J. of Multi. Fin. Manag. 15 (2005) 15–30
Foreign exchange exposure, risk management, and quarterly earnings announcements Niclas Hagelin, Bengt Pramborg∗ School of Business, Stockholm University, SE-106 91 Stockholm, Sweden Received 10 March 2003; accepted 18 November 2003 Available online 23 July 2004
Abstract This paper investigates the effects of foreign exchange (FX) exposure and hedging activities on the abnormal stock price volatility surrounding quarterly earnings announcements. The findings show that abnormal volatility is positively correlated with foreign exchange exposure, suggesting that investors use the information in earnings announcements to assess the impact of foreign exchange rate changes on firm performance. Further, abnormal volatility is positively correlated with currency derivatives hedging, but not significantly correlated with the use of foreign denominated debt. This may stem from investors’ inability to correctly evaluate information on currency derivatives usage or that investors lack the information needed to correctly judge the effect of derivatives on firm performance. We investigate whether abnormal volatility on earnings announcement days for firms with foreign exchange exposure might be caused by systematic mispricing, but find no evidence for this. © 2004 Published by Elsevier B.V. JEL classification: F31; M41 Keywords: Hedging; Earnings announcements; Derivatives; Risk management; Foreign exchange exposure
1. Introduction Firms are exposed to foreign exchange (FX) risk when their cash flows and therefore market values are influenced by unexpected changes in exchange rates. Such exposure includes the impact of unexpected exchange rate changes on contractual transactions as well as on the uncertain future cash flows generated by the firm’s income producing real ∗
Corresponding author. Tel.: +46 8 674 7427; fax: +46 8 674 7440. E-mail address:
[email protected] (B. Pramborg).
1042-444X/$ – see front matter © 2004 Published by Elsevier B.V. doi:10.1016/j.mulfin.2003.11.001
16
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
assets. In a highly internationalized and competitive economy, a significant number of firms can be expected to be subject to exchange rate exposure. For this reason, the exposure of firms to exchange rate changes has received attention in the literature. Modern capital market theory suggests that the effects of any unexpected changes in exchange rates on firm value should occur instantaneously. While empirical studies relying on this assumption have failed in identifying a strong connection between changes in foreign exchange rates and firm value (see Jorion, 1990, 1991; Amihud, 1993; Bodnar and Gentry, 1993), Bartov and Bodnar (1994) showed that lagged changes in exchange rates can explain changes in firm values. Bartov and Bodnar (1994) suggested that this puzzling result may arise because of the complex set of issues investors face in modeling and estimating the relationship between changes in exchange rates and firm value. For instance, investors are not always aware of firms’ activities to hedge FX exposures and how their real internal activities will be altered in response to the new competitive conditions. Because of these difficulties, Bartov and Bodnar (1994) suggested that investors learn of the full impact of changes in exchange rates on firm value only as information on the firm’s past performance is made available, leading to a lagged relationship between changes in exchange rates and firm value. They found support for such a lagged relationship from analyzing data on a sample of US exporting firms, and showed that investors use the information associated with earnings announcements to gauge the impact of past exchange rate changes on firm performance. In a recent paper, Guay et al. (2003) studied the influence of corporate risk exposures on the accuracy of earnings forecasts. Their results are consistent with arguments that corporate financial risk exposures are not transparent to investors or analysts. This study’s main contribution to earlier research is that it investigates the role of quarterly earnings announcements in informing investors of the impact of hedging activity. To the best of the authors’ knowledge this is the first attempt to document this relationship. Successful hedging activities should reduce the FX exposure, and thus potentially reduce the effect of FX rate changes on earnings. Accordingly, recent empirical studies have found that hedging decreases FX exposure (see e.g. Allayannnis and Ofek, 1998; Hagelin and Pramborg, 2004). On the other hand, if investors are not adequately informed of or find it difficult to understand hedging activities, it may increase their difficulty in understanding the effects of FX rate changes on firm value. Therefore, successful FX exposure hedging may be able to stabilize the value of the firm, and yet increase the volatility of the firm value at times when inadequately informed investors learn of the full impact of the hedges. For hedging to affect the volatility on earnings announcement days, the earnings announcement should not only reveal new information unknown to investors and analysts, but it should also be of economic significance. The study by Guay and Kothari (2003) casts doubt on this. They investigated the value and cash flow effects from derivatives use and found that these were very small in comparison to the overall risk profile of the investigated firms. If this is so, it can be expected that any new information on earnings announcement days concerning effects from hedging will be negligable, and thus not significantly affect volatility.1 1 The results of Guay and Kothari (2003) conflict with the evidence from e.g. Allayannnis and Ofek (1998), Allayannis and Weston (2001), Graham and Rogers (2002), and Hagelin and Pramborg (2004) who find significant effects from financial hedging activity.
