SEC Market Risk Disclosures: Implications for Judgment and Decision Making.

SYNOPSIS: In this paper, we draw on Judgment and decision-making research to examine the behavioral implications of the SEC's Financial Reporting Release No. 48 on market risk disclosures. While these disclosures have been examined using archival data, no research has investigated how these

disclosures might-affect individual users of financial statements. The purpose of our paper is to draw on research in the judgment and decision-making arena to identify and analyze the behavioral implications of the new risk disclosures. We offer three conclusions. First, FRR No. 48 users may have more complex evaluations of risk than perhaps anticipated by the SEC. Second, the flexibility accorded firms in FRR No. 48 will adversely affect users' risk judgments. Third, because the Release does not require disclosure of certain quantitative information that is important to risk assessments, inappropriate risk assessments may result. We believe our insights can help others conduct research in this important area and can help the SEC when they revisit the disclosure requirements in FRR No. 48.

INTRODUCTION

In 1997, the Securities and Exchange Commission (SEC) issued Financial Reporting Release No. 48, Disclosure of Accounting Policies for Derivative Financial Instruments and Derivative Commodity Instruments and Disclosure of Quantitative and Qualitative In formation about Market Risk Inherent in Derivative Financial Instruments, Other Financial Instruments, and Derivative Commodity Instruments (FRR No. 48). FRR No. 48 requires that companies disclose both qualitative and quantitative market risk information for risks of loss arising from adverse changes in interest rates, foreign currency rates, commodity prices, and equity prices. This reporting requirement is one of the few that requires companies to disclose forward-looking information about risk.

Understanding how these risk disclosures are likely to be interpreted by financial statement users is important because risk disclosures are not always interpreted as intended (Slovic 1987). Research in many domains documents the difficulties in communicating risk information (Fischhoff et al. 1998). Because prior research shows that risk perceptions influence behavior (Slovic et al. 1980; Viscusi et al. 1986), unintended interpretations of the FRR No. 48 risk disclosures could adversely affect investor and analyst judgment and decision-making (e.g., valuation judgments; buy/sell/hold decisions).

In this paper, we draw on judgment and decision-making research to examine two related issues. First, we summarize research from other domains that addresses how people perceive risk and conjecture as to how FRR No. 48 users might think about risk. Second, we examine how FRR No. 48's reporting requirements could adversely affect users' perceptions of the riskiness of companies' financial and derivative instruments. While existing research documents incidents of noncompliance with FRR No. 48 (Elmy et al. 1998; Roulstone 1999), we focus primarily on how the design of the FRR No. 48 disclosures (rather than noncompliance with them) causes judgment difficulties. [1] Throughout the paper, we draw on actual disclosures to illustrate our points about FRR No. 48. The purpose is to provide a descriptive understanding of behavioral implications of the new risk disclosures so that other researchers may draw on our insights when conducting research in this area. Our scholarly insights also could be used as input when th e SEC revisits the disclosure requirements. Finally, companies can use our insights to better understand the possible behavioral implications of their risk disclosures.

We draw three conclusions in the paper. First, we conclude FRR No. 48 users may have more complex evaluations of risk than perhaps the SEC anticipated. Psychology research indicates that when judging risk, individuals often consider additional risk factors beyond those in the SEC's risk disclosures. Such research further indicates that risk perceptions are context specific (Fischhoff et al. 1998). That is, how individuals think about risk in one domain may not be how they consider it in another, underscoring the need for research to learn how FRR No. 48 users view risk. Second, we conclude that the flexibility accorded by FRR No. 48 will adversely affect users' risk judgments. Specifically, the large number of choices in presenting market risk information can cause users to form inconsistent risk perceptions for the same underlying economic situation. Third, because FRR No. 48 does not require disclosure of certain quantitative information that is important to risk assessments, inappropriate risk assessments can occur.

We believe that our insights are applicable to a wide range of FRR No. 48 users including individual and institutional investors, financial analysts, bankers, and regulators. While one might think that only unsophisticated users will experience difficulties with the FRR No. 48 disclosures, research in psychology and accounting shows that expertise does not always overcome judgment problems and biases (Arkes 1991). Specifically, such research indicates that while expertise will overcome knowledge deficiencies, it will not necessarily compensate for deficiencies in available information. Because the latter deficiencies (e.g., noncomparable disclosures and insufficient quantitative information) are at the root of our concerns with FRR No. 48, the conclusions we draw herein are important for both unsophisticated and sophisticated FRR No. 48 users.

OVERVIEW OF FRR NO. 48

Before addressing the judgment and decision-making implications of FRR No. 48, we provide an overview of its reporting requirements (also see Linsmeier and Pearson 1997) as well as a brief review of the academic research on the topic. FRR No. 48, issued in response to perceived shortcomings of existing SEC and GAAP derivative disclosures, requires that SEC registrants disclose market risk information for off-balance-sheet derivatives and on-balance-sheet financial instruments. In particular, FRR No. 48 requires disclosure of market risk exposures for the items outlined in the left column of Table 1 and encourages disclosure for those in the right column.

FRR No. 48 requires that firms disclose two types of information about market risks: qualitative and quantitative information. Qualitative information includes the identification of primary market risks (i.e., interest rate, foreign exchange, commodity and other market risks such as equity price risks) and the particular markets in which the company is exposed. Firms also must discuss their derivative accounting policies, risk management goals, objectives, and controls. Finally, firms must reveal how risk is managed in general and the specific types of instruments used. Exhibit 1 shows several examples of the qualitative disclosures required under FRR No. 48.

To disclose quantitative information about their market risks, firms must select among three disclosure formats (tabular, sensitivity analysis, and value at risk [hereafter, VaR]) and three measurement bases (cash flows, earnings, or fair values). [2] Under the tabular format, no summary measure of risk is disclosed and thus no measurement basis is chosen. Rather, a table presents information concerning assets and liabilities subject to market risk. In this format, firms disclose fair values and average rates for market-risk-sensitive instruments, all categorized for each of the five years following the balance sheet date, and for the remaining years in aggregate. The table must include sufficient detail to allow users to estimate future cash flows from the financial instruments. An example of the tabular format is shown for Peoples Financial Corporation in Panel A of Exhibit 2. This disclosure provides no summary measure of risk, but does reveal certain attributes of the financial instruments that influence interest rate risk for Peoples Financial.

A sensitivity analysis quantifies the estimated potential loss in cash flows, earnings, or fair value resulting from a hypothetical change in a financial instrument's underlying market price. The hypothetical change in market prices must be a reasonably possible near-term change and, unless otherwise economically justified, is to be no less than 10 percent of period-end rates. For example, General Motors discloses in 1999 that a 10 percent change in equity prices would result in a $323 million loss in fair value of its equity investments (see Panel B of Exhibit 2).

VaR has been characterized as the worst-case scenario for a portfolio's performance under normal market conditions (Linsmeier and Pearson 1997, 117). More specifically, VaR is the maximum loss in terms of cash flows, earnings, or fair value that the firm could experience from its market-risk-sensitive instruments, over a given holding period, with a given probability. Unless otherwise economically justified, the confidence interval used to measure the probability of loss for FRR No. 48 purposes must be no less than 95 percent. For example, in 1999 Dow Chemical discloses that the one-day VaR for its foreign exchange risk is $5 million (see Panel C of Exhibit 2).

To date, only a handful of academic studies have examined FRR No. 48 disclosures. The descriptive research shows variation in the formats and measurement bases firms use to report market risk information. Roulstone (1999) reports that for a 1997 sample of 25 large firms, approximately 60 percent use sensitivity, 30 percent use VaR, and 10 percent use tabular presentation. Elmy et al. (1998) report that for a 1997 sample of 33 nonfinancial firms, 62 percent use fair value, 8 percent use cash flow, and 30 percent use earnings as the risk measurement basis. [3] Other studies have investigated the relation between FRR No. 48 disclosures and various surrogates for risk. Specifically, Thornton and Welker (1999) report that firms that quantify commodity exposure using the sensitivity or VaR format experience larger changes in their oil and gas betas than do firms that use tabular format. Rajgopal (1999) reports that firms' oil and gas betas are related to commodity sensitivities and tabular information (also see Ra jgopal and Venkatachalam 1999). Using interest rate and foreign-exchange risk disclosures, Linsmeier et al. (2000) document a decrease in stock trading volume when a firm uses either sensitivity or VaR formats. Finally, Ahmed et al. (1999) report that interest rate risk information presented in tabular format by banks may be useful in assessing one type of interest rate risk, namely volatility of future interest income. Thus, the literature indicates that FRR No. 48 disclosures capture variations in risk. What the existing research does not address is whether FRR No. 48 users assess risk in the same fashion as the SEC intends and whether the FRR No. 48's disclosures cause difficulties in assessing risk. These issues are the focus of our paper.

