Hedging Foreign Exchange Exposure: Risk Reduction from Transaction and Translation Hedging

Transcription

1 Journal of International Financial Management and Accounting 15: Hedging Foreign Exchange Exposure: Risk Reduction from Transaction and Translation Hedging Niclas Hagelin and Bengt Pramborg School of Business, Stockholm University, SE Stockholm, Sweden Abstract Using a sample of Swedish firms we investigate the risk reducing effect of foreign exchange exposure hedging. Further, we investigate risk reduction from using different hedging instruments, and particular interest is directed towards the impact of transaction exposure hedges and translation exposure hedges respectively. We find that firms foreign exchange exposure is increasing with the level of inherent exposure, measured as the difference between revenues and costs denominated in foreign currency, and that it is decreasing with firm size. We find a significant reduction in foreign exchange exposure from the use of financial hedges. The evidence suggests that the usage of foreign denominated debt as well as currency derivatives reduce firms foreign exchange exposure. Further, we find that transaction exposure hedges significantly reduce exposure, and that translation exposure hedges also reduce exposure. A possible explanation for the latter is that translation exposure approximates the exposed value of future cash flows from operations in foreign subsidiaries (i.e. economic exposure). If so, by hedging translation exposure, economic exposure is reduced. 1. Introduction The purpose of this paper is to investigate the relationship between foreign exchange (FX) exposure and hedging activity using financial instruments. Our study is motivated by the possibility that exposure management may be costly, but yet ineffective in reducing total risk. Copeland and Joshi (1996) argued that anticipating the consequences of hedging is difficult since so many other economic factors change when FX rates change. As a consequence, hedging activity risks being wasteful to the firm s shareholders, and may actually increase exposure. Survey evidence suggests that This paper has benefited substantially from suggestions by two anonymous referees and the editor, Richard M. Levich, and we are grateful for their contributions. We would also like to thank Johan Adolphson, Niclas Andre n, Hossein Asgharian, David Burnie, Martin Holmen, J. Austin Murphy, Lars Norde n, Fredrik Stenberg, Carsten Srensen, Clas Wihlborg, and Martin R. Young for valuable comments. We also acknowledge comments from participants in the EFMA Conference 2001 in Lugano, the 13th Annual PACAP/FMA Finance Conference 2001 in Seoul, the 31st FMA Annual Meeting 2001 in Toronto, and comments from workshop participants at School of Business, Stockholm University, and at the Arne Ryde Workshop in Empirical Finance 2001, Lund University. r Blackwell Publishing Ltd. 2004, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

2 2 Niclas Hagelin and Bengt Pramborg firms management perceive hedging as complicated (see Alkebäck and Hagelin, 1999). Hence, knowledge about whether or not hedging is successful is of importance to shareholders. 1 Until recently, little effort has been directed to analyze whether firms are successful or not in reducing risk pertaining to FX exposure. A few recent studies have, however, attempted to answer this question. The FX exposure of 171 Japanese multinationals was examined by He and Ng (1998). They documented that 25 percent of the firms experienced significant FX exposure. The extent to which a firm was exposed to FX risk was linked to the firm s export ratio. He and Ng (1998) also examined the relationship between FX exposure and variables that were assumed to reflect derivatives usage. The results showed that firms that were predicted to hedge had lower FX exposure, on average, than comparable sample firms. Nydahl (1999) investigated 47 Swedish firms FX exposure. The evidence showed that exposure increased with the fraction of sales classified as foreign. Further, using survey evidence on firms hedging of foreign assets, Nydahl (1999) examined the association between translation exposure hedging and FX exposure. The results indicate that translation hedging reduced the sample firms FX exposure. Allayannis and Ofek (2001) analyzed the link between FX exposure and the use of currency derivatives for a sample of nonfinancial S&P 500 firms. They used recently reported information on financial instruments with offbalance sheet risk, which firms in the U.S. have been required to report under SFAS No 105 since 1991, to investigate whether currency derivatives usage reduced firms FX exposure. Their evidence suggested that derivatives usage reduced FX exposure. Guay (1999) also used U.S. data on derivatives usage that was made available due to the issue of SFAS No 105. In contrast to the cross-sectional study of Allayannis and Ofek (2001), Guay (1999) examined the change in the risk level for firms surrounding each firm s inception of derivatives usage. Firm risk was found to decrease following the inception. Wong (2000) investigated the FX exposure of manufacturing firms in the U.S. to test for an association between FX exposure and derivative disclosures required by the SFAS No 119 (and its predecessors SFAS No 105 and 107). He documented weak associations between derivative disclosures and FX exposure and suggested that this can be due to inability in controlling for firms inherent exposures and shortcomings of the accounting disclosures. Nevertheless, taken together the results of the five studies presented above suggest that the use of currency derivatives may help to reduce firms FX exposure.

