The Accounting Onion

Making Revenue Recognition Simple and Informative

One thing that everyone should be able to agree on is that the accounting rules for measuring and reporting revenue ("revenue recognition") are a huge mess. There are too many of them, and far too many inconsistencies amongst them.

One of my favorite examples for illustrating how dysfunctional the rules can be is this illustrative example taken from a 1999 SEC Staff Accounting Bulletin (No. 101):

"A registrant sells a lifetime membership in a health club. After paying a nonrefundable "initiation fee," the customer is permitted to use the health club indefinitely, so long as the customer also pays an additional usage fee each month. The monthly usage fees collected from all customers are adequate to cover the operating costs of the health club. … Question: when should the revenue relating to nonrefundable, up-front fees in [this type of an arrangement] be recognized?"

If all you knew about accounting was that "assets" are what you have and "liabilities" are what you owe, then you should respond that the revenue should be recognized "when cash is received"; the health club has received an asset (from a non-owner), and it has no obligation to return cash at a later date, or even to provide services, under any circumstances.

The SEC, however, enunciated a different point of view in SAB 101, which can now be found in Topic 13 of the SAB Codification:

"Unless the up-front fee is in exchange for products delivered or services performed that represent the culmination of a separate earnings process, the deferral of revenue is appropriate. … [T]he staff does not view the activities completed by the [registrant (i.e., selling the membership) as a discrete earnings event.]"

The implication of this illustration is that revenue recognition under GAAP involves more than just the receipt of assets that you don't have to pay back; an implicit performance obligation to provide the customer with future services is deemed to exist, even though the customer has no legally enforceable right to receive the services.

In other words, current revenue recognition practices are dependent on factors that don't play into whether or not the definition of an asset or liability has been met for recognition, its subsequent measurement, or conditions for its derecognition. This disconnect of revenue recognition from the balance sheet, has led to a jagged path of positions hatched industry-by-industry, over decades of political wrangling with issuers and auditors. The resulting complexities, subtleties and inconsistencies within the revenue recognition rules are, at their very best, second only to the tangled web that is financial instruments accounting.

The Epic Revenue Recognition Project

The FASB/IASB joint project on revenue recognition seems to have been initiated as a good faith attempt by them to recapture the high ground from the SEC post-SAB 101 on a topic that arguably is the most consequential to the most financial statement issuers. After seven years (yes, the project was proposed back in January 2002) of consideration, the Boards' position has evolved essentially to these major positions as expressed in a discussion paper for which the comment deadline expired six months ago:

• The scope is limited to arrangements whereby an entity exchanges promises with a customer, leading to rights and obligations.  
• Increases in rights or decreases in obligations to customers are recognized as revenue.
• Performance obligations are to be initially measured by the transaction price, or an allocation of the arrangement consideration; subsequent measurement is still unsettled, but is likely to be a function of the initial measurement (as opposed to a market-based valuation).

The good news is that a new standard would supplant all extant revenue recognition guidance, and become the sole source of guidance on revenue recognition. The bad news is that it won't change much else. I see three major flaws.

First, revenue recognition should be nothing more than a question of income statement classification. If the FASB were fully committed to a balance sheet approach to financial reporting, it would plainly acknowledge and implement the principle that net income is merely a number derived from the changes in assets and liabilities during a period that resulted from transactions with non-owners. Hence, the only revenue recognition problem is determining which of those changes should be reported on the income statement as "revenues."   As it is invariably the case with both Boards, a piecemeal approach is being taken to avoid the larger contentious debate of accounting for rights and obligations in general.

Second, nothing worthwhile can come from including constructive performance obligations in the conceptual definition of a liability. The health club initiation fee example is the most straightforward example I can think of to illustrate this. The timing of services provided will cease to be a criteria, but the Boards are not willing to throw out the notion of a constructive obligation from their respective conceptual frameworks:

 "Sometimes an entity establishes a practice of providng particular goods or services, such as a warranty service.  Even if neither the contract nor the law explicity requires such a service, the entity by its customary business practice may have implicitly or constructively created an obligation that would be enforceable." (para. 3.6]

An "enforceable" constructive obligation?  I don't know what that means.  I imagine that what they are trying to say, but too embarassed to say directly is this: even though no legal obligation exists, an entity would harm itself by more than it would benefit if it just took a customer's money and ran.  Thus, economic forces make a constructive obligation "enforceable" through some sort of market mechanism.

