flows funded by borrowing or by issuing equity, and there is nothing wrong with that. An established company would normally generate positive operating cash flow and fund its necessary investment and its returns to capital providers, whether dividends, buybacks or debt repayments, out of it. For an established company, high-quality cash flow generally shows most or all of four features:

operating cash flow is positive;
it comes from sustainable sources;
it is sufficient to cover capital expenditure, dividends and debt repayments; and
its volatility is relatively low compared with others in the industry.

Those four are results quality. Reporting quality is separate and equally necessary: the reported flows must be relevant and must faithfully represent what the company did. Presenting a financing inflow as an operating inflow misrepresents economic reality however accurate the arithmetic is.

Where the manipulation happens

Operating cash flow is generally harder to manipulate than operating income or net income, which is why large differences between earnings and operating cash flow, or a widening of those differences, can point to earnings manipulation. That same importance, however, creates the incentive to manage the cash flow figure itself.

Two issues recur. The first is timing. Selling receivables to a third party, or delaying payment of payables, lifts operating cash flow without changing anything real. Both leave marks: days of sales outstanding falls and days of payables rises. The analyst can therefore detect choices to reduce current assets or increase current liabilities by watching asset utilisation ratios, movements in balance sheet accounts and the note disclosures. The second issue is classification. Management may try to shift positive items out of investing or financing and into operating.

Figure 5: A classification shift changes the mix, not the total

Nothing about the business changes. What changes is the number investors watch most closely, and the expectations they form about future cash generation.

Reading the statement for signs of misreporting

The Satyam fraud shows both the value and the limits of the approach. A model-based screen missed it entirely. An analyst using a computer model in September to examine the 500 largest Indian public companies found that more than 20 percent of them were potentially engaged in aggressive accounting, and Satyam was not among them. The reason is that automated screens start by looking for large gaps between reported earnings and cash flow, and at Satyam the cash appeared to keep pace with the profits. A screen for companies whose operating cash flow is persistently below earnings would never have flagged it.

A qualitative reading of the same statement performs better.

Satyam, IFRS consolidated interim cash flow statement excerpt ($ millions)

Line	Quarter to 30 June 2008 (unaudited)	Quarter to 30 June 2007 (unaudited)	Year to 31 March 2008 (audited)
Profit before income tax	143.1	107.1	474.3
Share-based payment expense	4.3	5.9	23.0
Financial costs	1.3	0.8	7.0
Finance income	(16.2)	(16.4)	(67.4)
Depreciation and amortisation	11.5	9.3	40.3
Loss or gain on disposal of premises and equipment	0.1	0.1	0.6
Value change on preference shares at fair value through profit or loss	0.0	0.0	(1.6)
Gain or loss on foreign exchange forward and option contracts	53.0	(21.1)	(7.4)
Share of results of joint ventures, net of taxes	(0.1)	0.0	(0.1)
Subtotal after adjustments	197.0	85.7	468.7
Working capital: trade and other receivables	(81.4)	(64.9)	(184.3)
Working capital: unbilled revenue	(23.5)	(6.0)	(39.9)
Working capital: trade and other payables	34.1	2.2	48.8
Working capital: unearned revenue	5.8	2.4	11.4
Working capital: other liabilities	(6.3)	30.3	61.2
Working capital: obligations for retirement benefits	3.7	1.3	17.8
Cash the operations generated	129.4	51.0	383.7
Taxes on income, paid	−3.8	−9.8	−49.4
Operating activities, net cash provided	125.6	41.2	334.3

Based on information from the Satyam Form 6-K filed on 25 July 2008.

One line does not belong. The $53.0 million labelled as a gain or loss on foreign exchange forward and option contracts in the June 2008 quarter is a non-cash item, and the label suggests a gain. Adding back a gain in a reconciliation from profit before tax to operating cash flow is wrong, because the gain is already inside profit before tax. Asked about the item on the quarterly conference call, the company representative could not answer and undertook to come back to the questioner. An item representing almost 40 percent of pre-tax profit for the quarter, since $53.0 divided by $143.1 is 37 percent, that senior executives cannot explain on the call is a signal in its own right.

The second signal in the same statement is the steady growth in receivables, which is best read through the ratios.

Satyam, selected annual receivables data ($ millions)

Measure	2008	2007	2006	2005
Total revenue	$2,138.1	$1,461.4	$1,096.3	$793.6
Change on prior year	46.3%	33.3%	38.1%	Not shown
Gross accounts receivable	$539.1	$386.9	$238.1	$178.3
Change on prior year	39.3%	62.5%	33.5%	Not shown
Allowance for doubtful debts	$31.0	$22.8	$19.1	$17.5
Change on prior year	36.0%	19.4%	9.1%	Not shown
Gross receivables as a share of revenue	25.21%	26.47%	21.72%	22.47%
Change in that share	−4.8%	21.9%	−3.3%	Not shown
Days of sales outstanding	92.0	96.6	79.3	82.0
Accounts receivable turnover	4.0	3.8	4.6	4.5

Based on data from the Satyam Form 20-F filings.

