A View Inside Corporate Risk Management*

Transcription

1 A View Inside Corporate Risk Management* Gordon M. Bodnar Johns Hopkins University John R. Graham Duke University & NBER Erasmo Giambona University of Amsterdam Campbell R. Harvey Duke University & NBER This Draft: June 7, 2014 Abstract A number of theories have been proposed to explain why firms hedge. Unfortunately, these theories are hard to test: While we might observe the hedges, it is hard to answer the question of why hedging occurs. Our paper attacks the why by directly questioning the managers that make the risk management decisions. Our results present a fresh, inside view of corporate risk management. Rather than hedging being conducted solely by firms, our results suggest that personal risk aversion in combination with other executive traits plays a key role in whether a company hedges. As such, our results suggest an important deficiency in many modern theories of risk management which ignore the role of the individual manager. Key words: Risk Management, Hedging, Managerial Risk Aversion, Behavioral Finance, Manager fixed-effects, Interest rate risk, Credit risk, Commodity risk, Foreign exchange risk. JEL classification: G02; G30; G32. *We are grateful for comments from John Buley, Jason Buser, Adriano Rampini, Henri Servaes, René Stulz, and Vish Viswanathan. We thank CFO magazine, the International Swaps and Derivatives Association (ISDA), and the Global Association of Risk Professionals (GARP) for helping us conduct the survey, though we note that our analysis and conclusions do not necessarily reflect those of CFO, ISDA, or GARP.

2 Introduction Why do firms hedge? It is very difficult to answer this basic question. Traditional economic theory suggests that firms rely on risk management to mitigate financial constraints or other market frictions due to informational asymmetries or agency problems. However, across the world, large, rated, dividend-paying firms (arguably firms that are less affected by market imperfections) are significantly more likely to hedge than their small, unrated, non-dividend paying counterparts (see Figure 1). 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 76% 63% 62% Sample Average: 52% 41% 42% 40% Large: vs. No Rated: vs. No Dividend Payer: vs. No Figure 1 Risk Management and Firm Characteristics This figure reports the percentage of firms with a formal risk management program in place. The data are from our Corporate Risk Management Survey, which was conducted in the first quarter of The presented sample includes non-financial firms from around the globe. Firms are defined as Large: if their sales are above $1 billion, Rated: if they have a credit rating for their debt, and Dividend Payer: if they pay regular dividends. These simple empirical facts motivate the central questions of our paper: What explains the wide variation in risk management practices across firms, and are these explanations captured by extant theories of risk management? The limited availability of data on corporate risk management has presented a major challenge to answering these questions. To assess current models of risk management, the empiricist needs detailed data on the firm s propensity to hedge, the extent of risk management, the motivation for hedging, the role played by executives in the decision to set up the risk management program, and managerial characteristics including executives attitudes towards risk. At an even more basic level, one also needs to identify whether the firm is facing any material hedgeable risks. This type of information is not available in standard archival data sources (such as COMPUSTAT), which only collect data on observed outcomes as recorded in annual reports or other corporate documents. To mitigate these limitations, we gather information from 690 CFOs from around the world. In this 2010 survey, we ask CFOs whether their companies have a formal risk management program in place, 1 their inside views on their firms growth prospects, and their views of the interconnected roles played by lines of credit and cash in hedging and liquidity management. We also ask the CFOs to reveal the motivations behind why their firms do (or do not) hedge; that is, we ask them to 1 Throughout the study we use risk management and hedging interchangeably. 1

3 answer the central questions of our paper. For instance, we ask the CFOs about the extent to which risk management is used as an instrument to reduce cash flow volatility or whether accounting standards might limit their firms usage of hedging instruments. We use the firm s ex ante motivation for hedging to study the relation between market frictions (credit rationing, information asymmetry, or agency problems) and the firm s decision to hedge. There is an important advantage to studying hedging intentions rather than ex post hedge outcomes. A firm might intend to hedge but due to (potentially) extraordinary economic circumstances, not follow through on their intentions. We observe the intentions or ex ante plans, which should not be contaminated by ex post factors that impact the execution of risk management. By asking about motivations for hedging, our goal is to shed light on the relation between the type of frictions that the firm faces and the role of risk management in mitigating these frictions. In addition to evaluating existing theories, we administer a psychometric test to gauge managerial risk aversion. We combine this information with demographic data on executive characteristics related to compensation age, experience, and education. This allows us to study the nexus between personal-risk aversion and corporate-risk management decisions. To our knowledge, ours is the first paper to empirically examine the link between personal characteristics and corporate risk management. While the survey method can be used to gather unique information, there are limitations to any survey-data based study. For instance, it is possible that the sample of respondents is not representative of the underlying population. Similarly, it is possible that some of the questions are misunderstood or otherwise produce noisy measures of risk management decisions, firm characteristics, or managerial traits. In addition, when interpreting field studies, one needs to consider that market participants do not necessarily have to understand the reason they do what they do in order to make (close to) optimal decisions. Finally, we are limited to a single cross-section of firms. As a result, we cannot make any causal inference. To alleviate some of these concerns, we consulted with design experts, conducted focus groups, and performed beta tests to make sure that our questions are unambiguous. Further, we confirm that our sample is representative of data from standard archival databases in terms of basic demographics related to size, dividends, profitability, leverage, and cash holdings. While these findings are reassuring, they do not eliminate all the limitations of survey-based research. To recap, our global survey is well suited to study why firms hedge. There are, however, limitations with this approach. In this sense, our approach complements existing empirical papers (such as Tufano, 1996; or more recently, Rampini, Sufi, and Viswanathan, 2014) that study risk management practices using detailed data for a limited number of firms from one industry. We find that 52% of the firms in our sample have a formal risk management program in place. 2 Our analysis also shows that there is significant variation in risk management practice across regions (71% of European companies hedge compared to 45% of North American firms) and ownership forms (74% of public companies have a risk management program versus 38% of private firms). These differences persist when we control for firm characteristics such as size. 2 Bodnar et al. (2013) detail the results on many different types of risk management: credit risk, interest rate risk, foreign exchange risk, commodity risk, and geopolitical risk management. 2

