Financial Statement Presentation and Nonprofessional Investors’ Interpretation of Fair Value Information
Shana M. Clor-Proell*
[email protected] (608) 262-8602
Terry D. Warfield
[email protected] (608) 262-1028
July 2009
Wisconsin School of Business University of Wisconsin - Madison Madison, WI 53706
We thank Sarah Bonner, Larry Brown, Damon Fleming, Stacy Hawkins, Karla Johnstone, Brian Mayhew, Mark Nelson, Chad Proell, Kristina Rennekamp, Nick Seybert, Hollis Skaife, Dan Wangerin, and workshop participants at Boston College, Georgia State University, San Diego State University, the University of Southern California, and the University of Wisconsin Madison for helpful comments. We thank the Wisconsin School of Business at the University of Wisconsin – Madison for its generous financial support. Terry Warfield acknowledges the financial support of the Arthur Andersen Center. We thank Matt Junemann and Kristina Rennekamp for their research assistance. *Corresponding author.
Financial Statement Presentation and Nonprofessional Investors’ Interpretation of Fair Value Information ABSTRACT This research investigates how the Financial Accounting Standards Board’s proposed changes to the presentation of the financial statements affect nonprofessional investors’ interpretation of fair value information. Based on prior accounting and psychology research, we predict and find that when the financial statements contain an additional column that highlights the effect of changes in fair value on net income, nonprofessional investors better integrate differences in the reliability of fair value information into their earnings growth assessments and P/E judgments. Additional analyses examine the intentionality of participants’ responses and examine how investors interpret information about fair value losses. Our results are relevant to both the debate about fair value accounting and to the Financial Accounting Standards Board’s financial statement presentation project. The results also have implications for regulators concerned with the judgments of nonprofessional investors. Keywords: Financial statement presentation, fair value measurement, heuristic-systematic model, nonprofessional investors. Data Availability: Contact the authors.
I. INTRODUCTION The Financial Accounting Standards Board (FASB) instituted Statement of Financial Accounting Standards (SFAS) No. 157, Fair Value Measurements, in 2006 to provide a single definition of fair value, to establish a framework for measuring fair value, and to expand disclosure about fair value measurements in the financial statements (FASB 2006). One of the goals of this standard is to provide enhanced information to financial statement users about the relevance, reliability, and risk inherent in fair value measurements and their potential effects on the persistence of net income. However, to date, little academic research has examined whether SFAS No. 157 is achieving this goal. Although, standard setters have recently expressed an interest in research that addresses this question (e.g., Barth 2008), it remains uncertain as to whether the additional disclosures mandated by SFAS No. 157 are useful in terms of helping users assess differences in fair value measurements. Accordingly, we conduct a controlled experiment using Masters of Business Administration (MBA) students to examine how nonprofessional investors interpret these disclosures. We also examine an additional factor that may affect the extent to which users incorporate fair value information into their judgments. Currently, the FASB and IASB are engaged in a joint project on financial statement presentation that is designed to improve the usefulness of the financial statements. As part of this project, the FASB/IASB has proposed the introduction of a new reconciliation schedule that, among other things, would highlight to users how recurring fair value changes affect net income (FASB 2008a). Thus, our research holds constant the presence of recurring fair value changes and examines how highlighting this information in the proposed reconciliation schedule affects nonprofessional investors’ interpretation of fair value information. Further, our research speaks to recent calls for ex ante
1
research on changes proposed as part of the Financial Statement Presentation project (Linsmeier 2007). Based on the framework in Maines and McDaniel (2000), we predict that nonprofessional investors’ judgments will more heavily weight differences in the fair value measurement inputs under the proposed presentation format than under the current presentation format because the proposed presentation format more clearly isolates and labels fair value information. Further, we link the dimensions of the Maines and McDaniel (2000) framework to prior psychology research on the heuristic-systematic model (HSM) to predict that nonprofessional investors will employ more systematic judgment processes when presented with the proposed presentation format than when presented with the current presentation format. We test our predictions by conducting an experiment using MBA students who estimate future earnings growth based on information about a hypothetical firm. The experiment utilizes a 2 x 2 between-subjects, pretest-posttest design that varies financial statement presentation (current vs. proposed) and the reliability of inputs used to value gains associated with the firm’s trading securities (high vs. low). All participants receive background information, an income statement, and selected footnotes. In the current treatment, all revenue and expense items appear in a single column. This is generally consistent with current income statement format. The income statement provided in the proposed treatment differs from that provided in the current treatment in that it contains an additional column that highlights how changes in fair value affect net income. In the high reliability treatment, footnote disclosures indicate that gains from changes in the fair value of the firm’s trading securities are based on more reliable inputs (i.e., observable market prices). In the low reliability treatment, footnote disclosures indicate that gains are based on less reliable inputs (i.e., model-derived valuations).
2
The results of our experiment show that, under the current presentation format, nonprofessional investors do not attribute stronger potential for future earnings growth to a firm with gains based on more reliable inputs than for a firm with gains based on less reliable inputs. These results indicate that the fair value disclosures mandated by SFAS No. 157 may not be effective in the current environment because users are not assessing the reliability differences that SFAS No. 157 intends to convey. In contrast, we find that, under the proposed presentation format, nonprofessional investors do assess stronger potential for future earnings growth for a firm with gains based on more reliable inputs than for a firm with gains based on less reliable inputs. This suggests that the FASB’s proposed changes to the financial statements can affect users’ reliance on the mandated fair value disclosures. Additional analyses reveal that participants in the proposed presentation format employ a more systematic judgment process than do participants in the current presentation format. Thus, participants in the proposed presentation format are more likely to think about differences in the reliability of fair value inputs when making their assessments of future earnings growth. Further, we find that differential assessments of future earnings growth affect the P/E multiple that users ultimately assign to the firm. We also use a within-subjects debriefing question to examine the intentionality of users’ judgments. We find that, although users think that gains based on less reliable inputs are less relevant and more risky than gains based on more reliable inputs, they do not use this information when making an assessment of earnings growth. Finally, we use a within-subjects debriefing question to examine how users interpret fair value losses based on more reliable inputs relative to fair value losses based on less reliable inputs. We find that, in contrast to differences in the reliability of inputs for fair value gains, differences in the reliability of inputs
3
for fair value losses do not affect users’ assessments of the firm’s potential for future earnings growth. Our research contributes to the extant accounting literature that has examined factors affecting the use of additional disclosures (e.g., Clor-Proell 2009; Hodge et al. 2004). Specifically, we find that the presentation format within the financial statements can impact the extent to which nonprofessional investors utilize information disclosed in the footnotes. This finding is also relevant to prior accounting literature that has examined factors that affect financial reporting transparency (e.g., Hunton et al. 2006) in that it demonstrates how the structure of the financial statements affects users’ willingness to extract information about the reliability of estimates underlying reported numbers. We also contribute to theory about presentation format effects in accounting by extending the Maines and McDaniel (2000) framework. While the framework specifies factors that affect users’ information weighting, it does not specify how differences in weighting occur. By linking the Maines and McDaniel (2000) framework to the heuristic-systematic model from psychology, we are able to provide evidence that greater weighting of information occurs via more systematic judgment processes. This has implications for future research on format effects in accounting. Our research also has implications for standard setters and regulators concerned with the judgments of nonprofessional investors. Specifically, our results demonstrate how two distinct financial accounting issues, the implementation of SFAS No. 157 and the financial statement presentation project, actually depend critically on one another. One goal of SFAS No. 157 is to increase the disclosures associated with the use of fair value accounting. However, our results indicate that nonprofessional investors do not incorporate this additional information into their judgments under the current presentation of the financial statements.