Days of sales outstanding jumped from 79.3 in 2006 to 96.6 in 2007, and the management commentary in the Form 20-F attributed the increase in net accounts receivable to higher revenues and a longer collection period. A lengthening collection period raises three questions at once: about the creditworthiness of the customers, about the efficiency of the collection effort, and about the quality of the revenue that was recognised.

A third signal never touched the cash flow statement at all. Satyam reported growing amounts held in current accounts. On a later conference call an analyst asked why roughly $500 million was parked in current accounts earning no interest, and, when the answer pointed to deposit accounts, followed up by observing that the deposit balances had been flat for four quarters while the incremental cash kept going into current accounts. The reply was that the balances sat in different countries and would be brought to India as needed, and that the amounts would move into deposits from the next quarter. In the resignation letter of January 2009 the chief executive confessed that the balance sheet at 30 September 2008 carried inflated, non-existent cash and bank balances of Rs. 5,040 crore against Rs. 5,361 crore shown in the books, a crore being ten million. More than 90 percent of the reported cash did not exist, and some of what had existed is understood to have been moved to a web of companies controlled by the chief executive and his family.

Example 13 · Worked

Return to Sunbeam. The excerpt below is from its consolidated statement of cash flows, prepared under the indirect method.

Sunbeam, consolidated statement of cash flows excerpt ($ thousands)

Fiscal year ended	28 Dec 1997	29 Dec 1996	31 Dec 1995
Net earnings or loss	109,415	(228,262)	50,511
Depreciation and amortisation	38,577	47,429	44,174
Restructuring, impairment and other costs	Nil	154,869	Nil
Other non-cash special charges	Nil	128,800	Nil
Loss on sale of discontinued operations, after tax	13,713	32,430	Nil
Deferred income taxes	57,783	(77,828)	25,146
Working capital: receivables, net	(84,576)	(13,829)	(4,499)
Working capital: inventories	(100,810)	(11,651)	(4,874)
Working capital: accounts payable	(1,585)	14,735	9,245
Working capital: restructuring accrual	(43,378)	Nil	Nil
Working capital: prepaid and other current items	(9,004)	2,737	(8,821)
Working capital: income taxes payable	52,844	(21,942)	(18,452)
Payments on other long-term and non-operating liabilities	(14,682)	(27,089)	(21,719)
Other, net	(26,546)	13,764	10,805
Net cash provided by or used in operating activities	(8,249)	14,163	81,516

Under the indirect method an increase in receivables appears as a negative figure, because sales recognised in income for which no cash has yet arrived have to be reversed.

1. Sunbeam inflated restructuring charges and later reversed them. How does that appear here?

Solution. The indirect method reconciles reported net income to operating cash flow, so every non-cash charge is visible. In 1996 the non-cash charges recorded for restructuring, impairment and other costs totalled about $284 million, being $154.869 million plus $128.8 million. The following year the restructuring accrual was reversed by $43 million, shown as $(43,378) thousand. Inflating the charge in 1996 and releasing it in 1997 made the reported performance appear to improve sharply after the arrival of the new chief executive in the middle of 1996, which is exactly what the reversal was for.

2. Which lines would improper revenue recognition affect first?

Solution. Net income, receivables and inventories. Net income and receivables are both overstated, and the statement, in which an increase in receivables is a negative figure, shows receivables continuing to grow: $(84,576) thousand in 1997 against $(13,829) thousand in 1996. Inventories tell the second half of the story. Sunbeam recorded sales with no economic substance because the buyer held the goods across a period end and then returned all of them, and the goods coming back show up as the substantial 1997 inventory increase of $(100,810) thousand. The company reported net earnings of $109,415 thousand in 1997 and used $8,249 thousand of cash in operations in the same year.

Classification shifting inside the cash flow statement

Shifting positive items from investing or financing into operating does not change total cash flow, but it changes how investors evaluate the company and what they expect next. Standards leave room for it. IFRS allows interest paid to be classified as operating or as financing, and interest and dividends received as operating or as investing. US GAAP requires interest paid, interest received and dividends received all to be operating. An analyst comparing an IFRS reporter with a US GAAP reporter therefore has to check that the classifications match and adjust if they do not, and an analyst following a single IFRS reporter has to watch for year-to-year changes. A company that moved interest paid from operating to financing would report higher operating cash flow than in the prior period even if nothing else had happened at all.

