Article Reviews

Fair Value Accounting and Managers’ Hedging Decisions Essay

Posted on

DOI: 10. 1111/j. 1475-679X. 2012. 00468. x Journal of Accounting Research Vol. 51 No. 1 March 2013 Printed in U. S. A. Fair Value Accounting and Managers’ Hedging Decisions W E I C H E N , ? H U N – T O N G TA N , † A N D E L A I N E Y I N G WA N G ‡ Received 6 September 2011; accepted 6 July 2012 ABSTRACT We conduct two experiments with experienced accountants to investigate how fair value accounting affects managers’ real economic decisions. In experiment 1, we ? nd that participants are more likely to make suboptimal decisions (e. g. forgo economically sound hedging opportunities) when both the economic and fair value accounting impact information is presented than when only the economic impact information is presented, or when both the economic and historical cost accounting impact information is presented. This adverse effect of fair value accounting is more likely when the price volatility of the hedged asset is higher, which is a situation where, paradoxically, hedging is more bene? cial. We ? nd that the effect is mediated by participants’ relative considerations of economic factors versus accounting factors (e. . , earnings volatility). Experiment 2 shows that enhancing salience of economic information or separately presenting net income not from fair value remeasurements reduces the adverse effect of fair value accounting. Our ? ndings are informative to standard setters in their debate on the ef? cacy of fair value accounting. ? Australian School of Business, University of New South Wales; † Nanyang Business School, Nanyang Technological University; ‡ Isenberg School of Business, University of Massachusetts Amherst. Accepted by Philip Berger.

We appreciate helpful comments from an anonymous reviewer, Jun Han, Lisa Koonce, Eng Juan Ng, Terence Ng, Seet-Koh Tan, and workshop participants at Nanyang Technological University. We are grateful to the Institute of Certi? ed Public Accountants, Nanyang Technological University, University of Massachusetts Amherst, and University of New South Wales for ? nancial support. We thank Clarence Goh, Jeffrey Pickerd, Yao Yu, and Bo Zhou for research assistance. 67 Copyright C , University of Chicago on behalf of the Accounting Research Center, 2012 68 W. CHEN, H. -T. TAN, AND E. Y. WANG 1.

Introduction Standard setters and regulators have implemented fair value accounting across a wide range of ? nancial instruments and other non? nancial items (SFAS No. 157 and 159). While standard setters and proponents of fair value accounting believe that it provides the most transparent and relevant information for investors’ decision making (Ahmed, Kilic, and Lobo [2011]), fair value accounting has been criticized and resisted by various user groups (Hodder and Hopkins [2012]). Recently, it has also been blamed for causing and exaggerating the 2008 ? nancial crisis (Dontoh et al. 2012]; see Laux and Leuz [2009] for a discussion). A key criticism is that fair value accounting results in excessive volatility when markets become illiquid and market prices are volatile. 1 In particular, it has been argued that, because the volatility may not properly re? ect the underlying economic fundamentals, it can distort managerial decisions (Plantin, Sapra, and Shin [2008]). Such allegations, along with general criticisms about the role of fair value accounting in the ? nancial crisis, have led to intense lobbying and even moves to eliminate fair value accounting (Gordon [2009]).

Despite the importance of this issue, there has been little empirical evidence on whether managers’ real economic decisions are actually adversely affected by fair value accounting. 2 Prior studies have found that managers opportunistically use fair value estimates to manage earnings (Dietrich, Harris, and Muller [2001]) and achieve higher compensation bene? ts (Dechow, Myers, and Shakespeare [2010]), but none of them has documented a distortion of real decisions in terms of sacri? cing of economic value caused by fair value accounting. Documenting any adverse real effect of fair value ccounting on managerial decisions is important because those decisions directly in? uence ? rms’ economic growth and value creation (Kanodia [2007]). We examine the real effect of fair value accounting in the context of derivatives because this is an area where the use and alleged bene? ts/costs of fair value accounting have been especially controversial, and the economic consequences of any distortion of managerial decisions (e. g. , not taking up a hedge despite good economic reasons) can be signi? cant. Anecdotal evidence (e. g. , McKay and Niedzielski [2000]) and surveys (Lins, Servaes, and Tamayo [2010]) document claims by chief ? ancial of? cers that the earnings volatility arising from fair value accounting adversely in? uences managers’ decisions, suggesting a cost arising from fair value accounting. An empirical-archival study by Zhang [2009] suggests the opposite. She 1 PricewaterhouseCoopers—Point of view: Fair value accounting. Please refer to the Web site: http://www. pwc. com/us/en/point-of-view/fair-value-accounting-? nance-proposal. jhtml. 2 Following the term used in Kanodia [2007], “real effect” of accounting disclosure refers to the impact of accounting measurement and ? nancial reporting on ? ms’ real decisions and resource allocation in the economy. We consider the sacri? cing of economic value to report intertemporally smooth earnings as a “real effect” of fair value accounting. REAL EFFECT OF FAIR VALUE ACCOUNTING 69 concludes that SFAS No. 133 bene? ted (in terms of risk reduction) a sample of ? rms that initiated derivative programs but did not reduce their risk exposures before the implementation of SFAS No. 133; however, this evidence is ambiguous because a signi? cant proportion of these ? rms actually stopped derivative activities after the implementation of SFAS No. 33. Our study employs two experiments to address the following research questions. We investigate three related questions in the ? rst experiment. First, we examine whether there is a causal link between managers’ foregoing of economically bene? cial hedging opportunities and fair value accounting. This issue is important because the current fair value accounting standards, intended to lead to better decisions, have been alleged to cause managers to underhedge or completely avoid hedging to reduce reported earnings volatility, exposing companies to larger risk factors (Leib [2001]).