4 We use information on the propensity to hedge to test existing theories of risk management. To the extent that large, rated, dividend-paying companies are less financially constrained than their small, unrated, non-dividend paying counterparts, evidence like that in Figure 1 is inconsistent with the credit rationing prediction that constrained firms hedge more (Froot, Scharfstein, and Stein, 1993) and consistent with Rampini and Viswanathan (2010, 2013). Importantly, we find that these patterns persist even if we focus on companies with high growth prospects, those more concerned with the effects of financial constraints on their ability to fund future investment. Theory also predicts that companies may substitute credit lines or cash holdings for risk management to deal with cash flow shortfalls (Froot, Scharfstein, and Stein, 1993; and Holmström and Tirole, 2000). If financially constrained companies were more likely to have access to credit lines or had more cash, this could explain why they are less likely to hedge. However, we find that financially constrained (small, unrated, non-dividend paying) firms are significantly less likely to have a formal risk management program (than their unconstrained counterparts), even after we control for access to credit lines or the availability of cash. Other studies have argued that economies of scale associated with setting up a risk management program (e.g., Mian, 1996; Bodnar, Hayt, and Marston, 1998) could explain why we do not find support for the credit rationing hypothesis. That is, smaller firms are often financially constrained but they may not have the scale to set up a risk management program. To deal with this concern, we rely on a testing strategy that bypasses the effect of economies of scale. An important part of credit rationing argument is that financially constrained companies hedge to reduce cash flow volatility. Therefore, in this part of our analysis we focus exclusively on firms with a risk management program in place and ask their CFOs to give us a qualitative assessment of the importance of risk management as a tool to reduce cash flow volatility. Across the board, we find that the CFOs indicate that risk management is an important reason to reduce the volatility of cash flows. However, we do not find that the desire to reduce the volatility of cash flows is more important for financially constrained firms; therefore, we do not find support for a key feature of the credit rationing hypothesis. To the extent that small firms face higher information asymmetry, our finding that small firms are less likely to establish a risk management program (and whether or not they have promising investment prospects) is also inconsistent with information asymmetry models of risk management such as DeMarzo and Duffie (1995) who predict that managers rely on risk management to signal private information about the investment prospects of the firm. Rampini and Viswanathan (2010, 2013) build a model that explains why financially constrained firms hedge less than unconstrained firms. They argue that financially constrained (e.g., small) companies are less likely to hedge because they focus their resources directly on investment rather than hedging. Specifically, they model a collateral constraint as the friction that leads to financially constrained firms hedging less. The collateral channel affects hedging as follows: Lenders require that firms pledge collateral to borrow and therefore, because they use collateral to borrow (and fund investment), these constrained firms lack the additional collateral that is required by hedging counterparties. Consistent with the models of Rampini and Viswanathan (2010, 2013) and previous empirical work (e.g., Nance, Smith, and Smithson, 1993; Mian, 1996; and Geczy, Minton, and Schrand, 1997), we find that financially constrained companies are less likely to hedge (see our Figure 1). Like Rampini and Viswanathan, we also find that the propensity to hedge declines with net worth. In terms of whether the collateral channel is the root cause of these effects, our evidence is 3