A second source of flexibility is the trading and non-trading distinction. Cash flows from non-trading securities are classified as investing, while cash flows from trading securities are typically operating, and each company decides what counts as trading given how it manages its holdings. That discretion is itself an opportunity.

Example 14 · Worked

Nautica Enterprises, an apparel manufacturer, published the statement of cash flows for the year ended 4 March 2000 in its annual report filed on 27 May 2000, and published the same year again as a comparative in its annual report filed on 29 May 2001.

Nautica Enterprises, cash flow excerpts ($ thousands)

Line	Year to 4 March 2000, as first filed	Year to 4 March 2000, as shown a year later	Year to 3 March 2001
Net earnings	$46,163	46,163	46,103
Minority interest in loss of consolidated subsidiary	Nil	Nil	Nil
Deferred income taxes	(1,035)	(1,035)	(2,478)
Depreciation and amortisation	17,072	17,072	22,968
Provision for bad debts	1,424	1,424	1,451
Operating movement: short-term investments	Not shown here	21,116	28,445
Operating movement: accounts receivable	(6,562)	(768)	(17,935)
Operating movement: inventories	(3,667)	(3,667)	(24,142)
Operating movement: prepaid and other current assets	(20)	(20)	(2,024)
Operating movement: other assets	(2,686)	(2,686)	(36)
Operating movement: trade accounts payable	(548)	(548)	14,833
Operating movement: accrued expenses and other current liabilities	9,086	3,292	7,054
Operating movement: income taxes payable	3,458	3,458	3,779
Net cash provided by operating activities	62,685	83,801	78,018
Purchase of property, plant and equipment	(33,289)	(33,289)	(41,712)
Sale or purchase of short-term investments	21,116	Nil	Nil
Payments to register trademark	(277)	(277)	(199)
Net cash used in investing activities	(12,450)	(33,566)	(41,911)

1. What operating cash flow does each report show for the year ended 4 March 2000?

Solution. The report filed in 2000 shows $62,685 thousand. The report filed in 2001 shows $83,801 thousand for the same year.

2. The first report showed investing cash flows of $21,116 thousand from the sale of short-term investments in that year. Where does the amount go in the later report?

Solution. Into operating. The $21,116 thousand is exactly the difference between the two operating figures, $83,801 minus $62,685, and in the later report it appears among the changes in operating assets and liabilities rather than in investing activities. The investing total for the year moves correspondingly, from $(12,450) thousand to $(33,566) thousand.

3. As reported, how did fiscal 2001 operating cash flow compare with fiscal 2000, and how would it have compared without the change in classification?

Solution. As reported, operating cash flow fell about 7 percent, since 78,018 divided by 83,801 minus 1 is −7 percent. Without the reclassification the fiscal 2001 figure would have been 78,018 − 28,445 = $49,573 thousand, and the comparison would have been against the originally reported $62,685 thousand, giving 49,573 divided by 62,685 minus 1, or a decline of 21 percent. A fall of 7 percent and a fall of 21 percent invite very different conclusions about the business.

4. How could the shift have been detected at the time?

Solution. By comparing the statement of cash flows for a year as first published with the same year as presented in the following annual report. The technique generalises. Whenever a company restates prior years because of an error, recasts them because of a change in accounting policy, drops information it used to disclose voluntarily, or adds something new such as a risk disclosure, the analyst should establish why.

For the balance sheet, high reporting quality has three components: completeness, unbiased measurement and clear presentation. High results quality, meaning a strong balance sheet, is a different matter and shows up as an optimal amount of leverage, adequate liquidity and an economically successful allocation of assets. Strength is assessed with ratio analysis, including common-size statements. There are no absolute values that mark adequate strength; the assessment only means something in the context of the earnings and cash flow outlook of the firm and the environment it operates in. The focus in this section is reporting quality.

Completeness

Obligations kept off the balance sheet in significant amounts understate leverage, and that is a direct threat to completeness. Purchase contracts are a common source, particularly where they are structured as take-or-pay contracts. The standard remedy is constructive capitalisation: the analyst estimates the obligation as the present value of the future purchase obligation payments and adds it to both the reported assets and the reported liabilities.

Unconsolidated joint ventures and equity-method investees can hide liabilities in the same way. They also distort profitability. Return on sales, also called net profit margin, can be overstated because the consolidated statements of the parent include its share of the profits of the investee but not its share of the sales of the investee. Where the disclosures allow it, the analyst can adjust to reflect the combined sales, assets and liabilities. A company running numerous or material unconsolidated subsidiaries at ownership levels close to 50 percent deserves particular attention, though the concern can often be settled by understanding why the structure exists, whether industry practice or a genuine need for strategic alliances in certain businesses or geographies.

Unbiased measurement

Measurement bias matters most where valuation is subjective. Four cases recur.