WHAT IS RISK?

Measures of Risk

Because scholars do not agree about the measurement of risk, multiple measures exist (Slovic 1966; Weber 1988). Common measures of risk in accounting and finance involve either the variance of returns, cash flows, or earnings or the covariance of returns with the market (Lipe 1998). Firms that have higher variances in their cash flows or earnings are considered riskier (Froot et al. 1993). Under the capital-asset-pricing model, a firm is considered riskier to the extent that its stock returns move more than the underlying market factors; that is, the covariance of a firm's returns with market factors is viewed as a measure of risk that, in turn, affects asset prices (Markowitz 1959). Despite the widespread acceptance of these measures, experimental research with individuals shows weak support for variance and covariance as descriptive measures of risk (Payne and Braunstein 1971; Slovic and Lichtenstein 1968; however, see Lipe 1998).

Substantial research in psychology and applied fields shows greater support for risk construed as loss or downside potential (Dreman 1998; March and Shapira 1987; Yates and Stone 1992). For example, risk in medicine is perceived as the chance of some adverse outcome, such as death or the contraction of some disease (Kleinbaum et al. 1982). In the area of technological hazards, risk is defined as the existence of threats to life or health (Fischhoff et al. 1981). Chartered financial analysts and professional bond traders believe risk involves the potential for monetary loss (Heisler 1994; Olsen 1997). This research further defines risk as primarily involving loss magnitude, probability, or some combination of both, and secondarily involving three other factors: dread, unknown, and gains. We consider each of these components of risk below.

Although it is generally agreed that risk is primarily determined by loss magnitude and probabilities, there is disagreement about how these two factors are considered. For example, some scholars believe that risky situations are best described by only the uncertainty measure, holding constant (or ignoring) the outcome. For example, a 20 percent chance of losing money is considered riskier than a 5 percent chance. Others argue that the magnitude or significance of possible outcomes is sufficiently descriptive, holding constant (or ignoring) the probabilities. For example, risk is greater when the potential loss is $100,000 compared to $1,000. Finally, some believe that risk consists of a combination of uncertainty and magnitude, such as the calculation of an expected loss, which would be the sum of all losses multiplied by their corresponding probabilities (Yates and Stone 1992). Others argue for using higher moments of the loss distribution, such as variance or skewness.

Prompted by concerns that probabilities and outcomes may not completely capture individuals' conceptions of risk, Slovic et al. (1980) explore the emotional bases of risk judgments. Using the psychometric paradigm, their research finds that people's perceptions of risk are a function of two dimensions: the extent to which outcomes are (1) unknown cognitively (the extent to which a hazard is judged to be unobservable, new, or delayed in producing harmful impacts) and (2) dreaded emotionally (perceived lack of control, worry, and catastrophic potential). These results replicate across groups of laypeople and experts judging large and diverse sets of situations (e.g., Olsen 1997; Slovic 1987). Although one might argue that unknown and dread loosely correspond to loss probability and loss significance/magnitude, research indicates that dread has explanatory power for understanding risk judgments even after statistically controlling for the effect of probabilities and outcomes (Holtgrave and Weber 1993; Koonce et al. 2000).

Consequently, Loewenstein et al. (1999) develop a model positing that people react to the prospect of risk at two levels: they evaluate risk cognitively (using, in some fashion, probabilities and outcomes), and they react to risk emotionally (in the form of dread).

Finally, the literature provides some support for potential gains being a component of risk assessments, although not in the form of variance as noted earlier. Specifically, Luce and Weber's (1986) conjoint expected risk model posits that risk judgments are a combination of the probabilities of winning, losing, and receiving nothing in addition to conditional expectations of losses and gains weighted by probabilities. Their model accounts for the fact that symmetric changes in good and bad outcomes do not affect judged risk equally and, thus, it can accommodate the findings from the risky-choice literature that people weight negative outcomes more heavily than positive ones (Kahneman and Tversky 1979a). Furthermore, Luce and Weber's (1986) model posits that information about gains decreases perceived risk (also see Alhakami and Slovic 1994). This perspective is at odds with the accounting/finance notion of risk as variance where, ceteris paribus, larger potential gains increase risk. Weber (1988) and Weber an d Bottom (1989, 1990) provide empirical support for the model, noting that the relative weights placed on gains generally are smaller than those on losses.

The SEC's Definition of Risk

In FRR No. 48, the SEC defines risk in terms of loss, consistent with much of the existing psychology literature. Both the sensitivity and VaR disclosures report risk in terms of a single loss outcome from the distribution of possible loss outcomes. Sensitivity requires companies to disclose the potential loss amount in response to a hypothetical market price change. VaR takes this a step further by requiring disclosure of the probability associated with the potential loss amount. Specifically, FRR No. 48 requires that the probability of a loss greater than the disclosed amount be 5 percent or less. Finally, tabular disclosure reports neither possible loss outcomes nor probabilities, but provides information allowing users to compute their own estimates of loss outcomes.

While the sensitivity and VaR disclosures present probability and outcome data related to losses, the psychology research discussed earlier suggests that these dimensions may not completely capture an individual's perception of a firm's market risk. That is, if the risk perceptions of FRR No. 48 users are affected by information about gains and by feelings of dread and unknown, risk assessments may differ in ways not anticipated by the SEC. For example, although FRR No. 48 does not require disclosure about possible gain outcomes related to financial instruments and derivatives, there are instances where companies voluntarily disclose information about potential gains. Consequently, two otherwise identical companies reporting the same sensitivity or VaR risk measure may be judged as differentially risky if one of those companies voluntarily discloses information about potential gains from a financial or derivative instrument.

Dread and unknown also may be considered by investors judging market risk. Both may be heightened by negative publicity associated with large derivative losses and investors' lack of perfect knowledge or understanding of derivatives (Loomis 1995). The dread dimension seems particularly plausible for derivatives because the media tends to focus on big-loss situations (e.g., Orange County, Barings Bank, Procter & Gamble). Substantial research in psychology shows that perceptions of risk are determined, at least in part, by how easily a person can imagine or recall instances of an event or risk (Slovic et al. 1980). Such a process for judging risks, referred to as availability, is valid in most instances because more frequent events generally are easier to recall and their potential occurrence is easier to imagine. However, ease of recall and imagination are related to factors other than just the statistical frequency with which events occur. The media focus on large derivative losses, therefore, could contribut e to overestimation of risk associated with derivative use, because these situations are easier to recall (even though they may not be more frequent) than those where derivative use leads to zero or small losses.

While feelings of dread are most likely to occur in situations where companies use derivatives to speculate, dread also could occur in situations where companies use derivatives to hedge business risks. In the latter case, FRR No. 48 requires companies to report only the potential loss from the derivative instrument, and not the offsetting benefit on the hedged item. For example, a company using a forward contract to hedge the cash flows associated with selling its inventory would quantify only the loss associated with a potential decrease in the value of the forward contract. However, the corresponding increase in the value of the inventory, which led to the loss on the forward, would not be quantified. Because of this focus on the loss component, consideration of dread may result, increasing perceptions of risk where arguably it is not warranted. Even if the qualitative FRR No. 48 disclosures specifically state that the derivatives are being used for hedging, the effect of the "one-sided" quantitative risk information may not be eliminated or even reduced because individuals have difficulty simulating alternative outcomes (Yates and Stone 1992). Further, because not all hedges are perfect, the amount of the offsetting benefit is not known.

