Foreign currency derivatives use in the Eurozone; does it increase firm value?

(1)

Foreign currency derivatives use in the Eurozone; does it increase firm value?

Author: Ardin Bosboom Student number: 2491788 Study Programme: MSc IFM Faculty of Economics and Business University of Groningen

Supervisor: Dr. Peter Smid

Abstract

This research focuses on the usage of foreign currency derivatives (FCDs) in large Eurozone firms. The main objective is to find out if there is a relationship between the usage of foreign currency derivatives and firm value. To this end, a sample of 146 large multinational firms headquartered in Eurozone countries is used. The aggregate value of underlying positions of firm FCD usage was collected and contrasted with firm value, proxied by Tobin’s Q, with and without controls. In the end, it was found that there exists an average hedging premium of 11.23% of market value for firms that have over 5% of their total assets worth in underlying FCD positions.

(2)

2 1. Introduction

Firms have many reasons and many strategies to hedge. The main reason for the application of hedging policies is to counteract the effects of different forms of exposure. Many forms of exposure exist, such as interest exposure and exchange rate exposure, which can form a risk to the firm, as they disturb projected cash flows. Hedging policies are therefore put in place as a means to decrease the variability of cash flows, thus mitigating exposure risk. Over the last few decades, many firms have turned to the usage of derivatives as a means of hedging. Even though Warren Buffet described derivatives as ‘financial weapons of mass destruction’, usage rates have not been affected as empirical research has shown that a growing majority of multinational firms have embraced derivative hedging as part of their risk management strategy. In this particular field of research, the largest contributions have been made over a decade ago (Allayannis and Weston, 2001; Jin and Jorion, 2006). The major studies in this field have taken an American perspective to this topic, which makes it interesting to transfer the research framework to a European setting.

(3)

3

Firms have many different choices and reasons with regard to hedging strategies, as they can be active on futures, forwards and option markets as well as engaging in currency swaps, and there are increasingly more complex derivative hedging instruments being developed. The controversy surrounding the use of derivatives is that while it is generally used purely for hedging purposes, derivatives can also be used for speculation purposes. For example, a firm can use derivatives to bet on the expectation that the underlying asset will increase in value, thus yielding a profit on the derivative, the risk of course being that the asset will not reach the expected value and the firm incurs a loss. Allayannis and Ofek (2001) researched this phenomenon and found that the majority of foreign currency derivative usage is for hedging purposes. Furthermore, in general, larger firms have more reason to use derivatives as they face exposure to foreign exchange rates and to interest rates. Geczy et al. (1997) confirm this hypothesis empirically, finding that users of derivatives are generally larger than nonusers, and have higher growth opportunities. Smaller firms, especially exporters, can also benefit from the use of FCDs, however for data availability and external validity reasons, this research will encompass a sample of large firms.

(4)

4 2. Literature review

Literature on hedging has been around for a few decades, Smith and Stulz (1985) identified the underlying reasons of firms’ hedging policies, stating firm value maximization as the main driver for hedging operations. Maximization of firm value can be achieved through hedging in several ways, through tax effects, greater leverage leading to tax advantage (Leland, 1998), or simply through the smoothing of cash flows (Smith and Stulz, 1985). Empirical literature regarding foreign currency derivative use and its relationship with firm value has shown differing results, with studies showing that there is indeed a hedging premium (Allayannis and Weston, 2001; Clark and Judge, 2009) and studies stating that hedging foreign exchange exposure has no statistically significant premium attached (Jin and Jorion, 2006; Jin and Jorion, 2007). It has to be noted that the two studies by Jin and Jorion were industry-specific, namely on the oil and gas industry and the gold mining industry, respectively. Allayannis and Weston researched US large nonfinancial firms, while Clark and Judge sampled large UK nonfinancial firms. Most of the studies on foreign currency

derivatives hedging have employed a U.S. based perspective, which begs the question if the proposed effects of hedging have the same effect in the Eurozone as they do in the U.S. Studies of this phenomenon in a European context have been scarce; Nguyen et al. (2007) find that after the introduction of the euro, foreign exchange rate exposure and the usage of foreign currency derivatives in France has decreased, while Hutson and O’Driscoll (2010) sample 7 Eurozone, including France, and 4 non-Eurozone countries after the introduction of the euro and find an increase in the level of foreign exchange exposure comparing it to the pre-euro period. Specific empirical literature on the link between foreign currency