Understated impairment of inventory, of property, plant and equipment, or of other assets overstates profit on the income statement and overstates the asset on the balance sheet at the same time. A company reporting substantial goodwill while its market value of equity is below the book value of shareholders equity may not have taken the goodwill impairments it should have.
Understated valuation allowance for deferred tax assets understates tax expense and overstates the asset. Overstating the allowance does the reverse. Significant variation in the valuation account that cannot be explained is a signal of biased measurement.
Investments in the debt or equity securities of other companies should ideally be measured against observable market data. For some there is none, and the valuation rests entirely on management estimates. A balance sheet with a substantial share of its assets valued on non-observable inputs warrants closer scrutiny.
Pension liabilities depend on estimates, including the discount rate used to present value future obligations. Where pension obligations exist, both the level of the discount rate and the changes in it should be examined.

Example 15 · Worked

In August 2012 a Wall Street Journal article listed six companies carrying more goodwill on their balance sheets than their entire market values. At the head of the list was Sealed Air Corporation, which operates in packaging and containers. The statements below cover the year after the article appeared.

Sealed Air Corporation, consolidated statements of operations ($ millions)

Year ended 31 December	2012	2011	2010
Net sales	$7,648.1	$5,550.9	$4,490.1
Cost of sales	5,103.8	3,950.6	3,237.3
Gross profit	2,544.3	1,600.3	1,252.8
Marketing, administrative and development expenses	1,785.2	1,014.4	699.0
Amortisation of acquired intangible assets	134.0	39.5	11.2
Impairment of goodwill and other intangible assets	1,892.3	Nil	Nil
Costs of acquiring and integrating Diversey	7.4	64.8	Nil
Restructuring and other charges	142.5	52.2	7.6
Operating profit or loss	(1,417.1)	429.4	535.0
Interest expense	(384.7)	(216.6)	(161.6)
Loss on debt redemption	(36.9)	Nil	(38.5)
Impairment of equity method investment	(23.5)	Nil	Nil
Currency losses or gains on Venezuelan subsidiaries	(0.4)	(0.3)	5.5
Net gains on sale of available-for-sale securities	Nil	Nil	5.9
Other expense, net	(9.4)	(14.5)	(2.9)
Earnings or loss from continuing operations before tax	(1,872.0)	198.0	343.4
Income tax provision or benefit	(261.9)	59.5	87.5
Net earnings or loss from continuing operations	(1,610.1)	138.5	255.9
Net earnings from discontinued operations	20.9	10.6	Nil
Net gain on sale of discontinued operations	178.9	Nil	Nil
Net earnings or loss available to common stockholders	$(1,410.3)	$149.1	$255.9

Sealed Air Corporation, consolidated balance sheet excerpt ($ millions)

Year ended 31 December	2012	2011
Cash and cash equivalents	$679.6	$703.6
Receivables, net of an allowance of $25.9 in 2012 and $16.2 in 2011	1,326.0	1,314.2
Inventories	736.4	777.5
Deferred tax assets, current	393.0	156.2
Assets held for sale	Nil	279.0
Prepaid expenses and other current assets	87.4	119.7
Current assets in total	$3,222.4	$3,350.2
Property and equipment, net	$1,212.8	$1,269.2
Goodwill	3,191.4	4,209.6
Intangible assets, net	1,139.7	2,035.7
Deferred tax assets, non-current	255.8	112.3
Other assets, net	415.1	455.0
Total assets	$9,437.2	$11,432.0

1. The statements show 192,062,185 common shares issued and outstanding in 2011. The share price was around $18 in December 2011 and around $14 in August 2012, when the article appeared. What was the market value of the company?

Solution. About $3,457 million in December 2011, being 192,062,185 shares × $18, and about $2,689 million in August 2012, being 192,062,185 shares × $14.

2. How does goodwill at 31 December 2011 compare with that market value?

Solution. Goodwill stood at $4,209.6 million, which exceeds the market value on either date. Goodwill and other intangible assets together represented about 55 percent of total assets at that date, being $4,209.6 million plus $2,035.7 million against $11,432.0 million.

3. Why does goodwill above market value point to a future write-off?

Solution. Because of what the arithmetic implies about everything else the company owns. If market capitalisation exactly equalled reported goodwill, the market would be implicitly valuing every other asset of the company at zero. Here market capitalisation is below reported goodwill, so the value implicitly attributed to all the other assets is less than zero, which is not a sustainable position. The natural inference is that goodwill is overvalued and that a write-off is likely.

4. Did that inference prove correct?

Solution. Yes. In the fiscal year ending 31 December 2012, the year after the article, Sealed Air recorded an impairment of goodwill and other intangible assets of $1,892.3 million. Goodwill fell from $4,209.6 million to $3,191.4 million and intangible assets from $2,035.7 million to $1,139.7 million, and the operating result swung from a profit of $429.4 million to a loss of $1,417.1 million.