4
Instead, it may be the case that some of the changes proposed as part of the FASB’s financial statement presentation project are necessary to highlight to users the importance of SFAS No. 157 disclosures for their judgments. These results are especially important given the recent criticism of fair value measurements and calls for suspending implementation of fair value in some contexts (Goldman Sachs 2008). Furthermore, the results of our within- versus between-subjects analysis indicate that the mandated SFAS No. 157 disclosures are not communicating information about risk in a manner that is useful for nonprofessional users’ judgments. Thus, standard setters may want to examine this issue more closely (also see Lo and Runkle (2009)). The next section provides background information and describes our predictions. The third section describes the experiment used to test our predictions. In section IV we present our results and in section V we discuss the implications of our research. II. BACKGROUND AND HYPOTHESES Fair Value Disclosures The FASB has called for increased use of fair value information in the financial statements as a way to provide more relevant and understandable information to financial statement users and has concluded that fair values are preferable to cost-based measures for financial assets and liabilities (FASB 2007). However, the FASB has also noted that GAAP contains a number of inconsistencies with respect to the way that fair value is defined, measured, and reported in the financial statements. To eliminate these inconsistencies, the FASB issued SFAS No. 157, Fair Value Measurements, in September 2006 (FASB 2006). SFAS No. 157
5
provides a single definition of fair value, 1 establishes a framework for measuring fair value, and expands disclosures about fair value measurements. The framework for reporting and measuring fair value is based on observable and unobservable inputs which are organized into a three-level hierarchy. Level 1 inputs are directly observable, and measure fair value based on quoted prices in active markets for identical assets or liabilities (FASB 2006). For example, a price quote from the New York Stock Exchange would constitute a Level 1 input. Level 2 inputs are indirectly, rather than directly, observable. For example, a price quote for a similar asset would constitute a Level 2 input. When observable inputs are not available, fair value is determined based on Level 3 inputs. Level 3 inputs are unobservable, and measure fair value using the entity’s own assumptions about how market participants would value the asset or liability. For example, pricing stock options would involve Level 3 inputs. The FASB places priority on the use of Level 1 inputs when possible, but also acknowledges that in certain circumstances Level 2 or 3 inputs may be more appropriate (FASB 2008b). SFAS No. 157 mandates that firms disclose the input level used for all assets and liabilities measured at fair value. In addition, because Level 3 measurements involve greater managerial discretion, firms are required to provide a reconciliation of beginning and ending balances for all recurring Level 3 items. This provides information to users about the extent to which Level 3 gains and losses are reported in income. Appendix A provides an example of the disclosure mandated by SFAS No. 157. One of the goals of SFAS No. 157 is to improve the transparency associated with the use of fair value measurements in the financial statements. The FASB argues that disclosing the
1
SFAS No. 157 defines fair value as “the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between market participants at the measurement date.” 6
input level associated with assets and liabilities measured at fair value provides information to users about the reliability, risk, and relevance of the measurement (FASB 2006). However, to date there is no direct evidence that nonprofessional investors consider reliability, risk, and relevance when evaluating SFAS No. 157 disclosures. 2 Thus, one purpose of our study is to provide evidence on the extent to which the mandated disclosures achieve the intended goal. Further, our study is designed to examine how reliability, risk, and relevance judgments can affect users’ subsequent assessments of the firm’s earnings growth. Along these lines, Figure 1 depicts the relationships among these variables. By their design, Level 1 inputs are more representative of the underlying economics of assets or liabilities. Thus, as indicated in the figure, Level 1 inputs are viewed as more relevant and reliable than Level 3 inputs (Links 1 and 2). 3 This suggests that Level 1 gains in the current period are viewed as relatively more sustainable and, therefore, are relatively more predictive of future performance than are Level 3 gains. Thus, Level 1 gains in earnings will lead to higher assessments of future earnings growth than will Level 3 gains in earnings (Links 4 and 5). Similarly, because Level 1 inputs are publicly observable and objectively determined, they contain less information risk than do Level 3 inputs that are based on private information (Link 3). Thus, assessments of future earnings growth will exhibit a lower risk discount when based on Level 1 gains than when based on Level 3 gains (Link 6). Insert Figure 1 Here Alternatively, users may not perceive there to be differences between measurements based on Level 1 versus Level 3 inputs. Because this information is disclosed in the footnotes, 2
A recent working paper by Song et al. (2009) examines the value relevance of fair value hierarchy information and argues that Level 1 and Level 2 fair value measurements are more value relevant than Level 3 fair value measurements. 3 The relevance link may not be unambiguously positive. In cases where observable market prices are not available, Level 3 information may be considered equally relevant by users. 7
users may be less likely to rely on it because they perceive it to be less useful (SFAC No. 5, 1984) and/or more likely to contain errors (Libby et al. 2006). To the extent that users fail to incorporate differences between Level 1 and Level 3 measurements, this is problematic for standard setters because it suggests that their attempt to increase the transparency associated with the use of fair value estimates may not be effective. Accordingly, it is important to understand what other factors may increase the likelihood that the additional disclosures mandated by SFAS No. 157 are having their intended effect. Recently proposed changes to the presentation of the financial statements may hold one answer to this question. Financial Statement Presentation Project In 2004 the Financial Accounting Standards Board (FASB) began joint work with the International Accounting Standards Board (IASB) on a financial statement presentation project intended to improve the usefulness of the financial statements (FASB 2008a). As described in the recently issued discussion paper, many of the proposed changes are intended to improve the cohesiveness of the financial statements by reorganizing the content of the Statement of Financial Position, Statement of Comprehensive Income, and Statement of Cash Flows into four distinct categories that group the information within each statement based on major business activities (FASB 2008a). One of the most important proposed changes to the financial statements is the introduction of a new reconciliation schedule that reconciles cash flows to comprehensive income. This schedule disaggregates the components of income by showing in separate columns changes in cash, accruals, recurring fair value changes, and other nonrecurring remeasurements. The intended purpose of this schedule is to help users understand the components of comprehensive income that differ in terms of their persistence and measurement subjectivity
8
(FASB 2008a). Because one column of this schedule highlights fair value changes, it seems reasonable to conclude that this column may serve to complement additional fair value disclosures mandated by SFAS No. 157. Indeed, because the proposed change is in effect a format change, we turn to prior accounting research on format effects to understand how this change will affect nonprofessional users’ judgments. Prior Research on Presentation Format Prior research on accounting presentation format has investigated a number of reporting contexts and has generally found that both sophisticated and unsophisticated financial statement users are affected by differences in financial statement format (e.g., Hirst and Hopkins 1998; Maines and McDaniel 2000). Some of the prior research in this area has examined how presentation format can affect information acquisition and has found that format can affect information acquisition for both professional (Hirst and Hopkins 1998) and nonprofessional users (e.g., Hodge et al. 2004). Much of the prior research in this area has also examined how differences in presentation format affect information weighting (e.g., Maines and McDaniel 2000). By creating simplified decision contexts for participants, prior experimental research has been able to examine how differences in judgment can persist even after information acquisition has occurred (e.g., Hirst and Hopkins 1998; Hirst et al. 2004; Hodge et al. 2004; Hopkins et al. 2000; Maines and McDaniel 2000). Thus, research that has examined information weighting provides evidence about some of the more subtle issues that arise due to differences in presentation format. Consistent with this prior research, we create a simplified setting in which information acquisition is likely to occur so that we can investigate whether differences in presentation
9
format affect the weight that nonprofessional investors place on fair value information. 4 Specifically, we rely on the framework proposed by Maines and McDaniel (2000) to predict that fair value information will be weighted more heavily under the proposed presentation format than under the current presentation format. The Maines and McDaniel (2000) framework is comprised of five dimensions of financial statement format that can affect users’ judgments because each dimension either provides a signal about the importance of the information or affects the cognitive costs associated with processing the information. Three of the five dimensions provide a signal about information importance. The first signaling dimension, information placement in the financial statements, conveys information importance by placing it in a performance statement (e.g., the income statement) rather than in a non-performance statement (e.g., the statement of stockholders’ equity). The second signaling dimension, information labeling, conveys information importance by assigning a clear label to the information (e.g., income). The third signaling dimension, linkage, conveys information importance creating a direct association between two items (e.g., placing two items in the same category on the financial statements). The remaining two dimensions in the Maines and McDaniel (2000) framework affect the cognitive cost of information processing. The first cognitive cost dimension, information isolation, eases information processing by separating certain types of information from other types of information (e.g., comprehensive income is easiest to process when it is in its own separate statement). The second cognitive cost dimension, information aggregation, eases
4
We acknowledge than in more complex settings, acquisition of SFAS No. 157 disclosures will be a greater concern. However, we argue that providing an empirical test to confirm this assertion is unnecessary. Rather, the more substantial question of interest is to understand whether efforts to improve acquisition will eliminate errors in judgment. By providing evidence about information weighting, our research is able to address this question. Specifically, we provide evidence that, even if acquisition concerns are eliminated, judgmental errors will still persist under the current presentation format. 10