derivatives and firm value in the Eurozone is difficult to come by. Belghitar, Clark and Mefteh (2012) researched a sample on currency exposure in French firms and tried to find a link between shareholder value and hedging against currency exposure, but found no statistically significant effect on shareholder value. Clark and Judge (2009) tested the effect of different forms of hedging on different forms of exposure on a sample of UK firms, which is not a Eurozone country, but nevertheless it is a study with a European perspective and found a significant hedging premium. Relevant empirical research on this topic is quite scarce, given that for a long time, research had to be done using survey data until FASB standards

(5)

5

reporting regulations in the U.S. and Europe require firms to report on their derivatives usage, there is ample opportunity to research this topic. Since the introduction of these regulations, there have been a few empirical studies on this topic that generally give support to the hypothesis that the usage of foreign currency derivatives leads to an increased firm value (Allayannis and Ofek, 2001; Allayannis and Weston, 2001; Allayannis et al., 2011; Carter et al., 2006). On the other hand, Jin and Jorion (2006), testing U.S. oil and gas producers, found no significant premium between users and nonusers of foreign currency derivative hedges and found the same result in Jin and Jorion (2007) when the gold industry was tested. The results of previous empirical literature thus have been mixed, but it seems that the larger part agrees that hedging is value-adding and consequently that hedging against foreign exchange exposure using foreign currency derivatives presents a hedging premium to the user. Thus the main hypothesis is: Users of foreign currency derivatives on average have a higher firm value compared to nonusers.

3. Data and methodology

This section deals with the relevant data and the methodology used for analysis. Firstly, the main analytical framework will be introduced and explained. After that, the sample will be delimited and the methodology will be discussed. Finally, a description of the dataset will be provided.

The analytical section of this study is twofold. Firstly, a univariate OLS analysis will be performed with the independent variable FCD usage regressed on the dependent variable, which is the proxy for market value, Tobin’s Q. For this univariate analysis, a model has been created, which is shown in equation (1).

Tobin’s Q = α + β1 (Gross notional FCD usage/Total Assets) (1)

(6)

6

divided by the book value of total assets. Different measures of Tobin’s Q are available and will be part of the robustness analysis.

The independent variable reflects the usage of foreign currency derivatives of a firm. Under IFRS 7, firms are required to disclose any financial instrument related to the entity with IAS 39 stipulating the recognition and measurement of these instruments. Thus, calculation of the hedging variable will be done through manual scanning of the annual reports. The composition of the hedging variable will be the aggregated gross notional value of

(7)

7

Tobin’s Q = α + β1(Gross notional FCD usage/Total Assets) + β2ln(Assets) + β3(ROA) +

β4(Dividend Dummy) + β5(Foreign Sales/Total Sales) + β6(Debt Ratio) + β7(Capex/Total Sales)

(2)

1) Size

Size is used as a control variable since larger firms will in general have higher usage of derivatives and have perhaps a higher profitability than smaller firms, thus it is important to control for that. Lang and Stulz (1994) and Allayannis and Weston (2001) found a negative direction for size, thus this is also the hypothesized direction. The calculation for size will be done by computing the natural log of the book value of a firm’s total assets at the end of each book year.

2) Profitability

If firms have higher levels of profitability than other firms, their q ratio will usually be higher due to the higher market value investors will attach to the profitable firm, which makes it necessary to control for profitability. Profitability is defined by the efficient use of resources, thus the return on assets is used as the proxy for profitability. Return on assets will be calculated as the ratio of a firm’s year-end net income to book value of assets.

3) Access to financial markets

The q ratio may be influenced by limited access to financial markets. If a firm has limited or no access to financial markets, it is more susceptible to risk due to lack of financing and will thus only undertake investments that yield guaranteed positive results. This can skew the q ratio upward compared to firms with more access to financing who are able to undertake more risky investments. To control for this, based on Allayannis and Weston (2001) a

(8)

8 4) Leverage

The capital structure of the firm might have an influence on its market value. In the classic Modigliani and Miller paradigm, with taxes, it is argued that the ideal firm capital structure contains a high amount of debt. However, other theories, such as trade-off and pecking order also have their own ideas about capital structure and market value, thus it is

important to control for differences in capital structures. To that end, based on Allayannis and Weston (2001), the proxy will be the debt ratio of the firm, calculated as the ratio of the long-term debt of a firm divided by the book value of its shareholder’s equity.