Clear presentation

Standards specify a great deal about what appears on the balance sheet, but companies retain discretion over which line items are shown separately and which are aggregated into a single total. Where an item is aggregated, the notes will usually break it down. Reading the inventory note, for example, may reveal that inventory is carried on a last-in, first-out basis, which in an inflationary environment means the balance sheet carries it at a cost significantly below current cost. That is a useful thing to know, and in this case it is reassuring, because inventory measured that way is unlikely to be overstated.

The financial statements themselves signal several kinds of risk. High leverage ratios, or equivalently low coverage ratios, signal financial risk. The analytical models described earlier signal bankruptcy risk and reporting risk. Operating risk shows up in highly variable operating cash flows or in a negative trend in profit margins. Beyond the statements there are four further sources worth using, and one that is worth much less than its reputation suggests.

The auditor opinion, and why it arrives too late

An audit opinion on the financial statements, and where required on internal control over financial reporting, does convey information about reporting risk. It rarely conveys it in time. A clean opinion on the statements says that they present the information fairly and in conformity with the relevant accounting principles; a clean opinion on internal controls says the company maintained effective controls over financial reporting. A negative or going-concern opinion, or a report identifying an internal control weakness, is clearly a warning. But the opinion addresses historical information, so it is almost never the first place an analyst learns anything.

Eastman Kodak filed for bankruptcy on 19 January 2012. The audit opinion for fiscal 2011 is dated 28 February 2012, six weeks later. It is identical to the prior-year opinion except that the years were updated and one paragraph was added, stating that the statements were prepared on the going-concern assumption, that the company and its US subsidiaries had filed petitions for relief under Chapter 11 of the United States Bankruptcy Code on 19 January 2012, that the uncertainties inherent in the bankruptcy process raise substantial doubt about the ability of the company to continue as a going concern, and that the accompanying statements include no adjustments that might result from the outcome. An analyst had known about the bankruptcy since 19 January. The opinion neither delivered the news nor covered statements adjusted for it, so on both counts it was of little use. Financial and market data would have flagged the difficulty far earlier.

Groupon makes the same point from a different angle. The table below is the sequence of events around a material weakness in its internal controls.

Groupon: a control weakness that no audit opinion reported

Date	Event
November 2011	The company completes its initial public offering
March 2012	Results are revised and management discloses a material weakness in internal control over financial reporting as of 31 December. The shares fall 17 percent. An exemption for newly public companies meant no external auditor opinion on the effectiveness of internal controls was required
May 2012	The first-quarter filing says the company is taking steps to correct the weaknesses but can give no assurance that controls will be effective by the year end
August 2012	The second-quarter filing carries a similar disclosure
November 2012	The third-quarter filing carries a similar disclosure
February 2013	The full-year filing states that the previously identified material weakness had been remediated as of 31 December 2012. This filing carries the first external auditor opinion on the effectiveness of internal controls, and the opinion is clean

No adverse external opinion appeared before or during the period in which the weakness existed: none was required for the first annual filing, and by the second the weakness had been fixed. Other information was far more useful. The company had been required to change its revenue recognition policy and to restate the revenue reported in its offering filing, which is an unambiguous sign of reporting difficulty. So was the sheer pace of expansion. Reported revenue for 2009 was more than 300 times the 2008 figure, and 2010 revenue was 23 times the 2009 figure. A commentator writing in August 2011 observed that a company which had completed 17 acquisitions in little over a year, expanded into 45 countries, grown its merchant base from 212 to 78,466 and its workforce from 37 to 9,625 in two years could hardly be expected to have effective internal control over financial reporting. Coupled with the disclosures in the offering filing about management inexperience, the growth data were a warning many months before the company disclosed the weakness, and the weakness never appeared in an audit opinion at all.

What the audit relationship does supply is a different signal. A change of auditor, and especially repeated changes, can indicate reporting problems. One of the largest feeder funds for the Ponzi scheme run by Bernie Madoff used three different auditors in the three years from 2004 to 2006, a fact identified in congressional testimony as a major warning sign of auditor shopping. An auditor whose capabilities look inadequate for the complexity of the client is equally informative: the accounting and auditing firm that audited the $50 billion Madoff operation consisted of three people, two principals and a secretary. Any question over the independence of the auditor belongs in the same category, whether the auditor and management appear unusually close or the client represents a substantial share of the revenue of the audit firm.

The notes

The notes are an integral part of the statements and are usually the richest source on risk. Both IFRS and US GAAP require specific disclosures about contingent obligations, about pension and post-employment benefits, and about the risks arising from financial instruments.