information processing by disaggregating the information (e.g., presenting gross changes rather than net changes). The proposed change to the financial statements that we focus on relates to two of the five dimensions described above. Specifically, we focus on the proposed reconciliation schedule which creates a separate column for changes in the fair value of assets and liabilities. By creating a separate column, changes in fair value that affect net income are explicitly labeled for the user, thus signaling the importance of this information. In addition, creating a separate column decreases the cost of processing this information by isolating it for the user. Thus, we predict that nonprofessional investors will more heavily weight gains associated with changes in fair value under the proposed format than under the current format. Because they are attending to this information more, they will be more sensitive to the inputs used to determine fair value and will be more likely to assess differences in reliability, risk, and relevance under the proposed presentation format than under the current presentation format. Further, we expect that these differential assessments will lead to differential assessments of future earnings growth. We restate this reasoning in the following hypothesis: H1: Nonprofessional investors exhibit a greater difference in earnings growth assessments for Level 1 compared to Level 3 gains when provided with the proposed presentation format than when provided with the current presentation format. While the weighting differences that we hypothesize in H1 above are commonly examined in the prior accounting literature, examination of the process by which these weighting differences occur has not been studied. Thus, one of the significant contributions of our research is to examine the weighting process by relying on similarities between the Maines and McDaniel (2000) framework and prior psychology literature. Specifically, the signaling and cognitive cost dimensions of the Maines and McDaniel (2000) framework dovetail with the critical dimensions
11
of the Heuristic-Systematic Model (HSM) used in prior psychology literature. Thus, we now turn to a discussion of prior research on the HSM, which can provide insights into the processes that underlie H1 above. Prior Research on the Heuristic-Systematic Model (HSM) The HSM is a theoretical model which posits that individuals use two different methods to process information – a heuristic method and a systematic method. The extent to which information processing is heuristic or systematic varies along a continuum, with the possibility that the two methods can co-occur. When a more heuristic method is used, judgments are based on simple decision rules that do not require much cognitive effort. In contrast, when a more systematic method is used, greater cognitive effort is expended to process the available information. The model posits that individuals must have both sufficient motivation and ability to engage in more systematic processing. If motivation and/or ability are lacking, then individuals will rely more on simple decision rules (heuristics) to process information (Eagly and Chaiken 1993). 5 With respect to relationship between the Maines and McDaniel (2000) framework and the HSM, we posit that the signaling dimensions of the Maines and McDaniel (2000) framework correspond to the motivational dimensions of the HSM. Similarly, we argue that the cognitive cost dimensions of the Maines and McDaniel (2000) framework correspond to the ability dimensions of the HSM. Thus, format differences that affect the signaled importance of the information can affect individuals’ motivation to process the information, while format differences that affect the cognitive costs associated with the information can affect individuals’ ability to process information. Because the proposed presentation format that we examine differs 5
The Elaboration Likelihood Model (ELM) is an alternative model that has been used in prior psychology research. The HSM and the ELM are closely related in many respects. Eagly and Chaiken (1993) provide a synthesis of the similarities and differences between the HSM and ELM. 12
from the current presentation format along both a signaling dimension and a cognitive cost dimension, we expect the proposed presentation format to affect both motivation and ability to engage in systematic processing. Thus, we rely on the HSM to predict that individuals who view the income statement in the proposed format will engage in more systematic processing than individuals who view the income statement in the current format. To the extent that individuals in the proposed format conditions use more systematic processing, we would expect that they would take longer to complete the task than individuals in the current format conditions. We restate this reasoning in the following hypothesis: H2: Nonprofessional investors provided with the proposed presentation format spend more time on the judgment task than do nonprofessional investors provided with the current presentation format. Further, we expect that individuals who use a more systematic process will be more likely to think about the implications of input levels for their judgments than will participants who use a more heuristic process. Thus, we expect that the links 1, 2, and 3 specified in Figure 1 are more likely to be significant for individuals in the proposed format conditions than for individuals in the current format conditions. We restate this reasoning in the following hypothesis: H3: Nonprofessional investors are more likely to assess differences in the relevance, risk, and reliability of fair value inputs in the proposed format conditions than in the current format conditions. Below we describe the experiment used to test our predictions. III. METHOD Participants The participants in our experiment are 59 Masters of Business Administration (MBA) students from two universities. Participation occurred during October and November of 2008.
13
As compensation for their participation, the MBA students were each paid $10 for their time and each group of participants was entered in a drawing for one of two $100 gift certificates. On average, participants had 3.6 years of work experience, had taken 2.9 accounting classes, and were 27 years old. 24 percent of the participants are female. We chose MBA students as participants because prior research has found that they are a good proxy for nonprofessional investors (Elliott et al. 2007). We focus on nonprofessional investors because they represent a significant proportion of the user community (Bogle 2005). Further, regulators have expressed concern that financial statements be useful to this investor group (Pitt 2001). Thus, making nonprofessional investors the focus of our research allows us to speak directly to this concern. Design The experiment utilizes a 2 x 2 between-subjects, pretest-posttest design that varies financial statement presentation format (current vs. proposed) and the reliability of inputs used to measure gains associated with the fair value of a firm’s trading securities (Level 1 vs. Level 3).6 The income statement in the current treatment is presented in the currently mandated format in which all revenue and expense items are presented in a single column. The income statement in the proposed treatment differs from that provided in the current treatment in that it contains an additional column that highlights how changes in fair value affect net income. 7 The income statements provided in the current and proposed treatments are shown in Appendix B, Panels A and B. In the Level 1 treatment, gains from changes in the fair value of the firm’s trading 6
We elect to examine differences between Level 1 and Level 3 inputs because these represent the end-points of the fair value hierarchy, thus increasing the likelihood of observing treatment effects. As discussed in section V, future research could also examine users’ responses to Level 2 inputs. 7 The FASB’s discussion paper proposes a column that highlights changes in fair value as part of a reconciliation schedule that would appear in the firm’s footnotes. In our experiment, we chose to place the additional column in the income statement, rather than in the footnotes, for two reasons. First, Maines and McDaniel (2000) found that information placement in the financial statements can signal information importance. Thus, by holding constant the placement of the information in the income statement in both the current and proposed treatments we are able to hold constant this dimension of the Maines and McDaniel framework. Second, placing this column in the income statement is consistent with one alternative that is still under consideration by the FASB. 14
securities are based on Level 1 inputs. In the Level 3 treatment, gains are based on Level 3 inputs. The fair value footnotes provided to participants in the Level 1 and Level 3 treatments are provided in Appendix C, Panels A and B. 8 Participants were randomly assigned to one of these four treatments. Materials and Procedure The experiment is administered via a three-part web-based instrument. The task requires participants to evaluate information about Trans-Global Exports, Ltd., a small specialty manufacturer of tools. The first part of the instrument contains background information, which includes a brief description of the company, industry information, a press release that reports Trans-Global’s 2007 annual results, and an income statement. The presentation of the income statement varies based on the treatment condition. Half of the participants receive an income statement in the current format, while the remaining participants receive an income statement in the proposed format. Based on this information, participants are asked to rate Trans-Global’s potential for future earnings growth and to select a price-to-earnings (P/E) multiple for TransGlobal. These measures serve as initial judgments and allow us to understand how differences in the presentation of the income statement affect participants. After completing the initial judgments, participants move to the next part of the instrument. The second part of the instrument includes the information that was provided in part one and also contains a footnote that contains information about Trans-Global’s fair value accounting. The footnote information varies based on the treatment condition. Half of the participants receive a footnote indicating that gains in income are based on Level 1 inputs, while 8
As is shown in Appendices B and C, the fair value gain reported in income represents a smaller percentage of the Level 1 assets than the Level 3 assets. Thus, one concern is that participants in the Level 1 conditions will view the experimental materials as more believable than will participants in the Level 3 conditions. To the extent that this occurs it will result in a main effect of input level, but will not result in an input level by presentation format interaction, which is the focus of our research. 15
the remaining participants receive a footnote indicating that gains in income are based on Level 3 inputs. Based on this information, participants are again asked to rate Trans-Global’s potential for future earnings growth and to estimate a P/E multiple for Trans-Global. These measures serve as the final judgments which allow us to understand how fair-value information, in conjunction with differences in the income statement presentation format, affects participants’ judgments. Furthermore, analyzing the change in judgments caused by the information provided in the fair value footnote provides increased power (Libby et al. 2002). In addition to responding to the dependent variable questions, participants respond to a series of questions that are described in more detail in the next section. Upon completion of these questions, participants proceed to the third and final part of the instrument. Part three consists of manipulation check questions, within-subjects manipulation questions, and demographic questions. Completion of the questions in this part concludes the experiment. Dependent Variables and Mediators As the primary dependent variable, participants are asked to assess Trans-Global’s potential for future earnings growth on a 15-point scale with endpoints labeled “very weak potential” and “very strong potential” and the midpoint labeled “neutral.” As the secondary dependent variable, participants are asked to provide a P/E multiple for Trans-Global. They are told that firms in Trans-Global’s industry have historically had trailing P/E multiples ranging from 10 to 20 times reported net income. Thus, they are asked to select a P/E multiple between 10 and 20 with the knowledge that, all else equal, higher P/E multiples result in higher stock prices.