We also assess the moderating role of price volatility. In our experiment, higher (lower) price volatility results in higher (lower) reported earnings volatility when managers choose to hedge and fair value accounting is used. We examine this factor because reported earnings volatility, which does not necessarily correspond to economic fundamentals, has been expressly singled out as a major concern about fair value accounting (e. g. , McKay and Niedzielski [2000], Barth [2004]). However, there is no evidence that the reported earnings volatility actually in? ences managers’ decisions. Second, we investigate the process by which fair value accounting in? uences managers’ decisions. Speci? cally, we posit that this occurs because the fair value accounting information shifts managers’ emphasis away from economic factors to accounting factors. Third, we investigate whether managers’ decisions will also be impeded if historical cost accounting is used to account for derivatives, in order to provide further causal evidence regarding the role of fair value accounting.

In experiment 2, we investigate the effectiveness of two simple debiasing mechanisms—altering the salience of accounting versus economic impact, and separately presenting net income not from fair value remeasurements—to mitigate any adverse impact of fair value accounting on managers’ decisions. Managers likely consider both economic and accounting factors when they make hedging decisions, but separating out managers’ considerations of economic versus accounting factors is dif? cult using archival data because managers’ considerations of these factors are not observable.

We capitalize on the comparative advantage of experiments by holding constant the anticipated economic impact of a hedging decision, and varying the presence/absence of accounting-related information (experiment 1), or the order/manner in which the accounting information is presented (experiment 2). In experiment 1, we manipulate whether managers are provided with information on the economic impact alone (the economiconly condition) or both the economic and accounting impact information (the economic-plus-accounting condition) when they make their isk hedging decisions. The economic impact of hedging relates to the effect of hedging on the company’s expected future cash ? ow, while the accounting 70 W. CHEN, H. -T. TAN, AND E. Y. WANG impact of hedging relates to the effect of hedging on the company’s reported earnings volatility. We create a context where it is economically desirable to hedge across all conditions. Speci? cally, based on the economic impact of hedging versus not hedging, managers bene? t from employing derivatives to hedge corporate risk exposures.

However, reporting derivatives and hedging activities will result in reported earnings volatility on the ? nancial statements. A lower propensity by participants to hedge in the economic-plus-accounting condition compared to the economic-only condition will indicate that this is due to their consideration of the accounting impact. Experiment 1 also manipulates the price volatility (low, high) of the hedged asset. High price volatility implies greater uncertainty over the future cash outlays, which suggests a greater bene? t of hedging by locking into an agreed-upon future price.

However, in our setting, despite the greater economic bene? t of hedging, higher price volatility also results in higher reported earnings volatility when a derivative contract is taken up and fair value accounting is applied. In addition, to compare the effects between fair value accounting and historical cost accounting, we also design a “historical cost accounting” control condition where participants are provided with information on both the economic and historical cost accounting impact when price volatility is high.

We further manipulate two presentation formats in experiment 2: (1) the order in which information is presented, either economic impact followed by accounting impact or the reverse order, and (2) whether the net income not from fair value remeasurements is reported in a separate column. Results in experiment 1 show that participants in the economic-plusaccounting condition are less likely to hedge the price risk than those in the economic-only condition only when fair value accounting is applied, but not when historical cost accounting is applied. This negative effect of fair value accounting on managers’ hedging decisions is magni? d when price volatility of the hedged asset is higher (vs. lower). We further document that participants in the economic-only condition are more likely to hedge when price volatility is higher than lower, consistent with the economic rationale for undertaking risk management. However, the reverse occurs in the economic-plus-accounting condition: participants are less likely to hedge when price volatility is higher than lower. Furthermore, based on participants’ rationales for their decisions, we ? nd that this effect is fully mediated by managers’ relative emphasis on economic versus accounting considerations related to hedging.