I might be less cynical about the reasons for wanting to recognize constructive obligations if we actually had effective financial reporting gatekeepers. How many auditors will call a potential constructive obligation the way they see it – as opposed to the way management sees it? Investors don't need that kind of financial reporting any more, and their returns don't need to be burdened with the considerable costs of amortizing upfront fees over some fuzzy measure of the expected relationship period. 

Third, measurement of customer-related assets and obligations by allocating arrangement consideration to them is a step backwards from the FASB's putative goal of measuring financial assets and liabilities at their fair values. I presume that the FASB is legitimately concerned about introducing more subjectivity into measurement of assets and liabilities; and more fundamentally, they are rightly concerned that exit values for liabilities will inevitably generate Day 1 gains for vendors.

I can't say that there is a sure-fire solution to fair value measurement's subjectivity, for that will always be present in any approach to current valuation.  But, there is a straightforward way around the problem of Day 1 gains. Consider, for example, a one-year fire insurance policy with a $1,000 up-front premium. If an exit value approach were taken to measure the performance obligation on Day 1 (i.e., the obligation to stand ready to pay for the insured's loss should a fire occur), it would likely cost the insurance company less than $1,000 to cede the liability to a reinsurer; hence, the Day 1 gain the Boards are worried about.

But, here is, yet one more time, the compelling economic logic of replacement cost measurements. As I have stated in numerous prior posts, a liability is most appropriately measured by the replacement cost of the corresponding insurance asset in the possession of the policyholder. That number would, by definition, still be $1,000 on Day 1. Presto change-o, the problem of Day 1 gains disappears!

Summing Up

Revenue recognition is arguably the most significant project the FASB and IASB can undertake. But momentum, such as it was, appears to have stalled following the issuance a full year ago of that discussion paper purporting to contain the Boards' "preliminary views."  I believe this is because the standards setters keep bumping up against inevitable logical inconsistencies. First and foremost, is that the predilection for exit values is birthing a frightening brood of financial reporting anomalies. Here are five that come to mind without thinking about it for more than 30 seconds:

• Day 1 expensing of transaction costs when acquiring financial assets or applying the "acquisition method" to a business combination;
• Assigning an exit value to a leased asset that, by contract, cannot be transferred.
• Recording gains on the completion of a manufactured good before it is sold (specifically scoped out of the discussion paper).
• Liabilities that cannot be transferred.
• Asset markets that are either illiquid, or temporarily taken over by distressed sellers. 

I have been told by someone with access to inside information that FASB chair Bob Herz is a replacement cost guy deep down; but, he can't seem to muster the political momentum to alter course away from exit prices. The way I see it, if we continue to reject replacement cost, we are doomed to chasing our tails: (1) try exit values; (2) become discouraged as the logical inconsistencies destroy representational faithfulness; (3) give up and settle for reporting computer-generated random numbers seeded by an historic cost.

As significant as the revenue project is, a reasoned solution is being lost by a destructive problem solving strategy that the late Nobel laureate Herbert Simon and his co-author Allen Newell called "means-ends analysis"  in their 1972 treatise, Human Problem Solving. The danger is that dogged "forward progress" on a portion of a problem may actually set back the solution of the larger problem to which it is apart. After eight years of hemming and hawing, the FASB surely must realize that its approach is clearly not working. Rather than continue to propose partial fixes to flawed models, it would be better off postponing the project until it establishes more pragmatic definitions and measurements for assets and liabilities.   

If it could only solve the larger problem, the problem of revenue recognition would practically solve itself.

Going to School on Revenue Recognition

I'm a night owl, but once I hit the sack, I'm out like a light for 8-9 hours. In fact, the two things I would say that I do best are type fast and sleep like a log. One recent night was a rare exception, though. I woke up only about two hours into my hibernation and couldn't fall back asleep. After about another hour, I gave up. It was too late to have a toodle, so I decamped to my office and turned on the computer.

The first thing on the web to catch my eye was a blurb in the Chronicle of Higher Education, in which it was reported that the revenue recognition policies of the Apollo Group Inc., the parent company of the University of Phoenix, were the subject of an "informal inquiry" by the SEC's Division of Enforcement. Apollo has declined to provide any further specifics, but their share price declined about 18% around the time of the announcement. Hmm.

I decided to look further into this for three reasons: (1) I thought it might help me get to sleep; (2) I was in the midst of preparing to lead a one-day workshop on revenue recognition, so I could actually benefit by a review of some of the rules; and (3) Apollo's headquarters are in Phoenix, where I live.