5 inconclusive (with coefficients often having the correct sign but being insignificant). Relatedly, we do not detect a significant hedging role for cash holdings, even though cash is likely the primary form of hedging collateral for financially constrained firms (e.g., the International Swaps and Derivatives Association (ISDA) reports that in 2009 cash is the primary form of collateral for about 95% of OTC derivatives that might be used to hedge). 3 Moreover, we do not find a significant link between hedging and investment prospects. Next we study the human element of corporate risk management. Aside from Smith and Stulz (1985), the human element is not explicitly considered in most theories of risk management and is also understudied empirically. We find that more risk averse managers are more likely to work at firms that have established a risk management program. Furthermore, we find that the effects of personal risk aversion vary conditional on other personal characteristics of the risk manager: Compensation, age, experience, and education. These findings are consistent with a core argument in cognitive psychology and economics that personal traits and aspects of human experience can alter the effect of manager risk aversion on corporate policies (e.g., Johnson and Tversky, 1983; Slovic, 1987). Recent finance theory has also embraced the implications of this argument for corporate decisions (e.g., Gervais, Heaton, and Odean, 2011; Palomino and Sadrieh, 2011). Our analysis suggests that ignoring the role of the individual manager might in part explain the limited ability of modern risk management theories to explain why firms hedge. To recap, we provide new insights regarding why firms hedge and test the predictions of modern risk management theories. Our results are consistent with aspects of some theories but inconsistent with others. Importantly, our survey allows us to go beyond traditional predictions and allows us to examine the nexus between personal-risk aversion, managerial traits, and corporate-risk management decisions. Our analysis suggests that the human factor executive risk aversion in combination with personal characteristics related to compensation, age, experience, and education plays a crucial role in corporate risk management decisions. Future research could benefit from incorporating this important human element into risk management models. The paper is organized as follows. The next section provides a review of risk management theories and summarizes the empirical predictions. We review the empirical literature in the appendix. We describe the survey data in Section 2. Section 3 presents our findings. The final section offers some conclusions. 1. Theories of Risk Management 1.1. The Neoclassical View of Risk Management In the absence of transaction costs and other frictions, hedging can affect the variability of cash flows but not necessarily their expected value. In this case, risk-neutral firms do not need to hedge because shareholders can hedge on their own without incurring any additional costs. This is known as the neoclassical view of risk management and is based on the same assumptions as Modigliani and Miller (1958). Several theories of risk management have been developed over the last 30 years. These theories depart from the neoclassical view by considering the effect of credit frictions and other market 3 ISDA reports very similar figures for the years

6 imperfections on the firm s decision to hedge. We summarize the key insights from these theories in the following subsections. To shorten the exposition, we briefly review some of the main empirical studies in the appendix, where we also highlight when the empirical predictions from risk management models are not fully testable with archival data The Credit Rationing Hypothesis of Risk Management One of the classic motivations for corporate risk management is the necessity to mitigate the effects of credit rationing on the firm s ability to invest. This is known as the credit rationing hypothesis of risk management (Froot, Scharfstein, and Stein, 1993; Holmström and Tirole, 2000). 4 Risk management helps mitigate the effects of credit rationing because it reduces the volatility of cash flows, which in turn improves the ability to fund investment projects when access to credit is limited or very costly. Froot, Scharfstein, and Stein (1993) and Holmström and Tirole (2000) also argue that prearranged lines of credit can function as a substitute for risk management in mitigating credit rationing. They argue that if companies have access to a credit line, they will be less compelled to rely on hedging as a tool to mitigate credit rationing because they could draw down from the credit facility to cover cash flow shortfalls. We examine three empirical predictions related to the credit rationing hypothesis of risk management. The first prediction is that firms are more likely to hedge if they face credit rationing. Given that the importance of risk management as an instrument to mitigate financial constraints is related to the firm s need to fund future investment, the hypothesis also predicts that credit rationed companies are more likely to rely on risk management if they have significant investment prospects that need funding. A related effect is that financially constrained firms should rank reduce the volatility of cash flows by hedging as more important compared to financially unconstrained firms. The third prediction is that companies hedge more if they do not have access to credit lines (or cash). Our data are well suited to test the predictions from the credit rationing hypothesis of risk management. We have information on the insiders views about the investment prospects of the firm, the degree of financial constraints, whether the firm has access to large credit lines or cash balances, and why the firm hedges Financial Constraints, Access to Collateral and Hedging Rampini and Viswanathan (2010, 2013) build a model that explains why financially constrained (e.g., small, non-dividend paying, unrated) firms hedge less than their unconstrained counterparts. In their model, the friction of limited availability of collateral leads to financially constrained firms hedging less. Financially constrained companies face a tradeoff. They can pledge collateral to lenders to borrow so they can increase (or maintain) spending on valuable capital and labor, or they can pledge collateral to hedging counterparties to set up a risk management program (e.g., a hedging program based on over-the-counter forward contracts). This tradeoff implies that when the firm needs to fund a new investment project, the financing need prevails over the hedging concern. Therefore, all 4 Mello and Parsons (2000) develop a dynamic model to show that hedging mitigates financial constraints by reducing the costs of financial distress and increasing financial flexibility. 5