In summary, investors' risk perceptions may be more complex than anticipated by the SEC when FRR No. 48 was drafted. Based on research in other domains, it is possible that investors may consider dimensions beyond loss probabilities and outcomes--the focus of the FRR No. 48 disclosures. Specifically, if investors selectively consider possible gain outcomes, dread, and unknown, then their risk assessments may be systematically different from what the SEC intends.

PROBLEMS WITH FRR NO. 48

Flexibility Concerns

FRR No. 48 allows companies to use one of three possible reporting formats (tabular, sensitivity, and VaR) for each market risk category. Under the sensitivity and VaR formats, companies have discretion regarding the measurement basis chosen for calculating risk of loss--fair value, cash flow, or net income. In combination, this yields seven distinct (and often noncomparable) risk disclosures that can be used both within and across firms. For example, Dell Computer's 1998 market risk information reveals a sensitivity measure for interest rate risk and a VaR measure for foreign exchange risk. In 1997, USX Marathon Oil used fair value for measuring interest rate risk and net income for commodity and equity risks.

FRR No. 48 has been criticized previously for the flexibility it allows in reporting information about market risks (Elmy et al. 1998; Schrand and Elliott 1998). The SEC's decision to permit choice in FRR No. 48 disclosures was reportedly motivated by cost-benefit trade-offs. Opponents of uniformity argued that sophisticated risk-measurement models, such as VaR, would prove too costly for smaller registrants. Further, flexibility allows firms to tailor their disclosures and provide risk information in the most meaningful way. Despite these potential advantages, we argue that the disclosure flexibility in FRR No. 48 can cause users to make dramatically different risk assessments even when the underlying economics are the same. [4]

To illustrate, we develop an example of a bank with assets of $1.6 billion and shareholders' equity of $118 million. Using data for this bank along with actual historical interest rate data, we provide in Table 2 the quantitative interest rate risk disclosures under the three possible reporting formats and measurement bases. Panel A shows the tabular disclosure, Panels B and C show the sensitivity and VaR measures, respectively. Our example illustrates how differences among the three formats and risk measurement bases can affect risk assessments, holding constant economic attributes. In fact, because of the maturity schedules of this bank's assets and liabilities, the cash flow and net income sensitivities are of the opposite sign from the fair value sensitivities. FRR No. 48 allows the company to report any material loss number shown in Panels B and C of Table 2, and does not require that the company report the largest loss number. [5]

Comparisons are difficult across the risk disclosures using different measurement bases--cash flows, earnings, or fair value--because these three bases represent different sources and consequences of loss. For example, losses in cash flow may prevent the company from meeting financial obligations, while losses in fair value may impair economic capital. In addition, loss magnitudes often are not directly comparable due to the different assumptions about horizons over which the losses could occur. For example, a fair-value risk disclosure is a point-in-time measure that captures the immediate effect of a hypothetical price change over the remaining life of an instrument, such as a bond liability. In contrast, a cash-flow risk disclosure captures the effect of a hypothetical price change on the cash flows from that bond over a stated time period (e.g., one day, one month, one year). Referring to Panel C of Table 2, our example shows that the one-year cash flow VaR is a $1,715 loss, while the corresponding fair value VaR is a $5,507 loss. Unless users understand that these two measurement bases provide different information, they may erroneously infer that this company is less risky if the cash flow, instead of fair value, VaR is reported.

Even if the measurement basis is held constant, comparisons across formats (e.g., sensitivity vs. VaR) are problematic for several reasons. First, the underlying time period typically differs when comparing sensitivity and VaR. Sensitivities usually are reported as yearly (near-term) effects. In contrast, most VaRs represent daily losses, even though VaR also can be reported over a weekly, monthly, or yearly time period. [6] Our hypothetical bank data in Table 2 highlight the large dollar difference between the one-day VaRs (Panel C) and the one-year sensitivities (Panel B). Because these are typical disclosures, such a comparison could lead a user to view risk as lower when VaR is disclosed. Research has shown that behavior can vary dramatically depending on whether risk information is described for a short time period (like a one-day VaR) or a longer time period (like a one-year sensitivity) (Hutton and Wilkie 1980). Comparing is disclosed. Research has shown that behavior can vary dramatically depending o n whether risk information is described for a short time period (like a one-day VaR) or a longer time period (like a one-year sensitivity) (Hutton and Wilkie 1980). Comparing the one-year VaR's (Panel C) to the one-year sensitivities (Panel B) shows that they are much more comparable when computed over similar time periods.

Second, directly comparing sensitivity and VaR measures is inappropriate, because they do not measure outcomes of equal probability. VaR is the maximum loss that will occur with a specified probability, which FRR No. 48 caps at 5 percent. For sensitivity, companies are required to calculate the loss associated with a hypothetical change in rates of at least 10 percent, but they are not required to state the probability associated with that hypothetical rate change. The probability associated with the hypothetical rate change depends on the time period over which the rate change is assumed to happen. That is, the likelihood of an instantaneous 10 percent rate change on any given day is very small. In contrast, the probability of a 10 percent rate change over a year is much more likely, and it is this more likely scenario that FRR No. 48 requires for sensitivity disclosure. Thus, comparing the one-year sensitivities and VaRs reported in Panels B and C of Table 2, respectively, yields two measures that differ n ot only in dollar amounts but also in underlying probability. Our concern is that FRR No. 48 users may not realize that the underlying probability levels differ between the two disclosure formats and conclude that the smaller of the sensitivity or VaR measure means less risk.

Although FRR No. 48 readers might attempt to remedy these problems by converting the available information to comparable measurement bases, time periods, or formats, such conversions are unlikely for two reasons. First, because of a lack of underlying information, it is difficult, if not impossible, to convert risk information from one disclosure format or measurement base to another. For example, Beder (1995) argues that it is computationally difficult to extrapolate a one-day VaR to a one-year time period to render it comparable to sensitivity. The least problematic conversion is to move from the tabular format to either sensitivity or VaR. However, to do this, an individual must make assumptions about prepayment rates, price elasticities, and strike prices, project future cash flows, ascertain historical rates, calculate time-series variances for each price and co-variances among prices, make present-value calculations, and calculate confidence intervals. Even then the results would only approximate the e stimates that management would have generated using their own private information.

Second, even in the limited cases where a user could move from one time period, disclosure format, or measurement basis to another, research indicates that decision makers are unlikely to do so (Slovic 1972), because the perceived costs (in terms of effort) associated with making the conversion are often deemed greater than the perceived benefits (Payne et al. 1990). Substantial research shows that information that is explicitly provided will be used more than information that must be inferred, estimated, or calculated (Johnson et al. 1988; Russo 1977; Anderson and Koonce 1998). Recent experimental evidence in accounting (Hirst and Hopkins 1998) shows that even sophisticated CFAs are susceptible to this effect in a familiar, financial-reporting domain. Hirst and Hopkins (1998) find that unless information about the trading activities of marketable securities is clearly displayed (and, thus, the costs of analysis are relatively low), analysts do not detect earnings management via opportunistic sales and repur chases of securities.

The latitude in FRR No. 48 not only causes difficulties for users, but also increases the potential for company manipulation. One consequence is that flexibility will allow firms to disclose the most favorable perspective or fail to depict the entire risk picture. To illustrate, in 1997 General Motors discloses a fair-value loss sensitivity to an interest rate decrease of $257 million arising from the company's net fixed-rate liability position. The company also has variable-rate liabilities of $21 billion that are excluded from its fair-value sensitivity measure because, by definition, the fair value of variable rate debt is always the face amount. However, the variable rate debt does have considerable cash flow sensitivity: a 10 percent interest rate increase would have cost General Motors an additional $210 million in interest expense. Because General Motors discloses one sensitivity number only, users may believe that General Motors is only sensitive to loss from interest rate decreases.

Insufficient Quantitative Information about Risk

Some of FRR No. 48's requirements are deficient because they provide insufficient quantitative information for users to fully understand the riskiness of companies' financial instruments, derivatives, and other positions. These disclosure deficiencies are described below.