What I Found – Before I Went Back to Bed

The first thing I did was to download Apollo's most recent 10-K and to read their description of critical accounting policies on revenue recognition. I also pulled 15 years worth of financial statements in spreadsheet format from a data service.

Here is what I found after a few minutes of perusal:

• Students are billed on a course-by-course basis. But, judging by the ratio of the allowance for doubtful accounts to gross student accounts receivable, 29%, it appears that a significant number of student accounts are eventually written off as uncollectible.
• Upon the first day of attendance, Apollo records a receivable and deferred revenue in the amount of the billing. As I will explain later, I was surprised to learn this; 'executory contracts' are usually not recognized (with the notable exception of capital lease accounting).
• Tuition revenue is recognized pro rata over the duration of the course, which is generally 6 - 9 weeks. As we say in the trade, it appears that Apollo has adopted a 'proportional performance' revenue recognition model. A more conservative choice would be a 'completed performance' model.
• Apollo recently changed its refund policy whereby students who attend 60% or less of a course are eligible for a refund for the portion of the course they did not attend.
• Apollo prepared its statement of cash flows under the direct method through 1997. They switched to the indirect method in 1998. I would have thought that a 'preferability letter' for such a change would have been included in the 1998 10-K, but I was unable to locate one.

What Could the SEC Be Looking At?

It appears that the SEC Enforcement Division received a referral from the Division of Corporation Finance, which reviewed Apollo's most recently filed 10-K, issued a comment letter, and received a reply from Apollo. Corp Fin's comments addressed, among other things, Apollo's revenue recognition policy for refunds, and whether bad debt expense and revenue were both overstated (i.e., certain amounts of bad debt expense should have been treated as reductions in revenue).

My own questions start at a much more basic level than Corp Fin's comments: when, if ever, it would be appropriate for Apollo to recognize revenue prior to the receipt of payment?

Accounting for the Students Who Pay in Arrears

The general rule in GAAP is that revenue cannot be recognized until it is earned, and realized or realizable (see Statement of Financial Accounting Concepts No. 5). The SEC staff has interpreted this general rule in Topic 13 of the Codification of Staff Accounting Bulletins (SAB Topic 13) to mean that four criteria must be met in order for revenue to be recognized. I won't go into all of them, but the last criteria is that "collectibility is reasonably assured." If the probability is 29% that a student won't pay Apollo, can it be said that collectibility is reasonably assured?

Apollo and its auditors might respond by stating that 71% of the billings to students are reasonably assured; moreover, Apollo has accumulated an extensive history of course delivery that enables them to reliably estimate the allowance at 29%.

But, there is no language in SAB Topic 13 that specifically allows Apollo to combine similar arrangements for the purpose of determining whether collectability is reasonably assured. SAB Topic 13 does provides that one can estimate future liabilities for warranties and returns by aggregating similar customer arrangements and estimating an average for the group; however, it did not specifically provide that those procedures were available for non-payments of enforceable claims. Notwithstanding, historic experience may not be all that helpful in determining the non-payment rate in the current economic environment. For companies with low non-payment rates, maybe, but for companies with a 29% payment rate, perhaps not.

I suppose it would have been clearer if SAB Topic 13 stated what was to be accomplished by providing that collectibility must be reasonably assured; and/or had specifically prohibited combining accounts when evaluating the criteria. Nonetheless, the following example may serve to illuminate the SEC's intent.

Take two companies, A and B; they are equally profitable and differ principally in collectibility rate of accounts receivable. Company A estimates its allowance for doubtful accounts to be 2% of gross accounts receivable, and B's allowance is 30%. Both companies discover, after the fact, that the real allowance was only two-thirds of what it should have been: that is, 3% for A and 45% for B. The premature recognition of earnings by Company A may or may not be material, but for B, it will be as cataclysmic to the income statement as was the AZ Cardinals loss last Sunday on a final-play touchdown pass to my son. It could be permissible to estimate an allowance for doubtful accounts for a group of similar arrangements, but it does not seem appropriate to determine that collectibility is reasonably assured on the same basis.

When is Revenue from a Course Earned?

While collectability may be an issue for some arrangements with students, many of Apollo's students pay in advance. The comment in this section apply to all arrangements, regardless of the timing and/or uncertainty of cash flows.

