The Strategic Motives for Corporate Risk Management

Transcription

1 April 2004 The Strategic Motives for Corporate Risk Management Amrita Nain* Abstract This paper investigates how the benefits of hedging currency risk and the incentives of a firm to hedge are affected by the hedging strategies of competing firms. In competitive industries, if currency hedging is uncommon, prices are expected to positively co-vary with exchange rate related cost shocks providing firms with a natural hedge. As the extent of currency hedging in an industry rises, prices become less sensitive to exchange rates, making profits more random. This suggests that, insofar as randomness in profits is undesirable, the decision to remain unhedged hurts expected profits more if a higher fraction of competitors are hedged. Consistent with this, I find that a firm s decision to remain unhedged hurts its value more if a higher fraction of competitors hedge exchange rate risk. Consequently, firms are more likely to hedge currency risk if a higher fraction of their competitors do so, even after controlling for industry level exposures to currency risk. These strategic incentives are robust and appear to be more important for currency hedging decisions than some of the firm-specific factors highlighted by existing theory. Ph.D. Candidate, Department of Finance, University of Michigan, 701 Tappan Street, Ann Arbor, MI Tel: (734) Fax: (734) anain@umich.edu. The author thanks Sugato Bhattacharyya, Amy Dittmar, Adair Morse and Vikram Nanda for helpful comments 1

2 A large body of research, both theoretical and empirical, examines firms incentives to engage in risk-management. Existing theories suggest that firms undertake risk-management in order to reduce expected taxes, financial distress costs and underinvestment problems. However, cross-sectional evidence regarding most theoretical predictions about the incentives to hedge is mixed. Empirical studies on these firm-specific reasons for hedging find that hedgers do not clearly fit the profile of any one theory. Recent empirical work, however, suggests that industry factors may be relevant for firms decisions to engage in foreign currency risk management. Allayannis and Weston (1999) show that firms that operate in industries with lower mark-ups are more likely to engage in foreign currency risk management in order to remain competitive during foreign exchange shocks. In a case study of a large foreign exchange hedger, Brown (2001) finds that traditional explanations for why a firm would use derivatives to manage hedgeable risk do not capture the primary incentives of the firm under question to hedge. Rather, a stated goal of the firm s currency hedging program is to reduce negative impacts from currency movements on competitiveness. If competitive considerations affect a firm s need to hedge, it is natural to expect that the benefits of hedging and thus, the decision to hedge, will be contingent on whether the firm s competitors also face foreign exchange exposure and whether they engage in foreign currency risk management. However, neither the current theoretical nor empirical literature provides an analysis of how the hedging behavior of competitors affects a firm s own decision to hedge. This paper shows that when industry prices are allowed to be endogenous in the basic Froot, Scharfstein, Stein (1993) model, henceforth FSS, a firm s decision to hedge and the benefits of hedging depend on the extent of hedging in its industry. The motivation for endogenous prices comes from DeMeza and Von Ungern Sternberg (1980) and De Meza (1986) who show that when firms in competitive industries face common cost shocks, prices co-vary positively with costs. The correlation between prices and costs reduces randomness of profits and provides firms with a natural hedge. The modified FSS framework predicts that, provided the 2

3 profit function is concave in investment, a firm s incentive to hedge will be increasing in the frequency of hedging by competitors. When hedging in an industry is uncommon, many firms are exposed to cost shocks and industry prices are positively correlated with costs. This reduces the need for an unhedged firm to adopt measures to counteract cost shocks. As more competitors choose to hedge the common shock, correlation between prices and costs declines, exposing the firm to greater randomness in profits. Insofar as firms dislike randomness in profits, the incentive to hedge will be higher when more competitors choose to hedge. This argument has value implications. The FSS model implies that, given a concave profit function, remaining unhedged hurts firm value since randomness in profits is undesirable. This paper shows that remaining unhedged is less harmful to firm value if many competitors are also unhedged because covarying prices reduce the concavity of the profit function. I test these predictions using comprehensive, hand-collected data on the usage of foreign currency derivatives in the U.S. that allow me to determine how many firms in an industry face exchange rate exposure, how many of those use foreign currency derivatives and the size of their derivatives portfolios. As in previous studies, I use corporate derivatives usage as a proxy for risk-management. I find that the decision to remain unhedged results in a greater value loss if more competitors choose to hedge. For the median frequency of hedging by competitors, a firm s decision to remain unhedged results in a value loss of about 5%. This is in line with the finding of Allayannis and Weston (2001) that foreign currency hedging increases firm value by 4.8%. Noticeably however, the effect of the hedging decision on firm value is asymmetric in the extent of hedging by competitors. When the frequency of derivatives usage by competitors is at its 25 th percentile, a firm s decision to remain unhedged results in a value loss of only 3.7%. Noticeably, when none of a firm s competitors hedge, a firm s decision to remain unhedged does not significantly reduce firm value. Turning to the decision to hedge, I find that a one percent increase in the fraction of competitors who hedge results in a 15.4 percent increase in the probability that a firm engages in 3