Reporting Just One Outcome from the Distribution

FRR No. 48 requires disclosure of one measure for each kind of market risk the company faces. [7] Typical is Dow Chemical's disclosure, "the VaR for one day, using a 95 percent confidence level at December 31, 1999, for foreign exchange, interest rate, and equity exposures, net of hedges was: foreign exchange--$5 million; interest rate--$40 million; and equity--$13 million." Existing research warns of the pitfalls of using summary risk statistics such as the ones above. Specifically, Slovic et ad. (1980) note that when there is uncertainty, a point estimate of risk can lead to overconfidence because people anchor on that estimate and fail to adequately consider other outcomes from the distribution (Kahneman and Tversky 1979b). Consequently, individuals may become overconfident about the single statistic presented and, thus, consider that outcome to be more likely than it actually is. For example, users may consider sensitivity or VaR measures to be expected values or even certain outcomes. Such behavior is c onsistent with Koonce (1992) who reports that when experienced auditors are asked to focus on one particular event or outcome, they come to believe more strongly that the event or outcome will occur.

Research shows that one way to avoid the pitfalls of a single summary statistic is to use multiple-scenario analysis to understand a range of future potential events (Kuhn and Sniezek 1996). Describing different points on the outcome distribution reduces the overconfidence typically observed for one specific prediction (because the individual contemplates multiple views of the

future) and reduces the element of the unknown. Empirically, though, only a small percentage of firms report multiple risk measures, presumably because FRR No. 48 does not require it. Of these firms, some report multiple outcomes on the loss side of the distribution. For example, USX Marathon voluntarily reports sensitivities to commodity price changes of both 10 and 25 percent. While this moves users away from an anchor point, the disclosure still falls short of presenting users with a full range of possible outcomes. Other firms report measures from both the loss and gain side. For example, in 1997, Rohm & Haas expresses its foreign exchange sensitivity in terms of a 10 percent appreciation in rates (which would result in a $9 million increase in the fair value of options and forwards) and of a 10 percent depreciation in rates (which would result in a $6 million decrease in the fair value of options and forwards).

One likely motivation for reporting gains is to reduce perceived risk (Luce and Weber 1986). As noted earlier, experimental evidence indicates that presentation of information about potential gains can moderate risk assessments (Luce and Weber 1986; Lopes 1984). That is, companies may communicate upside potential to alleviate the expected reaction to downside-only loss disclosures. Companies would have an incentive to do this particularly when the possible gain exceeds the loss, as in the Rohm & Haas example above. Because FRR No. 48 only requires loss disclosures, two otherwise identical companies may be perceived as differentially risky depending on whether they voluntarily report possible gains in their risk disclosures.

Because most firms report only one risk measure, individuals may attempt to compensate for this by developing their own range of possible outcomes. Such a task is complicated because in trying to self-generate additional outcomes, individuals will likely start from the provided risk number and linearly extrapolate to other possible outcomes (Dawes 1979). For example, if a company discloses that a 10 percent interest rate increase leads to a $2,000,000 loss, a reader who believes that interest rates will increase by 20 percent is likely to infer a $4,000,000 loss. Assuming linearity is inappropriate because gains or losses on derivative contracts with option-like characteristics increase exponentially with underlying prices. [8] Bancwest's 1998 10--K report illustrates this. The company reports that a 100- (200-) basis-point increase in interest rates results in a $3.6 million ($10.8 million) decline in net interest income. Had Bancwest not disclosed the 200-basis-point sensitivity, users might linearly extra polate from the 100-basis-point sensitivity and predict a $7.2 million decline in net interest income that understates the actual loss by about 50 percent.

Similarly, because prior research suggests that individuals tend to believe that outcome distributions are symmetric (Ashton 1982; Kahneman and Tversky 1973), FRR No. 48 users may believe that losses arising from price movements in one direction will be offset by gains if prices were to move the other way. Again, because of the option-like characteristics of many derivative contracts, there is no assurance that possible losses are symmetric with possible gains. In 1998, Bostonfed Bancorp disclosed that "a 200 basis point increase in interest rates will result in a decrease in fair value of $11 million, while a 200 basis point decrease in interest rates will result in an increase in fair value of only $1.4 million." Given the tendency for financial information users to assume outcome symmetry, Bostonfed's disclosure could impact investors' risk perceptions because it reveals that the potential gain is not as large as the potential loss in fair value.

FRR No. 48 recognizes limitations of the quantitative loss disclosures and requires that companies qualitatively describe those instruments that have leverage, option, and prepayment features. Even with these additional disclosures, though, highly knowledgeable users still will be unable to calculate other possible gain or loss outcomes, because sufficient detailed quantitative information is unavailable for this purpose. Consequently, the user is left either to rely on the single risk of loss disclosure or to perform an inappropriate linear extrapolation by ignoring the leverage, option, or prepayment features. Such behavior is consistent with research in psychology showing that individuals reduce complex problems to a manageable level by relying on simplifying heuristics (Payne et al. 1990).

Incomplete Coverage of Risk Disclosures

While FRR No. 48 covers many financial instruments and derivatives, it fails to cover all positions that lead to market risk exposure. As shown in Table 1, certain unhedged positions do not require quantitative loss disclosures under FRR No. 48. In particular, commodity positions and commodity contracts that settle in other-than-cash are exempt from the disclosure requirements. Thus, a company could be exposed to large market risks from its commodity positions but because the underlying positions are outside the scope of FRR No. 48 (i.e., they are not financial or derivative instruments), the company is not required to quantify market (commodity) risk. However, if the same firm uses derivative instruments to hedge some or all of its commodity position, then the company must comply with FRR No. 48 requirements, but only for the hedging instruments. This leads to a counterintuitive situation where the risks associated with unhedged positions go unquantified, but perfectly hedged positions, where commodity price risk is negligible, require possible derivative losses to be quantified.

To illustrate, DuPont in its 1997 MD&A does not disclose any market risk information about its existing commodity inventory because the risk of loss from that inventory is not being hedged. Thus, DuPont could be judged by some readers as riskless whereas more-knowledgeable users might realize that there is considerable risk in an unhedged position. Because sufficient detailed information is not presented, more-knowledgeable users will not be able to accurately quantify the risk from the unhedged position. If risk of the unknown affects these knowledgeable users' risk judgments, as suggested by Slovic (1987), they would likely increase their perception of DuPont's riskiness because of the company's unhedged commodity position. However, it is unclear whether the resulting risk assessments would be appropriate or match those that would result if DuPont had actually quantified the risk arising from the unhedged position.

FRR No. 48 indicates that, in situations like this, companies can voluntarily report quantitative market risks [9] and if they do not, then they must disclose qualitative information about these positions. However, even if companies do disclose this information, it is very difficult, if not impossible, for individuals to use this qualitative information to generate their own quantitative risk assessments. We maintain that the qualitative information about unhedged commodity positions does not compensate for the lack of quantitative market risk information.

Aggregating and Disaggregating Risks

Under FRR No. 48, companies are required to report separate risk information for interest rates, foreign currency rates, commodity prices, and equity prices. However, there is no requirement that companies disclose the component risks within each market-risk type. For example, if a company has financial assets denominated in Korean won and liabilities denominated in Japanese yen, only the risk associated with the total position (assets less liabilities) need be quantified. In sensitivity disclosures, for example, the company is permitted to assume an identical parallel shift in each exchange rate (e.g., a uniform weakening or strengthening of U.S. dollars by 10 percent relative to each currency), even though it is unlikely that both the Korean and the Japanese currencies would appreciate or depreciate simultaneously and/or by equal percentages. For both sensitivity and VaR, the company is allowed to report one aggregate measure; that is, won gains may be netted against yen losses or vice versa. To arrive at t he disclosed net loss, the company separately determines how a weakening and a strengthening of the U.S. dollar would affect their net position and then reports the loss number.

Such aggregate reporting has several consequences. First, it can obscure important information about the component risks. Research demonstrates improvements injudgment accuracy and reliability when complex problems are decomposed (Kleinmuntz et al. 1996; Webby and O'Connor 1996). The idea is that breaking down a problem into its component parts may help the decision maker reason through the problem and recognize the uncertainties in each component. Research by Russo et al. (1986) shows that consumers who were provided with all of the nutritional information of a product used that information more than they used a summary measure. The benefits of disaggregation also have been shown in accounting; components of net income provide information over and above what is contained in the aggregated net income figure (Lipe 1986). By analogy, users of FRR No. 48 information likely would find that disaggregated components of sensitivity or VaR provide information incremental to the current aggregate numbers. For example , if the company in our won-yen example reveals its near-term intentions to refinance the yen-denominated debt with U.S. dollar debt, disaggregated VaRs or sensitivities would allow users to discount the market risk associated with the yen exposure. An aggregate risk measure would make such an analysis impossible.