Finally, in experiment 2, we show that notwithstanding managers’ concerns about the accounting impact of hedging, their propensity to hedge is increased by making them attend to the economic impact of hedging prior to their decisions, or by separately presenting net income not arising from fair value remeasurements. Our paper contributes toward an improved understanding of the effect of fair value accounting on managers’ economic decisions. Prior studies on fair value accounting mostly focus on how fair value measurements in? uence the value relevance or risk relevance of accounting numbers

REAL EFFECT OF FAIR VALUE ACCOUNTING 71 (Barth [1994], Barth and Beaver [1996], Nelson [1996], Venkatachalam [1996], Ahmed, Kilic, and Lobo [2011]). Rather than focusing on external users’ perspective (e. g. , investors or creditors), we examine the effect of fair value accounting on internal managers’ decision making. We provide empirical evidence that, despite substantial economic bene? ts, managers actually abstain from hedging the risk because of their concerns over the fair value accounting impact (e. g. , increased earnings volatility).

We also provide evidence that this effect is magni? ed when price volatility of the hedged asset is higher (with expected higher earnings volatility), where the case for hedging is actually stronger from a risk management perspective. These ? ndings have implications for the existing debate about the pros and cons of fair value accounting (Laux and Leuz [2009]). While there have been demonstrated bene? ts of fair value accounting, such as increased relevance (e. g. , Hirst, Hopkins, and Wahlen [2004]), we provide evidence on an unintended effect of fair value accounting. Our ? dings are particularly relevant to concerns that fair value accounting distorts real decisions by inducing reported earnings volatility that is purely a consequence of the accounting treatment rather than something that re? ects the underlying fundamentals. Our paper also contributes to the literature on the comparison between fair value accounting and historical cost accounting. Prior studies have compared these two different reporting regimes in terms of their different impact on market transparency (Bleck and Liu [2007]), their respective information content in terms of providing early warning of potential ? ancial distress (Gigler, Kanodia, and Venugopalan [2007]), or their relative pros and cons in measuring different types of balance sheet items (Plantin, Sapra, and Shin [2008]). Our study compares the two reporting regimes from a different perspective: their effects on managerial decisions that have direct impact on the ? rm’s pro? tability. We show that participants tend to forgo an economically desirable hedging opportunity only when fair value accounting impact is considered.

When historical cost accounting impact is taken into consideration, participants’ decisions are similar to their decisions when only the economic impact is considered. Furthermore, we provide further evidence that the fair value accounting impact shifts managers’ concerns over the economic considerations toward accounting considerations, which in turn results in suboptimal economic decisions. This ? nding illustrates another potential downside of fair value accounting compared to historical cost accounting, and will be useful for regulators who are moving from a historical cost accounting regime to a fair value accounting regime.

Finally, our ? nding on the remedial effect of making the economic considerations more salient should be helpful to managers and regulators. Fair value accounting is likely to be the de facto reporting norm for many asset and liability classes in the future, and our ? nding suggests that a simple intervention (i. e. , making decision makers attend to the economic impact of hedging prior to their decisions) can possibly help managers make better economic decisions in this reporting regime. In addition, our ? nding that 72 W. CHEN, H. -T. TAN, AND E.

Y. WANG separate presentation of net income not from fair value remeasurements can reduce managers’ concerns over reported earnings volatility should also be of interest to regulators. Speci? cally, we provide some ex ante evidence on the effect of separately presenting net income changes due to fair value remeasurements from all other changes, which has been proposed by the Financial Accounting Standards Board (FASB) and the International Accounting Standards Board (IASB) in 2008 (FASB/IASB Discussion Paper [2008], Staff Draft [2010]).

Although the purpose of this proposal is targeted toward improving users’ welfare in their use of ? nancial information, we show that this proposed presentation format has the added effect of improving managers’ decisions as well. In the next section, we review the related literature and develop hypotheses. Section 3 describes the research design, experimental procedure, and results in experiment 1. Section 4 discusses experiment 2, where we examine the effects of two debiasing mechanisms. Finally, we conclude the study in section 5. 2. Literature Review and Hypothesis Development

Prior studies demonstrate that fair value measurements lead to more volatile earnings and stock returns. Hodder, Hopkin, and Wahlen [2006] ? nd that the volatility of full-fair-value income is more than three times that of comprehensive income and more than ? ve times that of net income for their sample. Their ? ndings also suggest that full-fair-value income volatility relates more closely to stock price volatility that is not captured by net income or comprehensive income. More recently, Zhou [2009] documents a similar ? nding in a more speci? c context—fair value accounting for derivatives.

She shows that an earnings measure in the post-SFAS 133 period that includes the fair-value-based hedging performance has a positive association with idiosyncratic stock return volatility. These ? ndings support managers’ argument that fair value accounting leads to more volatility on their income statements and stock prices. Managers are averse to reporting volatile earnings. Smooth earnings help managers avoid negative earnings surprises, and thus are associated with less negative stock price reactions (Bartov, Givoly, and Hayn [2002]) and lower management turnover (DeFond and Park [1997]).

Furthermore, investors associate higher earnings volatility with lower disclosure quality and lower earnings quality, resulting in an increase in required risk premium and cost of capital (Sengupta [1998], Francis et al. [2002], Easley and O’Hara [2004]). As reported in Graham, Harvey, and Rajgopal [2005], 96. 9% of surveyed CFOs prefer a smooth earnings path and 78% of them report that they would sacri? ce economic value to achieve that target. In this study, we propose that this aversion to reporting earnings volatility arising from fair value accounting can cause managers to make suboptimal economic decisions.