3 Hedging Foreign Exchange Exposure 3 This study contributes to the existing literature mainly by analyzing how firms FX exposure is affected by different types of hedges. In contrast to earlier studies, we investigate not only the use of currency derivatives but also the use of debt denominated in foreign currencies. Further, we investigate the effects from hedging transaction exposure and translation exposure. This is of interest because transaction exposure and translation exposure tend to affect firms differently. Transaction exposure to currency risk refers to potential changes in the value of future cash flows as a result of unexpected changes in exchange rates. 2 If competently executed, transaction exposure hedges should reduce the variability of cash flows and consequently the variability of firm value. Translation exposure, on the other hand, arises as the financial accounting statements of foreign affiliates are translated into the currency of the parent firm. 3 The general recommendation of the finance literature is not to worry about this type of exposure and thus not to hedge it (see e.g. Butler, 1999). The reasons for this are that translation gains (losses) tend to be (i) unrealized and have little direct impact on firms cash flows, and (ii) poor estimators of real changes in firm value. If this is true, translation exposure management should be inefficient in reducing firms FX exposure. The closer examination of FX exposure hedging contained in this study compared with previous studies is made possible due to the use of questionnaire data. In addition, while earlier studies assign all derivatives usage as hedges, the use of questionnaire data in this study controls for the possibility that financial instruments are not used for hedging. This may be important, given Leland s (1998) results indicating that firms may use risk management tools for speculative purposes. We employ cross-sectional regressions to investigate a sample of Swedish firms FX exposures and hedging practices. We find that FX exposure is increasing with the level of inherent exposure and decreasing with firm size and with the use of financial hedges. The evidence suggests that currency derivatives hedging as well as hedges with foreign denominated debt reduce firms FX exposure. Further, we find that transaction exposure hedges as well as translation exposure hedges reduce FX exposure. One possible explanation for the result concerning translation exposure is that it approximates the exposed value of future cash flows from operations in foreign subsidiaries (i.e. economic exposure). If so, by hedging translation exposure, economic exposure is reduced. The paper is organized as follows. The next section describes the sample, the estimation of FX betas, and concludes with a description of

4 4 Niclas Hagelin and Bengt Pramborg determinants of firms inherent exposure and hedging variables. Crosssectional results are presented in the third section, which is followed by the conclusions. 2. Sample Selection, Exposure, and Hedging Practices In this section we describe the sample and the criteria for including firms in the sample. We discuss how the FX exposure is estimated and report the estimations made. Finally, we discuss the inherent exposure of the sample firms and classify firms according to their hedging strategy. 2.1 Sample We study the relationship between a sample of Swedish firms hedging practices and FX exposure for the sample period January 1997 through December Because detailed public data are not available on firms inherent FX exposures and hedging practices, we employ three questionnaires to determine exposures and firms usage of financial hedges. 4 The first questionnaire was sent to 160 firms in October 1997, and contained questions concerning the respondents inherent exposures and hedging policies for the year The second questionnaire was sent to 275 firms in March 2000, followed by the third questionnaire, sent to 261 firms in September The three questionnaires are nearly identical. The difference is that, in the two latter, we added one question regarding the types of exposure hedged with foreign denominated debt. The questionnaires of 2000 and 2001 asked questions concerning the years , and , respectively. In the Appendix, the questions we use in this study are presented. The questionnaires were sent to firms that met the following three criteria: (i) the firm was listed at the Stockholm Stock Exchange; (ii) the firm was a nonfinancial firm; and (iii) the firm s headquarter was located in Sweden. Financial firms were excluded because the focus of the study is on end-users rather than producers of financial services. Foreign firms were also excluded (firms with headquarters located outside Sweden) to eliminate potential differences between firms that arise due to differences in accounting standards between countries. One hundred and one, 130, and 128 usable responses were obtained from the first, second, and third questionnaire. This represents response rates of 63 percent, 47 percent, and 49 percent, respectively. This can be

5 Hedging Foreign Exchange Exposure 5 compared with Bodnar et al. s (1998) 21 percent. The three surveys render a total of 617 firm year observations. However, the final sample is reduced to 462 observations due to missing data and the exclusion of observations with no inherent FX exposure, as we discuss below. To check for potential non-response bias, the sample firms were compared with firms that did not return the questionnaires. This comparison indicated that no response bias exists Stock Price Exposure Dumas (1978) and Adler and Dumas (1984) defined FX exposure as the effect of exchange rate changes on the value of a firm. Jorion (1990) proposed that estimates of the stock price exposure to FX rate changes could be obtained from the following time series regression R it ¼ a i þ b Mi R Mt þ b FXi R FXt þ e it ; ð1þ where R it is the return on stock i from t 1tot, R Mt is the return on a market portfolio, and R FXt is the return on one unit of a trade-weighted basket of foreign currencies to the local currency conditioned on R Mt. Specifically, the return series of the trade weighted basket of foreign currencies (the TCW index) 6 is orthogonalized to R Mt by regressing it on R Mt and the residuals are then used as the conditional returns. The inclusion of R Mt implicitly controls for value-relevant market-wide factors that are correlated with the exchange rate over the estimation period. The coefficient of interest to us is b FXi, firms FX betas, the measure of firms stock price exposure to currency fluctuations. Ultimately, we are interested in firms asset FX betas, but have only access to stock data. Therefore, we first apply equation (1) to estimate each firm s stock FX beta. We then compute each firm s asset FX beta by multiplying the stock FX beta with the firm s equity-to-value ratio. 7 In the cross-sectional regressions below we exclusively use firms asset FX betas. As Griffin and Stulz (2001) argued, including a market portfolio, as in equation (1), represents the case where the market exogenously determines changes in the FX index. The opposite view is that changes in the FX index exogenously determine market returns which would suggest that the market index should not be included. Also, ignoring causality, if the market is positively correlated with the FX rate we may underestimate positive FX exposure, and overestimate negative exposure, since some of the exposure may be captured by the market index betas. As a result, we