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
17
We perform an event study on a sample of Swedish firms covering the period from January 1997 through March 2000. In line with earlier studies, we document an increase in standardized abnormal stock price volatility on the earnings announcement day, reflecting that quarterly earnings announcements contain new information vital to investors. Further, the evidence indicates that abnormal volatility is negatively correlated to firm size and positively correlated to firms’ growth potential and diversification. The first result is in line with earlier research and suggests that large firms are characterized by more pre-disclosure information while the two latter results suggest that it may be difficult for investors to correctly predict performance for growth firms and for diversified firms. As expected from Bartov and Bodnar (1994), and Guay et al. (2003), we find that abnormal volatility is positively correlated with FX exposure. Moreover, we find abnormal volatility to be positively correlated with hedging. Specifically, such increased volatility is associated with currency derivatives use, while the use of foreign denominated debt has no significant effect. This may result from investors’ inability to correctly evaluate information on currency derivatives use, which includes non-linear payoffs from options. It could also be argued that, since derivatives are not included on the balance sheet, investors lack the necessary information to make correct judgments of the effect of derivatives on earnings. Further, we investigate whether the increased abnormal stock price volatility for firms with FX exposure is due to systematic mispricing, but find no evidence for this. Finally, it should be noted that the study uses relatively simple variables to classify firms’ hedging behavior. For this reason we suggest that the results should be interpreted with some caution and that further research is warranted. The paper is organized as follows. The next section covers the research methodology, followed by a section presenting the results. The final section offers our conclusions. 2. Research methodology 2.1. Abnormal volatility measure Beaver (1968) characterized an earnings announcement as conveying information on a security’s value if its release changes investor beliefs concerning attributes they value, such as claims to future dividends. He argued that earnings announcements possess information content if stock price volatility and/or trading volume increase around the time of the announcement (see also Atiase, 1985; Bamber, 1987; Holthausen and Verrecchia, 1990; Barron, 1995; Bamber et al., 1997). According to this view, stock prices reflect the market’s aggregated (or average) interpretation of the information, while trading volume measures investor activity, and thus reflects differential beliefs among investors. This paper investigates how the information content of earnings announcements may be explained on an aggregate level by FX exposure and hedging activity. Therefore, we have focused on stock price volatility as opposed to trading volume changes. To investigate stock price volatility on earnings announcement days, we conduct an event study.2 We employ the standardized abnormal volatility (SAV) measure developed by 2
See Campbell et al. (1997, chapter 4), for a comprehensive discussion on event studies.
18
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
Beaver (1968) and refined in subsequent research (see Francis et al., 2002). We define an event window around the announcement day (day t = 0) as the interval t ∈ [−10, +10]. Further, we use a window of 200 days, the interval t ∈ [−210, −11], for making estimates.3 To compute the SAV, the absolute value of a standardized abnormal return for each event day, we use the market model as our measure of expected returns as follows: The expected return for each firm, i, and each earnings announcement, k, is computed as E[Rik,t ] = aˆ ik + bˆ ik RMk,t ,
t = −10, . . . , 10
(1)
where Rik,t is the return for stock i and announcement k on the event day t, RMk,t is the return on a broad stock market index on the event day t, and t represents the day relative to the announcement day.4 The coefficients aˆ ik and bˆ ik are estimated for each event, ik, by running the following regression: Rik,t = aik + bik RMk,t + eik,t ,
t = −210, . . . , −11
(2)
where eik,t is the residual, and t represents the day relative to the announcement day. The abnormal return is calculated as the difference between the return on the event day and the expected value from Eq. (1). Thus, ARik,t = Rik,t − (ˆaik + bˆ ik RMk,t ),
t = −10, . . . , 10.
(3)
We standardize the abnormal return with the sample error adjusted standard deviation of the residuals from Eq. (2). The adjustment follows Boehmer et al. (1991), used by e.g. Friedrich et al. (2002), and is computed according to 1/2 2 RMk,t − µMk 1 σˆ ik = sˆik 1 + + −11 , (4) 2 Lest τ=−210 (RMk,τ − µMk ) for each event day, t. In Eq. (4) sˆik is the (unadjusted) standard deviation of the residuals from Eq. (2), Lest is the number of days in the estimation period, and µMk is the average market return over the estimation window. The standardized abnormal return can be written SARik,t =
ARik,t . σˆ ik
(5)
and is distributed N(0, 1) under the null hypothesis of no informational content in earnings announcements. Since we are not investigating the sign of the earnings surprise but rather its magnitude, we use the absolute value of the standardized abnormal return as our measure of standardized abnormal volatility (SAV). Formally, SAVik,t = |SARik,t |, 3
t = −10, . . . , 10.
(6)
We also used alternative lengths of the estimation window (250 and 350 days, respectively) with no change in the results. 4 We used a value weighted market index, the AFGX. However, as a robustness test, we also used a small-cap and a mid-cap index. These different specifications do not change the results of the study.