Fair value accounting cuts across many domains and has the potential to in? uence a myriad of economic decisions. For instance, REAL EFFECT OF FAIR VALUE ACCOUNTING 73 managers’ investing or ? nancing decisions may change when their investments or long-term liabilities are recorded at fair value, due to their expectations of greater ? nancial statement volatility with the use of fair value accounting. Furthermore, when earnings incorporating fair value changes are used to evaluate ? rm performance, managers’ operating decisions are likely to change in response to a more volatile performance benchmark.

A recent study by Bhat, Frankel, and Martin [2011] documents that fair value accounting exacerbated banks’ tendency to sell mortgage-backed securities when these assets faced liquidity-driven price declines during the 2007 ? nancial crisis, and that relaxation of these fair accounting rules in April 2009 reduced this effect. While this ? nding demonstrates a real effect of fair value accounting, it does not provide causal evidence of the distortional effects of fair value accounting as liquidity-driven selling makes economic sense.

In particular, as the authors acknowledge, their results are also consistent with a regulatory forbearance effect. Speci? cally, analytical models suggest that this selling driven by feedback effects is economically justi? able because it reduces the prospect of a reduction in bank regulatory capital (Allen and Carletti [2008]). 3 In this study, we provide a more direct demonstration of the distortionary effect of fair value accounting using fair value accounting for derivatives as a context. Current accounting standards for derivative instruments and hedging activities (e. g. , SFAS No. 33) require ? rms to account for derivatives as assets or liabilities at fair value on the balance sheet, with ? uctuations in fair value of the derivatives re? ected on either the income statement or comprehensive income. We use fair value accounting for derivatives as a setting because it is one that has been particularly controversial, with users claiming that it distorts managers’ decisions (McKay and Niedzielski [2000], Lins, Servaes, and Tamayo [2010]). Also, suboptimal risk management decisions can expose ? rms to substantial economic losses (Adam and Fernando [2006]).

It is also a setting where standard setters are suf? ciently concerned that they have permitted the hedge accounting approach to be adopted to reduce the full impact of fair value accounting. However, there are many occasions where this hedge accounting approach either does not apply or provides incomplete “protection” against fair value accounting effects. Empirical evidence regarding the impact of fair value accounting on managers’ risk management decisions is limited. The only study we are aware of is Zhang [2009]. She identi? es a sample of ? ms that initiated derivative programs during the four-year period between 1996 and 1999. 4 3 Speci? cally, in Allen and Carletti [2008], bank regulators refer to accounting numbers to assess capital adequacy, and banks that hold on to and do not sell these securities face the prospect of recognition of unrealized losses that are other than temporary, which can reduce capital adequacy. 4 Firms that have a prior history of engaging in derivative transactions (comprising 77% of the total sample), and arguably the more representative of the population, are eliminated. 4 W. CHEN, H. -T. TAN, AND E. Y. WANG She classi? es 125 ? rms as effective hedgers if their risk exposure reduces after initiation of the programs, and 87 ? rms as joint ineffective hedgers or speculators (because it is not possible to empirically separate these groups) if their risk exposure increases. She ? nds that the volatility of cash ? ows for the speculators/ineffective hedgers reduces after the introduction of SFAS No. 133, but there is no change in risk exposures for effective hedgers. She concludes that SFAS No. 133 is bene? ial for the speculative/ineffective hedgers. However, Zhang [2009] also documents that a signi? cant percentage (16%) of speculative/ineffective hedgers stopped using derivative instruments post-SFAS No. 133, while a smaller percentage (6%) of the effective hedgers did so. Because it is not possible to empirically separate out speculative from ineffective hedgers,5 this result can be alternatively interpreted to mean that hedging activities may have been curtailed for ineffective hedgers who cannot apply hedge accounting, a cost associated with SFAS No. 33. We develop our hypotheses in the following sections. 2. 1 REAL EFFECT OF FAIR VALUE ACCOUNTING ON MANAGERS’ DECISIONS Hedging with appropriate derivative instruments can reduce the ? rm’s risk exposure, an economically desirable outcome. Suppose a manager expects to purchase oil in the future and forecasts that the oil price will increase. The manager wants to lock in an agreed-upon purchase price, especially when the oil price is highly volatile. In order to hedge the price risk, the manager should enter into a derivative contract so that he/she can ? the future purchase price, irrespective of the rise or decline of oil price. If the manager does not hedge the price risk, he/she has to purchase the oil at a future market value, and will be subjected to the volatility of the future oil price. Thus, employing a hedging instrument can create an economic bene? t by protecting ? rms from volatile price. This bene? t from hedging increases with increasing volatility of the hedged asset. On the other hand, reporting derivatives using fair value accounting may result in additional earnings volatility on the ? nancial statements.