6 6 Niclas Hagelin and Bengt Pramborg might give too much weight to market changes and not enough to FX rate changes. We repeated all our regressions (i) without the inclusion of a market portfolio, and (ii) using market excess returns instead of raw returns, implicitly assuming a market beta of one, and estimated the FX exposure using only the FX index. The results are robust across these different specifications. Bodnar and Wong (2000) suggested that a size based control portfolio should be used instead of a common market index, which helps to control for both the macroeconomic factors that influence firm value and the size effect in FX exposure. They demonstrated that a size controlled portfolio leads to more reliable results for the U.S. market than a common market portfolio. In line with Bodnar and Wong (2000) we use a size controlled portfolio. 8 To estimate equation (1) we need to determine the sampling frequency. He and Ng (1998), Guay (1999), Wong (2000), and Allayannis and Ofek (2001) used monthly observations while Nydahl (1999) used weekly observations. We use weekly, bi-weekly, and monthly overlapping returns, using daily observations, and apply the Newey and West (1987) GMMmethodology for robust standard errors. The results are robust to the choice of sampling frequency. To conserve space we only report the results for the weekly observations. The choice of sampling frequency is also related to how information is impounded into stock prices. In fact, earlier studies, such as Jorion (1990) and Amihud (1993), relying on the assumption that capital markets react instantaneously to exchange rate changes had difficulties in identifying significant contemporaneous FX exposure coefficients for U.S. firms. Bartov and Bodnar (1994) showed that lagged changes in exchange rates demonstrated a significant effect on abnormal stock performance for U.S. firms. We check for this, and find support for lagged effects. However, using contemporary and lagged FX betas in our cross-sectional analyses produce similar results. For this reason, we only report results pertaining to the model with contemporary FX betas. This is in line with He and Ng (1998), Nydahl (1999), Guay (1999), Wong (2000), and Allayannis and Ofek (2001). In Table 1, Panel A, the number and percentages of significant stock FX betas are reported. The sample in Table 1 is reduced to 462 observations, from a possible 617, due to (i) missing data, which include too few data points for beta estimations, and incomplete survey responses (73 observations deleted), and (ii) the firm s survey response indicated no revenues and no costs denominated in foreign currency for a particular year (82 observations deleted).

7 Hedging Foreign Exchange Exposure 7 Table 1. Foreign Exchange Betas The table contains descriptive statistics of the estimated stock foreign exchange (FX) betas and the absolute value of the asset FX betas. The stock FX betas are estimated using the regression model R it 5 a i 1b Mi R Mt 1 b FXi R FXt 1e it, where R it is the return of firm no. i s stock, R Mt is the return on the market index, R FXt is the return on the orthogonalized trade weighted currency index, and e it is the residual. In Panel A, the number and percentage of stock FX betas, ^b FXi, that are significant on a five percent level of significance (a ), and on a ten percent level of significance (a ) are reported for each year in the sample period. For each year returns are calculated as overlapping, weekly observations, using daily closing prices. The standard errors of the estimates are corrected for serial correlation using the Newey- West (1987) methodology. We require a minimum of three months of data to estimate the stock FX beta for a firm. In Panel B, descriptive statistics for the absolute asset FX betas, j^b A FXij, are presented, where the observations are divided into three separate groups based on the book value of total assets (the average of beginning-of-year and end-of-year value) for each firm year observation. Panel A. Number of Significant ^b FXi No. of estimated a No. of significant (%) a No. of significant (%) (7.6) 12 (15.2) (15.6) 22 (24.4) (17.4) 29 (31.5) (16.5) 30 (29.1) (9.2) 17 (17.3) Total (13.4) 110 (23.8) Panel B. Descriptive Statistics, j^b A FXi j Large Mid Small All obs Observations Mean Max Q Median Q Min About 13 (24) percent of the FX betas are significant at the five (ten) percent level, which is similar to earlier studies. Several reasons may explain why few firms have significant FX exposures. First, the TCW index may capture the average FX exposure of all firms, but for many individual firms, the FX exposure may not be well represented by the index. Second,