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
19
√ Under the null hypothesis, since SAR ∼ N(0, 1), the expected value of SAV is (2/π), or approximately 0.8. For each event day, we calculate the mean SAV over all events, ik, and use bootstrapping to create confidence intervals for the mean SAV.5 As a robustness test, we used an alternative measure for expected returns, the market excess return model.6 This does not change our results, so to conserve space, we only report findings using the market model. We also used the squared SAV as a dependent variable, as originally suggested by Beaver (1968) and applied by Landsman and Maydew (2002), in the cross-sectional regressions. The results are robust for this specification. It is important to keep in mind that the SAV is a relative measure. It indicates for each event the deviation from an expected volatility. Since the abnormal return is standardized in Eq. (5) a high risk stock has the same expected abnormal volatility as a low risk stock. Therefore, a result that one firm is associated with a higher SAV as compared to another firm does not indicate that the former firm is riskier per se, but, as Beaver (1968) suggested, that the information content of earnings announcements for the former firm is higher. 2.2. Sample description and variable definitions To study the relationship between FX exposure, hedging practices, and earnings surprises, we use a sample of Swedish firms covering the period from January 1997 through March 2000. Since public data on firms’ FX exposure and hedging practices are not available in tractable form, we employed two subsequent questionnaires to determine the extent of firms’ FX exposure and whether specific firms used financial hedges.7 The first questionnaire was sent to 160 firms in October 1997, and the second questionnaire was sent to 275 firms in March 2000. The first questionnaire asked about each respondent’s inherent exposure and hedging policy for 1997, while the second asked about 1998 and 1999 (both questionnaires are available on request). The questionnaires were sent 5 Specifically, for each event day, we randomly draw 2000 samples with replacements from the data and calculate a mean value for each of these. These 2000 mean values are then sorted (low to high) in a vector and used to estimate confidence intervals, where observations number 10 and number 1990 in the sorted vector represent the 0.005 and 0.995 percentile, respectively. For an account of the bootstrapping methodology, see Efron and Tibshirani (1993). 6 Using the market excess return model, the expected return is simply
E[Rik,t ] = RMk,t ,
t = −10, . . . , 10.
Thus, the expected return for the stock equals the market return on day t. The calculation of abnormal returns follows from the earlier case. The abnormal returns are standardized using the standard deviation of the difference between stock i’s return and the market’s return, Rik,t − RMk,t , during the estimation period t ∈ [−210, −11]. 7 The accounting practices regarding FX exposures and hedging for Swedish firms did not follow strict rules during the sample period. Recommendations from Bokföringsämnden (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. Some firms provided detailed information at the country level, while others simply reported this as foreign. Also, firms should, but were not required to, disclose in footnotes the details of their derivatives positions (most firms simply reported a net position, not providing details as to the currency, types of exposures hedged, or exposures partitioned over time). It is also noteworthy that practices varied widely among firms. Recent accounting standards, such as FAS133 and IAS39, will change this by requiring that all derivatives be included on the balance sheet. However, these standards were not in effect in our sample period.
20
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
to firms that met the following criteria: (1) the firm was listed on the Stockholm Stock Exchange; (2) the firm was a non-financial firm; and (3) the firm’s headquarters were located in Sweden. The reason for excluding financial firms is that the focus of the study is on end-users rather than producers of financial services. The reason for excluding foreign firms (firms with headquarters located outside Sweden) was to eliminate differences between firms that could potentially arise due to differences in accounting standards between countries. One hundred and one (130) usable responses were obtained for the year(s) 1997 (1998– 1999), representing response rates of 63% (47). The two surveys supplied a total of 361 firm year observations, which translates into a total of 1109 earnings announcements over the period.8 However, due to data availability at the firm level, the sample in the final cross-sectional analysis in the paper was reduced to 1008 observations distributed over 159 firms. The questionnaires asked firms to specify how much of their revenues and costs were denominated in foreign currency for each year. Also, the respondents were asked whether or not they hedged FX exposure with currency derivatives and/or foreign denominated debt. The questionnaire can be found in Hagelin and Pramborg (2004), the 1997 responses are described in detail in Hagelin (2003), and the 1998–1999 responses are described in detail in Hagelin and Pramborg (2003). One advantage of using survey data is that, while most other research assign all derivatives use as hedging (see e.g. Allayannis and Weston, 2001), the survey questions concern hedging only.9 Using answers from these questionnaires in combination with publicly available data, we are able to define variables that reflect firms’ FX exposures, hedging strategies, as well as a set of firm characteristic variables. We describe below the various variables representing firm characteristics used in our tests and the reasons that led us to use them. 2.2.1. FX exposure (abs(NE), FA) According to the above discussion, we would expect to find a positive relationship between the FX exposure of a firm and abnormal volatility at the time of the earnings announcement of that firm. We use two different measures of firms’ FX exposure. The first is the absolute value of the difference between the percentage of revenues and the percentage of costs denominated in foreign currency (abs(NE)). This measure draws on that suggested by Marston (2001) and captures the potential effect of a net position (long or short) in foreign currency. We note however that firms may be subject to FX exposure even if abs(NE) is zero. This is because firms typically are exposed to more than two currencies and that perfect matching between revenues and costs denominated in foreign currency may not be the case. Therefore, we use a second measure of FX exposure that should be able to reflect these sources for exposure. This latter measure is the average of the percentages of revenues and costs that are denominated in foreign currency (FA) and may be interpreted as the degree of multinationality. 8
We checked for potential non-response bias and our findings suggested that no such bias exists (see Pramborg, 2002; Hagelin, 2003 for details). 9 This however, does not exclude that the responding firms also speculate. The normal disadvantages with surveys also apply, such as respondents misinterpretating questions.