Volatile market prices of the hedged asset in the interim period from the inception of a derivative contract to its expiration date will result in fair value changes of the derivatives and reported earnings volatility. 6 The reported earnings 5 Prior research points out that a challenge with the use of archival data to examine this issue is that, although the change (or no change) in ? rms’ ex post risk exposure is observable, the purposes for their decisions to hedge or not hedge are not observable (Geczy, Minton, and Schrand [2007]). The argument made by these researchers is that it is dif? ult to determine whether the change is due to managers’ ex ante risk management decisions, speculation, accounting implications, or luck. In the context of Zhang’s [2009] study, there are also likely differences in risk setting associated with each derivative usage pre- and post-implementation of SFAS No. 133, and effective versus ineffective hedgers/speculators also systematically differ in ? rm size, leverage, and inherent risk exposure. 6 Since the transaction (e. g. , oil purchase) is forecasted to happen in the future, no asset (e. g. , oil) will be currently booked on the balance sheet.

However, derivatives are booked at fair value when ? rms enter into derivative contracts, which is well before the forecasted REAL EFFECT OF FAIR VALUE ACCOUNTING 75 volatility purely arises from fair value accounting for derivatives rather than the company’s real economic activities. Since the use of a hedging derivative ? xes the future delivery price, the price volatility in the interim period will not result in any real monetary effect. The extent to which reported earnings volatility is induced depends on the hedging effectiveness and whether the derivative transaction quali? es for hedge accounting.

Given managers’ aversion to reporting volatile earnings (Graham, Harvey, and Rajgopal [2005]), we expect that managers will be more likely to forego economically sound hedging opportunities if they factor in the fair value accounting impact of hedging (i. e. , increased earnings volatility). In other words, managers will engage in derivative transactions to manage their risk exposures when only the economic impact of hedging is presented to managers (i. e. , the economic-only condition in our experiment; henceforth, E-only). In contrast, when managers are also provided with information about he fair value accounting impact of hedging (i. e. , the economic-plus-accounting condition; henceforth, E-plus-A), they are less likely to hedge the risk once the increased earnings volatility arising from fair value accounting for derivatives is highlighted. Moreover, in a setting where higher price volatility induces higher reported earnings volatility, the possibility of a smooth earnings path is greatly reduced with the adoption of a hedge. Thus, the difference between managers in the E-only and E-plus-A conditions will be larger when price volatility of the hedged asset is higher than when it is lower.

Speci? cally, with high price volatility, managers in the E-plus-A condition are more likely to forgo a bene? cial hedge than those in the E-only condition in order to obtain a smooth earnings path. Paradoxically, this high price volatility situation is likely one where the company can bene? t more from hedging. On the other hand, when the price is only slightly volatile and the induced earnings volatility is relatively low, managers can still expect a relatively smooth earnings path even with the adoption of a hedge.

Thus, the accounting impact will not be a primary concern, and we expect that the difference in managers’ decisions between the E-only condition and the E-plus-A condition is smaller when price volatility is lower. Our ? rst hypothesis is formally stated as follows: H1: Managers are more likely to hedge risk exposure with derivatives when only the economic impact information is presented than when both the economic impact and the fair value accounting impact information is presented; this effect is more likely when the price volatility of the hedged asset is higher but less likely when it is lower. ransaction happens. Hence, from the inception of the derivative contract to the point when the forecasted transaction happens, the recognized fair value gains or losses on the derivatives cannot be offset by the fair value changes on the hedged asset (e. g. , oil), resulting in greater earnings volatility. 76 W. CHEN, H. -T. TAN, AND E. Y. WANG 2. 2 MEDIATING EFFECT OF ACCOUNTING VERSUS ECONOMIC CONSIDERATIONS The premise underlying our hypotheses is that, notwithstanding signi? cant economic bene? ts from adopting the hedge, managers are concerned about the fair value accounting impact when they hedge.

Such concerns are strengthened when the fair value accounting impact information is copresented with the economic impact information but reduced when only the economic impact information is presented. In other words, the consideration of economic factors (e. g. , future expected cash out? ow, ? rms’ risk exposures) becomes diluted, and the consideration of accounting factors related to fair value measurements (e. g. , the application of hedge accounting, reported earnings volatility, and balance sheet volatility) becomes salient in the presence of information about the fair value accounting impact of hedging.