16
Participants are asked to assess three potentially mediating variables using 15-point scales. 9 First, they assess the risk of an investment in Trans-Global with scale endpoints labeled “very low” and “very high” and the midpoint labeled “neutral.” Second, they are told that reliable information is information that is reasonably free from error and bias and faithfully represents what it purports to represent. All participants rate the reliability of Trans-Global’s financial statements with scale endpoints labeled “not reliable” and “highly reliable” and the midpoint labeled “somewhat reliable.” Finally, participants are told that relevant information is information that makes a difference in a decision by helping users to form predictions about the outcomes of past, present, and future events. All participants rate the relevance of TransGlobal’s financial statements with scale endpoints labeled “not relevant” and “highly relevant” and the midpoint labeled “somewhat relevant.” IV. RESULTS Manipulation Checks Two participants did not complete the manipulation check questions. Of the remaining fifty-seven participants, responses to the input level manipulation check question indicate that five participants (four participants) in the Level 1 (Level 3) treatment incorrectly indicated the input level associated with the fair value gains included in income. 10 Responses to the presentation format manipulation check question indicate that five participants (seven participants) in the Current (Proposed) treatment incorrectly indicated how changes in fair value were reported on the income statement. Dropping participants who failed one or more of the
9
Using 15-point scales is consistent with prior related experimental research (e.g., Hirst et al. 2004). Of the nine total participants who failed the input level manipulation check, four were in the current format conditions and five were in the proposed format conditions. Thus, consistent with what we expected in our simplified environment, information acquisition did not differ depending on presentation format. 10
17
manipulation check questions, or did not respond to the manipulation check questions, does not change the results. Therefore, all analyses include these participants. Hypotheses Tests Due to the small sample size, all hypothesis tests are conducted using the difference between the initial and final earnings growth judgments as the dependent variable. As described above, this allows us to increase statistical power (Libby et al. 2002). Insert Table 1 Here H1 predicts that participants will weight fair value information more heavily under the proposed presentation format than under the current presentation format. Thus, we expect that differences in the measurement of fair value gains will have a greater impact on assessments of future earnings growth under the proposed format than under the current format. 11 Because we expect an ordinal interaction, we test H1 using a single, planned contrast. Whereas an ANOVA does not specify the pattern of relationships among cell means, contrast coding enables us to specify a set of contrast weights that correspond to the shape of our theoretically predicted interaction, thus increasing statistical power without increasing the likelihood of Type I error (Buckless and Ravenscroft 1990). We derive contrast weights based on the procedure described by Buckless and Ravenscroft (1990). Our specific contrast weights are as follows: Current Format/Level 1 Gains condition Current Format/Level 3 Gains condition Proposed Format/Level 1 Gains condition Proposed Format/Level 3 Gains condition
+1 -1 +3 -3
11
At the time of their initial judgment, participants do not have any information about the level of the fair value gains in income. Thus, we expected no difference in initial earnings growth assessments across the Level 1 and Level 3 conditions. Consistent with this expectation we find an insignificant main effect for gain level (F = 0.35, p = 0.56) based on an ANOVA with initial earnings growth assessment as the dependent variable. 18
The first two weights reflect our expectation that participants in the current format conditions will exhibit at least some reaction to the input level associated with the fair value gains. The third and fourth weights reflect our expectation that participants in the proposed format conditions will exhibit a greater reaction to the input level associated with the fair value gains. 12 As shown in Table 1, Panel B, the contrast hypothesized in H1 is supported (F=3.71, p=0.06). 13 Simple effects tests reported in Table 1, Panel C reveal that this difference is driven by participants in the proposed format conditions (t = 1.60, p = 0.06, one-tailed). When examining only those participants in the current format conditions, we find that revisions to assessments of earnings growth in the Level 1 condition are not significantly different than revisions to assessments of earnings growth in the Level 3 condition (t = 1.30, p = 0.10, onetailed). Thus, reactions to differences in inputs for fair value gains are greater under the proposed presentation format than under the current presentation format. Furthermore, our results indicate that, under the current presentation format, the additional disclosures mandated by SFAS No. 157 do not affect the judgments of nonprofessional investors. 14 Based on the HSM, we argue that participants who received the proposed presentation format would engage in a more systematic judgment process than participants who received the
12
An alternative would be to assign weights to the current format conditions based on the reasoning that participants will not rely on the SFAS No. 157 disclosures in these conditions. Weights of 0, 0, +1, -1 would be consistent with this reasoning. Our results are robust to this alternative specification. Also, note that weights of 0, 0, +1, -1 are less extreme than the weights we selected; hence, we purposely chose weights that bias against rejecting the null hypothesis. 13 Analyzing the data using the default coefficients of ANOVA reveals an insignificant effect of presentation format (F = 0.11, p = 0.74), a significant effect of level (F = 4.71, p = 0.03), and an insignificant format by level interaction (F = 0.06, p = 0.82). As noted by Buckless and Ravenscroft (1990), using default coefficients, rather than coefficients that are dictated by theory, can lead to misleading main effects that obscure the hypothesized interactive effects. 14 It is important to note that our focus is on examining how the proposed presentation format can increase the extent to which users differentiate between Level 1 and Level 3 inputs. However, we cannot speak to the extent to which users either over or under react to differences in input levels. While participants in the current presentation format seem to be under reacting relative to what standard setters would anticipate, it cannot be determined from our experiment whether participants in the proposed presentation format have achieved the “correct” level of differentiation between Level 1 and Level 3. 19
current presentation format. To test this reasoning, H2 predicts that users engaged in a more systematic process will spend more time on the judgment task. As described in Section III, the experiment consists of three parts. Part one elicits the initial judgments, part two elicits the final judgments, and part three contains the debriefing questions. We test H2 by examining both the total time spent on the task and the total time spent in parts one and two of the task. After eliminating two outliers who took longer than 100 minutes to complete the task, we find that the remaining participants spent significantly more time on the task (t = 2.58, p = 0.01, one-tailed) in the proposed format conditions (19.6 minutes) than in the current format conditions (12.2 minutes). Further, we find that the remaining participants spent significantly longer completing parts one and two (t = 2.07, p = 0.02, one-tailed) in the proposed format conditions (12.8 minutes) than in the current format conditions (7.9 minutes).15 Importantly, when asked to rate on a 15-point scale the difficulty associated with the judgment task, participants in the current (mean = 7.00) and proposed (mean = 6.70) format conditions rated the task as equally difficult (t = 0.39, p = 0.70). Thus, the additional time spent on the task by participants in the proposed format conditions is not driven by a perception that the task was more challenging. Thus, H2 is supported. To further examine the extent to which participants in the proposed format conditions engaged in systematic processing (H3), we conduct a mediation analysis that assesses the extent to which participants in the current and proposed format conditions consider the reliability, risk, and relevance of fair value inputs when making their earnings growth assessments. Figure 2, Panel A, summarizes the results for participants in the proposed format conditions. The results 15
Results are inferentially identical if we do not eliminate the two participants that spent more than 100 minutes on the task. Specifically, we find that participants spent significantly longer on the task (t = 2.06, p = 0.02, one-tailed) in the proposed format conditions (35.7 minutes) than in the current format conditions (12.2 minutes). Further, we find that participants spent significantly longer completing parts one and two (t = 1.86, p = 0.03, one-tailed) in the proposed format conditions (28.9 minutes) than in the current format conditions (7.9 minutes). 20