Second, individuals' reactions may differ when component risks are detailed, rather than aggregated. Research on "mental accounting" suggests that how gains and losses are disclosed can have a significant impact on behavior. Based on the fundamental notions of prospect theory where people react to losses more severely than they do to gains,[10] Thaler (1999) argues that whether losses and gains are aggregated or disaggregated can affect individual behavior. More favorable risk assessments are likely to occur when: (1) multiple gains are segregated rather than aggregated (because the gain function is concave), (2) multiple losses are combined (because the loss function is convex), (3) smaller losses are combined with larger gains (because the larger gain completely offsets the loss), and (4) smaller gains are segregated from larger losses (because the gain function is steepest at the origin, the utility of a small gain can exceed the utility of slightly reducing a large loss). Thus, in the foreign-currency ex ample above, a company that follows the guidelines of FRR No. 48 and discloses one aggregate immaterial net won-induced loss likely will cause investors to consider there to be less risk than had the company disaggregated the net outcome into the yen-induced gain and the won-induced loss.

Tabular data can provide sufficient information to overcome many of the problems associated with aggregate risk measures. Because the tabular format requires that companies group instruments based on common characteristics, there is greater detail provided about these instruments. By making various assumptions, users could construct risk measures at a more disaggregated level than that provided by the sensitivity or VaR disclosures (e.g., Ahmed et al. 1999; Hodder and McAnally 2000). However, as noted earlier, tabular disclosure, even when supplemented with qualitative information about instrument characteristics, is insufficient for determining the quantitative effects of embedded options or other nonlinear positions when these positions are significant. Also, as noted earlier, conversions from one format to another are costly in terms of effort and time, and individuals may not deem their benefits worth these costs.

While there are benefits from disaggregating component risks, there are benefits associated with aggregation, as well. [11] Aggregate risk measures can allow users to identify large risks that are the result of positive correlation among component market risks. To illustrate, many argued that the DC-10 aircraft failed in several early flights because designers had not realized how the systems functioned as a whole. That is, the designers failed to realize that decompression of the cargo compartment would increase the risk associated with vital control systems, thereby increasing the overall risk of aircraft failure (Slovic et al. 1980).

Under FRR No. 48, companies are not required to report an overall (company-wide) VaR or sensitivity. Because of this, users may attempt to infer a firm's overall level of market risk by combining components additively across market risk categories. Such aggregation would lead to erroneous risk assessments because it is not possible to combine sensitivity numbers across market risk categories, and individual VaRs cannot be added if the component risks co-vary. [12] Research in psychology underscores the difficulty that individuals have in combining multiple component risks to form an assessment of overall risk (Bettman et al. 1986).

If FRR No. 48 were to require companies to report an aggregate risk measure, then it also would need to require them to disclose their risk-management objectives and goals quantitatively so that overall risk measures could be evaluated appropriately. Currently, FRR No. 48 requires companies to discuss risk-management goals and practices in qualitative terms, but it does not require them to communicate this information in quantitative terms. [13] For example, Bank One's 1998 annual report discloses that they have limits on their overall market risk exposure. These limits are approved by the board of directors and monitored on a daily basis by management. Bank One further voluntarily discloses that their actual overall VaR for 1998 is $29 million. Without quantitative information about Bank One's risk limits, interpreting Bank One's overall VaR is difficult.

In summary, FRR No. 48 requirements are deficient because they do not require sufficient quantitative information for users to form valid risk judgments. We argue that information about other possible outcomes, unhedged positions, and component as well as overall risks would facilitate users' risk assessments.

CONCLUSIONS

As Barth notes (in Schrand and Elliott 1998), we know little about how investors assess risk. In this paper, we draw on judgment and decision-making research to conjecture as to how users of FRR No. 48 disclosures might think about risk and how certain aspects of FRR No. 48 might adversely influence those risk assessments. Such adverse effects at the individual user level are important, because considerable research suggests that the potential judgment and decision-making problems we document in this paper will not be mitigated by market forces. Libby (1989) argues that while highly efficient markets can protect less effective decision makers from themselves, they cannot correct systematic judgment biases where people tend to err in the same direction (e.g., Ganguly et al. 1994; Tuttle et al. 1997). Because many of the judgment problems identified in this paper are systematic due to lack of comparable information or lack of detailed quantitative information, market forces and/or the presence of highly knowle dgeable users such as financial analysts, will likely not remedy the shortcomings of FRR No. 48.

FRR No. 48 will soon come under review by the SEC. We believe that our conclusions could assist the SEC in determining whether and how to modify the Release. First, we conclude that the construct of risk is more complex and multidimensional than perhaps anticipated by the SEC when they issued FRR No. 48. The existing literature indicates that risk perceptions can be influenced by gains, dread, and unknown, even after controlling for probabilities and outcomes--the focus of the SE C's risk disclosures. Risk scholars have concluded that risk is context specific (Fischhoff et al. 1998). That is, how individuals think about risk in one domain may not be how they consider it in another. Once researchers understand how FRR No. 48 users view risk in the financial instruments and derivatives context, then they can explore the issues associated with the various disclosure formats, measurement bases, and other features of FRR No. 48.

If FRR No. 48 users do consider the dimensions of dread and unknown when evaluating the market risk disclosures, it is possible that more detailed quantitative disclosures could mitigate the effects of dread and unknown. That is, by quantifying the extent to which risk management could be successful in achieving the company's risk management objectives, FRR No. 48 users may perceive less dread. Similarly, by disclosing risk for all risk-sensitive positions, including unhedged commodities, the element of unknown might be reduced.

Second, we conclude that FRR No. 48's flexibility of application will adversely affect users' risk judgments. Specifically, alternative disclosure formats and measurement bases are not substitutes and, to the extent they are viewed as such, investors will form inconsistent risk perceptions for the same underlying economic situation. If the SEC wants investors to be able to compare risk management strategies across companies, then they should mandate just one type of disclosure format. If the SEC is concerned that all material risks be disclosed, then each market risk (as currently defined by FRR No. 48) should be quantified using all three measurement bases, rather than a single one.

Third, we conclude that because FRR No. 48 does not require the disclosure of certain quantitative information known to be important to risk assessments, additional quantitative disclosures might be useful. For example, the presentation of the complete distribution for VaR or the likelihoods of a wide range of hypothetical rate changes for sensitivity may improve decision making. Expanded quantitative disclosure also may serve to overcome users' inappropriate fixation on single-point loss outcomes and their use of inappropriate extrapolation techniques. Mandated disclosure of potential gains would avoid problems associated with investors forming different risk assessments for similar economic circumstances. Finally, the presentation of both component and aggregate risk information would allow users to better evaluate the causes and consequences of market risks. Because the FRR No. 48 disclosures are complex, research is needed to test these recommendations.

We also believe that our paper has implications beyond FRR No. 48. As we move toward more comprehensive risk disclosures (as noted by Schrand and Elliot [1998]) and as SFAS No. 133 is adopted, the insights in this paper can be used as a basis for developing and understanding other risk disclosures. Because the large literature on risk in the judgment and decision-making arena has demonstrated a number of consistencies across tasks and domain, that literature is a very useful starting point for understanding risk in financial reporting.

Leslie Hodder is a Doctoral Candidate, Lisa Koonce is an Associate Professor, and Mary Lea McAnally is an Assistant Professor, all at The University of Texas at Austin.

(1.) Noncompliance may be a temporary phenomenon as companies become more familiar with FRR No. 48 requirements and the SEC steps up disclosure enforcement. The types of noncompliance documented by Elmy et al. (1998) and Roulstone (1999) exacerbate the problems that we identify in the paper.

(2.) Separate disclosures also are required for "trading" and "other than trading" instruments.