This relative emphasis on accounting factors over economic factors in managers’ considerations, in turn, leads them to withhold hedging decisions. Our theory therefore predicts that the relative economic versus accounting considerations by managers mediate the joint effect of hedging impact (economic alone vs. both economic and accounting impact) and price volatility on managers’ decisions to hedge. Our second hypothesis is formally stated as follows: H2: The joint effect of hedging impact and price volatility on managers’ hedging decisions is mediated by their relative considerations of economic versus accounting factors. 2. 3

FAIR VALUE ACCOUNTING VERSUS HISTORICAL COST ACCOUNTING Public dissent against fair value accounting generally pits fair value accounting against historical cost accounting. Analytical models also pit fair value accounting against historical cost accounting (Bleck and Liu [2007], Gigler, Kanodia, and Venugopalan [2007], Allen and Carletti [2008], Plantin, Sapra, and Shin [2008]). For example, Plantin, Sapra, and Shin [2008] present an analytical model that compares conditions (speci? cally, short-lived/long-lived assets, liquid/illiquid assets, and junior/senior assets) under which historical cost measurement systems result in lower inef? iencies than fair value accounting measurement systems. Similarly, archival studies also compare the value relevance of historical cost accounting versus fair value accounting (e. g. , Danbolt and Rees [2008], Christensen and Nikolaev [2010]). However, prior research provides little empirical evidence on the effect of fair value versus historical cost accounting on managerial decisions. 7 7 There has been no empirical test of the model in Plantin, Sapra, and Shin [2008], which compares the trading decision of a manager operating under a fair value accounting regime versus a historical cost accounting regime.

Note that the trading decision in their model is biased due to managers’ overreaction to the fair value changes on the balance sheet, and not REAL EFFECT OF FAIR VALUE ACCOUNTING 77 As we discussed in H1, we expect managers’ risk hedging decisions to be affected by fair value accounting information, as their concerns over earnings volatility can impede rational hedging decisions. In contrast, under historical cost accounting, no accounting entry (zero cost) is recorded for entering a derivative contract because derivatives only re? ect a mutual exchange of promises at their inceptions.

In addition, no value changes of derivatives are recorded on the balance sheet until the ? nal settlement of the derivative contract. As a result, no derivative gain (or loss) is recorded and the use of derivatives will not affect the income statement. Hence, we expect that managers’ risk hedging decisions will not be in? uenced by accounting information if historical cost accounting is applied. Thus, our hypothesis is stated as follows: H3: Managers are more likely to hedge risk exposure with derivatives when historical cost accounting is applied than when fair value accounting is applied.

We test H1 to H3 in experiment 1, and discuss experiment 1 in section 3. In section 4, we identify and develop our expectations about the effect of two debiasing mechanisms, along with a discussion of experiment 2 that is designed to test the ef? cacies of these two debiasing mechanisms. 3. Experiment 1 3. 1 PARTICIPANTS We conduct an experiment with experienced accountants attending continuing professional education sessions conducted by the national accountancy academy in Singapore. A total of 126 accountants with a mean (median) working experience of 10. 01 (8. 0) years participate in our study. As shown in table 1, the participants indicate their highest positions in their career, such as CEOs or CFOs (8. 7%), executives (10. 3%), managers (41. 3%), and other positions varying from accountants (12. 7%), auditors (10. 3%), controllers (6. 3%), consultants or analysts (4. 0%) to selfemployed practitioners (1. 6%). We ask participants to assess their knowledge in accounting for derivatives/familiarity with hedge accounting on a 15-point scale that varies from 0 (extremely low knowledge/unfamiliar) to 14 (extremely high knowledge/familiar).

The results show that our participants are somewhat knowledgeable about accounting for derivatives (mean = 5. 70) and somewhat familiar with hedge accounting (mean = 5. 55). They also indicate some familiarity with risk-hedging strategies using ? nancial derivatives (mean = 5. 59 on the 15-point scale, as stated above). because of managers’ concerns over volatile earnings associated with fair value accounting (as proposed in our study). 78 W. CHEN, H. -T. TAN, AND E. Y. WANG TABLE 1 Composition of Participants Number 52 16 13 13 11 8 5 2 6 126 Percentage 41. % 12. 7% 10. 3% 10. 3% 8. 7% 6. 3% 4. 0% 1. 6% 4. 8% 100. 0% Managers (including ? nance/accounting/ tax/assistant/senior manager) Accountant Executives Auditors (including senior, manager, and partner) Top Executives (including CEO/CFO/VP/AVP) Controllers Consultant/Analysts Self-employed Unknown Total This table reports the components of participants. Participants are asked to indicate their highest positions in their career. 3. 2 DESIGN We use a 2? 2 + 1 between-subjects design to test our hypotheses. The ? st manipulated factor, hedging impact, relates to whether participants are provided with information relating to only the economic impact, or both the economic impact and the accounting impact of hedging when they make their hedging decisions. In the E-plus-A condition, participants are shown both the economic impact and the accounting impact information, and then asked to make hedging decisions. In the E-only condition, participants are shown only the economic impact (without the accounting impact) information, and asked to make hedging decisions. 8 The second factor we manipulate is the price volatility of the hedged asset (low vs. igh). In our experiment, higher (lower) volatility in the oil prices is associated with higher (lower) reported earnings volatility during the period from the inception of the derivative contract to its expiration date. Changes in the oil prices in this interim period do not alter the ? rm’s cash ? ows or the fact that it has locked in the future cash outlay. However, these oil price changes in the interim result in greater/lower changes in the fair value of the hedging derivative (i. e. , West Texas Intermediate (WTI) Oil forward contract in our case) and the hedged item (i. e. forecasted purchase of Nigeria Oil in our case), leading to higher/lower volatility in reported earnings (see exhibit 1 of appendix A). 9 In our experiment, participants 8 In the E-only condition, after participants have made their ? rst hedging decisions, we also provide them with the accounting impact information and ask them to make a second/revised hedging decision. In addition, we have another between-subjects condition, in which participants ? rst make their initial hedging choices after the accounting impact information is presented, and then make their updated hedging choices after the economic impact information is presented.