indicate that participants in the proposed format conditions do not rate Trans-Global’s financial statements as differentially relevant or differentially risky based on the use of Level 1 or Level 3 inputs. However, participants do rate Trans-Global’s financial statements as significantly more reliable (t = 1.73, p = 0.05, one-tailed) if fair value measurements are based on Level 1 gains (mean = 9.0) than if fair value measurements are based on Level 3 gains (mean = 7.38). Moreover, these differences in reliability assessments are significantly correlated with participants assessments of earnings growth (t = 2.32, p = 0.01, one-tailed). Thus, when provided with SFAS No. 157 disclosures, participants in the proposed format conditions are focusing on one of the factors that the FASB views as a critical difference between Level 1 and Level 3 inputs. In contrast, as is summarized in Figure 2, Panel B, participants in the current format conditions do not rate Trans-Global’s financial statements as differentially relevant, reliable, or risky based on the use of Level 1 or Level 3 inputs. Thus, participants in the current format condition do not seem to be considering the key differences between Level 1 and Level 3 inputs when evaluating SFAS No. 157 disclosures. We view this evidence as support for H3 and for the argument that participants in the proposed format conditions are engaging in more systematic processing than are participants in the current format conditions. 16 Additional Analyses Effect on Price-Earnings Multiple
16
Note that an alternative method of testing H3 is to conduct a mediation analysis to determine whether assessments of relevance, reliability, and risk mediate the effect of the input level by presentation format interaction on assessments of earnings growth. Conducting this analysis using the 4-step procedure specified by Baron and Kenny (1986) and Kenny et al. (1998) reveals that, consistent with the results presented above, reliability assessments mediate this interaction. 21
Participants in our experiment were also asked to assess a P/E multiple for TransGlobal’s common stock. By eliciting this judgment, we are able to tie our results to externally verifiable data, thus increasing the external validity of our results. To make this assessment, all participants are provided with information about Trans-Global’s trailing P/E multiple. Thus, differences in participants’ responses are driven by differential views of Trans-Global’s risk and growth opportunities. As shown in Figure 2, risk assessments are not significant for either the current or proposed formats. Thus, we can focus on the link between earnings growth and the P/E assessment. We find that the change in earnings growth is significantly associated with the change in P/E assessments for all participants (t = 7.91, p = 0.00, one-tailed, untabulated). Two additional analyses are useful when understanding this effect. First, because reliability assessments are an important factor in participants’ judgments, it is important to confirm that these assessments do not affect the P/E multiple directly, but instead only operate through the effect on earnings growth. We find that reliability assessments are significantly correlated with P/E judgments (t = 3.48, p = 0.00, untabulated). However, when examining the effect of reliability assessments on P/E judgments in the presence of earnings growth assessments, we find that reliability assessments are insignificant (F = 1.28, p = 0.26, untabulated) while earnings growth assessments remain significant (F = 43.07, p < 0.00, untabulated). Thus, it appears that any differences in reliability assessments are affecting earnings growth assessments, but are not affecting P/E judgments directly. Second, to further establish the validity of the links shown in Figure 1, we confirm that our manipulated independent variables are significantly associated with the change in P/E judgments. The results presented in Table 2, Panel B reveal that conducting the same planned contrast as was used for the change in earnings growth assessments yields significant results (F =
22
5.10, p = 0.03). Importantly, simple effects test reported in Table 2, Panel C reveal that this is driven only by participants in the proposed format conditions. Thus, the pattern of results hypothesized in H1 extends to participants’ P/E judgments. Insert Table 2 Here Intentionality of Input Level Judgments In our experiment, the level of inputs used to determine fair value is manipulated between-subjects. In our post-experimental questionnaire we use a within-subjects manipulation of input level to examine the intentionality associated with participants’ interpretation of the difference in input levels. 17 Finding a significant difference using a within-subjects approach while finding a nonsignificant difference using a between-subjects approach suggests that participants are aware of the implications of the difference in input levels, but that participants are not able to access or use this information in their natural reasoning process (Kahneman and Tversky 1996; Libby et al. 2002). Insert Table 3 Here Table 3 provides the text of the question used to elicit the within-subjects assessment of input level and provides participants’ responses. Chi-square tests indicate that participants view a firm with Level 1 gains as warranting a higher P/E multiple (χ2 = 43.37, p < 0.00), as having higher future earnings growth (χ2 = 12.74, p = 0.00), as being less risky (χ2 = 49.37, p < 0.00), and as having more reliable (χ2 = 46.42, p < 0.00) and relevant (χ2 = 13.58, p = 0.00) financial statements than a firm with Level 3 gains. These results are consistent with the results from the between-subjects approach, with two exceptions. On a between-subjects basis, we found no difference in relevance and risk
17
Two participants did not complete the post-experimental questionnaire. Thus, this analysis is based on only fiftyseven participants. 23
assessments between the Level 1 conditions and the Level 3 conditions. Thus, although participants think that a firm with Level 3 gains has less relevant financial statements and is riskier than a firm with Level 1 gains, they are not able to use this information in their natural reasoning process. This finding is consistent with prior research that has examined the difficulty associated with providing information about risk to users in a way that is meaningful (e.g., Fischhoff et al. 1998 in psychology; Hodder et al. 2001 in accounting). Interpretation of Fair Value Losses Financial statement users may be naturally skeptical of a firm that reports Level 3 gains in income because this choice suggests that management is strategically using the discretion inherent in Level 3 estimates to report higher net income. Users may not have the same reaction to a firm that reports Level 3 losses because reporting lower net income is less likely to be consistent with management’s incentives. 18 Thus, we use a within-subjects debriefing question to investigate how participants interpret the difference between losses measured using Level 1 inputs and losses measured using Level 3 inputs. The text of the within-subjects debriefing question is provided in Table 4. Insert Table 4 Here We find that participants view a firm with Level 3 losses as marginally riskier (χ2 = 5.16, p = 0.08), and as having less reliable (χ2 = 13.05, p = 0.00) and less relevant (χ2 = 12.74, p = 0.00) financial statements than a firm with Level 1 losses in income. However, we find that participants’ P/E assessments and assessments of future earnings growth do not differ between a firm that reports Level 1 losses in income and a firm that reports Level 3 losses in income. Thus, 18
It is for this reason that we elected to focus on the gain domain in the between-subjects portion of our experiment. That is, it is easier to make directional predictions in the gain domain than in the loss domain. An alternative choice would have been to test both the gain domain and the loss domain in the between-subjects portion of the experiment. We elected not to take this approach as the time required for the task would likely exceed the amount of time that our participants were willing to give. 24
it appears that the additional disclosures mandated by SFAS No. 157 may have a differential impact on judgment depending on whether the inputs used to measure fair value result in gains or losses. 19 V. DISCUSSION This research investigates how nonprofessional investors respond to the fair value disclosures mandated by SFAS No. 157. Further, it investigates how proposed changes to the presentation of the financial statements can affect nonprofessional investors’ interpretation of fair value information. We find that, under the current income statement format, nonprofessional investors in our setting did not incorporate differences in fair value disclosures into their earnings growth assessments for the firm. In contrast, under the proposed format, in which the income statement isolates and labels changes in fair value that affect net income, we find that users employ more systematic judgment processes and, therefore, place greater weight on the reliability differences in inputs used to measure fair value. This, in turn, affects users’ earnings growth assessments. Additional analyses reveal that these effects carry over to users’ P/E judgments. Our findings indicate the importance of providing fair value measurement guidance in combination with changes to the format of the financial statements. Further, our research identifies areas that standard setters should consider in their fair value and presentation projects. For example, additional analyses reveal that users are aware that different input levels within the fair value hierarchy have differing levels of relevance and differing levels of risk for the firm.