7 else equal, we should expect financially constrained firms with low collateral capacity to be less likely to hedge relative to financially constrained companies with high collateral capacity. We should also expect these patterns to be stronger for firms with good investment prospects. Testing the effects of the collateral channel is complicated by the difficulty of obtaining good measures of the amount of collateral usable in hedging agreements and of the companies growth prospects (e.g., Rampini, Sufi, and Viswanathan, 2014). In our study, we ask CFOs whether the size of cash surplus would influence the intensity of risk management activities at their firms. Similarly, rather than using market-based measures to assess growth prospects, we directly ask CFOs to give us their personal view on the investment prospects for their firms. The richness of our data allows us to study the collateral channel of risk management in a way that is not possible with the ex post archival databases commonly used in risk management studies Risk Management and Agency Problems The key assumption of the agency models of risk management is that managers are risk averse (Smith and Stulz, 1985; Holmström and Ricart i Costa, 1986). This is an important difference from the other neoclassical theories of risk management, which assume that managers are aligned with shareholders and act as risk neutral agents. However, agency models of risk management do not explicitly account for heterogeneity in the degree of risk aversion across managers. The assumption of managerial risk aversion has important implications for risk management. Given that managerial claims to the firm are not easily diversifiable, risk averse managers can reduce the effect of the non-diversifiable risk of their claims by hedging, even when this decision is not valuemaximizing from the perspective of well-diversified shareholders. (For example, executives in the oil industry could reduce their personal exposure to oil price fluctuations by selling oil price forward contracts as a part of their firm s hedging program). The first prediction is that firms in which risk averse managers have a relatively large share of their wealth invested in their firm s shares will be more likely to manage risk to reduce the effect of low diversification. Relatedly, holding constant the percentage of their wealth invested in their firm s shares, more risk averse executives are more likely to manage risk. 5 In our survey, we estimate managerial attitude towards risk using a psychometric-test design following Graham, Harvey, and Puri (2013). This allows us to test the predictions from agency models of risk management using a direct measure of managerial risk aversion. Importantly, we go beyond testing agency models of risk management. Building on recent developments in cognitive psychology and economics, recent finance theory (e.g., Gervais, Heaton, and Odean, 2011; Palomino and Sadrieh, 2011) embraces the argument that personal traits and aspects of personal experience can modify the effect of risk aversion on corporate policy. We are able to gather detailed information on managerial risk aversion and other personal characteristics related to compensation structure, age, experience, and education. This allows us to study whether the effect of personal risk aversion on corporate risk management changes in relation to other personal characteristics. 5 Risk management has also implications for debt capacity. If hedging reduces the variability of expected cash flows, then lenders should be willing to provide more financing to firms with a risk management program in place (Mayers and Smith, 1982; Smith and Stulz, 1985; and Leland, 1998; Graham and Rogers, 2002). 6

8 1.4. Information Asymmetry and Risk Management Breeden and Viswanathan (1999), DeMarzo and Duffie (1991, 1995), and Raposo (1997) argue that when it is difficult for non-controlling shareholders (outsiders) to assess the quality of the management, higher quality managers can signal their type by hedging. The premise of their argument is that firm performance depends on managerial ability and other contingencies that are not directly controllable by management (e.g., currency fluctuations). In the presence of information asymmetry, outsiders cannot separate managerial ability from external contingencies. Higher ability managers hedge to mitigate the effect of hedgeable risks on firm performance and signal their type. Lower ability managers do not hedge because having a risk management program is costly. The main prediction from this signaling argument is that firms are more likely to install a risk management program when information asymmetry is high Testing the Theories Table 1 presents a summary of the main empirical predictions from the current theories of risk management. Empirical evidence related to the predictions from existing models is limited (see the Appendix). Table 1 Theories of Risk Management: Summary of Empirical Predictions Theories Neoclassical View Main theory reference Modigliani and Miller (1958) Key Assumptions Role of Managerial risk aversion Main Predictions Testable w/archival data Testable w/ survey data Absence of frictions No No risk management Credit Rationing Froot, Scharfstein, and Stein (1993) Credit rationing; firm is risk neutral No 1. Risk management more likely by credit rationed firms 2. Risk management more likely by credit rationed/growth firms Partially 2.1. Risk management more likely to reduce cash flow volatility by credit rationed/growth firms No 3. Risk management more likely by firms w/out credit line Partially Access to Collateral Rampini and Viswanathan (2010, 2013) Credit rationing; limited collateral; firm is risk neutral No Risk management less likely by credit rationed/growth firms w/ limited collateral Agency Problems Smith and Stulz (1985) Shareholder-manager; shareholder-bondholder conflicts; firm is risk neutral 1. Risk management more likely if risk averse manager has large stock ownership No Partially 2. Risk management more likely if manager risk aversion is high (holding stock ownership constant) No Information Asymmetry DeMarzo and Duffie (1995) Managerial ability unobservable; firm is risk neutral No Risk management more likely when information asymmetry is higher No Partially 7