The results for participants’ revised hedging decisions in these two conditions are similar to those in experiment 2 (reported later). 9 In our case where the hedge is not 100% effective, price volatility leads to earnings volatility when hedge accounting is applied. In cases where the hedging instrument perfectly REAL EFFECT OF FAIR VALUE ACCOUNTING TABLE 2 Experimental Design of Experiment 1 Fair Value Accounting Price Volatility Low High Economic (E-only) Condition 1 Condition 2 Economic-plusaccounting (E-plus-A) Condition 3 Condition 4 79

Historical Cost Accounting Economic-plusaccounting (E-plus-A) Condition 5 This table shows the design of experiment 1. We use a 2? 2 + 1 between-subjects design. In order to test our H1 and H2, we use a 2? 2 between-subjects design (i. e. , conditions 1/2/3/4); we manipulate whether participants are provided with only the economic impact information or both economic and accounting impact information when they make their hedging decisions. In the E-only condition, participants are shown the economic impact (without the accounting impact) information, and asked to make hedging decisions (conditions 1 and 2).

In the E-plus-A condition, participants are shown both the economic impact and the accounting impact information, and then asked to make hedging decisions (conditions 3 and 4). The second factor we manipulate is the price volatility of the hedged asset (low versus high). In order to test our H3, we hold price volatility as high and participants are provided with both the economic impact and the accounting impact information (i. e. , similar to the “E-plus-A/high volatility” condition). Participants are told that historical cost accounting is applied to account for the hedging instrument. cross all conditions are provided with information on the oil price volatility (either high or low), but information on reported earnings volatility (either high or low) is only presented to participants receiving the accounting impact information. We include a control condition to examine managers’ hedging decisions when the historical cost accounting approach is used. In the control condition, we hold the price volatility as high, and participants are provided with information relating to both the economic and accounting impact of the hedging decision (in other words, the same context as the “E-plus-A/high volatility” condition).

In addition, participants are told that historical cost accounting is applied to account for the hedging instrument. The accounting impact indicates that, if the management chooses to hedge, the company’s quarterly earnings will not change since historical cost accounting is used to account for the forward contract. The accounting treatment here is similar to that applied prior to the issuance of SFAS No. 133, where many derivative instruments were carried off-balance sheet. Table 2 summarizes our experimental design. 3. PROCEDURE All participants are instructed to assume the role of a manager for a listed company. They are given the same background information and ? nancial data about an oil company (ABC Company) that we develop from an actual company. Participants are told that the company forecasts a purchase of 2 billion barrels of Nigeria Oil in one year. Management of the company is concerned that the rise in the price of Nigeria Oil will result in additional cash payment in the future, and therefore is considering using a forward atches the hedged item, price volatility has no impact on earnings volatility, as the gains or losses of hedging instrument will exactly offset those of hedged item. 80 W. CHEN, H. -T. TAN, AND E. Y. WANG contract to hedge the risk. 10 Then, participants are provided with information indicating that the hedging derivatives may not be perfectly effective due to the limited supply of derivative contracts written on Nigeria Oil. The company can only enter into a one-year forward contract written on WTI Oil, the price of which is highly correlated with that of Nigeria Oil.

For those participants in the low (high) price volatility condition, they are informed that the volatility of oil price is relatively low (high). 11 Following the information about the hedged item and the hedging instrument, participants are given a short description about the accounting treatment for the forward contract if the company decides to hedge. In all conditions except for the control condition, participants are told that the forward contract is measured at its fair value and the derivative transaction is recorded using hedge accounting.

If the management decides to hedge, it will be designated as a cash ? ow hedge. 12 They are also informed that, due to the mismatch between hedging instrument (WTI Oil forward contract) and hedged item (Nigeria Oil), the hedge may not be perfectly (100%) effective, and this ineffectiveness will be recognized immediately in the company’s earnings. In the control condition, the participants are told to assume that the company applies historical cost accounting to recognize derivatives. Therefore, the hedging decision will not in? uence the company’s reported quarterly earnings.

After the background information about the company and the hedging case, participants are shown information relating to the economic and/or accounting impact of undertaking a hedge. The economic (cash ? ow) impact indicates that, if the management does not hedge, it is estimated that the company will pay an additional 3. 5 billion dollars next year for the same amount of Nigeria Oil. This is held constant across conditions. The accounting impact shows that, if the management chooses to hedge, the company’s reported quarterly earnings are expected to become slightly (highly) volatile in the low (high) volatility condition.