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
21
2.2.2. Hedging (H) A firm’s hedging activity may reduce the FX exposure of the firm and consequently reduce abnormal volatility on announcement days. On the other hand, if investors do not correctly assess the effect of hedging, or if investors are not adequately informed about how the firm hedges, the reduction in abnormal volatility could be lessened or even reversed. Therefore, successful FX exposure hedging may cause the abnormal volatility to increase at the announcement day even if it stabilizes the return process at other days. To capture the effect of hedging, we use a dummy, H, that is set to one if the firm hedges (using currency derivatives or foreign denominated debt), and to zero otherwise. 2.2.3. Hedging instruments (DD,CD,FD) Currency derivatives and foreign denominated debt may result in quite different pay-off structures. Currency derivatives include options and swaps, with pay-off structures which investors may find more difficult to understand than the relatively simpler structure of foreign denominated debt. Also, the accounting treatment of these instruments differs: currency derivatives are off-balance sheet items, while foreign denominated debt is kept on the balance sheet. To investigate whether hedging with different types of instruments affects the abnormal volatility on the announcement day, we use three dummy variables. The first, DD, is set to one if the firm uses currency derivatives and foreign denominated debt, and to zero otherwise. The second, CD, is set to one if the firm uses currency derivatives but not foreign denominated debt, and to zero otherwise, while the third, FD, is set to one if the firm uses foreign denominated debt but not currency derivatives, and to zero otherwise. 2.2.4. Size (SIZE) Earlier research has shown that firm size, used as a proxy for the availability of predisclosure information, is negatively correlated with abnormal volatility (see Atiase, 1985; Bamber, 1987). This is sensible since, for example, more analysts follow large firms than small firms. We expect size to be negatively correlated with abnormal volatility on announcement days in our study. The logarithm of the market value of equity, SIZE, at the beginning of each year is used as a proxy for firm size. 2.2.5. Intangibles (BM) Landsman and Maydew (2002) found that the stock of firms in intangible intensive industries displayed significantly greater abnormal announcement day volatility, suggesting that earnings announcements are highly informative in the case of intangible intensive firms (growth firms) and less informative for other firms. We expect this to translate into a negative correlation between abnormal volatility and our proxy for intangibles, the book-to-market ratio of equity at the beginning of each year (BM), where a lower value indicates more intangibles. 2.2.6. Diversification (DVN) The reporting requirements for a diversified firm, as compared to a comparable set of single sector firms, are lower since many transactions that would be reported as trades between firms are internalized in the diversified firm, and thus not reported to external parties. In accordance with Dolde and Mishra (2003), we hypothesize that this makes it
22
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
difficult for investors to scrutinize diversified firms’ operations and, as a consequence, forecast firm performance. Thus, we expect diversification to be positively correlated to abnormal volatility on the announcement days. We use a dummy variable, DVN, that is set to one if a firm operates in more than one industrial sector, and to zero otherwise. The data used in creating the dependent variable were obtained from the Stockholm Stock Exchange, and the data for the explanatory variables (a)–(c) were obtained from the survey responses. The accounting data required to calculate the variables (d) and (e) were collected from the Nordbanken Aktieguide Sommar 2000 stock market guide, while the market values of equity were collected from the Owners and Power stock market ownership guides (Sundin and Sundqvist, 1997, 1998, 1999, 2000). For variable (f) we use the industry classification from the SCB (Statistics Sweden) standard SNI 92.
3. Results 3.1. Abnormal volatility Fig. 1 presents the estimated effect of earnings announcements on abnormal stock price volatility. The solid line shows the mean value of SAV for each day in the event window. The dashed lines represent bootstrapped 99% confidence intervals for the means. As is evident from the figure, there is positive and significant abnormal volatility of stock returns on both the announcement day (t = 0) and on day t = +1. This finding is consistent with Landsman and Maydew (2002), and suggests that earnings announcements have strong informational content. Table 1 presents statistics describing firm-specific variables for the 361 firm year observations. The FA level is above 25% for about half of the firm year observations, and above 59% for over a quarter of the firm year observations. Thus, the sample firms are characterized by a high degree of multinationality, which is to be expected in a small and open Table 1 Descriptive statistics survey responses
Mean Median Q3 Q1
FA
abs(NE)
H
DD
CD
FD
ME
BM
DVN
33.0 25.0 59.0 2.0
19.7 10.0 30.0 2.0
0.56
0.30
0.19
0.08
6268 1060 4080 320
0.65 0.58 0.85 0.33
0.24
The table displays descriptive statistics on firm characteristics, where the variables are FA, the average of the percentage of revenues and the percentage of costs denominated in foreign currency; abs(NE), the absolute value of the difference between the percentage of revenues and percentage of costs denominated in foreign currency; H, a dummy that has the value of one if a firm hedges and zero otherwise; DD, a dummy that has the value of one if a firm hedges with currency derivatives or foreign denominated debt and zero otherwise; CD, a dummy that has the value of one if a firm hedges with currency derivatives but not with foreign debt and zero otherwise; FD, a dummy that has the value of one if a firm hedges with foreign debt but not with currency derivatives and zero otherwise; ME, measured as market value of equity (MSEK); BM, proxy for growth opportunities, measured as the book-to-market ratio for the firm’s equity; and DVN, a dummy that has the value of one if a firm is diversified and zero otherwise. There are a total of 361 firm year observations.