9 2. Data 2.1. The Data Gathering Process To obtain our data, we contacted members of the Duke-CFO Magazine Survey panel, International Swaps and Derivatives Association (ISDA), and the Global Association of Risk Professionals (GARP). We invited CFOs to take part in the survey via in the last week of February Reminder s were sent out throughout March The survey closed at the end of April We sent out about 29,000 invitations to non-financial companies in North America, Europe, Asia, and other regions (including Australia, New Zealand, Latin America, Middle East, and Africa). In total, we gathered 690 responses (from both public and private companies), which, to our knowledge constitute the largest survey sample on risk management assembled to date. The response rate is 2.5% which seems low. However, the response rate is misleading. We are interested in surveying only the officers with detailed knowledge or who are in charge of risk management. Many of the ed organizations, particularly for GARP and ISDA, had a very low proportion of senior financial officers. In the end, our sample includes only senior financial officers. We refer to the survey sample as CFOs though some have titles such as Treasurers, Vice-President of Finance, or Chief Risk Officer. We ask CFOs about their companies risk management practices and demographics, including sales, ratings, dividend policies, investment prospects, and several other characteristics. As previously mentioned, we also collect information about risk aversion and other personal traits. We use this information below in tests to study the nexus between personal-risk aversion and corporate-risk management decisions Descriptive Statistics Table 2 reports descriptive statistics for the firms in our sample. About 52% of the companies report having a formal risk management program (Risk Management indicator variable). 6 Table 2 also shows significant regional variation in risk management practices. About 71% of the firms headquartered in Europe have a risk management program relative to 45% of the North America firms and 52% of the Asian firms. Risk management also varies significantly with respect to ownership form. 74% of the public companies have a risk management program compared to 38% of the private companies. In our regressions, we control for these sources of heterogeneity by including region- and ownership-form dummies. 7 Table 2 reports also descriptive statistics for several measures of firm characteristics. About 30% of the firms have revenues above $1 billion (Large). There is also some demographic variation across regions and ownership form. For instance, 15% of the Asian firms are classified as Large, compared to 35% and 42% respectively of the North America and Europe companies. We also report summary statistics on dividend policy, investment prospects, leverage, and whether the firm has a credit rating. 6 All our findings are qualitatively very similar (albeit in some cases with lower statistical power) if we replace the Risk Management indicator variable with an indicator for whether the firm actively manages financial risk (foreign exchange rate, interest rate, and commodity risk) with derivatives. 7 There is also some industry variation. The lowest percentage of hedgers is in the retail industry with 38% of the firms hedging compared to the maximum of 60% of firms hedging in the tech industry. 8

10 [Table 2 Here] Table 2 shows that 37% of the firms are public, which indicates that our sample has a good balance of public and private companies. The table also provides information on the distribution of the sample across regions for the firms for which headquarters information is not missing. 53% of the firms in the sample are from North America, while the companies from Europe and Asia are respectively 18% and 25% of the total sample. The remaining 4% of companies (25 observations) are headquartered in Australia, New Zealand, Latin America, the Middle East, and Africa. We collectively categorize these companies as Other Region firms Comparing the Survey Sample to COMPUSTAT Table 3 compares our sample of public firms with data from the COMPUSTAT Global database (which only includes public firms). This allows us to assess whether our sample characteristics are similar to standard databases used in corporate finance research. There are 252 non-financial publicly listed firms in our survey sample, which we compare with a sample of about 22,700 non-financial companies in COMPUSTAT with a fiscal year ending in May 2010 or the 11 months prior. [Table 3 Here] The evidence in Table 3 suggests that our sample is broadly comparable to the COMPUSTAT sample. There are about 45% of firms with annual sales of less than $1 billion ( small ) in our public sample versus 52% in COMPUSTAT. Our data indicate that about 34% of the survey firms do not pay dividends regularly, relative to 24% of the companies in Global COMPUSTAT. 8 The two samples are quite similar in terms of profitability. 3. Results 3.1. Why Firms Hedge We ask CFOs to tell us why their firms do (or do not) have a risk management program in place. We explore how risk management objectives relate to cash flow variability, access and cost of finance (debt and equity), ratings, firm value, and decision making. We tailor the questions to the theories of risk management that we test in this study. For example, the credit rationing hypothesis of risk management predicts that firms hedge to reduce the volatility of cash flows that can be used to fund new investment projects when they are financially constrained and funding access is limited. Therefore, it is important to know whether firms value risk management as a tool to reduce the volatility of cash flows, improve access to finance, and invest, all conditional on the degree of financial constraint. Figure 2 summarizes the factors that CFOs rank as important or very important (a 3 or 4 respectively, on a scale from 1 to 4). The primary factors include a desire for lower unexpected losses, as well as decreased volatility and surprises. More than 80% of the CFOs in our sample say that Increase Expected Cash Flows and Decrease Unexpected Losses are important determinants of risk management policies. Relatedly, about 80% (75%) of the CFOs say that Reduce Cash Flow 8 For U.S. firms in COMPUSTAT, the percentage of non-dividend paying firms is about 70%. For North America firms in our sample, the percentage of non-dividend paying firms is 47%. 9