4 foreign currency risk-management. The observed positive relation between a firm s hedging decision and that of its competitors could reflect common industry level exposures rather than the strategic response of firm to its competitors hedging decisions. To minimize the effect of industry factors, I examine the change in a firm s hedging policy in response to a significant firm-specific increase in foreign sales. I find that a one percent increase in the fraction of hedged competitors results in an 8.9 percent increase in the probability that a previously unhedged firm will hedge in response to a foreign sales shock. Moreover, a previously unhedged firm is three percent more likely to begin hedging in response to a foreign sales shock if at least one of its competitors starts hedging during the same period. These findings are robust and strongly suggest that competitors hedging decisions affect a firm s incentive to hedge in the manner predicted by the endogenous-price model described above. The predictions of the model hinge on the premise that industry prices are more sensitive to cost shocks when fewer firms in an industry are hedged. Although empirical results are consistent with the predictions, I test the validity of the premise by examining how the sensitivity of domestic producer prices to exchange rate shocks varies across industries depending on the extent of foreign exchange hedging. Results indicate that a depreciation of 10 percent in the real external value of the dollar results in an increase in domestic producer prices relative to overall inflation by 1.2 percent on average by the following quarter. Noticeably, however, when a higher fraction of firms are hedged against foreign exchange shocks, industry prices are less likely to rise in response to a depreciating dollar. That is, when a depreciating dollar increases the cost of inputs purchased abroad, industry prices tend to rise, but rise less when more firms in the industry are insulated against exchange rate fluctuations. This finding supports the notion that industry prices are less sensitive to common cost shocks when more firms are hedged and lends credence to the channel through which strategic considerations are introduced into the basic FSS riskmanagement framework. 4

5 In contemporaneous research, Adam, Dasgupta and Titman (2004), henceforth ADT, model corporate hedging in an industry equilibrium. They show that for sufficiently small levels of investment, the profit function may be convex in investment and the incentives of a firm to hedge may actually decrease with the extent of hedging by competitors. This result stands in contrast to the predictions of my model which assumes concavity of the profit function and shows that a firm s incentive to hedge increases with the extent of hedging by competitors. In order to distinguish my prediction from that of ADT, I attempt to empirically identify firms that have low levels of investment and possibly operate in the convex region of their profit functions. If the profit functions faced by these firms are not concave, a positive relation between a firm s decision to hedge and hedging by competitors should not be observed. In fact, if the sub-sample captures sufficiently low levels of investment, the opposite prediction of ADT should be observed. I find that for the low-investment sub-sample, the extent of hedging by competitors has no effect on a firm s decision to hedge. Although this result does not directly support the ADT prediction, it does indicate that the level of investment matters in the direction suggested by ADT - for low levels of investment, firms do not respond to hedging by competitors in a manner suggested by concave profit functions. Finally, I address the concerns of Guay and Kothari (2003) that derivatives hedging constitutes too small a part of firms risk-management activity to have a significant affect on firm value. Guay and Kothari show that the median firm s derivatives portfolio pays off a small amount relative to the firm s operating and investing cash flows and hence appears to be a small piece of a non-financial firm s overall risk profile. Thus, they call into question previous findings that derivatives usage has economically significant effects on firm value. I explicitly control for the importance of a firm s currency derivatives portfolio using Guay and Kothari s (2003) portfolio sensitivity measure and find that, as before, on average remaining unhedged results in a value loss and the value loss is lower if fewer competitors hedge. 5

6 The paper is organized as follows. Section 1 provides an overview of past research on risk-management. Section 2 presents a simple model, based on the FSS framework, of strategic considerations in risk-management. Section 3 describes the data, variables and methodology. Section 4 describes univariate and multivariate tests. Section 5 addresses robustness of results. Section 6 concludes. 1. Prior Research on the Use of Derivatives Research on derivatives usage largely focuses on understanding why firms hedge. Prior research suggests different theories, most of which rely on capital market imperfections, to explain a firm s incentives to hedge. Corporate hedging can be optimal if it reduces the risk premium demanded by managers, and likewise reduces required compensation. Smith and Stulz (1985) argue that a risk-averse manager who owns a large number of the firm s shares will direct the firm to hedge when he believes that it is cheaper for the firm to hedge than it is for him to hedge on his own account. In contrast, when managers own stock options, the value of which increases with firm volatility, they are less likely to engage in risk-management. Tufano (1996), Schrand and Unal (1998) and Knopf, Nam and Thornton (2002) find evidence consistent with managerial incentives. In contrast, Geczy, Minton and Schrand (1997) and Haushalter (2000) do not find evidence that managerial risk aversion or shareholdings affect corporate hedging. Other theories (e.g. Froot, Scharfstein and Stein (1993)), suggest that hedging can reduce underinvestment problems (Myers (1977)). Evidence on the underinvestment theory is also mixed. Since the underinvestment problem is the most severe for firms with valuable investment opportunities, studies have used research and development (R&D) expense and market-to-book as explanatory variables. Mian (1996) and Allayannis and Ofek (2001) find no relation between market-to-book and hedging. Several papers, however, find that R&D expense increases a firm s incentive to hedge (e.g. Geczy, Minton and Schrand (1997), Dolde (1995)). Smith and Stulz (1985) argue that hedging can increase firm value by reducing the probability of financial distress. Many papers use the debt ratio to measure deadweight costs of 6