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
23
1.5
1.4
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6 -10
-8
-6
-4
-2
0
2
4
6
8
10
Fig. 1. The figure presents the mean standardized abnormal volatility (SAV) for earnings announcements days over the period from January 1997 through March 2000. The horizontal axis represents days relative to the announcement day, t = 0, and the vertical axis shows the mean value of the SAV for the observations on each day, t, respectively. The solid line represents the mean value for each event day, while the dashed lines represent the 99% confidence interval for the mean values. The dashed horizontal line crosses the vertical axis at the value of SAV expected in the event that earnings announcements contain no information.
economy such as Sweden’s. The variable representing long and short positions, abs(NE), is lower in magnitude than is FA, but half of the sample observations still exceed 10%. Fifty-six percent of the observations are of firms that engage in hedging, and it was most common to hedge using both currency derivatives and foreign denominated debt (30%). Finally, 24% of the observations pertain to diversified firms. Table 2 presents summary statistics for the 1008 SAV observations on day t = 0, where the sample is divided into two groups based on the median value of SAV.10 The table shows that higher SAV values are associated with firms characterized by higher values for FA, hedging, intangibles, and diversification. Interestingly, the use of currency derivatives (DD, CD) seems to be positively associated with SAV, while the use of foreign denominated debt (FD) seems to have the opposite, if any, effect. 10
Note that the percentages of observations of, for example, hedgers (H) are not equal to the percentages in Table 1. This is because of missing data pertaining to SAV or to the explanatory variables for some firms.
24
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
Table 2 Statistics on earnings announcement observations FA
abs(NE)
H
DD
CD
FD
ME
BM
DVN
Below median SAV Mean 32.5 Median 25.0
13.6 4.0
0.58
0.32
0.17
0.08
6618 1300
0.72 0.63
0.20
Above median SAV Mean 42.2 Median 41.0
14.7 7.0
0.66
0.39
0.21
0.06
8104 1371
0.62 0.55
0.29
622 0.006
358 0.035
190 0.107
74 0.228
– 0.000
250 0.001
No. of dos. = 1 P-value
– 0.000
– 0.404
– 0.175
The table displays summary statistics for the 1008 earnings announcements where the sample is divided based on the median value of abnormal volatility (SAV). Means and medians are reported for the following variables: FA, the average of the percentage of revenues and the percentage of costs denominated in foreign currency; abs(NE), the absolute value of the difference between the percentage of revenues and percentage of costs denominated in foreign currency; H, a dummy that has the value of one if a firm hedges and zero otherwise; DD, a dummy that has the value of one if a firm hedges with currency derivatives or foreign denominated debt and zero otherwise; CD, a dummy that has the value of one if a firm hedges with currency derivatives but not with foreign debt and zero otherwise; FD, a dummy that has the value of one if a firm hedges with foreign debt but not with currency derivatives and zero otherwise; ME, measured as market value of equity (MSEK); BM, proxy for growth opportunities, measured as the book-to-market ratio for the firm’s equity; and DVN, a dummy that has the value of one if a firm is diversified and zero otherwise. The last row presents the P-values from t-tests for differences in means between the observations below the median and the observations above the median.
The relationships suggested in Table 2 do not take into account possible interactions between the explanatory variables. To investigate formally the effects from FX exposure and hedging on abnormal volatility, we use pooled OLS regressions. The first model, model (a), includes the effects of FX exposure, and is written SAVik = α + β1 FAik + β2 abs(NE)ik + β3 SIZEik + β4 BMik + β5 DVNik + εik
(7)
for all events ik.11 All variables are as explained in Section 2 and the results are displayed in Table 3, Panel A. The coefficient of FA in model (a) is positive and statistically significant, while the coefficient of abs(NE) is insignificant. This suggests that investors have difficulties understanding the operations of firms with high degrees of multinationality, but that the modeling of effects of a net long or short position may cause less difficulties. The result for FA supports the findings of Guay et al. (2003) that corporate risk exposures are not transparent to investors or analysts. The coefficients of the other explanatory variables have the expected signs. The SAV decreases along with SIZE and BM, which parallels earlier research. It is noteworthy that diversification, as captured by the dummy variable DVN, is associated with an increase in SAV. The latter conforms to our expectations and may be explained by the relative difficulty in analyzing the performance of diversified firms as compared with single sector firms. 11
The subscript t is suppressed for notational convenience. We only report findings for t = 0.
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
25
Table 3 Cross-sectional regressions (SAV) Model
a Coefficient
Panel A: pooled OLS regressions FA 0.0072 abs(NE) −0.0023 H DD CD FD SIZE −0.0764 BM −0.2492 DVN 0.2181 AdjustedR2 0.033 Panel B: between effects regressions FA 0.0077 abs(NE) −0.0030 H DD CD FD SIZE −0.0786 BM −0.3107 DVN 0.2129 AdjustedR2 0.038
b
c
P-value
Coefficient
P-value
Coefficient
P-value
0.000 0.289
0.0058 −0.0026 0.2407
0.001 0.222 0.033
0.0059 −0.0031
0.001 0.174
0.2097 0.3299 0.1298 −0.0978 −0.2876 0.2305 0.036
0.077 0.034 0.527 0.001 0.006 0.025
0.0064 −0.0039
0.001 0.149
0.1920 0.3445 0.1035 −0.0985 −0.3602 0.2262 0.043
0.215 0.052 0.685 0.006 0.007 0.053
0.003 0.005 0.032
−0.0988 −0.3058 0.2306 0.036
0.001 0.001 0.023
0.000 0.244
0.0063 −0.0033 0.2303
0.002 0.204 0.105
0.014 0.010 0.068
−0.0999 −0.3762 0.2257 0.042
0.004 0.003 0.052
The table displays the results from cross-sectional regressions. The dependent variable is the standardized abnormal volatility, denoted SAV, at day t = 0. Panel A displays the results from pooled OLS regressions and Panel B displays panel data regressions (between effects regressions). The independent variables are: FA, the average of the percentage of revenues and the percentage of costs denominated in foreign currency; abs(NE), the absolute value of the difference between the percentage of revenues and the percentage of costs denominated in foreign currency; H, a dummy that has the value of one if a firm hedges and zero otherwise; DD, a dummy that has the value of one if a firm hedges with currency derivatives or foreign denominated debt and zero otherwise; CD, a dummy that has the value of one if a firm hedges with currency derivatives but not with foreign debt and zero otherwise; FD, a dummy that has the value of one if a firm hedges with foreign debt but not with currency derivatives and zero otherwise; SIZE, proxy for size, measured as the logarithm of market value of equity; BM, proxy for growth opportunities, measured as the book-to-market ratio for the firm’s stock; and DVN, a dummy that has the value of one if a firm is diversified and zero otherwise. In Panel A, OLS estimates and P-values using White-adjusted standard errors are reported. There are a total of 1008 observations for the pooled OLS regressions. In Panel B, WLS estimates are reported. There are a total of 159 observations for the panel data regressions.