Disclosures about contingent obligations cover the nature of the obligation, the estimated amounts, the timing of required payments and the related uncertainties. Contingent losses are recognised when the loss is probable and the amount can be reasonably estimated, and disclosed without recognition when a loss is less than probable but more than remote, or when the amount cannot be reliably estimated. IFRS distinguishes provisions, which are recognised as liabilities because they meet the definition of a liability, from contingent liabilities, which are disclosed only. The concepts are similar under US GAAP despite the difference in terminology.

Royal Dutch Shell: decommissioning and other provisions ($ millions)

Provision	Current, 31 Dec 2012	Current, 31 Dec 2011	Non-current, 31 Dec 2012	Non-current, 31 Dec 2011	Total, 31 Dec 2012	Total, 31 Dec 2011
Decommissioning and restoration	1,356	894	14,715	13,072	16,071	13,966
Environmental	366	357	1,032	1,078	1,398	1,435
Redundancy	228	406	275	297	503	703
Litigation	390	256	307	330	697	586
Other	881	1,195	1,106	854	1,987	2,049
Total	3,221	3,108	17,435	15,631	20,656	18,739

The company states that both the timing and the amounts eventually settled are uncertain and depend on factors that are not always within the control of management. Of the decommissioning and restoration provision at 31 December 2012, an estimated $4,666 million was expected to be used within one to five years and $3,483 million within six to ten years, with the remainder later. The annual review of estimated decommissioning and restoration costs produced an increase of $1,586 million in 2012.

Year-on-year movements in estimates of that kind carry information for risk assessment: the decommissioning and restoration provision rose from $13,966 million to $16,071 million in a single year. The same notes also describe legal proceedings, in the Shell case including groundwater contamination claims brought by water purveyors and governmental entities against Shell Oil Company over releases of gasoline containing oxygenate additives, of which fewer than ten cases remained at 31 December 2012, and various environmental and contractual disputes in Nigeria at different stages of litigation, some already decided against the company on appeal. In both cases management stated that it did not expect a material effect, and in both cases the disclosure gives the analyst the facts on which to form an independent view.

Disclosures about pensions and post-employment benefits give information relevant to the actuarial risks that could make actual benefits differ from the reported obligations, and to the investment risks that could make actual assets differ from the estimated amounts. Disclosures about financial instruments cover credit risk, liquidity risk and the market risks arising from the instruments the company holds, together with how those risks have been managed.

Example 16 · Worked

The financial instruments note of Royal Dutch Shell explains that most debt is raised through central borrowing programmes, and that interest rate swaps and currency swaps have been used to convert most centrally issued debt to a floating rate linked to the dollar market reference rate, reflecting a policy of holding debt principally in dollars with a largely floating exposure profile. The financing of most subsidiaries is also structured on a floating-rate basis, and further interest rate risk management at subsidiary level is discouraged except in special cases.

On the basis of the floating rate net debt position at 31 December 2012, holding other factors constant and assuming no further management action, an increase in interest rates of 1 percent would reduce pre-tax income by $27 million, against $146 million on the 2011 position. On the same assumptions, a 10 percent appreciation against the dollar of the main currencies to which the group is exposed would have the pre-tax effects below.

Effect of a 10 percent appreciation against the dollar ($ millions)

Currency	Income effect, 2012	Income effect, 2011	Net asset effect, 2012	Net asset effect, 2011
Sterling	(185)	(58)	1,214	1,042
Canadian dollar	131	(360)	1,384	1,364
Euro	30	458	1,883	1,768
Australian dollar	246	153	142	120

The sensitivities are computed from the carrying amounts of assets and liabilities at 31 December only. The income effect arises on monetary balances held in currencies other than the functional currency of the relevant entity; the net asset effect arises principally on translating the assets and liabilities of entities that are not dollar-functional.

1. Does the group manage interest rate risk centrally or locally?

Solution. Centrally. Most debt is raised centrally, and swaps are used to convert most of the exposure to the dollar market reference rate. Beyond structuring subsidiary financing on a floating-rate basis, the group discourages further interest rate risk management at the subsidiary level.

2. For the year ended 31 December 2012 the group reported pre-tax income of $50,289 million. How significant is the exposure to a 1 percent rise in interest rates?

Solution. It is close to immaterial. A 1 percent rise would reduce pre-tax income by $27 million, and $27 million against $50,289 million is less than 0.1 percent of reported pre-tax income.

3. What would a 10 percent appreciation of the Australian dollar against the US dollar do to pre-tax income in the same year?

Solution. It would increase pre-tax income by $246 million, which is about 0.5 percent of the reported pre-tax income of $50,289 million. Notice that the same movement raises net assets by only $142 million, so the income and balance sheet exposures to a single currency need not be of similar size or even in the same relative order.