7 financial distress and find that hedging increases with the debt ratio (e.g. Graham and Rogers (2002), Dolde (1995), Purnanandam (2004)). Others, however, find no evidence or mixed evidence for the relationship between hedging and leverage (e.g. Nance, Smith and Smithson (1993), Geczy, Minton and Schrand (1997)). Smith and Stulz (1985) also suggest that firms might hedge in response to tax function convexity. Evidence regarding this incentive is also unclear. For example, Nance, Smith and Smithson (1993) find that firms that hedge face more convex tax functions. Graham and Rogers (2002), on the other hand, use a more refined measure of tax function convexity and find no relation between hedging and tax function convexity. Finally, Lel (2003) finds that governance structures that mitigate agency conflicts, such as the existence of non-managerial blockholders and better country-level governance structures increase corporate hedging activities. While evidence about why firms hedge is mixed and does not support any one theory, recent studies show that derivatives usage has significant effects on firm value. Allayannis and Weston (2001) find that the use of foreign currency derivatives increases total firm value by as much as 4.8 percent on average. Graham and Rogers (2002) document a positive relation between derivatives use and debt capacity and argue that derivatives-induced debt capacity increases firm value by 1.1 percent on average. Allayannis, Lel and Miller (2003) find that the hedging premium is statistically significant and economically large for firms that have strong internal and external corporate governance. Carter, Rogers and Simkins (2003) investigate jet fuel hedging and conclude that hedging is associated with higher firm value by 12 to 16 percent, possibly due to the reduction of underinvestment costs. Bartram, Brown and Fehle (2003) find that interest rate hedging increases firm value by 4 to 9 percent. Finally, Lookman (2003) suggests that the observed relation between firm value and hedging is related to agency costs between managers and shareholders. He shows that once these factors are controlled for, valuation effects associated with hedging become largely insignificant. 2. The role of competitors hedging decisions 7

8 Although recent finance literature focuses mostly on firm specific motives for corporate hedging, the notion that incentives to hedge may be affected by competition has been recognized in economics literature. De Meza and Von Ungern Sternberg (1980) demonstrate that in competitive industries, if a rise in input prices affects every constituent firm, they all tend to raise their price. Thus, the marginal revenue curve facing each individual firm shifts upwards and consequently, the profits of the firm are less sensitive to fluctuating input prices, and there is less incentive to adopt measures that counteract cost shock. DeMeza and Von Ungern Sternberg conclude that competitive firms are less likely to hold input stocks and participate in futures markets. De Meza (1986) studies the choice between constant cost and variable cost technologies in a market equilibrium setting. He notes that in the presence of competition, output prices are positively correlated with input prices. Thus, if most firms adopt the same technology, price fluctuations will compensate for cost changes, providing firms with a natural hedge. Since the extent of derivatives hedging in an industry determines how many firms are affected by cost shocks, it has implications for the natural hedge provided by covarying prices. Consider, for example, jet-fuel hedging by the airline industry. If many firms choose not to hedge against a rise in jet-fuel costs, they will all tend to raise output prices when jet-fuel costs rise. This correlation of prices changes with cost changes partially insulates an unhedged firm against jetfuel cost shocks and reduces the incentive of the firm to engage in derivatives hedging, which usually involves a high set-up cost. In contrast, if most firms in the industry are hedged, prices are less correlated with costs and an unhedged firm is exposed to randomness profits. Insofar as firms dislike variability in profits, the incentives to hedge jet-fuel costs will be higher as more competitors choose to hedge. This intuition can be applied to the general FSS framework, which serves as the basis for most recent research on corporate hedging. Consider the FSS model of the benefits of hedging generic cash flow risk. In a two-period investment decision, a firm uses a single input, capital, to produce a level of output given by the 8

9 production function f(.) where f > 0 and f < 0. In the first period, the firm has an amount of internal funds w. At this time, the firm chooses its capital expenditures, K, and external financing needs. Assuming that external financing is costly, the firm prefers to use internal funds first and raises an amount e= K w from external sources. 1 In the second period, the output from investment is realized, sold at the market price and outside investors are repaid. Total expected profit from capital investment in the first period is given by π ( w) = max p f ( K) K C( e) (1) K In equation (1), p is the market price at which the output is sold, which for the moment is assumed to be exogenous. C(e) is the cost of capital which, assuming that internal funds are costless, is equal to the cost of external financing. The first-order condition for this problem is p f K 1 = C e (2) In period zero, the firm chooses its hedging policy to maximize expected profits. The hedging policy is a choice about whether or not to enter into a contract that fixes the amount of internal cash flow available in the first period. Random fluctuations in internal cash flow, w, reduce expected profits if π (w) is a concave function. Using the first order condition in (2), the second derivative of profits with respect to w is given by 2 * * K K π 1 ww = p f KK Cee (3) w w 2 where f KK and C ee are evaluated at the optimal level of capital K. If the expression in equation (3) is globally negative, hedging raises expected profits. It is clear from equation (3) that concavity of the profit function may be driven by the concavity of the production function and/or convexity of the cost function. The production function will be concave as long as there are 1 Cost of external financing could arise from various sources like costs of financial distress, informational asymmetries between managers and investors, direct costs like legal fees etc. 9