11 Volatility ( Improve Earnings Predictability ) is an important determinant of risk management. Overall, these findings suggest that predictability of cash flows and earnings are among the main reasons that firms hedge. [Figure 2 Here] Other factors also play an important role in corporate hedging decisions. For example, of the firms with credit ratings, almost 75% use hedging to Increase/Maintain Ratings. Given its importance in the theoretical risk management literature, it is worth noting that Improve Investment in Difficult Times is important, but less so than the reasons listed above. About two-thirds of CFOs indicate that this investment-based motive to hedge is important or very important. As Figure 2 shows, other factors such as decreasing the cost of equity or share price volatility are less important. Overall, these results suggest that accounting considerations (i.e., related to financial statements and ratings) are as important as financial market considerations (such as stock volatility). In unreported tests, we find, for example, that earnings predictability/smoothing are significant determinants of why public firms hedge more than private firms. This finding is consistent with the argument that public firms are more compelled to hit earnings targets and they hedge to increase the predictability of earnings. In the following sections, we tie the CFOs responses to formal tests of risk management theories. While the predictions from some of these theories are quite intuitive, it is less obvious that they play a primary role in explaining corporate risk management in practice. We also provide evidence on interactions between various theoretical influences on hedging. For instance, one theory predicts that financially constrained firms cannot hedge because they are required by lenders to pledge collateral to borrow and lack the additional collateral that is needed to enter into a hedging agreement. However, another theory predicts that hedging can increase debt capacity of financially constrained firms by reducing the volatility of their cash flows. The interplay of these channels is not part of any existing model. Therefore, it is important to test the relation between these channels empirically Why Firms Do Not Hedge: The Role of Accounting Standards The accounting literature shows that managers have strong incentives to manage earnings. Earnings smoothing can be executed via accounting choices or real business decisions (such as the timing of capital projects, research and development, advertising and the decision to hedge). 9 There is an open debate as to whether accounting standards impede firm s usage of derivatives for hedging (see, e.g. Zhang, 2009). Our paper provides new insight on this issue. Hedge accounting is regulated under the International Accounting Standards (IAS) 39 (or the corresponding Financial Accounting Standards (FAS) 133 in the U.S.). Through hedge accounting firms can offset the profit and loss volatility in the derivatives (arising from marking-to-market the derivatives) with the reciprocal profits and losses in the hedges. The most stringent requirement for hedge accounting to be applicable is for the hedging firm to show that there is an offsetting relation between the value of the derivatives and the value of the hedged item ( hedge effectiveness 9 In the accounting literature, Graham, Harvey, and Rajgopal (2005) show that managers have a preference for using real actions (like capital investment) to smooth earnings rather than accounting choices. 10

12 tests 10 ), and that the offset is not the mere consequence of chance. However, this is often difficult to prove in practice because the derivative instrument and the hedged item often do not match in terms of commodity type, notional amounts, payment dates, and other basic terms. The accounting treatment of derivatives as well as regulatory requirements may provide a strong disincentive for hedging. In our survey, we ask CFOs to tell us whether the hedge effectiveness tests mandated by IAS39 or FAS133 have led to changes in the intensity of corporate hedging with derivatives. In unreported results, we find that 19% of firms with a risk management program reduced the use of derivatives as a result of these regulations. In the analysis that follows, we relate the CFOs descriptions of why firms hedge (or do not hedge) to firm-level characteristics as well as linking to modern theories of risk management Evidence on the Credit Rationing and Information Asymmetry Models of Risk Management The credit rationing hypothesis of risk management predicts that the probability of risk management increases with financial constraints and investment growth prospects. As a first step in the credit rationing analysis, we categorize firms as financially constrained if their sales are below $1 billion, their public debt is unrated, or they do not pay dividends regularly. 11 We follow the literature in this conditioning. We classify firms as high (low) investment prospect firms if the CFO rates their investment prospects above (below) the sample median. 12 Table 4 presents mean difference tests on the percentage of companies having a risk management program in place by financial constraints, investment prospects, and the combination of financial constraints and investment prospects. Panel A shows that a significantly lower percentage of financially constrained firms have a risk management program in place. For example, small firms are much more likely to have risk management programs in place relative to large companies (41% vs. 76%). Panel A also reports that companies in the high investment prospect group are more likely to manage risk relative to their low investment prospect counterparts (55% vs. 48%). However, this difference is not statistically significant. [Table 4 Here] 10 For example, consider the case of an airline entering a long futures contract to hedge against a possible increase in the price of jet fuel. Assume also that the minimum notional amount for standard futures contracts is 1 million gallons of jet fuel. Consider two scenarios: (1) The firm needs to hedge 1 million gallons; (2) The firms needs to hedge 500,000 gallons. If the fuel price decreases, marking-to-market the derivative requires the airline firm to report a loss. However, if the notional amount of the futures contract matches the firm s need for hedging as in (1), the firm meets the hedge effectiveness tests mandated by IAS39 or FAS133 and can offset the loss on the derivatives with the profit arising from a lower market price of jet fuel. On the other hand, the firm does not meet the hedge effectiveness tests in scenario (2) and must report the loss on the derivatives in the income statement without being able to offset it. 11 We follow the literature in this categorization. Gilchrist and Himmelberg (1995) and Fama and French (2002) argue that small firms are typically young, less well known, and therefore more exposed to credit frictions. The argument that rated companies are less likely to be financially constrained is proposed by Faulkender and Petersen (2006). Related approaches for characterizing financing constraints are used by Gilchrist and Himmelberg (1995) and Almeida, Campello, and Weisbach (2004). Fazzari, Hubbard, and Petersen (1988) argue that firms are more likely to pay dividends if they are less susceptible to credit rationing. 12 We obtain very similar results if we classify firms as high/low investment prospects according to whether the firms investment prospects are above/below median investment prospects for the industry. 11