19
Separating responses to both of the within-subjects questions based on presentation format reveals that the results are driven more by the participants who received the proposed format than by participants who received the current format. This further confirms the important role that presentation format plays in the interpretation of fair value information and again indicates that participants in the proposed format conditions are more systematic in their thought processes. 25
However, they do not access or use this knowledge when assessing the firm’s potential for earnings growth or when assigning a P/E multiple to the firm. 20 Indeed, SFAS No. 157 disclosures are intended to convey additional information about relevance and risk to investors. To the extent that nonprofessional investors are not incorporating this information into their judgments, then these additional disclosures are potentially rendered ineffective for a potentially large segment of the investing community. Thus, standard setters could examine more effective ways to communicate this information that would allow nonprofessional investors to incorporate relevance and risk differences into their judgments. We also find that users interpret differences in fair value inputs differently when those inputs result in losses, rather than gains, in income. Specifically, we find that, in the case of fair value losses, differences in inputs are not incorporated into users’ P/E assessments or into their assessments of the firm’s future earnings growth. Thus, the additional disclosures mandated by SFAS No. 157 that provide information about input levels seem to be used asymmetrically by nonprofessional investors. This finding supports the importance of the general GAAP prohibition against netting of assets and liabilities and gains and losses in financial statements (e.g., FASB Interpretation No. 39 (FASB 1992)) and has implications for the FASB as it continues to consider the use of fair value measurement in the financial statements. Our findings are subject to a number of limitations that provide opportunities for future research. First, we elected to present fair value changes in an additional column in the income statement. However, if the FASB elects to place this additional column in a reconciliation schedule that appears in the footnotes to the financial statements then this could dampen the
20
With respect to risk assessments, our findings are consistent with prior research that has examined the difficulty associated with providing information about risk to users in a way that is meaningful (e.g., Hodder et al. 2001). 26
effectiveness of this change to the financial statements because prior research has found that users are less likely to attend to information placed in the footnotes than to information placed in the financial statements (e.g., Hirst et al. 2004). Thus, future research may want to examine how footnote disclosure of this information affects the results that we observed in our study. Second, we elected to examine differential reactions to Level 1 and Level 3 inputs. However, future research could examine responses to Level 2 inputs as a way to more fully understand users’ reactions to SFAS No. 157 disclosures. Importantly, Level 2 inputs enable more managerial discretion than do Level 1 inputs, but do not require the supplemental reconciliation of beginning and ending balances that are required for Level 3 inputs. Thus, Level 2 inputs represent an important level of the hierarchy in that they may allow greater discretion without greater transparency for the user. Understanding users’ responses to this information could provide useful information for standard setters and regulators. Third, we hypothesized that placing changes in fair value in a separate column in the income statement would increase the weight that nonprofessional investors placed on this information because an additional column serves to both label and isolate the information. Thus, the information is signaled to be more important and is easier to access. While prior psychology research indicates that both of these dimensions must be affected to yield a change information processing, it is possible that in this setting only one of these dimensions led to the effects that we observed. For example, if users already view this information as sufficiently important, then only an increase in the ability to access the information may be necessary. Thus, future research could try to identify which of these two dimensions, labeling or isolation, resulted in the effects that we observed.
27
Finally, we elected to investigate the judgments of nonprofessional investors. However, this is just one user group of potential interest to standard setters and regulators. Future research could investigate the extent to which the results obtained in this study generalize to more sophisticated investors or to creditors.
28
REFERENCES Baron, R.M., and D.A. Kenny. 1986. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology 51 (6): 1173 – 1182. Barth, M.E. 2008. Global financial reporting: Implications for U.S. academics. The Accounting Review 83 (5): 1159 – 1179. Bogle, J. 2005. The Ownership of Corporate America – Rights and Responsibilities. Remarks by John C. Bogle, Founder and Former Chairman, The Vanguard Group, 20th Anniversary Meeting of the Council of Institutional Investors, April 11. http://www.vanguard.com/bogle_site/sp20050411.htm Buckless, F.A., and S.P. Ravenscroft. 1990. Contrast coding: A refinement of ANOVA in behavioral analysis. The Accounting Review 65 (4): 933 – 945. Clor-Proell, S.M. 2009. The effects of expected and actual accounting choices on judgments and decisions. The Accounting Review, forthcoming. Eagly, A.H, and S. Chaiken. 1993. The Psychology of Attitudes. Fort Worth, TX: Harcourt Brace Jovanovich College. Elliott, W.B., F. Hodge, J. Kennedy, and M. Pronk. 2007. Are M.B.A. students a good proxy for nonprofessional investors? The Accounting Review 82 (1): 139 – 168. Financial Accounting Standards Board (FASB). 1984. Recognition and Measurement in Financial Statements of Business Enterprises. Statement of Financial Accounting Concepts No. 5. Norwalk, CT: FASB. -----. 1992. Offsetting of Amounts Related to Certain Contracts, an Interpretation of APB Oponion No. 10 and FASB Statement No. 105. FASB Interpretation No. 39. Financial Accounting Series. Norwalk, CT: FASB. -----. 2006. Fair Value Measurements. Statement of Financial Accounting Standards No. 157. Norwalk, CT: FASB. -----. 2007. The Fair Value Option for Financial Assets and Financial Liabilities. Statement of Financial Accounting Standards No. 159. Norwalk, CT: FASB. -----. 2008a. Preliminary Views on Financial Statement Presentation. Discussion Paper. Norwalk, CT: FASB. -----. 2008b. Determining the Fair Value of a Financial Asset When the Market for that Asset is not Active. FASB Staff Position 157-3. Norwalk, CT: FASB.
29
Fischhoff, B., D. Riley, D. Kovacs, and M. Small. 1998. What information belongs in a warning? Psychology and Marketing 15 (7): 663 – 686. Goldman Sachs. 2008. Fair value accounting and economic challenges. Global Economic Weekly, July 9. Hirst, E., and P. Hopkins. 1998. Comprehensive income reporting and analysts’ valuation judgments. Journal of Accounting Research 36 (Supplement): 47 – 75. -----, -----, and J.M. Wahlen. 2004. Fair values, income measurement, and bank analysts’ risk and valuation judgments. The Accounting Review 79 (2): 453 – 472. Hodder, L., L. Koonce, and M.L. McAnally. 2001. SEC market risk disclosures: Implications for judgment and decision making. Accounting Horizons 15 (1): 49 – 70. Hodge, F.D., J.J. Kennedy, and L.M. Maines. 2004. Does search-facilitating technology improve the transparency of financial reporting? The Accounting Review 79 (3): 687 – 703. Hopkins, P., R.W. Houston, and M.F. Peters. 2000. Purchase, pooling, and equity analysts’ valuation judgments. The Accounting Review 75 (3): 257 – 281. Hunton, J.E., R. Libby, and C.L. Mazza. 2006. Financial reporting transparency and earnings management. The Accounting Review 81 (1): 135 – 157. Kahneman, D., and A. Tversky. 1996. On the reality of cognitive illusions. Psychological Review 103 (3): 582 – 588. Kenny, D. A., D.A. Kashy, and N. Bolger. 1998. Data analysis in social psychology. In The Handbook of Social Psychology 4th ed., Vol. 1. Edited by D.T. Gilbert, and S.T. Fiske, 233-265. New York, NY: McGraw-Hill. Libby, R., R. Bloomfield, and M.W. Nelson. 2002. Experimental research in financial accounting. Accounting, Organizations and Society 27: 775 – 810. -----, M.W. Nelson, and J.E. Hunton. 2006. Recognition v. disclosure, auditor tolerance for misstatement, and the reliability of stock-based compensation and lease information. Journal of Accounting Research 44 (3): 533 – 560. Linsmeier, T. 2007. Report on FASB research initiatives. American Accounting Association annual meeting, Chicago, IL, August 5-8. Lo, A., and D. Runkle. Crisis fuelled by accounting. Financial Times (April 1, 2009). Maines, L., and L. McDaniel. 2000. Effects of comprehensive income volatility on nonprofessional investors’ judgments: the role of presentation format. The Accounting Review 75 (2): 179 – 207. 30
Pitt, H. 2001. Remarks before the AICPA governing council. Miami Beach, FL, October 22. Song, C.J., W. Thomas, and H. Yi. 2009. Value relevance of FAS 157 fair value hierarchy information and the impact of corporate governance mechanisms. Working paper, Virginia Polytechnic Institute and State University.