10 diminishing returns to investment. FSS show that convexity of the cost function emerges naturally in optimizing modes with rational agents under a wide range of parameterizations. So far I have simply summarized the basic FSS framework for why hedging may be valuable. Given the concavity of the profit function in w, the benefits of hedging are given by ( E( w) ) E ( π ( w) ) = B1 > 0 π (4) Assuming that the fixed cost of setting up a hedging program isκ, a firm will hedge if B > κ. 2 1 Consider what happens if we allow the output price to vary depending on competitors hedging decisions. As more firms in the industry choose not to hedge at period zero, more firms will be subject to randomness in w. Thus, in case of a negative (positive) cash flow shock, a higher (lower) fraction of firms will have insufficient internal cash (w < K) and will resort to costly external financing. That is, a cash flow shock translates into an input cost shock. Unhedged firms facing a higher (lower) cost of required input capital K will tend to raise (lower) output prices. Thus, the randomness in output price offsets randomness in input costs, making profits less sensitive to cash flow shocks. When most firms in an industry are unhedged, the positive correlation of prices with costs provides firms with a natural hedge against cost shocks and reduces the firms incentives to engage in other forms of hedging. This is shown more formally below. As more firms in an industry choose to remain unhedged, the periods in which an unhedged firm experiences a negative (positive) cash flow shock are also the times when many other firms experience a similar shock and revise output price upwards (downwards). Thus, as more firms choose to remain unhedged, industry prices becomes more sensitive to cash flow/cost p shocks and < 0. Note that now market prices is no longer independent of the hedging w 2 Hedging costs arise from wages and fees paid to experts and other transactions costs. 10

11 decisions of the industry. I assume for simplicity that the relationship between output prices and 2 p internal cash is linear, that is = 0. Rewriting the expected profit function from (1) as 2 w we see that, π ( w) = max p ( w) f ( K) K C( e) (5) K 2 * K * K * K π ww = p( w) f KK C ee + pw f K w 1 (6) w w 2 where f KK, C ee and K f are evaluated at the optimal level of investment. Note that the expression for π ww in equation (6) differs from that in equation (3) by the last term. Therefore, if this term is negative, the profit function will still be concave in investment. Since f K > 0 and p w < 0, the last term in (6) is negative as long as the level of internal wealth has a positive K impact on the optimal level of investment, i.e. w * > 0. This condition is a common feature of models of external financing in the presence of information asymmetry or agency problems. 3 The last term in equation (6) makes concavity of the profit function sensitive to the extent of hedging in an industry. If all firms in an industry are hedged, prices are not sensitive to cost shocks, i.e. pw is zero, and we re back in the original setting where concavity of the profit function is given by (3) and benefits of hedging are B 1 as shown in equation (4). However, as more firms in an industry choose to remain unhedged, price becomes more correlated with cash flow shocks and p w is non-zero. Since π p ww w > 0, it is clear from equation (6) that as the sensitivity of price to cash flow shocks increases (because more firms in the industry choose to 3 Example, Fazzari, Hubbard and Petersen (1988), and Hoshi, Kashyap and Scharfstein (1991) 11

12 remain unhedged), profits become less concave in w. Thus, we can say that when a higher fraction of firms remain unhedged, the benefit of hedging is given by ( E ( w) ) E ( π ( w) ) = B2 < B1 π (7) That is, the benefit of hedging is lower when many firms in an industry are unhedged. Given the fixed cost κ of initiating a hedging program, fewer firms will find it worthwhile to hedge cash flow risk as more competitors choose to remain unhedged. Clearly, the assumption of costly external financing is crucial for hedging to be beneficial. If markets are frictionless and external financing is available costlessly whenever needed, hedging cash flow shocks is unnecessary since firms can always make up for internal shortfalls by tapping the capital markets. Moreover, in this general framework where firms decide whether or not to hedge randomness in cash flow, costly external financing is important for industry considerations to matter. If external funds are costless and available at all times, shocks in internal cash flow do not translate into cost shocks and there is no reason to expect prices to be sensitive to cash flow shocks. That is, pw is zero and strategic considerations are irrelevant. However, the key point of the analysis above is that whenever the decision to hedge affects firms costs whether it is the cost of capital, cost of jet-fuel, cost of imported inputs etc. output prices are not independent of hedging decisions, and industry considerations are relevant. The analysis above leads to the two primary hypotheses I test in this paper. First, insofar as the profit function is concave, randomness in cash flow reduces expected profits. Therefore, an unhedged firm is likely to be valued lower than a hedged firm. Existing studies, discussed in the previous section, have already documented that non-hedgers are valued lower. However, the analysis above suggests that this relation will be contingent on the fraction of competitors who choose to remain unhedged. If many firms in an industry are unhedged, co-varying prices insulate unhedged firms against randomness in cash flow and thus, remaining unhedged hurts less. Thus, the value loss from remaining unhedged will be smaller (greater) if many competitors are 12