(3.) Our own analysis of 40 firms from five industries--airlines, automotive, oil and gas, computer hardware, and chemical manufacturing--as well as 91 banking firms shows similar breakdowns in terms of reporting format and measurement bases used. The only exception is that banks tend to use tabular to a greater extent (33 percent) and VaR to a lesser extent (2 percent) than other industries. For these two samples, we also checked the number of times a firm switched from one reporting format to another and found no cases of switching in the 40-firm sample and only four incidents of switching in the banking sample. Thus, variability in disclosure formats is across companies and not across time.

(4.) Although the SEC is expected to revisit FRR No. 48 and possibly reduce disclosure flexibility, at least one industry (i.e., banking) has argued for retaining the flexibility in how risk is communicated (Basle Committee on Banking Supervision 1996).

(5.) FRR No. 48 does not explicitly define materiality. For purposes of our example, we use net income as a benchmark to gauge materiality.

(6.) The use of the VaR methodology requires a number of key assumptions including the selection of a confidence level for expected losses and the portfolio holding period over which losses are incurred. FRR No. 48 prescribes a minimum confidence level, but does not specify a holding period. One-day VaR is an artifact of the method's historical use as an internal risk measurement and control tool. VaR was originally designed and used by investment firms to measure portfolio risk and monitor trading activities. Because trading portfolios can usually be liquidated (or "unwound") within a day, a single day is an appropriate length of time over which to calculate VaR for active trading portfolios. For other contexts, such as those to which FRR No. 48 applies, it is less reasonable to assume that a company would quickly unwind its asset and liability positions. Thus, a one-day VaR may not accurately measure most companies' underlying market risk. The latter is of concern because psychology research shows that peop le fail to realize the cumulative effect of exposure to risks, sometimes overestimating the risk and at other times underestimating it (Linville et al. 1993; Svenson 1985).

(7) For the tabular display where no summary risk numbers are displayed, users are given the raw materials to (presumably) calculate their own risk measure.

(8.) When a portfolio includes instruments that are not linearly dependent on underlying market rates, VaRs will always be nonlinear and asymmetric (Duffie and Pan 1997). Even absent explicit or embedded option features, however, many financial instruments are nonlinear functions of underlying rates.

(9.) Far example, Mobil Oil discloses, in 1997, that the "value-at-risk analysis of commodity price risk includes managed physical commodities as well as hedging and trading derivatives because Mobil manages this risk on a combined basis."

(10.) Prospect theory defines value in terms of gains and losses rather than final wealth. Key to prospect theory is the idea that the value function is different for gains than for losses. Specifically, the value function is convex and quite steep for losses and concave and less steep for gains (Kahneman and Tversky 1979a).

(11.) We believe that it is not contradictory to call for an overall risk measure as well as disaggregated information about risk. An overall risk measure would allow an FRR No. 48 user to understand and monitor the combined risks of a company, while the disaggregated risk data allows him/her to assess the source and magnitude of the component risks.

(12.) Users could sum VaRs across risk types only if the company had calculated the individual VaRs by first calculating an overall VaR and then partitioning that total among the individual risk components (i.e., interest rate, commodity, foreign currency, and equity)

(13.) FRR No. 48 requires "contextual disclosures" that help users assess the validity of the market risk disclosures for one company over time. For example, for the VaR disclosures, companies are to report "either (1) the average or range in value-at-risk amounts for the current reporting period, (2) the average or range in actual changes in fair values, earnings, or cash flows from market risk sensitive instruments occurring during the current reporting period, or (3) the percentage of actual changes in fair values, earnings, or cash flows from market risk sensitive instruments that exceeded the reported value-at-risk amounts during the current reporting period." Although analogous disclosures are not required for sensitivity, they are strongly recommended. As more companies begin to provide these disclosures (see Roulstone [1999] for evidence that compliance on these disclosures is currently low), FRR No. 48 users will be better able to assess the validity of the VaR and sensitivity disclosures for one com pany over time. However, because these disclosures are to be made by market risk category (interest rate, foreign currency, commodity, and equity), they would not address the aggregation issue.

REFERENCES

Ahmed, A., A. Beatty, and B. Bettinghaus. 1999. Evidence on the efficacy of market risk disclosures by commercial banks. Working paper, Pennsylvania State University.

Alhakami, A., and P. Slovic. 1994. A psychological study of the inverse relationship between perceived risk and perceived benefits. Risk Analysis 14 (December): 1085-1096.

Anderson, U., and L. Koonce. 1998. Evaluating the sufficiency of causes in audit analytical procedures. Auditing: A Journal of Practice & Theory 17 (Spring): 1-12.

Arkes, H. 1991. Costs and benefits of judgment errors: Implications for debiasing. Psychological Bulletin 110 (November): 486-498.

Ashton, R. 1982. Human In formation Processing in Accounting. Studies in Accounting Research No. 17. Sarasota, FL: American Accounting Association.

Basle Committee on Banking Supervision (Bank for International Settlements). 1996. Compendium: Capital adequacy. Overview of the Amendment to the Capital Accord to Incorporate Market Risks, No. 23, January. Basle, Switzerland.

Beder, T. 1995. VAR: Seductive but dangerous. Financial Analysts Journal 51 (September/October): 12-24.

Bettman, J., J. Payne, and R. Staelin. 1986. Cognitive considerations in designing effective labels for presenting risk information. Journal of Public Policy and Marketing 5: 1-28.

Dawes, R. 1979. The robust beauty of improper linear models in decision making. American Psychologist 34 (July): 571-582.

Dreman, D. 1998. Contrarian Investment Strategies. New York, NY: Simon and Schuster.

Duffie, D., and J. Pan. 1997. An overview of value at risk. Journal of Derivatives 4 (Spring): 7-49.

Elmy, F., L. LeGuyader, and T. Linsmeier. 1998. A review of initial filings under the SEC's new market risk disclosure rules. Journal of Corporate Accounting and Finance (Summer): 33-45.

Fischhoff, B., S. Litchenstein, P. Slovic, S. Derby, and R. Keeney. 1981. Acceptable Risk. New York, NY: Cambridge University Press.

_____, D. Riley, D. Kovacs, and M. Small. 1998. What information belongs in a warning? Psychology and Marketing 15 (October): 663-686.

Froot, K., D. Scharfstein, and J. Stein. 1993. Risk management: Coordinating corporate investment and financing policies. The Journal of Finance 48 (December): 1629-1658.

Ganguly, A., J. Kagel, and D. Moser. 1994. The effects of biases in probability judgments on market prices. Accounting, Organizations and Society 19 (November): 675-700.

Heisler, J. 1994. Loss aversion in a futures market: An empirical test. Review of Futures Markets 13: 793-822.

Hirst, E., and P. Hopkins. 1998. Comprehensive income reporting and analysts' valuation judgments. Journal of Accounting Research 36 (Supplement): 47-75.

Hodder, L., and M. L,. McAnally. 2000. SEC market risk disclosures: Enhancing comparability. Working paper, The University of Texas at Austin.

Holtgrave, D., and E. Weber. 1993. Dimensions of risk perception for financial and health and safety risks. Risk Analysis 13 (October): 553-558.

Hutton, R., and W. Wilkie. 1980. Life cycle cost: A new form of consumer information. Journal of Consumer Research 6 (March): 349-360.

Johnson, E., J. Payne, and J. Bettman. 1988. Information displays and preference reversals. Organizational Behavior and Human Decision Performance 42 (August): 1-21.

Kahneman, D., and A. Tversky. 1973. On the psychology of prediction. Psychological Review 80 (July): 237-251.

_____. 1979a. Prospect theory: An analysis of decision under risk. Econometrica 47 (March): 263-291.

_____. 1979b. Intuitive prediction: Biases and corrective procedures. In Forecasting, edited by S. Makridakis, and S. Wheelright. TIMS, Studies in Management Science 12: 313-27.

Kleinbaum, D., L. Kupper, and H. Morgensten. 1982. Epidemiologic Research: Principles and Quantitative Methods. Belmont, CA: Lifetime Learning Publications.