Management commentary

The IFRS Practice Statement on management commentary, issued in December 2010, is a non-binding framework for commentary accompanying IFRS statements. One of its purposes is to help users understand the risk exposures of the company, its approach to managing them, and how effective that management is. Five elements are listed for inclusion: what the business is; what it is trying to achieve and how; the resources it holds, the risks it runs and the relationships it depends on; how it has performed and what lies ahead; and the measures and indicators by which performance is judged. It says management should disclose the principal strategic, commercial, operational and financial risks, meaning those that may significantly affect the strategies and the progress of the value of the entity, and that the description should cover exposures to negative consequences as well as potential opportunities. Principal risks and uncertainties can be either significant external or significant internal risks.

Public US companies must include a management discussion and analysis as Item 7 of Form 10-K, covering liquidity, capital resources, results of operations, off-balance-sheet arrangements and contractual arrangements. The last two allow an analyst to anticipate future effects on cash flow. Quantitative and qualitative information about exposure to market risks appears separately as Item 7A, which should let the analyst understand the impact of movements in interest rates, foreign exchange and commodity prices. Disclosures about risk factors relevant to the securities of the company are required as Item 1A.

Both frameworks ask for selectivity. The IFRS Practice Statement says companies should present only the principal risks rather than list every possible risk and uncertainty, and the internal Financial Reporting Manual of the SEC Division of Corporation Finance says the management discussion and analysis should not be generic or boilerplate but should reflect the facts and circumstances specific to the individual registrant. Practice does not always match the intent, and one of the recurring challenges is separating risks that are generic, and therefore relevant to every company, from risks that are specific to this one.

Autonomy Corporation illustrates the problem. Its 2010 annual report, the last before Hewlett-Packard acquired it for $11.1 billion in 2011, carried a key risks and uncertainties section running to two pages and set out as a table of risk, description, impact and mitigation. The entries covered dependence on the core technology, competing technology, quarterly variability and late-in-the-quarter purchasing cycles, expenditure rising without a matching rise in revenue, falling average selling prices, unfavourable economic and market conditions, reliance on resellers, the continued service of executive directors, the hiring and retention of qualified staff, errors or defects in products, problems arising from potential acquisitions, claims of intellectual property infringement, managing growth, international expansion and security breaches. Nearly all of those apply to any technology company. Hewlett-Packard later took a multi-billion-dollar write-down on the investment, attributing the majority of the impairment charge to serious accounting improprieties, disclosure failures and outright misrepresentations at Autonomy that occurred before the acquisition, and to the effect of those matters on the expected long-term performance of the business. A large volume of generic commentary makes the specific and important risks harder, not easier, to find.

Event-driven disclosures and the financial press

Required disclosures tied to a specific event can matter a great deal: capital raising, failure to file financial reports on time, changes in management, and mergers and acquisitions. In the United States these are reported to the SEC on Form 8-K, and on an NT filing, meaning notification of inability to file on time, where appropriate. Delays in filing usually come from accounting difficulties, whether internal disagreement over a principle or an estimate, inadequate finance staff, or the discovery of a fraud that needs further examination. An NT filing is highly likely to signal problems with reporting quality.

For public companies in Europe, guidance originally published by the Committee of European Securities Regulators, which has been replaced by the European Securities and Markets Authority, specifies which categories of inside information have to reach the market as they arise. The list runs to a change of control; a change in the composition of the management or supervisory board; a merger, a split or a spinoff; a dispute in the courts; and a newly obtained licence, patent or registered trademark. Companies use the disclosure mechanism specified by their national authority, which in the United Kingdom means an announcement released through an approved regulatory information service.

Each announcement has to be examined on its own facts. The sudden resignation of the most senior financial officer, or of the external auditor, is clearly a warning about reporting quality. A legal dispute over an important asset or product warrants attention because it may damage future earnings. An announcement of a merger or acquisition may point to positive developments and may equally point to a change in the risk profile of the company, particularly while the transaction is in progress.

The financial press is a genuine source when a reporter uncovers a reporting issue nobody had recognised. Jonathan Weil of the Wall Street Journal was among the first to identify the accounting problems at Enron, and at other companies using gain-on-sale accounting, an aggressive policy that allowed immediate revenue recognition on long-term contracts. James Chanos cites an article by Weil as the trigger for his own investigation. The discipline that follows matters as much as the tip. Chanos began by analysing the annual and quarterly SEC filings of Enron, then gathered information on insider stock sales, on the business strategy and tactics of the company, and on the views of stock analysts. An initial idea from a news article always needs confirmation from definitive sources, meaning the regulatory filings, and then support from other sources where they exist. The provenance of the article matters too, and will matter more as electronic media expand: a well-established financial news provider is more likely to be factual than a less-established source, and a story or blog written by a financial journalist is more likely to be unbiased than one written by somebody with a related product or service to sell.

