Selling-price estimates in revenue recognition and the usefulness of financial statements

According to studies on accounting frauds and earnings restatements, revenue is among the most frequently misreported items in firms’ financial statements (Dechow et al. 1996; COSO 1999; Palmrose et al. 2004).

Kothari et al. (Kothari et al. 2010, p 282) call for research that would be “useful to standard-setters as they grapple with questions of how to trade-off the different informational and contracting demands of various contracting parties.” This study might interest standard setters, as described later.

The vendor specific objective evidence (VSOE) condition of SOP 97-2 requires firms to establish the selling prices of each component based solely on the firm’s own history of standalone transactions for that component (Sondhi 2006). Recently, EITF 08-1[FASB 2009a]) and ASU 2009-13 [FASB 2009b]) have partially relaxed the objective-price constraint.

For example, firms could shift to real activities manipulation, defined as “a purposeful action to alter reported earnings in a particular direction, which is achieved by changing the timing or structuring of an operation, investment, or financing transaction, and which often has suboptimal business consequences” (Zang 2012, p 676). Examining this aspect, however, is beyond the scope of this study.

Watts and Zimmerman (1990), Healy and Wahlen (1999), Fields et al. (2001), Altamuro et al. (2005), Dechow et al. (2010), Badertscher et al. (2012).

I conduct two “placebo tests” in settings unaffected by SOP 97-2 to ascertain whether my findings are due to patterns unrelated to SOP 97-2 implementation: on software firms in the year before the implementation year and on control firms in the SOP 97-2 implementation year. In neither setting are abnormal deferrals priced as revenues.

For example, Microsoft’s deferred-revenue account reached $17 billion at the end of 2011.

This project was started at the behest of the SEC to address the revenue-recognition issues of multiple-element firms (Turner 2001). This project has been identified as among FASB’s highest priority projects by its advisory council in its previous four surveys (FASAC 2004, 2005, 2006, 2009).

Among the firms that restated their accounts, technology firms were disproportionately represented.

SOP 97-2 builds upon the concepts of earnings process completion and revenue realizability (SFAC No. 5). It requires compliance with four necessary conditions before revenue can be recognized: (1) a formal arrangement, (2) completion of delivery, (3) determinable fees, and (4) revenue realizability.

The criteria used in practice to establish VSOE are stringent. For example, a firm should provide evidence from at least 30 prior randomly selected transactions, and 85 % of such transactions should have been priced within 15 % of the median price (Sondhi 2006).

Typically software firms match variable expenses, such as the cost of goods sold and warranty/maintenance costs, with revenues, but immediately expense R&D, product development, and marketing costs.

Francis and Schipper (1999) interpret earnings informativeness, measured by long-window association tests, as earnings’ ability to capture or summarize information, regardless of source, that affects share values. This interpretation does not require that financial statements be the earliest source of information (Hanlon et al. 2008).

RECCHY and OANCFY are year-to-date variables. Thus, for each of those variables, I subtract the variable by its value in the previous quarter in fiscal quarters two through four. Moreover, I use the negative of RECCHY because a positive (negative) value in the statement of cash flows represents a decrease (increase) in accounts receivable.

I obtain similar results after controlling for firms’ incentives to meet or beat earnings benchmarks (results not tabulated). I control for firm size (Log of Assets [Compustat AT]), Debt (Compustat DLTISY), secondary equity offerings (Compustat SSTKY), and cash balances (Compustat CHEQ).

Altamuro et al. (2005) estimate a regression of changes in stock prices during a short time interval (that is, during a few days surrounding earnings-announcement dates) on seasonally adjusted changes in earnings. I find partial support for H2a using this model (results not tabulated). However, I do not use this model for my main tests for the following reasons. First, a change specification assumes that the investors’ earnings expectations at the beginning of the return window equal the prior-period performance. This is an unrealistic assumption in my setting because, as I show later, software firms in the late 1990s not only grew rapidly, they also varied widely in their growth rates. Second, Ball et al. (2012) caution against drawing inferences exclusively from market reactions around “announcement periods,” because audited financial reporting indirectly affects information released at other times and through other channels. Moreover, the shorter the window used to measure announcement period returns, the higher the “errors” because of the “prices lead earnings” phenomenon (Kothari 2001, p 129). Such errors are especially important in my setting, because in the late 1990s, software firms’ share prices routinely responded to events such as new product launches and strategic partnerships, long before the firms reported their earnings effects.