13 unhedged (hedged). This is the first hypothesis I test in the empirical section. The second hypothesis is that firms are more likely to hedge if a higher fraction of their rivals are hedging. This follows naturally from the discussion above. As more competitors choose to hedge cost shocks, the natural shield provided by co-varying prices declines and the need to engage in cash flow risk management increases. I test these hypotheses using comprehensive hand-collected data on foreign currency derivatives usage by U.S. non-public firms. In the empirical section, I also test the critical premise of the analysis above that, when the fraction of unhedged competitors is high, industry prices are more sensitive to cash flow shocks providing firms with a natural hedge. Since this paper centers on foreign exchange hedging, I test whether domestic industry prices are more sensitive to foreign exchange shocks if a smaller fraction of firms in the industry are hedged. My prediction that firms are more likely to hedge if a higher fraction of competitors choose to do so is the opposite of the conclusion reached by ADT. ADT examine corporate riskmanagement in an industry equilibrium setting. In their model, as in the analysis above, hedging by competitors becomes relevant because it affects the correlation of industry prices with cost shocks. However, unlike the FSS model which assumes that the production function is concave, ADT focus on firms that operate in the convex region of their production functions and thus have profit functions that are convex in the level of investment. Such firms prefer randomness in the level of investment and therefore have an incentive to increase the variability of cash flows. When most firms in an industry are hedged, prices are insensitive to cost shocks and thus, randomness in w results in random profits. Given the convexity assumption, randomness in w increases expected profits and therefore the benefits of remaining unhedged are high. However, as more competitors choose to remain unhedged, prices co-vary with costs, reducing the randomness of profits of an unhedged firm. Thus the benefits of remaining unhedged decline. In this case, a firm is better of hedging because by eliminating cost shocks when prices are variable, the firm exposes itself to greater randomness in profits. ADT conclude that an increase in the 13

14 fraction of unhedged firms decreases the attractiveness of remaining unhedged and thus firms are more likely to hedge when more competitors choose to remain unhedged. The difference in the predictions of my paper and ADT exists because my analysis assumes that, as in FSS, the production is concave in investment while ADT assume that the production function is convex in investment. In my setting, a firm s decision to hedge is driven by its desire to reduce randomness of profits whereas in ADT, a firm hedges to increase the randomness of profits. Since the two assumptions lead to diametrically opposite predictions, I attempt to distinguish between them empirically. Recent finance literature has touched upon the influence competition might have on riskmanagement by firms. Allayannis and Ihrig (2001) develop a model in which firms in more competitive industries have an increased exposure to exchange rates. In a related study, Allayannis and Weston (1999) find that firms in more competitive (low mark-up) industries are more likely to hedge foreign exchange risk. Although these papers suggest that competitive considerations affect hedging incentives, none of these papers have explored how the hedging decisions of competitors affect the advantage a firm derives from hedging or the incentives of a firm to hedge. This is partly due to the fact that previous studies have tended to use small, sometimes randomly selected samples which do not have complete information on hedging by all competitors. This paper uses hand-collected, comprehensive derivatives risk-management data for all publicly listed firms in the US. The data allow me to capture, fairly accurately, how many firms in an industry face exchange rate exposure, how many of those choose to hedge exchange rate risk, what types of instruments they use, and how much they hedge. Using this unique data, I address questions that arise from the discussion above. Are nonhedgers valued lower than hedgers as demonstrated by previous studies? Is this value effect conditional on the hedging decisions of competitors? If yes, what implications does this have for a firm s incentives to hedge? Finally, are industry prices more sensitive to cost shocks when a smaller fraction of firms are hedged? 14

15 3. Derivatives Data, Variable Description and Methodology 3.1 Sample Selection I obtain data on currency derivative holdings of public U.S. as of I do this by searching the financial footnotes and Management Discussion and Analysis of SEC 10-K filings for text strings such as hedg, swap, cap, forward etc. SFAS 105 requires all firms to report information about financial instruments with off balance sheet risk for fiscal years ending after June 15, In particular, firms are required to report the notional amounts of the financial instruments used. If a reference is made to any of the search terms and the firm is not a financial firm, I read the surrounding text to confirm that it refers to foreign currency derivatives holdings and classify the firm as foreign currency derivatives (FCD) user in that year. I collect the gross notional amounts of foreign exchange forwards, swaps and options outstanding as of fiscal year ending in In cases where there were no contracts outstanding as of fiscal year end, but the firm did engage in foreign exchange risk-management during the year, I take the notional amounts that expired during the year If there are no references to the keywords, I classify the firm as an FCD non-user in that year. I match this database with COMPUSTAT and retain all non-financial firms that have positive values for net sales, total assets and market value of equity. Out of the 6,389 firms that meet these criteria for the year , I study 3,259 firms that face ex-ante exchange rate exposure as defined below. Out of these 3,259 firms, 548 firms engage in foreign exchange risk management. An advantage of this comprehensive sample is that it enables me to determine, for each firm, how many competitors who face exchange rate exposure choose to hedge, as well as how much competitors hedge. Most previous studies on foreign exchange hedging have either focused on a single industry or have used sample-selection criteria that do not give a complete picture of hedging activity in any given industry. For example, Allayannis and Weston (2001) use a sample of non-financial firms that have total assets of more than 500 million in each year between 1990 and Geczy, Minton and Schrand (1997) study Fortune 500 non-financial firms. Graham 15