8 8 Niclas Hagelin and Bengt Pramborg many firms may have low inherent FX exposure in the first place, and third, even for firms with large inherent FX exposures, the usage of operational and/or financial hedges may reduce FX exposure. Table 1, Panel B, presents the absolute value of the asset FX betas, 9 where the sample is divided into three equally sized groups based on firms book value of total assets, which is motivated by the potential difference in hedging practices and multinationality between large and small firms. An interesting relationship between firm size and FX exposure is indicated, namely that the smaller the firm, the larger the exposure in absolute terms. There can be several reasons for this result. It could reflect the documented relationship between firm size and the usage of financial hedges (see e.g. Bodnar et al., 1998), where larger firms, due to economies of scale, are more likely to engage in hedging with financial instruments. Also, it may reflect that larger firms are more likely to be multinational corporations (MNCs) with production and sales in many currencies, which can reduce FX exposure, and also enable those firms to use sophisticated operational hedges. Smaller firms on the other hand, are more likely to be exporters or importers, with few opportunities to use operational hedging to reduce FX exposure. 2.3 Firms Inherent Exposure and Hedging Practices Firms were asked in the questionnaires to specify how much of their revenues and costs that were denominated in foreign currencies for each year. Also, the respondents were asked whether or not they hedged FX exposure with derivatives and if so, if their hedging concerned translation and/or transaction exposure. In addition, they were asked whether they used foreign denominated debt for hedging purposes, and, in the second and third questionnaire, if this concerned translation and/or transaction exposure. Based on these questionnaires, we were able to define variables that reflect each firm s inherent FX exposure as well as hedging strategy. Earlier studies have shown that a firm s FX exposure is significantly related to proxies for foreign involvement (see He and Ng, 1998; Nydahl, 1999; Wong, 2000; Allayannis and Ofek, 2001). Lack of better proxies has led most studies so far to use one-sided measures, such as export ratio (see e.g. He and Ng, 1998; Allayannis and Ofek, 2001). An alternative measure is the net exposure, suggested by Marston (2001). This measure, a net of long and short positions in foreign currency, supposedly captures exporters as well as importers net exposure in specific currencies. Using the unique data from our questionnaires, we can create a proxy for firms

9 Hedging Foreign Exchange Exposure 9 net exposure. This proxy, denoted NE, is defined as the difference between the percentage of revenues (FR) and the percentage of costs (FC) denominated in foreign currency. Since we are primarily interested in the magnitude of the FX exposure, as measured by the absolute value of firms asset FX betas, we use the absolute value of this measure for our analysis, denoted abs(ne). 10 Instead of this measure, the measure proposed by Marston (2001), NE, could be used, which would enable us to model the sign of the exposure (positive versus negative FX betas) using the net long or short position. However, our empirical investigation using NE suggests very little explanatory power. The problem lies with the large subset of FX beta observations with insignificant exposure, which makes the determination of sign noisy. This argument is supported by a test (not reported) where we used only significant FX betas, and the NE measure. For this subset we found the expected significant positive relationship between FX exposure and NE. 11 Using the full sample and our choice of variables, i.e. the absolute values of FX betas and abs(ne), focuses the analysis on the magnitude of the FX exposure, which is the main interest of the study, while allowing us to model the relationship without having to estimate the sign of the FX exposure. In Table 2, total assets, FR, FC, and abs(ne) for our sample are displayed. As in Table 1, the sample is divided into three equally sized groups based on book value of total assets. On average, the sample firms in Table 2 are characterized by higher FR than FC. Further, larger firms are characterized by higher levels of FR and FC than smaller firms are. Interestingly, larger firms have lower abs(ne) than smaller firms, which may partly explain the lower FX betas for larger firms reported in Table 1. However, the lower FX betas of larger firms can potentially be explained by other firm characteristics, such as degree of multinationality, and sophistication of risk management. We therefore include firm size as explanatory variable in our cross-sectional regressions. Specifically, our proxy for firm size is the logarithm of the average book value of total assets for each firm year observation, denoted SIZE. In addition to operational hedging, firms may hedge with different types of financial instruments. Also, firms may choose to hedge different types of FX exposure. In order to capture financial hedging practices we define three sets of dummy variables. First, to explore the impact on firms FX exposure from hedging with financial instruments, we classify the sample into firms that hedge with financial instruments versus those firms that do not. This results in a dummy, H, which is set to one if the firm hedges its foreign exchange exposure using financial instruments and zero otherwise.

10 10 Niclas Hagelin and Bengt Pramborg Table 2. Firm Size and Foreign Exchange Exposure The table shows descriptive statistics on size and foreign exchange exposure for all firm year observations where the firm indicated at least some exposure. The sample is divided into three equal-sized groups based on book value of total assets. Total assets (in million SEK) are calculated as the average of beginning-of-year and end-of-year total assets. Three exposure measures are reported. FR is the share of revenues that is denominated in foreign currency; FC is the share of costs that is denominated in foreign currency; and abs(ne)is the absolute value of net exposure, where net exposure is defined as, (FR FC), the difference between the percentage of revenues and the percentage of costs that are denominated in foreign currency. Observations Large Mid Small All obs Total Assets Mean Median FR Mean Median FC Mean Median abs(ne) Mean Median The second classification regards hedging of different types of FX exposure; namely transaction and translation exposure. For this purpose we classify our sample into four distinct groups, using three dummy variables. The first, denoted TRTL, is set to one for firms that hedge transaction exposure and translation exposure. 12 The second, TR, is set to one for firms that hedge transaction exposure only, and the third, TL, is set to one for firms that hedge translation exposure only. The base case consists of firms that do not hedge with financial instruments. Because we do not have information for this classification from the first questionnaire, we use observations for 1998 to 2001 only. From the questionnaires we have information on the proportion of transaction exposure and translation exposure that firms hedged with currency derivatives, but we do not have this information for foreign denominated debt. Admittedly, the dummy variables are crude measures of hedging, but, on the other hand, the variables indicating the proportions hedged are noisy proxies for hedging. This is because, although the questionnaire provides us with the proportions hedged, it does not include information regarding the explicit hedge horizon or how much of the total exposure that derives from each