31
APPENDIX A Example of SFAS No. 157 Disclosure – Abbott Laboratories 10-Q for March 2007 The following table summarizes the bases used to measure certain assets and liabilities at fair value on a recurring basis in the balance sheet:
Balance at March 31 2007
Assets: Trading securities Marketable available for sale securities Commodity contracts Foreign currency forward exchange contracts
Liabilities: Gain sharing derivative financial instrument liability Interest rate swap derivative financial instruments Fair value of hedged long-term debt Foreign currency forward exchange contracts
Basis of Fair Value Measurements Quoted Prices in Active Significant Markets for Other Significant Identical Observable Unobservable Items Inputs Inputs
$
880,450 $ — $ 106,176 106,176 3,090 3,090 28,288 — $ 1,018,004 $ 109,266 $
880,450 $ — — 28,288 908,738 $
$
— $ — $ — (67,782 ) — (1,432,218 ) — (26,277 ) — $ (1,526,277 ) $
(1,350) $ (67,782) (1,432,218) (26,277) $(1,527,627) $
— — — — —
(1,350) — — — (1,350)
The following table summarizes the activity for the gain sharing derivative financial instrument liability. The adjustment to record this liability at fair value has been recorded in Other (income) expense, net for the three months ended March 31, 2007. Balance at December 31, 2006 Adjustments to record item at fair value Balance at March 31, 2007
$(24,800) 23,450 $ (1,350)
For assets and liabilities that are measured using quoted prices in active markets, the total fair value is the published market price per unit multiplied by the number of units held without consideration of transaction costs. Assets and liabilities that are measured using significant other observable inputs are valued by reference to similar assets or liabilities, adjusted for contract restrictions and other terms specific to that asset or liability. For these items, a significant portion of fair value is derived by reference to quoted prices of similar assets or liabilities in active markets. For all remaining assets and liabilities, fair value is derived using a fair value model, such as a discounted cash flow model or Black-Scholes model.
32
APPENDIX B
Panel A: Income Statement Provided to Participants in the Current Treatment Trans-Global Exports, Ltd. Statement of Comprehensive Income For the year ended December 31, 2007 (Amounts in thousands except per share amounts) Net Sales Cost of goods sold Gross profit Selling, general & administrative expenses Income from operations Investment gains
$ 2,716,256 1,831,250 885,006 402,500
Interest expense Income before income taxes Income tax Net income
$
482,506 60,400 542,906 252,378 290,528 108,571 181,957
Earnings Per Share (EPS)
$
1.82
Shares Outstanding
100,000
33
Panel B: Income Statement Provided to Participants in the Proposed Treatment Trans-Global Exports, Ltd. Statement of Changes in Net Assets For the year ended December 31, 2007 (Amounts in thousands except per share amounts) Current-Period Transactions & Estimates Net Sales Cost of goods sold Selling, general & administrative expenses Net operating activities Investments: revaluation to fair value Net investing activities Interest expense Net financing activities Income tax Change in net assets Per Share Basis Shares Outstanding
Change in Fair Value of Related Assets & Liabilities
Change in Net Assets
$ 2,716,256 (1,831,250)
-
$ 2,716,256 (1,831,250)
(402,500)
-
(402,500)
482,506
-
482,506
-
$ 60,400 60,400
60,400 60,400
$ 60,400
(252,378) (252,378) (108,571) $ 181,957
(252,378) (252,378) (108,571) $ 121,557
$ 1.82 100,000
34
APPENDIX C Panel A: Fair Value Note Disclosure Provided to Participants in the Level 1 Treatment Note 9. Fair Value Accounting The Company uses fair value accounting for its trading securities. Fair value is measured based upon observable and unobservable inputs. Observable inputs reflect market data obtained from independent sources, while unobservable inputs reflect the Company’s assumptions. These two types of inputs create the following fair value hierarchy: •
Level 1—Quoted prices for identical instruments in active markets.
•
Level 2—Quoted prices for similar instruments in active markets; quoted prices for identical or similar instruments in markets that are not active; and model-derived valuations whose inputs are observable.
•
Level 3—Model-derived valuations that reflect the Company’s own assumptions and projections.
The following table presents, for each level of the fair value hierarchy, the Company’s trading securities at fair value as of December 31, 2007: Fair Value Measurements Using Fair Value at Description
(Level 1)
(Level 2)
(Level 3)
12/31/2007
Assets Trading Securities $ 211,200,000
$ 147,624,000
$
0
$
63,576,000
$ 211,200,000
$ 147,624,000
$
0
$
63,576,000
Total Assets
Changes in the Level 3 Fair Value Category As shown above, a portion of the Company’s trading securities are measured at fair value using the company’s assumptions (Level 3). There were no changes to the Level 3 category in 2007. Any gains or losses in income are due to changes in the Level 1 category.
35
Panel B: Fair Value Note Disclosure Provided to Participants in the Level 3 Treatment Note 9. Fair Value Accounting The Company uses fair value accounting for its trading securities. Fair value is measured based upon observable and unobservable inputs. Observable inputs reflect market data obtained from independent sources, while unobservable inputs reflect the Company’s assumptions. These two types of inputs create the following fair value hierarchy: •
Level 1—Quoted prices for identical instruments in active markets.
•
Level 2—Quoted prices for similar instruments in active markets; quoted prices for identical or similar instruments in markets that are not active; and model-derived valuations whose inputs are observable.
•
Level 3—Model-derived valuations that reflect the Company’s own assumptions and projections.
The following table presents, for each level of the fair value hierarchy, the Company’s trading securities at fair value as of December 31, 2007: Fair Value Measurements Using Fair Value at Description
(Level 1)
(Level 2)
(Level 3)
12/31/2007
Assets Trading Securities $ 211,200,000
$ 147,624,000
$
0
$
63,576,000
$ 211,200,000
$ 147,624,000
$
0
$
63,576,000
Total Assets
Changes in the Level 3 Fair Value Category As shown above, a portion of the Company’s trading securities are measured at fair value using the company’s assumptions (Level 3). The following table provides a reconciliation of the beginning and ending balances for the Level 3 category. Any gains or losses in income are due to changes in the Level 3 category. Beginning balance at December 31, 2006
$
60,400,000
Total Level 3 gains included in income Ending balance at December 31, 2007
3,176,000
$
63,576,000
36
FIGURE 1 Conceptualization of the Effects of Input Level on Earnings Growth Assessments
Relevance Link 1 (+)
Input Level
Link 2 (+)
Link 4 (+)
Reliability
Link 5 (+)
Earnings Growth
Link 7 (?)
P/E Multiple
Link 6 (-)
Link 3 (-)
Risk
Note: This figure illustrates how differences in the level of inputs used in fair value measurements affect assessments of earnings growth for the firm. Signs in (.) indicate relative effects of a Level 1 input compared to a Level 3 input.