16 and Rogers (2002) use a randomly selected sample of non-financial firms. My sample, on the other hand, is more representative of the universe of firms. Throughout this paper, I focus on firms that face ex-ante exchange rate exposure. This allows me to interpret the absence of derivatives usage as a choice not to use derivatives, rather than an indication of lack of exposure to foreign exchange risk. Following Graham and Rogers (2002), I define firms to have ex-ante currency exposure if they disclose foreign assets, sales or income in the COMPUSTAT Geographic segment file, or disclose positive values of foreign currency adjustment, exchange rate effect, foreign income, or deferred foreign taxes in the annual COMPUSTAT files. As mentioned above, 3,259 firms from the initial sample of 6,389 (fifty-one percent) face ex-ante exchange rate exposure. The gross notional amounts of foreign exchange swaps, forwards and options outstanding are summarized in Table I, Panel A. The descriptive statistics are comparable to previous studies. Graham and Rogers (2002) report a mean foreign currency derivatives notional amount of $558 million for the year which is on average 8.06 percent of total assets. The mean in my sample is $745 million, 8.86 percent of total assets. The mean notional amount of swaps, forwards and options scaled by total assets are 4.90 percent, 7.80 percent and 6.35 percent respectively. These numbers are comparable to those reported by Purnanandam (2004) who also uses a more comprehensive sample to study derivatives usage. Panel B of Table I reports the frequency and amount of derivatives usage by industry. Industry classification is based on Campbell (1996). Currency derivatives usage is the most frequent in the food and tobacco, textiles and capital goods industries where respectively 24 percent, 23 percent and 22 percent of companies are derivative users. The frequency of derivatives usage in my sample is lower than that reported by Geczy, Minton and Schrand (1997) and Allayannis and Weston (2001). These papers limit the sample to large firms where as mine includes many small and medium sized firms that do not use currency derivatives. Two caveats are necessary here. First, the gross notional value represents derivative ownership and may not accurately estimate derivatives hedging if a firm holds offsetting 16

17 contracts. Graham and Rogers (2002) collect both gross and net notional amounts for their sample and conclude that using net as opposed to gross positions is only marginally important in helping identify factors that affect corporate hedging decisions. Second, derivative holdings may measure speculative activity, not hedging. SFAS 119 requires firms to explicitly state whether they speculate with derivatives. I exclude firms that claim to use derivatives for trading purposes and thus classify any firm using foreign currency derivatives as a hedger. 3.2 Description of Variables In this section, I describe the primary variables of interest as well as all control variables. As in previous studies, I use Tobin s Q as a measure of firm value. It is equal to the market value of equity (price times shares outstanding from CRSP) plus assets minus the book value of equity, all divided by assets. Book value of equity is equal to common equity plus deferred taxes. A firm s hedging decision is captured by a derivatives non-user dummy that equals one if the firm does not disclose the use of foreign exchange swaps, forwards or options and zero otherwise. In some tests the hedging decision is captured by a derivatives user dummy that equals one if a firm uses foreign currency derivatives and zero otherwise. Competitors hedging activity is captured by the fraction of competitors with the same three-digit SIC code that disclose the use of foreign currency derivatives. Specifically, it is equal to the number of competitors who face exchange rate exposure and disclose the use currency swaps, forwards or options divided by the total number of competitors who face foreign exchange exposure. When estimating the effect of the hedging decision on firm value, I control for factors that have been known to impact firm value. Previous studies have shown that firm value is affected by growth opportunities, size, leverage, profitability and industrial diversification. I use research and development expense over sales and capital expenditures over sales as proxies for growth opportunities. Log of total assets serves as the measure of firm size. Leverage is calculated as total long-term debt divided by total assets. Return on assets serves as a proxy for profitability and is calculated as net income over total assets. Industrial diversification is captured 17