SAB Topic 13 provides that revenue should not be recognized for "delivered elements" (i.e., classes in Apollo's vernacular) if remaining elements to be delivered to the customer are "…essential to the functionality of the delivered … services." The staff created an exception to this rule for undelivered elements that are "inconsequential" or "perfunctory," but it is not applicable if failure to complete the activities would result in the customer receiving a full or partial refund … (or a right to a refund…)." Stated from a balance sheet perspective, the underlying principle is that one is generally precluded from recognizing a receivable that is not backed by an enforceable right to collect it.

I am unfamiliar with the way courses are conducted by Apollo, but I assume that they all end with an evaluation leading to a final grade. In my experience, students will not pay for a course that doesn't provide them with a grade. Grading is an essential function of the service provided; therefore, it would seem that SEC guidance would require Apollo to defer revenue related to a course until a grade is given to the student.

But, to be fair, I did check the revenue recognition policies of a number of other public companies in the same industry as Apollo, and they all recognize revenue in some ratable fashion as courses progress. For me, that is just one more reason why the SEC's investigation of Apollo's revenue recognition practice is significant. It could better the practices of an entire industry.  (And, by the way, numerous competitors have non-collectibility rates that are roughly the same as Apollo's.)

Accounting for Students Who Drop a Course

The question that the SEC seems to be homing in on is whether Apollo has properly allowed for refunds to students who may drop the course before the 60% point. That also happens to be the only revenue recognition issue that analysts were asking about in Apollo's fourth quarter earnings conference call.

It could be that the analysts and the SEC are both missing the boat. The SEC may believe that Apollo should allocate a portion of the deferred revenue to an estimated liability for refunds. That would initially affect balance sheet classification of liabilities, and it may affect the pattern by which revenue hits the income statement; but it doesn't seem to be that big a deal to me. Yet, it must be said that Apollo's stock price did take an 18% hit around the time of the announcement of the SEC investigation.  If the accounting for the new refund policy is the reason for the stock price drop, then so be it.   

Other Red Flags

I have two final thoughts regarding items that I noticed in my relatively brief perusal of the financial statements. First, even though the ratio of the allowance for doubtful accounts to accounts receivable is around 30%, the ratio of the allowance to receivables for which Apollo actually has enforceable rights could be significantly higher.

That's because Apollo has a practice of recognizing receivables for which it has no enforceable rights. Recall that Apollo recognizes a receivable and deferred revenue for the price of a course when the student shows up for the first day of class. I suppose Apollo is reasoning that both parties have gone down the road somewhat, but it pretty much looks like an executory contract to me. Be that as it may, the SEC should be asking whether the allowance for doubtful accounts is based on the total reported balance of accounts receivable, or just the portion representing enforceable rights. It makes no sense to me to create an allowance for doubtful accounts (and an offset to bad debt expense) on a 'receivable' that is not owed, and may never become owed if the student drops the course. If Apollo sees it the same way, the ratio of doubtful accounts to enforceable student receivables could be significantly higher than even the 29% reported.

Second, regarding the change in the method of presenting cash flows, it would be a pretty big stretch for an auditor to maintain that the switch in accounting was to a preferable method; the FASB has stated in SFAS 95 that the direct method is the approach they encourage issuers to employ as "the more comprehensive and presumably useful." (para. 119) I sure would like to see the SEC ask Apollo and their auditors about that one.

And, perhaps, here is a Hanukah present to the FASB: Apollo could be their poster child for why the direct method for presentation of cash flows should be required. Would 18% of total shareholder value have been destroyed in one fell swoop, had Apollo reported cash flows to investors using the direct method? Perhaps not, because trends in the amount of cash collected from customers would have been disclosed.  The direct method of presenting the statement of cash flows reduces the criticality that investors accurately evaluate the quality of an issuer's revenue recognition policies.  

Winding Up

My goal for this posting was simply to raise interesting questions about Apollo's revenue recognition policies. I want to explicitly state that my intention is not to pass judgment on any of Apollo's choices, even though the market may have spoken to that effect by devaluing Apollo's shares.

Indeed, there are many more questions suggested by this case, and they go beyond the specific effects on Apollo. For example, one could consider whether the revenue recognition rules applicable to Apollo's arrangements with students are themselves representationally faithful or appropriate. We might also ask whether a different result would obtain if the revenue recognition rules under IFRS were applied. Finally, and perhaps most interesting, we could ask how the revenue recognition project being undertaken by the FASB and the IASB jointly has the potential to improve the quality of financial reporting by companies like Apollo. Unfortunately, I am not confident that the proposed approach would be an improvement, but that's for another post.