Check yourself

1. A company reports strong revenue growth while its receivables grow considerably faster, and its days of sales outstanding rises for a third consecutive year against a flat industry median. Name three distinct explanations that are consistent with this pattern.

First, customers are being induced to order early through discounts, generous terms or return rights, which pulls future sales into the present. Second, the credit quality of the customer base has deteriorated, so genuine sales are simply not being collected. Third, revenue is partly or wholly fictitious, since revenue booked against a customer that does not exist raises receivables by exactly the same amount. All three produce the identical ratio signature, which is why the pattern is a reason to investigate rather than a conclusion.

2. Two companies both report operating cash flow well above net income. One is capitalising costs that should be expensed; the other is genuinely converting profit into cash. What single further check separates them?

Look at investing activities alongside the operating section. Improperly capitalised costs leave the operating section and reappear as capital expenditure, so the company that is capitalising will show investing outflows and a build-up in gross property, plant and equipment that its strategy does not justify. This is exactly the WorldCom pattern: cash from operating activities exceeded net income in 1999, 2000 and 2001, while gross property, plant and equipment climbed from 31 percent of total assets in 1998 to 47 percent in 2001.

3. An analyst computes a Beneish M-score of −1.60 for a company. Using the cut-off Beneish identified for investors, what does that imply, and what does it not imply?

The cut-off is an M-score of −1.78, corresponding to a probability of manipulation of 3.8 percent, and higher scores mean less negative ones. A score of −1.60 is above the cut-off, so the model flags the company as a likely manipulator and the analyst should treat the reports with particular care. It does not imply that manipulation has occurred. The model establishes association, not cause and effect, its predictive power has been declining as managers learn to test their own reporting against it, and the placement of the cut-off itself depends on the relative cost of Type I and Type II errors.

4. A company takes a large impairment and a large restructuring charge in the same year. Under what condition should the analyst normalise earnings by excluding the charge, and under what condition by spreading it?

Exclude the charge only if the item is genuinely one-off, the financial effect of a natural disaster being the clearest case. Spread it across past periods as well as the current one if similar events happen regularly enough that they belong in permanent earnings, since the underlying deterioration almost certainly occurred over several years rather than in the one year the charge was booked. The judgement rests on knowledge of the facts, and the practical consequence of spreading is a recognition that reported earnings in the earlier years were overstated.

5. An IFRS reporter moves interest paid from operating to financing activities. What happens to its reported cash flows, and how should the analyst respond?

Operating cash flow rises relative to the prior period even if nothing at all has changed in the business, while total cash flow is unchanged. IFRS permits the classification of interest paid as either operating or financing, so the change is legitimate. The analyst should restate the current period on the old basis, or the prior period on the new one, before drawing any conclusion about the trend, and should apply the same treatment when comparing an IFRS reporter with a US GAAP reporter, since US GAAP requires interest paid, interest received and dividends received all to be classified as operating.

6. A company carries goodwill of a size close to its own market capitalisation. Explain the inference an analyst should draw, and why it follows.

If market capitalisation exactly equalled reported goodwill, the market would be implicitly valuing every other asset of the company at zero; if goodwill exceeds market capitalisation, the implied value of the other assets is negative. Neither is a sustainable position, so the reasonable inference is that goodwill is overstated and that an impairment is likely. Sealed Air is the illustration: goodwill of $4,209.6 million at 31 December 2011 against a market value of roughly $3,457 million in December 2011, followed by an impairment of goodwill and other intangible assets of $1,892.3 million in 2012.

7. Why is a clean audit opinion on internal controls weak evidence about reporting risk, and what related fact about the auditor is stronger evidence?

The opinion addresses a historical period and arrives after the analyst has usually learned the same facts elsewhere. Eastman Kodak received an opinion dated 28 February 2012 covering fiscal 2011, six weeks after it filed for bankruptcy on 19 January 2012, and Groupon received a clean opinion on internal controls only in the filing that reported the weakness had already been remediated. The stronger evidence concerns the audit relationship rather than the opinion: repeated changes of auditor, an audit firm whose capabilities look inadequate for the complexity of the client, and anything that raises a question about the independence of the auditor.

8. A company reports the whole of a period profit from a single litigation settlement, measures it correctly and discloses it clearly. Where does it sit on the quality continuum, and why is this case worth remembering?

It sits one step below the top: reporting quality is high, earnings quality is low because the profit is not from an activity the company can repeat. The case is worth remembering because it is the cleanest demonstration that the two attributes are separate, and because it shows what the framework is for. The first question, on compliance and decision usefulness, is answered yes; the second question, on adequate and sustainable returns, is answered no; and it is only because the first answer is yes that the second can be answered at all.

End of lesson.

Back to 04 Financial Statement Analysis