I obtain similar results after controlling for fixed and interaction effects of the following ERC determinants: persistence, cost of capital (risk-free rate and beta), and firm size (Kormendi and Lipe 1987; Collins and Kothari 1989; Easton and Zmijewski 1989; Freeman 1987; Collins et al. 1987).

I use reported revenues to estimate abnormal revenue deferrals, as there is no better alternative. Prior studies on revenue manipulation (Stubben 2010) also use reported revenues to estimate the magnitude of manipulated revenues. Using this measure also enables me to isolate the component of deferred revenues that is unrelated to reported revenues.

Zhang (2005) ^and Altamuro et al. (2005) examine the implementation effects of SOP 91-1 and SAB 101 by capitalizing on reported cumulative effect adjustments. In contrast to SOP 91-1 and SAB 101, SOP 97-2 was applicable on a prospective basis and specifically prohibited retrospective application. Therefore, firms did not report cumulative effect adjustments upon SOP 97-2 implementation.

However, firms that defer reporting the first type of revenue record no corresponding revenues, assets, or liabilities because no transaction has yet occurred. Consequently, post-implementation financial statements would not reveal this effect of new rules. This effect can, however, be measured from decreases in an asset account called “revenues in excess of billings.” I attempted to hand-collect data on this asset account. I found that this account is economically insignificant, because very few firms report this account.

This SOP was effective for fiscal years beginning after December 15, 1997. Thus, for firms with fiscal years ending in December, the transactions in the year 1998 were recorded using SOP 97-2. Consequently, the SOP 97-2 implementation year for December year-end firms corresponds to Compustat fiscal year 1998. However, for all other firms, the first fiscal year with transactions recorded using SOP 97-2 ends in calendar year 1999. Of those firms, for firms with year-ends June through November, this year corresponds to Compustat fiscal year 1999. Nevertheless, for firms with year-ends January through May, this year corresponds to Compustat fiscal year 1998. I hand-collect data on actual implementation years.

I winsorize at 5 and 95 % levels. These levels are consistent with Zhang (2005), who excludes all regression observations with standard errors greater than two-sigma limits, because software firms display large variations in their characteristics. I obtain qualitatively similar results by winsorizing at 1 and 99 % levels.

As noted in Sect. 5, the average implementation-year stock return for software firms was 44 %. Efendi et al. (2007) show that abnormal increases in equity valuation in the late 1990s increased managers’ personal incentives (for example, to protect the value of their stock-option holdings) as well as firm incentives (for example, to conduct secondary equity offerings at favorable prices) to misreport earnings.

I use panel data with eight observations per firm. This model allows error variances to be correlated across firms and quarters.

The sensitivity of dependent variable to a unit change in independent variable (X_K) in a logistic regression is calculated by the following formula (LeClere 1992, Equation 6, p 771): \( \frac{\partial P}{{\partial X_{K} }} = \frac{{e^{ - (\alpha + \beta \times X)} }}{{(1 + e^{{ - \left( {\alpha + \beta \times X} \right)}} )^{2} }} \times \beta_{K} \)

The Control group for this test is different from the control groups for H1 and H2 tests, because this test requires data on deferred -revenue accounts. Most firms, however, do not report this account. Thus, I identify the control group for H3 tests by first identifying firms that reported deferred revenue accounts in year 2002 when Compustat started reporting details on this data item. I then hand-collect deferred amounts for those firms from their 1997 and 1998 10-K filings. For this test, I relax matching criteria for selecting control firms described in Sect. 5.

For this test, I hand-collect data on quarterly deferred-revenue accounts from software firms’ 10-Q filings. Then, I estimate rules-created deferrals on a quarterly basis by using a similar procedure as in Eq. (5), using similar sized unaffected firms’ same quarter’s average revenue growth as the instrumental variable for RevenueGrowth, and deflating dollar-denominated values by market value of equity.

I find that, by 2009, the number of firms reporting deferred-revenue accounts increased to 2,270, which constitutes 25.9 % of the listed firm population. This finding suggest that the number of firms using multi-element revenue arrangements has increased, consistent with the notion that more and more firms now provide a solution to their customers, as opposed to just selling a standalone product or service (Turner 1999a, b).