13 The cross-tabulations in Panel A shows that a significantly lower percentage of the financially constrained firms have a risk management program in place relative to their unconstrained counterparts for both the high and the low investment prospect groups. For instance, we find that only 45% of small firms with high investment prospects have a risk management program in place, relative to 76% of the large companies with the same investment prospects. Overall, these findings indicate that a firm s propensity to hedge varies with financial constraints, but not investment prospects. Panel A also shows that financially constrained companies (small, no rating, no dividend) are somewhat more likely to establish a risk management program when their investment prospects are high, although the numbers are generally not statistically different from low investment prospect firms. For instance, we find that 45% of small firms with high investment prospects have a risk management program relative to 39% of low investment prospect firms. We do not find any patterns in the propensity to hedge across investment prospect groups for financially unconstrained firms. 13 Panels B and C show similar patterns in risk management practice (to those documented in Panel A for the full sample) for different subsamples based on ownership form (public and private companies) and regions (North America, Europe, and Asia). The statistical power is lower in these tests due to the smaller number of observations in the subsamples. To sum up, the evidence in Table 4 shows that financially unconstrained firms (large, rated, dividend paying) are significantly more likely to have a risk management program in place relative to their constrained counterparts. These patterns are very similar for both firms with low investment prospects and firms with high investment prospects. Notably, these findings are inconsistent with the key prediction of the credit rationing hypothesis that financially constrained firms hedge to mitigate the effect of limited access to credit. However, in partial support of the credit rationing hypothesis, we find that within the financially constrained categories, companies with high investment prospects are somewhat more likely to hedge, although differences are only statistically significant in one of three categorizations (dividend paying status). The evidence in Table 4 can also be used to evaluate a prediction from the information asymmetry models of risk management (DeMarzo and Duffie, 1995). To the extent that information asymmetry is higher for small firms (alternatively, unrated, non-dividend paying firms), theory predicts these firms to be more likely to hedge than their large firm counterparts (alternatively, rated, dividend paying counterparts). The evidence in Table 4 suggests the opposite (see discussion above). 14 Overall, the evidence in Table 4 is inconsistent with the credit rationing hypothesis of risk management. Previous empirical studies have reported evidence consistent with ours (e.g., Nance, Smith, and Smithson, 1993; Geczy, Minton, and Schrand, 1997). We now focus on a specific channel of the credit rationing hypothesis: The link between risk management and cash flow volatility. The volatility of cash flows is at the core of the credit rationing hypothesis, which predicts that risk management helps financially constrained firms fund new investment opportunities when credit is 13 In unreported tests, we find very similar evidence in a regression framework after controlling for firm heterogeneity and regional variation. 14 In unreported tests, we compare the hedging practice of private and public firms. Studies (e.g., Brennan and Subrahmanyam, 1995; Easly, O Hara, and Paperman, 1998; Hong, Lim, and Stein, 2000) suggest that information asymmetry is lower for public firms because they are followed by analysts. In contrast to the prediction of information asymmetry models of risk management, we find that public firms with high investment prospects are significantly more likely to hedge than private firms in the same investment group. Results hold if we control for firm s size. 12