Next, we add hedging variables to the analysis. First, the hedge dummy, H, is added to regression Eq. (7) and the result is presented as model (b) in Table 3, Panel A. The coefficients of variables included in model (a) have unchanged signs and significance levels in this model. The main result from model (b) is that the SAV is significantly higher for firms that hedge than for non-hedgers. We interpret the results as possibly indicating that investors are unable to correctly assess the effects of hedging or that investors are inadequately informed of firms’ hedging activities. It is also noteworthy that the result concerning firms with high
26
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
degrees of multinationality (FA) is not only driven by hedging, corroborating the evidence presented by Bartov and Bodnar (1994) and Guay et al. (2003).12 Finally, to investigate possible differences between the effects of hedging with currency derivatives and hedging with foreign denominated debt, the last model in Panel A, model (c), incorporates variables serving as proxies for hedging with various financial instruments. In this model, hedgers are divided into three distinct groups based on the instruments used for hedging, as explained in Section 2.2. The coefficients and significance levels for the variables SIZE, BM, and DVN, as well as for the exposure variables FA and abs(NE), are similar to those reported in models (a) and (b). The coefficients of the dummy variables representing hedging with derivatives, CD and DD, are positive with low P-values (0.034, and 0.077, respectively), while the coefficient of FD lacks explanatory power (P-value = 0.527). This may be interpreted as evidence that investors find it more difficult to understand the effects of currency derivative use than the effects of the use of foreign denominated debt. However, the results for these dummy variables should be interpreted with some caution, for at least three reasons: (1) there are relatively few firms in the FD group, (2) the result for the DD dummy includes effects from both derivatives and debt, and for this group it is not possible to determine which of the two contributes most, and (3) the results from Guay and Kothari (2003) suggest that effects from currency derivatives use is very small, which questions our findings concerning the CD dummy. Although other findings suggest significant effects from derivatives use e.g. Allayannnis and Ofek (1998), it cannot be excluded that other factors, such as operational hedges, are correlated with these dummy variables and may drive our results. The sample consists of multi-year data for a core number of firms, which may cause high autocorrelation among the independent variables. Therefore, as a robustness test, we also performed panel data regressions. Table 3, Panel B, reports the estimates from between effects regressions. The between estimates use the mean values of each variable for each firm (reducing the sample to one observation for each firm), and are thus based solely on the cross-sectional component of the data. We present these estimates to address concerns that observations drawn repeatedly from the same sample of firms may not be independent.13 Overall, the results from the panel data regressions support the findings in Panel A, using the pooled OLS regressions. Throughout this analysis we use dummy variables to classify hedging behaviour. Since the dummy variables do not reflect the amount that each firm hedges, they are crude measures. Additional insights could be gained by using continuous variables of hedging. From 12 To investigate this further, we ran model (a) for hedgers and non-hedgers, respectively. The coefficients for abs(NE) and the control variables were practically unchanged, but for non-hedgers the FA was found to be insignificant. This may be explained by the fact that for non-hedgers the FA levels are low with little cross-sectional variation. The mean (median) FA is 15.7% (1.5) for non-hedgers compared to 49.7% (50.0) for hedgers. 13 Using panel data regressions, a choice of model has to be made. Fixed effects models adjust for the possibility that unobservable, firm specific factors influence the SAV for each individual company; the estimates are equivalent to estimating OLS models and including a dummy variable for each firm in the sample. However, the Hausman specification test does not support the use of fixed effect models. Further, shortcomings of the sample (short sample period and relatively static variables) reduce the possibility of finding within variation even if such effects exist. We therefore report between effects models, and it should be noted that the results using random effects models are similar to those reported in Panel B.