37
FIGURE 2 Results for Effects of Input Level on Earnings Growth Assessments Panel A: Results for Participants in the Proposed Format Conditions
t=0.17, p=0.44
Input Level
t=1.73, p=0.05 t=0.20, p=0.42
Relevance
Reliability
Risk
t=0.22, p=0.41 t=2.32, p=0.01
Earnings Growth
t=5.00, p=0.00
P/E Multiple
t=1.07, p=0.15
Panel B: Results for Participants in the Current Format Conditions
t=0.62, p=0.27
Input Level
t=0.69, p=0.25 t=0.22, p=0.41
Relevance
Reliability
Risk
t=0.00, p=0.50 t=3.06, p=0.00
Earnings Growth
t=6.22, p=0.00
P/E Multiple
t=0.14, p=0.44
Note: The figure above provides the empirical results from our experiment. t/p-values at each link refer to relative effects of a Level 1 input compared to a Level 3 input. Panel A reports the results for participants in the proposed format conditions, and Panel B reports the results for participants in the current format conditions. We find that differences in the input level for fair value measurements affect participants’ earnings growth assessments via perceptions of financial statement reliability. However, this effect occurs for only those participants in the proposed format conditions, and does not occur for those participants in the current format conditions. All p-values are one-tailed.
38
TABLE 1 Results for Earnings Growth Assessment Panel A: Initial, Final, and Change in Earnings Growth Assessment (Means with standard deviations in parentheses) Initial Earnings Growth Assessment Level Level Row 1 3 Means 10.06 8.86 9.46 (2.21) (2.57) (2.42) n = 16 n = 14
Final Earnings Growth Assessment Level Level Row 1 3 Means 9.44 6.93 8.18 (2.31) (2.62) (2.73) n = 16 n = 14
Change in Earnings Growth Assessment Level Level Row 1 3 Means -0.63 -1.93 -1.28 (1.93) (2.76) (2.40) n = 16 n = 14
Proposed
9.54 (2.93) n = 13
10.00 (2.00) n = 16
8.85 (2.19) n = 13
7.69 (3.53) n = 16
-0.69 (2.18) n = 13
-2.31 (3.20) n = 16
Column Means
9.80 (2.52)
9.43 (2.32)
9.14 (2.24)
7.31 (3.11)
-0.66 (2.00)
-2.12 (2.96)
Current
9.77 (2.43)
8.27 (3.02)
-1.50 (2.86)
Panel B: Hypothesized Contrast – H1 Source Model contrast Residual Error
d.f. 1 2 55
M.S. 28.02 0.58 6.63
F-value 3.71 0.08
p-value 0.06 0.92
t-value 1.30 1.60
p-value (one-tailed) 0.10 0.06
Panel C: Simple Effects Source Current format Proposed format
d.f. 1 1
M.S. 12.75 19.03
Note: Panel A provides the mean earnings growth assessment and standard deviation associated with each treatment. Larger numbers indicate greater potential for future earnings growth. Participants in the current format conditions view an income statement with all numbers in the same column. Participants in the proposed format conditions view an income statement with a separate column for fair value gains. In the Level 1 (Level 3) conditions the fair value gains reported in income are based on Level 1 (Level 3) inputs. At the time of their initial judgment, participants have not yet received the input level manipulation. We separate the initial judgments in Panel A by input level solely for ease of comparison with the final judgments. Panel B reports the test of H1 in which the change in earnings growth assessments is the dependent variable and the hypothesized contrast weights are +1, -1, +3, and -3. Panel C reports the corresponding simple effects tests. 39
TABLE 2 Results for Price-to-Earnings (P/E) Multiple Panel A: Initial, Final, and Change in P/E Multiple (Means with standard deviations in parentheses) Initial P/E Multiple Level Level Row 1 3 Means 16.50 14.57 15.54 (2.42) (2.41) (2.57) n = 16 n = 14
Final P/E Multiple Level Level Row 1 3 Means 15.75 13.50 14.63 (2.41) (2.44) (2.64) n = 16 n = 14
Change in P/E Multiple Level Level Row 1 3 Means -0.75 -1.07 -0.91 (1.34) (3.22) (2.37) n = 16 n = 14
Proposed
14.92 (2.63) n = 13
16.00 (1.75) n = 16
14.54 (2.15) n = 13
13.31 (2.15) n = 16
-0.38 (1.71) n = 13
-2.69 (2.80) n = 16
Column Means
15.71 (2.60)
15.29 (2.17)
15.14 (2.34)
13.40 (2.25)
-0.57 (1.50)
-1.88 (3.06)
Current
15.46 (2.21)
13.93 (2.20)
-1.54 (2.61)
Panel B: Planned Contrast Source Model contrast Residual Error
d.f. 1 2 55
M.S. 38.55 3.36 5.72
F-value 5.10 0.44
p-value 0.03 0.51
t-value 0.32 2.26
p-value (one-tailed) 0.38 0.01
Panel C: Simple Effects Source Current format Proposed format
d.f. 1 1
M.S. 0.77 38.45
Note: Panel A provides the mean P/E multiple and standard deviation associated with each treatment. Participants in the current format conditions view an income statement with all numbers in the same column. Participants in the proposed format conditions view an income statement with a separate column for fair value gains. In the Level 1 (Level 3) conditions the fair value gains reported in income are based on Level 1 (Level 3) inputs. At the time of their initial judgment, participants have not yet received the input level manipulation. We separate the initial judgments in Panel A by input level solely for ease of comparison with the final judgments. Panel B reports the results of a planned contrast in which the change in P/E multiple is the dependent variable and the contrast weights are +1, -1, +3, and -3. Panel C reports the corresponding simple effects tests. 40
TABLE 3 Intentionality of Input Level Judgments Consider two firms that only differ in terms of the investment gains they report in income. Firm A reports a gain of $10 million due to changes in Level 1 inputs. Firm B reports a gain of $10 million due to changes in Level 3 inputs. Please rate Firm A and Firm B on the following dimensions: Number (Percentage) Choosing Firm A 41 (72%)
Number (Percentage) Choosing Same 15 (26%)
Number (Percentage) Choosing Firm B 1 (2%)
Which firm has stronger future earnings potential?
30 (53%)
19 (33%)
8 (14%)
12.74 (0.00)
Which firm is a more risky investment?
7 (12%)
6 (11%)
44 (77%)
49.37 (0.00)
Which firm has more reliable financial statements?
43 (75%)
10 (18%)
4 (7%)
46.42 (0.00)
Which firm has more relevant financial statements?
27 (47%)
24 (42%)
6 (11%)
13.58 (0.00)
Question Which firm has a higher P/E multiple?
χ2 (p-value) 43.37 (0.00)
Note: This table reports the content of and responses to a hypothetical scenario and series of debriefing questions that manipulate input level associated with fair value gains on a withinsubjects basis. Chi-square tests reveal that differences in the input level associated with the gain affect assessments of P/E, earnings growth, risk, reliability, and relevance.
41
TABLE 4 Interpretation of Fair Value Losses Consider two firms that only differ in terms of the investment losses they report in income. Firm Y reports a loss of $10 million due to changes in Level 1 inputs. Firm Z reports a loss of $10 million due to changes in Level 3 inputs. Please rate Firm Y and Firm Z on the following dimensions: Number (Percentage) Choosing Firm Y 19 (33%)
Number (Percentage) Choosing Same 18 (32%)
Number (Percentage) Choosing Firm Z 20 (35%)
Which firm has stronger future earnings potential?
22 (39%)
19 (33%)
16 (28%)
0.95 (0.62)
Which firm is a more risky investment?
14 (25%)
16 (28%)
27 (47%)
5.16 (0.08)
Which firm has more reliable financial statements?
31 (54%)
17 (30%)
9 (16%)
13.05 (0.00)
Which firm has more relevant financial statements?
19 (33%)
30 (53%)
8 (14%)
12.74 (0.00)
Question Which firm has a higher P/E multiple?
χ2 (p-value) 0.11 (0.95)
Note: This table reports the content of and responses to a hypothetical scenario and series of debriefing questions that manipulate input level associated with fair value losses on a withinsubjects basis. Chi-square tests reveal that differences in the input level associated with the loss affect assessments of risk, reliability, and relevance, but do not affect assessments of P/E or earnings growth.
42