11 Hedging Foreign Exchange Exposure 11 type of exposure. As a consequence, the interpretation of the proportions hedged is not straightforward, and they cannot be added or easily compared. In order to simplify interpretation, we have chosen to classify firms using dummy variables. Nevertheless, as a robustness test we used the proportions for currency derivatives hedging in a separate set of crosssectional regressions. The results from these regressions support the findings from using dummy variables (not reported). 13 This may be explained by the fact that, for our sample, firms that hedged a particular exposure tended to hedge a substantial part of it. The median firm hedged 90 percent of its transaction exposure (mean value 79 percent), and 79 percent of its translation exposure (mean value 66 percent). We make a final classification in order to study the relative impact from currency derivative hedges and foreign denominated debt. As with the previous classification, we classify hedgers into four distinct groups, using three dummy variables. The first dummy variable is set to one if the firm use currency derivatives and foreign denominated debt (CDFD), and zero otherwise. The second dummy (CD) is set to one if the firm use currency derivatives but no foreign denominated debt to hedge, and the third dummy (FD) is set to one if the firm use foreign denominated debt but not currency derivatives to hedge. The base case contains firms that do not hedge with financial instruments. In Table 3, Panel A, the results from the classification on type of FX exposures hedged are summarized. About two-thirds (63 percent) of the firm year observations are for firms that hedged their FX exposure using financial instruments (H). Almost all large firms hedged, and the table confirms that smaller firms were less likely to hedge than larger firms were. The most common strategy were to hedge transaction exposure only, while very few observations are for firms that hedged translation exposure only. For large firms, it was most common to hedge transaction exposure as well as translation exposure, while smaller firms primarily hedged transaction exposure. Panel B of Table 3 contains evidence on the choice of instruments. Large firms most often used currency derivatives and foreign denominated debt to hedge, while it is noteworthy that very few of the smaller firms indicated that they used foreign denominated debt. 3. Cross-sectional Results In this section we investigate cross-sectionally whether firms that hedge are characterized by lower FX betas or not. Further, we investigate the impact

12 12 Niclas Hagelin and Bengt Pramborg Table 3. Hedging Practices The table shows descriptive statistics on hedging practices, reported for all firm year observations where the firm indicated at least some exposure. Also reported are statistics for subsamples where the original sample is divided into three equal-sized groups, based on book value of total assets (the average of beginning-of-year and end-of-year value). Panel A contains the number and percentage (in parenthesis) of observations that are for firms that hedge (H), and the numbers and percentages of observations that are for firms that hedge transaction exposure and translation exposure (TRTL), that hedge transaction exposure only (TR), and that hedge translation exposure only (TL). The sample in Panel A is reduced to include observations from 1998 to 2001 only. Panel B contains the number and percentage (in parenthesis) of observations that are for firms that hedge (H), and the numbers and percentages of observations that are for firms that use currency derivatives and foreign denominated debt to hedge (CDFD), that use currency derivative only (CD), and that use foreign denominated debt only (FD). Panel A. Hedging and Types of Exposure, Large (%) Mid (%) Small (%) All obs (%) Total H n 121 (95) 89 (70) 32 (25) 242 (63) TRTL 65 (51) 20 (16) 3 (2) 88 (23) TR 49 (38) 60 (47) 29 (23) 138 (36) TL 7 (5) 9 (7) 0 (0) 16 (4) Panel B. Hedging and Instruments, Large (%) Mid (%) Small (%) All obs (%) Total H n 145 (94) 115 (75) 47 (31) 307 (66) CDFD 93 (60) 48 (31) 3 (2) 144 (31) CD 35 (23) 42 (27) 41 (27) 118 (26) FD 16 (10) 24 (16) 3 (2) 43 (9) n In the table, when hedgers (H) are broken down into groups, the resulting number of firms may not add up to total number of hedgers due to incomplete responses. from hedging different types of FX exposures, as well as the impact from currency derivatives hedging and hedging with foreign denominated debt. In order to investigate the effect from inherent FX exposure and financial hedging on firms FX betas, we use cross-sectional regression models with the absolute value of the estimated FX beta as the dependent variable and the firm specific variables presented in the previous section as independent variables. First, to investigate the effect on firms FX betas from inherent FX exposure, we run the following cross-sectional

13 Hedging Foreign Exchange Exposure 13 regression ^b A FXi ¼ a 0 þ a 1 absðneþ i þ X4 d j YEAR j i þ X10 g k IND k i þ e i ; j¼1 k¼1 ð2þ where ^b A FXi is the estimated FX beta for firm i s assets; abs(ne) i is the absolute value of the net of firm i s proportion of revenues and expenditures denominated in foreign currency; YEAR i j is a year dummy for year j and firm i; IND i k is an industry dummy for industry k and firm i; 14 and e i is the residual. Table 4, model (a), presents the result from regression equation (2). The coefficient for abs(ne) is positive and significant on a five percent level, which is consistent with Marston (2001) that a larger difference between revenues and costs in foreign currency results in larger FX exposure. Model (b) in Table 4 is an augmented version of regression equation (2) to which the variable SIZE is added. The evidence confirms the expected negative relationship between firm size and FX exposure. Larger firms may have lower FX exposures than small firms in this setting because they are (i) more likely to be MNCs, and therefore have natural hedges built-in to their business, and (ii) that they are more likely to hedge, as shown in Table 3. In order to investigate this, we add variables representing financial hedges to model (b), which are reported in Table 4, models (c) (e). In model (c), it is shown that by adding the dummy for hedgers, H, to the model the FX exposure is as before increasing in abs(ne) and decreasing in SIZE. Further, the hedging dummy is significant and negative, suggesting that firms that use financial instruments to hedge reduce their FX exposure compared with firms that do not. It is interesting to note that SIZE is still significant when we control for exposure and hedging. This result may reflect that larger firms have more opportunities to use operational hedges than what is captured by abs(ne), our proxy for inherent FX exposure. While this proxy measures the mismatch between revenues and costs in foreign currency, it does not capture other characteristics of MNCs, for example how firms can move production or change the product mix in response to FX rate changes. Since larger firms on average are more likely to be MNCs with operations in many countries SIZE may capture such differences. Model (d) presents evidence regarding the risk reducing effect from different types of financial hedging. We find significant and negative coefficients for all three dummy variables, suggesting that FX exposure is reduced for transaction hedgers as well as for translation hedgers. 15