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
27
the questionnaires we have information on the proportion of committed transactions, anticipated transactions and translation exposure that firms hedged with currency derivatives, but we do not have this information for foreign denominated debt. Unfortunately, these variables are noisy proxies for hedging. This is because, although the questionnaires provide us with the proportions hedged, it does not include information regarding the explicit hedge horizon or how much of a firm’s total exposure that derives from each type of exposure. As a consequence, the interpretation of the proportions hedged is not straightforward, and they cannot be added or easily compared. Furthermore, the proportions of transaction and translation exposures hedged are highly correlated which makes inference difficult. In order to simplify interpretation and to avoid problems with multicollinearity, 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 cross-sectional regressions. The results from these regressions support the findings from using dummy variables. Specifically, the coefficients for the proportion variables have identical signs, but lower significance than the coefficients for the dummy variables. The lower significance is likely to be a result of multicollinearity and the similarity of the coefficients is expected given the fact that, for our sample, firms that hedged a particular exposure with derivatives tended to hedge a substantial part of it. For example, the median firm hedged 90% of its committed transactions and 79% of its translation exposure. 3.2. Systematic mispricing Bartov and Bodnar (1994) used a simple trading strategy and found evidence that investors systematically underestimated the effects of foreign exchange fluctuations on firm performance. To investigate whether such underestimation may systematically affect the firms in our sample, we use a trading strategy similar to that of Bartov and Bodnar (1994). The use of survey data enables us to use a direct proxy for firms’ net long or short positions in foreign currency. Specifically, our proxy, the net exposure (NE), is defined as the difference between the percentage of revenues and the percentage of costs of a firm that are denominated in foreign currency. A positive value indicates that the firm is long in foreign currency while a negative value indicates that the firm is short in foreign currency. We include only firms where the absolute value of NE is at least 5%.14 The trading strategy consists of shorting equal amounts of firms with net long positions (NE of at least 5%) when the Swedish krona (SEK) had appreciated in the previous quarter, and buying equal amounts of the same firms when the SEK had depreciated in the previous quarter. The same strategy, but with reversed positions, was applied in the case of firms with net short positions (NE less than or equal to −5%). We use two trading windows, a one-day trading window (day t = 0) and a three-day trading window (days t = −1, 0, +1), and calculate buy-and-hold returns. If there is any systematic mispricing that is corrected on earnings announcement days, we would expect there to be significant abnormal returns. We use standardized abnormal returns, SARik,0 , for the one-day trading window tests.15 For the three-day, buy-and-hold strategies, we use a standardized buy-and-hold abnormal return, 14 15
We also used abs(NE) values of at least 10%, and at least 20% with similar results. As a robustness test, we applied the strategies using abnormal returns, ARik,0 , with no change in results.
28
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
Table 4 Standardized abnormal returns (SAR) Firms included
Net exposure
Observations
Mean SAR(0) (standard error)
Mean SAR(−1,0,1) (standard error)
All firms
NE > 5 NE < −5
415 163
−0.118 (1.999) −0.028 (1.997)
−0.357 (6.691) 0.011 (6.297)
Hedgers
NE > 5 NE < −5
363 105
−0.217 (2.024) −0.107 (2.130)
−0.376 (6.092) −0.255 (6.801)
Non-hedgers
NE > 5 NE < −5
52 58
0.550 (1.690) 0.122 (1.727)
−0.232 (9.938) 0.527 (5.203)
The table displays the results from trading strategies based on the movement in the trade-weighted currency index in the quarter prior to the earnings announcement. The trading strategy consisted of shorting equal amounts of all firms with net exposures (NE) to foreign currencies of at least 5% (NE > 5) when the currency appreciated and buying equal amounts of all firms with a net exposure to foreign currencies of at least 5% when the currency depreciated. The same strategy, but with reversed positions, was also performed for firms that are short in foreign currency (NE < −5). Standardized abnormal stock performances are calculated over the following two intervals. The column SAR(0) reports on the trading strategy using a one-day window (t = 0), and the column SAR(−1,0,1) reports on the trading strategy using a three-day window (t = −1, 0, +1).
SARik,(−1,0,1) , where each observation is standardized as in the earlier case, but with an estimated three-day standard deviation. Further, to investigate whether hedging had a systematic effect on the abnormal returns for the sample firms, we apply the same trading strategies to hedgers and non-hedgers respectively. Table 4 displays the results of these tests. The lack of significant abnormal returns suggests that there is no systematic mispricing. This finding, that investors have unbiased expectations, is in contrast to the findings of Bartov and Bodnar (1994). However, they found systematic mispricing to be less pronounced in the latter part of the period they studied. They argued that this may indicate that investors were gradually improving their assessment of the effect of FX rate changes on firm performance. The sample period of our study is more recent than that of Bartov and Bodnar’s (1994) study which possibly can explain the difference between the studies.
4. Conclusions We investigate the effect of FX exposure and hedging activity on the abnormal stock price volatility associated with quarterly earnings announcements. We find that abnormal stock price volatility on earnings announcement days is negatively related to firm size. This is in line with earlier research, and suggests that large firms characteristically supply more pre-disclosure information. Evidence also indicates that it is relatively difficult for investors to correctly predict the earnings of growth firms and of diversified firms. We find, in line with Bartov and Bodnar (1994), and Guay et al. (2003), that investors need the information contained in earnings announcements to assess the impact of FX rate changes on firm performance. Furthermore, our results suggest that this is related to firms’ degree of multinationality, rather than to their net long or short positions. Moreover, in contrast to
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
29
Bartov and Bodnar (1994) we find no evidence that the increase in abnormal volatility is due to systematic mispricing. Importantly, we find that abnormal stock price volatility on earnings announcement days is positively correlated with hedging. More specifically, volatility is positively correlated with currency derivatives use, but not significantly correlated with the use of foreign denominated debt. This may result from investors’ inability to correctly evaluate information on currency derivatives use, which includes matters such as the non-linear payoffs from options. It may also stem from the fact that derivatives are not included on the balance sheet, so investors lack the information necessary to make correct judgments on the effect of derivatives on earnings. If the latter explanation is true, improving the reporting requirements, as required by the recent FASB statement FAS 133, may help investors in contemporaneously assessing the effect of FX rate changes on firm value and consequently reduce the abnormal volatility on earnings announcement days. In this study, we categorize firms’ hedging behavior in a simple fashion by using dummy variables, and are able to document several interesting results. However, we suggest that valuable contributions in this field can be made by exploring our findings further using improved measures of the determinants of abnormal volatility on earnings announcement days. This may include refined measures of financial hedging, but also other measures of firms’ exposure management, such as operational hedging techniques. Acknowledgements We would like to thank Raj Aggarwal, Hossein Asgharian, Martin Holmén, Baeyong Lee, Lars Nordén, Clas Wihlborg, and an anonomous referee for their valuable comments. We would also like to thank participants in the 27th Annual Meeting of the European International Business Academy (Paris, 2001), the 51st Annual Meeting of the Midwest Finance Association (Chicago, 2002), the 38th Annual Meeting of the Eastern Finance Association (Baltimore, 2002), and seminar participants at Stockholm University School of Business.