18 with a dummy that equals one if a firm operates in more than one segment and zero otherwise. Following the findings of Lookman (2003), I use managerial stock-ownership and institutional ownership as controls for potential agency conflicts between managers and shareholders. To reduce the influence of outliers, Q, long-term debt ratio, research and development expense, and return on assets are winsorized at the 1 percent level. In the hedging-decision regressions, I control for factors that, according to existing research, create incentives for a firm to hedge. Firms are more likely to hedge if they have a greater exposure to exchange rate fluctuations. The extent of foreign exchange exposure is captured by total foreign sales divided by net sales. Theory suggests that firms hedge to reduce tax function convexity. As in most previous research, I measure tax function convexity with net operating loss carryforwards scaled by total assets. More liquid firms are less likely to need riskmanagement. I control for this by including the quick ratio, measured as current assets minus inventories over current liabilities as an explanatory variable. Derivatives usage is associated with economies of scale in that larger firms more likely to engage in risk management. Thus, I include log of total assets as an explanatory variable. According to existing theory, firms hedge to reduce costs of underinvestment and financial distress. I use the long-term debt ratio as a proxy for financial distress. I also use the square of the long-term debt ratio as an explanatory variable because Purnanandam (2004) finds that leverage has non-linear affects on hedging. Since underinvestment costs are likely to be the most severe for firms that have more growth options, I include research and development expense as an explanatory factor. Following Geczy, Minton and Schrand, I also use the product of long-term debt ratio and market-to-book as a proxy for underinvestment costs. Smith and Stulz (1985) predict a positive relation between managerial stock holding and derivatives use, but a negative relation between managerial option holding and derivative use. This is because risk-averse managers who have a lot of wealth invested in the firm through stock ownership may direct the firm to hedge if it is more expensive to hedge on their own accounts. On the other hand, expected utility of managerial wealth can be a convex function 18

19 of the firm s expected profits when managers own stock options. In this case, managers may choose to increase the risk of the firm in order to increase the value of their options. To control for managerial incentives, I include two variables. The first is the market value of stocks owned by the executives of the firm scaled by total assets and the second is the Black-Scholes value of the options held by executives of a firm also scaled by total assets. DeMarzo and Duffie (1991) suggest that firms that face more information asymmetry are more likely to hedge. As in previous studies, I use institutional ownership as a proxy for information asymmetry. If firms owned primarily by institutional owners face less informational asymmetry, then the theory of DeMarzo and Duffie implies that firms with more institutional ownership should be less likely to engage in foreign exchange risk management. It is clear that some variables that are expected to affect the hedging decision also affect firm value. Thus, any regression with Q as the dependent variable and the hedging decision as an independent variable will suffer from endogeneity. The estimation methodology used to control for endogeneity in the Q regressions is described below. 3.3 Methodology I examine the effect of currency derivative hedging on firm value by modeling firm value as V = δ D. F + e, (8) i 0 + δ 1 X i + δ 2Di + δ 3 i i where X iis a set of exogenous observable characteristics of the firm, Di is a dummy variable that takes the value of 1 if the firm is currency derivatives non-user and 0 otherwise, F is a measure of the frequency of derivatives usage by competitors, { δ, δ, δ δ } δ = is a vector of 0 1 2, parameters to be estimated, and ei is the error term. The discussion in Section II suggests that δ 2 < 0 and more importantly δ 3 < 0. As the frequency of derivatives usage by competitors goes up (higher F), prices are less sensitive to cost shocks and do not provide a natural hedge. Profits 3 19

20 become more random and, given the concavity assumption, remaining unhedged hurts more. This is the intuition behind the prior thatδ 3 is less than zero. It is evident from the discussion in Section 3.3 that a firm s decision to engage in risk management may be correlated with some unobserved variables that also affect firm value. Thus, D i may be correlated with the error term in equation (8) rendering OLS estimates of δ 2 biased. To control for potential self-selection of firms that hedge, I use Heckman s (1979) two-stage procedure in which the hedging decision is modeled as a function of firm-specific variables that have been shown to affect a firm s incentives to hedge exchange rate risk, specifically, foreign sales, size, leverage, research and development expense, institutional ownership, managerial stock and options holdings. If the association of a firm s decision to hedge, D i, with the frequency of derivatives usage by its competitors, F, is correlated with unobserved variables that also affect the firm s value, then the interaction D i. F will be correlated with the error term, causing δ 3 to be biased as well. While the circumstances under which such endogeneity might arise are not clear, prudence dictates the need to control for it. Since the Heckman two-stage procedure does not correct for a potential bias inδ 3, I turn to the instrumental variable approach. For this, I need an instrument that is correlated with the hedging dummy but uncorrelated with firm value other than through the variables already controlled for in equation (8). Such an instrument, if it exists, can be used in place of D i in equation (8) to get consistent estimates of δ 2 and δ 3. One possible instrument is the lagged value of the derivatives non-user dummy. Recall that the derivatives non-user dummy equals one if a firm did not engage in currency hedging during the year 1999 and zero otherwise. I create another dummy variable, called lag_hedge that equals 1 if a firm did not engage in foreign currency hedging in the year 1997 and zero if it did. Not surprisingly, the derivatives nonuser dummy for 1999 is significantly positively correlated with lag_hedge. Firms that engaged in 20