14 14 Niclas Hagelin and Bengt Pramborg Table 4. Cross-sectional Regressions This table reports the results of multiple regression analyses on factors affecting the absolute value of the estimated foreign exchange (FX) asset betas. The dependent variable in all models is the absolute value of the asset FX beta, which is the stock FX beta estimated with weekly overlapping stock returns using daily data and corrected for leverage. For example, model (c) is j^b A FXi j¼a 0 þ a 1 absðneþ i þa 2 SIZE i þ a 3 H i þ P 4 j¼1 d jyear j i þ P 10 k¼1 g kind k i þe i, where i denotes firm year observation i. The independent variables capturing inherent exposure are abs(ne), the absolute value of the difference between the share of revenues and costs denominated in foreign currencies; and SIZE, the logarithm of average book value of total assets. Variables representing hedging practices are; H, a dummy set to one if a firm reported that it used derivatives and/or foreign denominated debt; CDFD, a dummy set to one if a firm used derivatives and foreign denominated debt; CD, a dummy set to one if a firm used derivatives but not foreign denominated debt; FD, a dummy set to one if a firm used foreign denominated debt but not derivatives; TRTL, a dummy set to one if a firm hedged transaction exposure and translation exposure; TR, a dummy set to one if a firm hedged transaction exposure only; and TL, a dummy set to one if a firm hedged translation exposure only. In the second column, the number of observations for which the dummy variable is equal to one is reported. The regressions are estimated with year dummies, YEAR j, and industry dummies, IND k, which are not reported. p-values, using Newey-West (1987) robust standard errors, from two-sided tests are reported in parentheses. Variable No. of Obs. Dummy 5 1 Model Model Model Model Model (a) (b) (c) (d) (e) Intercept (0.000) (0.000) (0.000) (0.000) (0.000) abs(ne) (0.024) (0.107) (0.048) (0.037) (0.098) SIZE (0.000) (0.009) (0.047) (0.011) H (0.002) TRTL (0.003) TR (0.022) TL (0.002) CDFD (0.001) CD (0.007) FD (0.015) Number of observations Adjusted R

15 Hedging Foreign Exchange Exposure 15 Although the result concerning translation hedging is in accordance with Nydahl (1999), it is somewhat surprising, and it raises an interesting question: Why does exposure management targeted at objectives other than foreign exchange betas influence observed betas so strongly? One possible explanation for this is provided by Oxelheim and Wihlborg (1997), who argued that if the translation FX rate is a close proxy for a weighted average of future FX rates at the time cash flows occur, then the economic exposure and the translation exposure are nearly identical. If so, by hedging translation exposure, firms may successfully reduce economic exposure. It could be that through experience firms develop proxies for economic exposure, which is being hedged, while being confined conceptually to translation and transaction exposure. 16 An interesting observation related to these arguments is Hagelin s (2003) results suggesting that transaction exposure is hedged in order to mitigate costs associated with market imperfections, but that translation exposure hedging cannot be explained by such arguments. Thus, if hedging different types of exposures are all proxies for economic exposure, then it seems that there may still be different rationales for hedging the different types of exposures. Finally, model (e) presents evidence regarding the risk reducing effect from different types of instruments: firms that use currency derivatives and foreign denominated debt (CDFD), firms that use foreign denominated debt only (FD), as well as firms that use currency derivatives only (CD). Firms in all classes have significantly lower FX beta estimates at any conventional significance level than non-hedgers, suggesting that firms are reducing exposure effectively using currency derivatives as well as with foreign denominated debt or with a combination of these financial instruments. 4. Conclusions In this study we investigate whether firms are successful or not in reducing their foreign exchange exposure with currency derivatives and foreign denominated debt. This is motivated by the possibility that exposure management may be costly, but yet ineffective in reducing total risk. Consequently, if management fails to reduce total risk by hedging, shareholder value may be eroded. On the other hand, a successful hedging program may increase shareholder value by reducing costs related to different market imperfections.