Kleinmuntz, D., M. Fennema, and M. Peecher. 1996. Conditioned assessment of subjective probabilities: Identifying the benefits of composition. Organizational Behavior and Human Decision Processes 66 (April): 1-15.

Koonce, L. 1992. Explanation and counterexplanation in audit analytical review. The Accounting Review 67 (January): 59-76.

_____, M. McAnally, and M. Mercer. 2000. How do investors judge the risk of financial and derivative instruments? Working paper, The University of Texas at Austin.

Kuhn, K., and J. Sniezek. 1996. Confidence and uncertainty in judgmental forecasting: Differential effects of scenario presentation. Journal of Behavioral Decision Making 9 (December): 231-247.

Libby, R. 1989. Experimental research and the distinctive features of accounting settings. In The State of Accounting Research as We Enter the 1990s, edited by T. Frecka, 126-147. Champaign, IL: University of Illinois.

Linsmeier, T. and N. Pearson. 1997. Quantitative disclosures of market risk in the SEC release. Accounting Horizons 11 (March): 107-135.

_____, D. Thornton, M. Venkatachalam, and M. Welker. 2000. Do market risk disclosures improve consensus about market risk exposures? Working paper, Queen's University.

Linville, P., G. Fischer, and B. Fischhoff. 1993. Perceived risk and decision making involving AIDS. In The Social Psychology of HIV Infection, edited by J. Pryor, and G. Reeder. Hillsdale, NJ: Erlbaum.

Lipe, M. 1998. Individual investors' risk judgments and investment decisions: The impact of accounting and market data. Accounting, Organizations and Society 23 (October): 625-640.

Lipe, R. 1986. The information contained in the components of earnings. Journal of Accounting Research 24 (Supplement): 37-68.

Loewenstein, G., E. Weber, C. Hsee, and E. Welch. 1999. Risk as feelings. Working paper, Carnegie Mellon University.

Loomis, C. 1995. Untangling the derivatives mess. Fortune 131 (March 20): 50.

Lopes, L. 1984. Risk and distributional inequity. Journal of Experimental Psychology: Human Perception and Performance 26 (August): 465-485.

Luce, D., and E. Weber. 1986. An axiomatic theory of conjoint, expected risk. Journal of Mathematical Psychology 30 (June): 188-205.

March, J., and Z. Shapira. 1987. Managerial perspectives on risk and risk taking. Management Science 33 (November): 1404-1418.

Markowitz, H. 1959. Portfolio Selection. New York, NY: Wiley.

Olsen, R. 1997. Investment risk: The experts' perspective. Financial Analysts Journal 53 (March/ April): 62-66.

Payne, J., and M. Braunstein. 1971. Preferences among gambles with equal underlying distributions. Journal of Experimental Psychology 87 (January): 13-18.

_____, J. Bettman, and E. Johnson. 1990. The adaptive decision maker: Effort and accuracy in choice. In Insights in Decision Making: A Tribute to Hillel J. Einhorn, edited by R. Hogarth, 129-153. New York, NY: Cambridge University Press.

Rajgopal, S. 1999. Early evidence on the informativeness of the SEC's market risk disclosures:

The case of commodity price risk exposure of oil and gas producers. The Accounting Review 74 (July): 251-280.

_____, and M. Venkatachalam. 1999. Are earnings sensitivity measures risk-relevant? The case of oil price risk for the petroleum refining industry. Working paper, University of Washington.

Roulstone, D. 1999. Effect of the SEC Financial Reporting Release No. 48 on derivative and market risk disclosures. Accounting Horizons 13 (December): 343-363.

Russo, J. 1977. The value of unit price information. Journal of Marketing Research 14 (May): 193-201.

_____, R. Staelin, C. Nolan, G. Russell, and B. Metcalf. 1986. Nutrition information in the supermarket. Journal of Consumer Research 13 (June): 48-70.

Schrand, C., and J. Elliott. 1998. Risk and financial reporting: A summary of the discussion at the 1997 AAA/FASB Conference. Accounting Horizons 12 (September): 271-282.

Securities and Exchange Commission (SEC). 1997. Disclosure of Accounting Policies for Derivative Financial Instruments, and Derivative Commodity Instruments and Disclosure of Quantitative and Qualitative In formation about Market Risk Inherent in Derivative Financial Instruments, Other Financial Instruments, and Derivative Commodity Instruments. Release Nos. 33-7386; 34-38223; IC-22487; FR 48; International Series No. 1047; File No. 27-35-95. Washington, D.C.: SEC.

Slovic, P. 1966. Trust, emotion, sex, politics, and science: Surveying the risk-assessment battlefield. In Psychological Perspectives to Environment and Ethics in Management, edited by M. Bazerman, D. Messick, A. Tenbrunsel, K. Wade-Benzoni, 217-313. San Francisco, CA: Jossey-Bass.

_____, and S. Lichtenstein. 1968. Importance of variance preferences in gambling decisions. Journal of Experimental Psychology 78: 646-654.

_____. 1972. From Shakespeare to Simon: Speculations--and some evidence--about man's ability to process information. In Oregon Research Institute Bulletin. Eugene, OR: Decision Research.

_____. 1987. Perception of risk. Science 236 (April): 280-285.

_____, B. Fischhoff, and S. Lichtenstein. 1980. Facts and fears: Understanding perceived risk. In Societal Risk Assessment: How Safe is Safe Enough?, edited by R. Schwing and W. Albers, Jr. San Francisco, CA: Jossey-Bass.

Svenson, O. 1985. Cognitive strategies in a complex judgment task: Analyses of concurrent verbal reports and judgments of cumulated risk over different exposure times. Organizational Behavior and Human Decision Processes 36 (August): 1-15.

Thaler, R. 1999. Mental accounting matters. Journal of Behavioral Decision Making 12 (September): 183-206.

Thornton, D., and M. Welker. 1999. Impact of market risk disclosures on stock price sensitivity to oil and gas prices. Working paper, Queen's University.

Tuttle, B., M. Coller, and E. Burton. 1997. An examination of market efficiency: Information order effects in a laboratory. Accounting, Organizations and Society 22 (January: 89-103.

Viscusi, W., W. Magat, and J. Huber. 1986. Informational regulation of consumer health risks: An empirical evaluation of hazard warnings. Rand Journal of Economics 17 (Autumn):351-365.

Webby, R., and M. O'Connor. 1996. Judgmental and statistical time series forecasting: A review of the literature. International Journal of Forecasting 12 (March): 91-118.

Weber, E. 1988. A descriptive measure of risk. Acta Psychologica 69 (November): 185-203.

_____, and W. Bottom. 1989. Axiomatic measures of perceived risk: Some tests and extensions. Journal of Behavioral Decision Making 2 (April-June): 113-131.

_____. 1990. An empirical evaluation of the transitivity, monotonicity, accounting, and conjoint axioms for perceived risk. Organizational Behavior and Human Decision Processes 45 (April): 253-275.

Yates, F. and E. Stone. 1992. The risk construct. In Risk-Taking Behavior, edited by F. Yates, 1-25. New York, NY: John Wiley & Sons.