21 foreign currency risk-management in the past are much more likely to be hedged in 1999 than firms that did not hedge previously. Moreover, once a firm s current (i.e. 1999) hedging decision is controlled for, it is unlikely that the firm s hedging policy in 1997 would affect current firm value. Thus, lag_hedge satisfies the requirements of a good instrument. Results of the Heckman procedure and the instrumental variable regressions are provided in Section Results 4.1 Univariate Tests Table II, presents summary statistics for variables described in the previous section and tests the differences between the means of these variables for users and nonusers of currency derivatives. Not surprisingly, foreign currency derivatives users belong to industries in which a significantly higher fraction of competitors also use foreign currency derivatives. Consistent with previous research, derivatives users are significantly larger, less liquid and have more leverage than non-users. Derivative users also face greater exchange rate exposure as measured by foreign sales over net sales. Unlike Geczy, Minton and Schrand (1997) derivatives users do not have significantly higher research and development expenditure. However, the product of leverage and market-to-book is higher for derivatives users than for non-users. Thus, univariate evidence for the underinvestment costs hypothesis is mixed. Managerial stock and option holding is significantly larger for derivatives non-users. Univariate tests do not support the notion that managers have an incentive to hedge if their stock ownership is higher and but do suggest that option ownership reduces the incentive to hedge. Net operating loss carry-forwards are lower for derivatives users than for non-users. This appears to go against tax incentives for hedging. As in previous empirical research institutional ownership is significantly higher for derivatives users than for non-users contradicting the predictions of DeMarzo and Duffie (1991) that firms with greater information asymmetry are more likely to hedge. 21

22 The distribution of Tobin s Q is skewed in my sample. For both derivative users and nonusers, the mean value of Tobin s Q is considerably larger than the median. Therefore, I use the natural log of Q for multivariate tests of firm value. This makes my results comparable to Allayannis and Weston (2001) and Allayannis, Miller and Lel (2003) who also use the natural log of Q to correct for skewness of Q. Univariate tests do not suggest that foreign exchange hedgers are valued higher. It should be noted, however, that derivatives users tend to be larger firms and size is generally associated with lower Q. Multivariate tests that control for other factors that affect Q are discussed in the next section. 4.2 Multivariate Tests Use of derivatives and firm value In order to document a relation between the use of derivatives and firm value, I need to control for other variables that could affect Q. In this section, I use a multivariate setting which controls for size, growth opportunities, leverage, industrial diversification, profitability, agency conflicts and industry characteristics. The primary explanatory variables of interest are the derivatives non-user dummy, which takes the value 1 if a firm does not use currency derivatives and zero otherwise, and the interaction of the derivatives non-user dummy with the fraction of competitors who use currency derivatives. Panel A of Table III presents results of a simple OLS regression. Since the characteristics that make a firm choose to hedge may also be correlated with firm value, I use Heckman s two step procedure and instrumental variable regressions to control for potential endogeneity. Panel B presents results of Heckman s two-step procedure. Panel C presents results of the IV regression. Panel B of Table III shows that the self-selection parameter, lambda, is 0.16 and significant at the 10 percent level. This indicates the prevalence of selfselection and suggests that characteristics that make firms choose to remain unhedged are positively correlated with firm value. The coefficient on the derivatives non-user dummy is insignificant in all specifications. However, the interaction of the derivatives non-user dummy and the fraction of competitors who 22

23 hedge is negative and significant in all specifications. This indicates that the decision to remain unhedged hurts firm value as more competitors choose to hedge. Coefficients from the Heckman specification suggest that, for the median frequency of hedging by competitors, a firm s decision to remain unhedged results in a value loss of about 5%. This is in line with the finding of Allayannis and Weston (2001) that foreign currency derivatives use increases firm value by 4.8%. Noticeably however, the effect of the hedging decision on firm value is asymmetric in the frequency of hedging by competitors. When the frequency of derivatives usage by competitors is at its 25 th percentile, a firm s decision to remain unhedged results in a value loss of only 3.7%. Noticeably, when none of a firm s competitors hedge, a firm s decision to remain unhedged does not hurt value. These results are consistent with the hypothesis outlined in Section II. If many competitors are hedged, prices are relatively less correlated with cost shocks and profits more random. Thus, the decision to not engage in derivatives hedging causes a greater loss in value. What implications would this have for the incentives of a firm to engage in currency risk management? When many competitors choose not to engage in derivatives risk-management, the natural hedge provided by co-varying prices reduces the firm s need to participate in a derivatives hedging program. Thus, one expects that the probability a firm engages in derivatives hedging will be higher when more firms in its industry hedge. I test this prediction in the next section. Before moving to the next section, I note that as expected, Q is higher for firms with greater research and development expenditure, for firms that are more profitable and for firms in which managers interests are aligned with those of shareholders. It is smaller for larger firms, firms with higher leverage and for diversified firms The decision to hedge In this section, I test the implications of the previous section. Particularly I test whether a firm is more likely to hedge if a higher fraction of its competitors hedge. Existing studies use two methods to study a firm s hedging decision - probability models where the dependent variable is a hedging dummy that equals 1 if a firm uses derivatives and censored tobit models where the 23