16 16 Niclas Hagelin and Bengt Pramborg The results demonstrate that foreign exchange exposure, as measured by firms foreign exchange betas, is increasing in inherent exposure. Also, exposure is decreasing in firm size. We suggest that larger firms may have lower inherent exposure due to their ability to use operational hedges. Further research efforts could extend our research in this respect by including variables that capture the usage of operational hedging and flexibility, for example firms ability to alter production and pricing policy in response to foreign exchange rate changes (see e.g. Bodnar et al., 1999; Allayannis and Ihrig, 2001). Further, we find evidence that financial hedges are effective in reducing firms foreign exchange exposure. Financial hedging is associated with risk reduction for firms that use currency derivatives and/or foreign denominated debt. Particular interest is directed towards the impact from transaction exposure hedges and translation exposure hedges respectively. This is of interest because translation exposure and transaction exposure tend to affect firms differently. The results suggest that there are risk reducing effects from transaction exposure hedges as well as from translation exposure hedges. A possible explanation for the latter is that translation exposure approximates the exposed value of future cash flows from operations in foreign subsidiaries (i.e. economic exposure). If so, by hedging translation exposure, economic exposure is reduced. We suggest that valuable contributions in this field can be made by using refined measures of exposures and hedging variables, as well as additional efforts to resolve why firms hedge translation exposure. Notes 1. In addition to reduce total risk, to add value hedging must be aimed at minimizing the advert impact from market imperfections, such as progressive tax rates, costs of financial distress, and agency problems (see Smith and Stulz, 1985). 2. Economic exposure can be divided in three: the exposure of committed transactions, the exposure of identifiable anticipated transactions, and competitive exposure (unidentifiable anticipated transactions). By transaction exposure hedging we refer to the two former. The reason for not investigating the use of derivatives for managing competitive exposure is that few firms tend to hedge this type of exposure with currency derivatives (see Bodnar et al., 1998). Instead, the longer-termed competitive exposure is typically managed by operational hedges. From an examination of annual reports, it is suggested that this is the case for our sample firms. Butler (1999) provides a discussion on the problems associated with using financial hedges to manage longer-termed competitive exposure. 3. Translation exposure depends on the translation method used. The International Accounting Standard 21 (IAS 21) suggests the use of the current rate method for selfcontained foreign affiliates and the use of the temporal method for integrated foreign affiliates and for foreign affiliates in countries with high inflation. The exposure under the

17 Hedging Foreign Exchange Exposure 17 current rate method is given by the equity of the foreign affiliate, whereas under the temporal method it is the net amount of assets and liabilities translated at the current exchange rate. Changes in exchange rates on foreign operations thus always cause changes in group equity and under the temporal method, these changes also affect group net income. We use a sample of Swedish firms. Firms in Sweden follow The Swedish Institute of Authorised Public Accountants proposal to recommendation, which builds on the IAS 21. In practice, the current rate method dominates among Swedish firms. 4. The accounting practices regarding FX exposures and hedging for Swedish firms did not follow strict rules during the sample period. New rules and regulations, based on IAS 39 will be introduced, but for the sample period this was not in effect. Recommendations from Bokfo ringsna mnden (BFN R9) stipulated that firms should report net revenues, investments, and employees for geographical markets, with considerable freedom in deciding what the geographical market was. Out of two accounting methods allowed for hedging, deferral accounting and mark-to-market, almost all firms used deferral accounting. Firms were required to disclose in footnotes derivatives positions, where most firms reported a net position and not per currency, type of exposures hedged, or exposures partitioned over time. Noteworthy is also that the practices were quite dispersed among firms. In sum, accounting practices for international operations and hedging seem to some extent to lag those in the US, at least for the sample period (see also Hung (2001), for a discussion on country differences of the value relevance of financial statements). 5. See Pramborg (2002) and Hagelin (2003) for details. Results are available by request from the authors. 6. The TCW index is the official trade currency weighted index provided by the Swedish central bank, Sveriges Riksbank. The index is based on the IMF Total Competitiveness Weights, and reflects export and import competitiveness. 7. Specifically, each firm s asset FX beta is calculated as ^b A FXi ¼ ^b E i FXi V i, where ^b A FXi is firm i s asset beta, ^b FXi is the estimated stock beta from equation (1), E i is the total market value of the firm s equity (at the beginning of the year) and V i is the total market value of the firm, calculated as (total book value of assets minus book value of equity plus market value of equity). One underlying assumption is that debt has no systematic risk. 8. To check for robustness, we repeated the tests in this study using the standard market index as well. For the sample of Swedish firms in this study, the choice of market index does not matter. 9. The absolute value of the FX betas captures the magnitude of exposure, which is the interest of this study. 10. Formally, this measure is defined as abs(ne) 5 abs(fr FC), where FR is the percentage of revenues denominated in foreign currency, and FC is the percentage of costs that is denominated in foreign currency. 11. We required an absolute t-value of at least 1.64 to include the observation in the sample. As a robustness check, we used this subsample for all cross-sectional regressions we discuss in Section 3 below. The results were confirmed with this subsample. 12. We define firms that hedged committed transactions or anticipated transactions (see the Appendix) as transaction hedgers. Almost all firms that hedged anticipated transactions also hedged committed transactions. 13. Specifically, the coefficients for the proportion variables had identical signs, but somewhat lower significance, as compared to the dummy variables. This is likely to be a result of a high degree of multicollinearity. 14. We add year dummies and industry dummies in order to account for differences in absolute FX beta levels between years and industries, but do not report them. 15. Note that only 16 observations are for firms that hedged translation exposure only. Thus, the result concerning translation exposure hedging should be interpreted with some caution.