                   FRR No.48 Disclosure Requirements
Disclosure Required For              Disclosure Encouraged But Not
                                     Required For
* Cash and foreign currency          * Leases
* Marketable debt and equity         * Pension assets and liabilities
  securities
* Short- and long-term debt          * Insurance contracts
  obligations
* Commodity derivative contracts     * Other post-retirement and
  that settle in cash or in a           post-employment benefit
  financial instrument                 obligations
* Derivative instruments including   * Underlying position in hedged
  swaps, futures and forward           commodity
  contracts, options                 * Anticipated transactions
* Other financial instruments,       * Commodity derivative contracts
  structured notes, mortgage-backed    that do not settle in cash or
  securities, IOs, or POs              in a financial instrument
                  Sample FRR No. 48 Disclosures for a
                       Hypothetical Banking Firm
                      Panel A: Tabular Disclosure
                   Assets and Liabilities Subject to
                    Interest Rate Risk (in thousands
                              of dollars)
                                      Maturity     Non-
              Maturity    Maturity    Greater    Interest
              Less than    One to       Than     Bearing
              One year   Five Years  Five Years   Items [a]
Assets        $967,621    $437,232    $178,172   $ 9,361
Liabilities    866,544     117,282     220,733   387,827
Net position   101,077     319,950    (42,561)
Average rate       6.5%        7.0%        7.4%
               Balance
                Sheet        Fair
                Total       Value
Assets        $1,592,386  $1,560,381
Liabilities    1,474,220   1,419,545
Net position     118,166
Average rate
                   Panel B: Calculations for Sensitivity
                         Disclosure [b] (in thousands
                                 of dollars)
             One-Year Sensitivity Analysis
                Interest Rate Increase      Interest Rate Decrease
Cash Flows             $1,768.82                 $(1,768.82)
Net Income              1,149.73                  (1,149.73)
Fair Values            (5,676.04)                  5,747.35
                  Panel C: Calculations for Value-at-
                     Risk (VaR) Disclosure [c] (in
                         thousands of dollars)
                   One-Day                 One-Year
             Value at Risk (VaR) [d]  Value at Risk (VaR) [e]
Cash Flows        $(106.13)               $(1,714.58)
Net Income          (68.99)                (1,114.48)
Fair Values        (340.89)                (5,507.25)

(a.)Noninterest bearing assets are primarily loan loss reserves and certain fixed assets. Noninterest bearing liabilities are demand deposits. The tabular presentation is condensed from what would be shown in an actual disclosure. Information for years two through five has been aggregated into one category to conserve space here.

(b.)We calculate sensitivities as the change in cash flows, net income, or fair values that would result from an instantaneous 100-basis-point increase or decrease in interest rates. Cash flow and net income sensitivities represent the near-term (i.e, one-year) impact of the instantaneous change. We do not explicitly consider prepayment risk. Under FRR No. 48, only one sensitivity amount is disclosed (the others are shown here for illustration purposes only). In particular, FRR No. 48 states that if the loss magnitudes from cash flows, earnings, and fair value are all material, then the firm can choose the one that is "most appropriate." If only one loss is material, then that particular risk measure is disclosed.

(c.)We calculate Value-at-Risk numbers using historical distribution of daily one-year Treasury Constant Maturity Rates over the year. Under FRR No. 48, only one VaR amount is disclosed (the others are shown here for illustration purposes only). In particular, FRR No. 48 states that if the loss magnitudes from cash flows, earnings, and fair value are all material, then the firm can choose the one that is "most appropriate." If only one loss is material, then that particular risk measure is disclosed.

(d.)VaR represents the maximum cash flow, income, or fair-value loss the bank would incur over a one-day period, using a 95 percent confidence interval, calculated using a nonparametric (historical) approach. VaRs calculated using a parametric approach do not differ significantly.

(e.)VaR represents the maximum cash flow, income, or fair-value loss the bank would incur over a one-year period, using a 95 percent confidence interval, calculated using a nonparametric (historical) approach. It should be noted that one-year VaRs are not typical of the disclosures firms provide. They are included here for illustrative purposes.

Examples of Qualitative FRR No. 48 Disclosures

Intel 1999

Foreign Currency Risk, Particular Markets, and Management Strategies

The company hedges currency risks of investments denominated in foreign currencies with foreign currency borrowings, currency forward contracts and currency interest rate swaps. Gains and losses on these foreign currency investments would generally be offset by corresponding losses and gains on the related hedging instruments, resulting in negligible net exposure to the company. A substantial majority of the company's revenue, expense, and capital purchasing activities are transacted in U.S. dollars. However, the company does enter into these transactions in other currencies, primarily Japanese yen and certain other Asian and European currencies. To protect against reductions in value and the volatility of future cash flows caused by changes in currency exchange rates, the company has established revenue, expense, and balance sheet hedging programs. Currency forward contracts and currency options are utilized in these hedging programs. The company's hedging programs reduce, but do not always entirely eliminat e, the impact of currency exchange rate movements.

American Airlines 1999

Managing Interest Risk and Accounting Treatment of Interest-Rate Derivatives

American enters into interest rate swap contracts to effectively convert a portion of its fixed-rate obligations to floating-rate obligations. These agreements involve the exchange of amounts based on a floating interest rate for amounts based on fixed interest rates over the life of the agreement without an exchange of the notional amount upon which the payments are based. The differential to be paid or received as interest rates change is accrued and recognized as an adjustment of interest expense related to the obligation. The related amount payable to or receivable from counterparties is included in current liabilities or assets. The fair values of the swap agreements are not recognized in the financial statements. Gains and losses on terminations of interest rate swap agreements are deferred as an adjustment to the carrying amount of the outstanding obligation and amortized as an adjustment to interest expense related to the obligation over the remaining term of the original contract life of the terminat ed swap agreement. In the event of the early extinguishment of a designated obligation, any realized or unrealized gain or loss from the swap would be recognized in income coincident with the extinguishment.

Loews Corporation 1999

Speculative Use of Derivative Contracts

The most significant areas of market risk in the Company's trading portfolio result from positions held in S&P futures contracts, short sales of certain equity securities, and put options purchased on the S&P 500 index. The Company enters into these positions primarily to benefit from anticipated future movements in the underlying markets that Company management expects to occur. If such movements do not occur or if the market moves in the opposite direction from what management expects, significant losses may occur. In 1998, the Company started to reduce its exposure in certain positions. At December 31, 1999, the Company continued to maintain certain of these positions.

Examples of Quantitative FRR No. 48 Disclosures

Panel A: Tabular Disclosure

Peoples Financial Corporation 1999: Interest Rate Risk

The tabular disclosure reflects contractual interest rate repricing dates and contractual maturity dates. Loan maturities have been adjusted for reserve for loan losses. There have been no adjustments for such factors as prepayment risk, early calls of investments, the effect of the maturity of balloon notes or the early withdrawal of deposits. The Company does not believe that the aforementioned factors have a significant impact on expected maturity.

                         Expected Maturity Date
                                 2000     2001     2002     2003
(thousands of U.S. dollars)
Loans, net                   $108,413  $37,037  $38,541  $64,847
 Average rate                   8.61%    8.61%    8.61%    8.28%
Investments                    41,385   27,189   16,979   21,208
 Average rate                    5.40     5.62     5.95     5.62
Deposits                      302,848    7,914    2,473    2,068
 Average rate                    4.38     4.87     5.49     5.15
Long-term funds                    14       14       15       15
 Average rate                    5.38     5.38     5.38     5.38
                                                             Fair
                                2004   Beyond     Total     Value
(thousands of U.S. dollars)
Loans, net                   $65,979  $13,354  $328,171  $327,962
 Average rate                  8.28%    8.58%     8.48%
Investments                   19,763   23,472   149,996   148,432
 Average rate                   6.10     5.53      5.67
Deposits                       1,379       19   316,701   316,636
 Average rate                   5.14     4.72      4.41
Long-term funds                  216    --0--       274       256
 Average rate                   5.38     5.38      5.38

Panel B: Sensitivity Disclosure

General Motors 1999: Equity Price Risk Using Fair Value as a Measurement Basis

GM holds investments in various available-for-sale equity securities, which are subject to price risk. The fair value of such investments, as of December 31, 1999 and 1998, was approximately $3.2 billion and $2.3 billion, respectively. The potential change in the fair value of these investments, assuming a 10 percent change in prices, would be approximately $323 million and $230 million for 1999 and 1998, respectively.

Panel C: Value-at-Risk Disclosure

Dow Chemical 1999: VaR Using Fair Value as a Measurement Basis

Dow uses value at risk (VAR), stress testing, and scenario analysis for risk measurement and control purposes. VAR estimates the potential gain or loss in fair market values, given a certain move in prices over a certain period of time, using specified confidence levels. On an ongoing basis, the Company estimates the maximum gain or loss that could arise in one day, given a two standard deviation move in the respective price levels. These amounts are relatively insignificant in comparison to the size of the equity and earnings of the Company. The VAR methodology used by Dow is based primarily on the variance/covariance statistical model. The following table is given as an example[w1]: [w1]

                          Average Daily VAR at
                              December 31
                                     1999      1998
                                 (in million)
Foreign exchange                     $ 5       $ 4
Interest rate                         40        23
Equity exposures, net of hedges       13         6