14 rationed by reducing the volatility of cash flows. We discuss tests related to this prediction in the remaining part of this section. In these tests, we focus exclusively on companies with a risk management program in place whose CFOs rate on a scale from 1 to 4 the importance of risk management as an instrument to decrease cash flow volatility. Figure 2 (discussed above) indicates that almost 80% of the CFOs say that reducing cash flow volatility is either important or very important for the decision to hedge. But does the importance of risk management as a tool to reduce cash flow volatility vary by financial constraints and investment prospects as predicted by the credit rationing hypothesis? To answer this question, Table 5 reports mean difference tests on Decrease Cash Flow Volatility by financial constraints, investment prospects, and the combination of financial constraints and investment prospects. Table 5 shows that the average unconditional response on Decrease Cash Flow Volatility is 3.00 on a scale from 1 to 4. Perhaps more interesting, we examine whether the importance of decreasing cash flow volatility varies with other key variables. We find that the importance of cash flow volatility does not vary in relation to financial constraints, investment prospects, or the combination of both financial constraints and investment prospects. For instance, the mean response is 3.10 for the companies in the large category relative to 3.06 for firms in the small category (row 1, columns 2 and 3) and the difference is not statistically different from zero. The mean response is 3.04 for companies in the high investment prospect group relative to 2.95 for firms in the low investment prospect group (column 1, rows 2 and 3). Similarly, the mean responses for high versus low investment prospect large firms, and the same difference for small firms, 15 are not significantly different from each other. 16 [Table 5 Here] Altogether, the findings in Table 5 are not consistent with the predictions of the credit rationing hypothesis of risk management. Our evidence suggests that the importance of risk management as an instrument to reduce cash flow volatility is uncorrelated with financial constraints, investment prospects, or the combination of financial constraints and investment prospects Evidence on the Substitution between Risk Management, Credit Lines, and Cash Holdings Froot, Scharfstein, and Stein (1993) and Holmström and Tirole (2000) argue that lines of credit can function as a substitute for risk management in mitigating credit rationing. The implication of this argument for our tests is that if the majority of our financially constrained firms have access to untapped funds from credit lines, this could explain why we do not find (in Table 4) that financially constrained firms are more likely to rely on risk management. Table 6 presents tests related to this prediction. In Panel A, we report mean difference tests on the percentage of companies having a formal risk management program in place conditional on financial constraints, credit lines, and the combination of the two. To mitigate the concern that having access to a line of credit could depend on whether the firm is financially constrained, we focus exclusively on firms with a certain proportion of the credit line that is untapped (and only look at firms with access to a credit line). We categorize a firm as having a high ( low ) percentage of the credit line 15 We find very similar patterns if we partition the sample by continent (results available upon request). 16 As an alternative to Reduce Cash Flow Volatility, we ask CFOs to rate risk management as an instrument to Improve Earnings Predictability and find patterns very similar to the ones documented in Table 5. 13

15 undrawn if the percentage of the credit line unutilized is above (below) the sample median. Arguably, firms with a high percentage of undrawn credit line are those that can substitute credit lines for risk management as predicted by the credit rationing hypothesis. Panel A shows that a significantly lower percentage of financially constrained (small, unrated, nondividend paying) firms have a risk management program in place relative to their unconstrained counterparts (large, rated, dividend paying) independently from the percentage of the credit line that is undrawn. Importantly, and opposite to the prediction from the credit rationing hypothesis, within each financial constraint category, we find that firms with a high percentage of unused credit line are generally more likely to have a risk management program in place (although differences are generally not statistically different from zero). For instance, we find that 45% of the small firms with a high percentage of their credit lines undrawn have a risk management program in place relative to 38% of the small companies with a low percentage undrawn. [Table 6 Here] The literature on liquidity management 17 suggests that lines of credit and cash holdings might work as substitutes. Therefore, we reevaluate the evidence in Panel A by sorting companies on the basis on their cash holdings. We say that a firm has high ( low ) cash holdings if the ratio of cash and marketable securities to total assets is above (below) the sample median. The evidence in Panel B is very similar to the credit line findings in Panel A. For example, we find that 45% of the small firms with high cash holdings have a risk management program in place relative to 37% of the small companies with low cash holdings. Overall, we do not find evidence that firms substitute either credit lines or cash for risk management The Role of Collateral The Rampini and Viswanathan (2010, 2013) model predicts that financially constrained firms hedge less due to the friction of having limited access to collateral. We investigate the broad question of whether financially constrained firms hedge less but focus this portion of our analysis on whether financial constraint effects are driven by the collateral channel. ISDA (2009) reports that between 93%-98% of the collateral used for Over-the-Counter (OTC) derivatives is in the form of cash or cashlike (government) securities. Therefore, in Panel B, Table 6 we focus our tests on the cash collateral channel by comparing firms with high cash (collateral) to firms with low cash. Across each of our financial constraint proxies, the propensity to hedge is higher for the firms with more collateral, consistent with the prediction of Rampini and Viswanathan. However, the differences are significant in only one of three categories (unrated firms with low investment prospects). In unreported tests, we find very similar results for both high and low investment prospect firms. Overall, the evidence in Panel A that financially constrained companies are less likely to hedge is consistent with Rampini and Viswanathan (2010, 2013). However, the statistical evidence in our analysis is weak that the effects of financial constraints are driven by limited access to collateral (when collateral is measured by the most common form of collateral: cash and marketable securities). In addition to examining the existence of a risk management program, we also ask CFOs whether the intensity of risk management would change when access to cash surplus is high. For 77% of CFOs, 17 See Sufi (2009), Lins, Servaes, and Tufano (2010), and Campello, Giambona, Graham, and Harvey (2011). 14