18 18 Niclas Hagelin and Bengt Pramborg 16. Translation exposure hedging that is strictly accounting based may reduce FX betas for firms that have more revenues than costs in foreign currency (in our case defined as NE40). This is because a translation hedge that is unrelated to economic exposure (of the specific foreign affiliate being hedged) may offset transaction exposure and thereby reduce the FX beta. For firms with short positions (NEo0) we would expect an opposite effect. If a sample contains a large proportion of observations with long positions, the results may suggest that translation exposure hedging reduces exposure, although it is aimed only at accounting numbers. Most firms in our sample that hedged translation exposure were long in foreign currency (81 out of 106 observations) so this explanation may be valid. In order to investigate this formally, we used an augmented version of model (d) that includes dummy variables to discriminate between firms with long and short positions. Our results suggest that firms with short positions as well as firms with long positions reduce their FX betas significantly through translation exposure hedging. However, the sample size is small and the results should therefore be interpreted with care. Detailed results are available on request. References Adler, M. and B. Dumas, Exposure to Currency risk: Definition and Measurement, Financial Management 13 (1984), pp Allayannis, G. and J. Ihrig, Exposure and Markups, Review of Financial Studies 14 (2001), pp Allayannis, G. and E. Ofek, Exchange Rate Exposure, Hedging and the Use of Foreign Currency Derivatives, Journal of International Money and Finance 20 (2001), pp Alkeba ck, P. and N. Hagelin, Derivative Usage by Nonfinancial Firms in Sweden with an International Comparison, Journal of International Financial Management and Accounting 10 (1999), pp Amihud, Y., Evidence on Exchange Rates and Valuation of Equity Shares, in Exchange Rates and Corporate Performance, Y. Amihud and R.M. Levich, eds. (Irwin Professional Publishing, 1993). Bartov, E. and G.M. Bodnar, Firm Valuation, Earnings Expectations, and the Exchange- Rate Exposure Effect, Journal of Finance 49 (1994), pp Bodnar, G.M., G.S. Hayt, and R.C. Marston, 1998 Wharton Survey of Financial Risk Management by U.S. Non-financial Firms, Financial Management 27 (1998), pp Bodnar, G.M., B. Dumas, and R.C. Marston, Pass-Through and Exposure, Working paper (Weiss Center for International Finance, 1999). Bodnar, G.M. and F.M.H. Wong, Estimating Exchange Rate Exposures: Some Weighty Issues, Working paper (NBER, 2000). Butler, K.C., Multinational Finance, 2nd ed. (South-Western College Publishing, 1999). Copeland, T.E. and Y. Joshi, Why Derivatives Don t Reduce FX Risk, McKinsey Quarterly No. 1 (1996), pp Dumas, B., The Theory of the Trading Firm Revisited, Journal of Finance 33 (1978), pp Griffin, J.M. and R.M. Stulz, International Competition and Exchange Rate Shocks: A Cross-Country Industry Analysis of Stock Returns, Review of Financial Studies 14 (2001), pp Guay, W.R., The Impact of Derivatives on Firm Risk: An Empirical Examination of New Derivative Users, Journal of Accounting and Economics 26 (1999), pp

19 Hedging Foreign Exchange Exposure 19 Hagelin, N., Why Firms Hedge with Currency Derivatives: An Examination of Transaction and Translation Exposure, Applied Financial Economics 13 (2003), pp He, J. and L.K. Ng, The Foreign Exchange Exposure of Japanese Multinational Corporations, Journal of Finance 53 (1998), pp Hung, M., Accounting Standards and Value Relevance of Financial Statements: An International Analysis, Journal of Accounting and Economics 30 (2001), pp Jorion, P., The Exchange Rate Exposure of U.S. Multinationals, Journal of Business 63 (1990), pp Leland, H.E., Agency Costs, Risk Management, and Capital Structure, Journal of Finance 53 (1998), pp Marston, R.C., The Effects of Industry Structure on Economic Exposure, Journal of International Money and Finance 20 (2001), pp Newey, W. and K. West, A Simple, Positive Semi-definite Heteroscedasticity and Autocorrelation Consistent Covariance Matrix, Econometrica 55 (1987), pp Nydahl, S., Exchange Rate Exposure, Foreign Involvement and Currency Hedging of Firms: Some Swedish Evidence, European Financial Management 5 (1999), pp Oxelheim, L. and C. Wihlborg, Managing in the Turbulent World Economy: Corporate Performance and Risk Exposure (John Wiley, 1997). Pramborg, B., Empirical Essays on Foreign Exchange Exposure, Doctoral Thesis (Stockholm University, 2002). Smith, C.W. and R. Stulz, The Determinants of Firms Hedging Policies, Journal of Financial and Quantitative Analysis 20 (1985), pp Wong, F.M.H., The Association Between SFAS 119 Derivatives Disclosures and the Foreign Exchange Risk Exposure of Manufacturing Firms, Journal of Accounting Research 38 (Autumn 2000), pp Appendix This appendix contains the questions that we use to construct inherent exposure and hedging variables for the empirical analysis. Questions for Observations Approximately what percentage of your firm s revenues and costs are denominated in foreign currencies? Revenues y % Costs y % 2. Does your firm use derivatives instruments to hedge currency exposure? Yes y No y 3. Approximately what percentage of the following exposures are you hedging with currency derivatives? a. Translation exposure y % b. Committed transactions y % c. Anticipated transactions y %