In the control (i. e. , historical cost accounting) condition, managers’ hedging decisions have no impact on the reported quarterly earnings. Exhibit 1 of appendix B summarizes the economic impact and the accounting impact information shown to the participants. Participants’ hedging decisions are captured using a binary option (A. Hedge commodity risk; B. Do not hedge commodity risk), and they also indicate the strength of preference for their hedging choices based on 10 Management’s emphasis on reducing the risk exposure strengthens the case for the participants to use (rather than not use) derivatives to hedge the risk. 1 The hedging effectiveness for the high price volatility condition is the same as that for the low price volatility condition. As illustrated in exhibit 2 of appendix A, the hedging effectiveness is measured as the ratio between the cumulative change in fair value of hedging instrument (column B) and the cumulative change in expected cash ? ow of hedged asset (column D). 12 We employ a cash ? ow hedge as our setting because it is at the root of the dispute between the accounting regulators and the industry (Sapra and Shin [2004]). REAL EFFECT OF FAIR VALUE ACCOUNTING 81 his information on a 15-point scale that varies from ? 7 (de? nitely hedge) to +7 (de? nitely not hedge). We also ask all participants to provide the reasons for their hedging decisions. Following this, participants are asked to return all materials to Envelope A and start with Envelope B, which contains manipulation check questions and other debrie? ng questions. In addition, participants are asked to assess the validity of the company’s accounting treatment for hedging activities, the extent that the case material is realistic, and the effort they put on analyzing the company’s accounting treatments.

They are also asked to evaluate their knowledge on accounting for derivatives and the extent to which they are familiar with risk hedging strategies. All these assessments use 15-point scales. Each participant is paid 20 Singapore dollars for participating in the experiment. 3. 4 RESULTS 3. 4. 1. Manipulation Checks. To check our manipulation of price volatility, participants are asked to indicate the volatility of the Nigeria Oil price for the period of 2010–2011 on a 15-point scale ranging from 0 (extremely low volatility) to 14 (extremely high volatility).

The mean assessment in the high price volatility condition (9. 47) is signi? cantly greater than that in the low price volatility condition (6. 48, F = 47. 38, p < 0. 01). 13 As a check on our manipulation of the hedging impact information provided (i. e. , whether the available information involves only economic or both economic and accounting impact), we ask participants to indicate the number of times they are asked to make hedging decisions and the information available for them to make hedging decisions. 4 About 79% of participants correctly answer at least one of these two questions, and there is no difference across conditions (p = 0. 83). Participants perceive the accounting treatment for hedging activities to be moderately valid (mean = 7. 48) and the case material to be fairly realistic (mean = 7. 13). 15 3. 4. 2. Test of H1. H1 predicts that participants are more likely to hedge the price risk when only the economic impact information is presented than when both the economic impact and the fair value accounting impact information is presented, and that this effect is larger when the price volatility is high.

Table 3, panel B reports the two-way categorical ANOVA with hedging decisions as dependent variable, and hedging impact and price volatility as independent variables. We ? nd a signi? cant main effect of hedging impact (? 2 = 7. 86, p = 0. 01); an insigni? cant main effect of price volatility (? 2 = 0. 56, p = 0. 46); and, more importantly, a signi? cant The p-value ? gures are all two-tailed, unless otherwise speci? ed. We ask the ? rst question because in the E-only condition, participants make two hedging decisions: ? rst with economic impact information only, and then with both economic and accounting impact information. 5 Our results are similar after we exclude those participants who fail one of the manipulation check questions or after we control for participants’ working experience, investment experience, and their familiarity with accounting for derivatives and hedge accounting. 14 13 82 W. CHEN, H. -T. TAN, AND E. Y. WANG TABLE 3 Descriptive Statistics and Analysis of Managers’ Hedging Decisions (DV = Managers’ Hedging Choice) Panel A: Descriptive Statistics: Mean (Standard Deviation) Hedging Impact Price Volatility E-only E-plus-A Row Mean Low 83% (0. 38) 80% (0. 41) 82% (0. 39) n = 24 n = 25 n = 49 (Condition 1) (Condition 3) High 96% (0. 20) 56% (0. 1) 76% (0. 43) n = 24 n = 25 n = 49 (Condition 2) (Condition 4) Column Mean 90% (0. 31) 68% (0. 47) n = 48 n = 50 Panel B: Categorical ANOVA Results for Managers’ Hedging Choice df Chi-Square Intercept 1 418. 99 Hedging Impact 1 7. 86 Price Volatility 1 0. 56 Hedging Impact ? Price Volatility 1 5. 62 Panel C: Planned Comparisons Hedging Choice: Contrast tests Low Volatility: E-only vs. E-plus-A (83% vs. 80%) High Volatility: E-only vs. E-plus-A (96% vs. 56%) High Volatility: Control vs. E-plus-A (88% vs. 56%) High Volatility: Control vs. E-only (88% vs. 96%) Chi-Square 0. 09 13. 77 7. 27 1. 04 Control 88% (0. 33) n = 25 (Condition 5) p-value