References Allayannnis, G., Ofek, E., 1998. Exchange rate exposure, hedging, and the use of foreign currency derivatives. Journal of International Money and Finance 20, 273–296. Allayannis, G., Weston, J.P., 2001. The use of foreign currency derivatives and firm market value. The Review of Financial Studies 14, 243–276. Amihud, Y., 1993. Evidence on exchange rates and valuation of equity shares. In: Amihud, Y., Levich, R.M. (Eds.), Exchange Rates and Corporate Performance. Irwin Professional Publishing, New York. Atiase, R.K., 1985. Predisclosure information, firm capitalization, and security price behavior around earnings announcements. Journal of Accounting Research 23, 21–36. Bamber, L.S., 1987. Unexpected earnings, firm size, and trading volume around quarterly earnings announcements. The Accounting Review 62, 510–532. Bamber, L.S., Barron, O.E., Stober, T.L., 1997. Trading volume and different aspects of disagreement coincident with earnings announcements. The Accounting Review 72, 575–597. Barron, O.E., 1995. Trading volume and belief revisions that differ among individual analysts. The Accounting Review 70, 581–597.
30
N. Hagelin, B. Pramborg / J. of Multi. Fin. Manag. 15 (2005) 15–30
Bartov, E., Bodnar, G.M., 1994. Firm valuation earnings expectations, and the exchange rate exposure effect. Journal of Finance 49, 1755–1785. Beaver, W.H., 1968. The information content of annual earnings announcements, Journal of Accounting Research 6 (Suppl.), 67–92. Bodnar, G.M., Gentry, W.M., 1993. Exchange rate exposure and industry characteristics: evidence from Canada, Japan, and the USA. Journal of International Money and Finance 12, 29–45. Boehmer, E., Musumeci, J., Poulsen, A., 1991. Event-study methodology under conditions of event induced variance. Journal of Financial Economics 30, 253–272. Campbell, J.Y., Lo, A., MacKinley, A.C., 1997. The Econometrics of Financial Markets. Princeton University Press, Princeton. Dolde, W., Mishra, D.R., 2003. Firm complexity and FX derivatives use. Working Paper, University of Connecticut. Efron, B., Tibshirani, R.J., 1993. An Introduction to the Bootstrap. Chapman and Hall, New York. Francis, J., Schipper, K., Vincent, L., 2002. Earnings announcements and competing information. Journal of Accounting and Economics 33, 313–342. Friedrich, S., Gregory, A., Matatko, J., Tonkis, I., 2002. Short-run returns around the trades of corporate insiders on the London Stock Exchange. European Financial Management 8, 7–30. Graham, J., Rogers, D., 2002. Do firms hedge in response to tax incentives? Journal of Finance 57, 815–839. Guay, W., Haushalter, D., Minton, B., 2003. The influence of corporate risk exposures on the accuracy of earnings forecasts. Working Paper, University of Pennsylvania. Guay, W., Kothari, S.P., 2003. How much do firms hedge with derivatives? Journal of Financial Economics 70, 423–461. Hagelin, N., 2003. Why firms hedge with currency derivatives: an examination of transaction and translation exposure. Applied Financial Economics 13, 55–69. Hagelin, N., Pramborg, B., 2004. Hedging foreign exchange exposure: risk reduction from transaction and translation hedging. Journal of International Financial Management and Accounting 15, 1–20. Hagelin, N., Pramborg, B., 2003. Empirical evidence on the incentives to hedge transaction and translation exposure. Working Paper, Stockholm University. Holthausen, R.W., Verrecchia, R.E., 1990. The effect of informedness and consensus on price and volume behavior. The Accounting Review 65, 191–208. Jorion, P., 1990. The exchange rate exposure of U.S. multinationals. Journal of Business 63, 331–345. Jorion, P., 1991. The pricing of exchange risk in the stock market. Journal of Financial and Quantitative Analysis 26, 353–376. Landsman, W.R., Maydew, E.L., 2002. Has the information content of annual earnings announcements declined in the past three decades? Journal of Accounting Research 40, 797–808. Marston, R.C., 2001. The effects of industry structure on the economic exposure. Journal of International Money and Finance 20, 149–164. Pramborg, B., 2002. Empirical essays on foreign exchange risk management. Doctoral Thesis, Stockholm University.