5) Investment growth

Firm value has been shown to also be dependent on the possibility of future investment opportunities. Hedgers in general have more investment opportunities due to their decreased variability in cash flows, thus higher amounts of hedging could lead to more investment opportunities and in the end a higher q. The ratio of capital expenditures over total sales is used as a proxy to define the investment opportunities of a firm and the expected direction is positive.

6) Country effects

Differing country regulations, such as tax laws, might have an effect on q. To control for this effect, country dummies will be inserted to control for any inter-country variation.

7) Industry effects

Q-ratios may be substantially different across industries, which needs to be controlled for. In this sample, categorizing the firms using SIC codes provides controllable samples by industry, much like the country effects.

(9)

9

worth of assets in all years between 2012 and 2016. The choice for large firms over small and medium firms has several advantages. Firstly, data availability is higher when

researching large firms, given that FCD usage has high start-up costs (Allayannis and Weston, 2001) and that FCD’s are mainly used in conjunction with international activity, mitigating exchange rate exposure, which is more a characteristic for large, listed firms. Finally, the choice for large firms concerns the reproducibility of earlier studies (Allayannis and Weston, 2001; Allayannis and Ofek, 2001; Clark and Judge, 2009) which also use large firms. The final sample, which includes all firms with non-missing data on Tobin’s Q and the hedging

variable, encompasses 146 firms or 730 firm-year observations across 10 Eurozone countries and 9 industries. France and Germany have the most representative firms in the sample, 57 and 42 respectively. Regarding industries, based on US primary SIC codes the manufacturing industry has 87 firms in the sample, which makes it the largest representative industry. Prior to OLS testing, it is necessary to find out if OLS is the best estimator to use. Firstly, the variables used in the multivariate regression equation (2) will be trimmed for significant outliers in order to improve normality within the variables. To test for any correlation between variables, a correlation table will be made and can be found in table 1.

Log assets FCD ratio Debt ratio

Div. dummy For. Sales% Inv. Growth ROA Tobin's Q

Log assets 1 -0.114 0.148 -0.003 0.069 -0.115 -0.113 -0.230 FCD ratio -0.114 1 -0.164 -0.065 0.099 -0.130 0.051 -0.002 Debt ratio 0.148 -0.164 1 -0.056 -0.108 0.109 -0.262 -0.128 Div. dummy -0.003 -0.065 -0.056 1 -0.032 0.023 0.298 0.204 For. Sales% 0.069 0.099 -0.108 -0.032 1 -0.273 0.046 0.024 Inv. Growth -0.115 -0.130 0.109 0.023 -0.273 1 -0.015 -0.023 ROA -0.113 0.051 -0.262 0.298 0.046 -0.015 1 0.544 Tobin's Q -0.230 -0.002 -0.128 0.204 0.024 -0.023 0.544 1

Table 1: Correlation table of the trimmed dependent and independent variables

(10)

10

therefore for the regressions, it has been transformed to its natural log in order to normalize.

Finally, the descriptive statistics of all the firms and variables used in the panel for the main analysis can be found in table 1.

Panel: All firms with disclosed notional amounts of FCD between 2012-2016

Observations Min Max Mean Std. Dev. Median

General firm descriptives

Total Assets (millions) 727 1950 280000 28600 42900 11500

Total Liabilities (millions) 727 471 239000 19000 31700 6580

Total Equity (millions) 727 -7340 93500 8850 12600 3710

Dividend Payment

(millions) 730 0 16200 501 1050 501

Tobin's Q and FCD usage

Tobin´s Q

730 0.277 3.852 1.029 0.544 0.906

Log Tobin’s Q 730 -1.284 1.349 -0.089 0.476 -0.099

Gross FCD value (millions) 730 0 83600 3590 8860 965

FCD to assets ratio 730 0 1.14 0.14 0.18 0.06

Control variables

Log assets 727 21.393 26.358 23.311 1.205 23.165

Foreign Sales/Total Sales 696 0 100 62.99 27.57 67.75

Return on Assets 721 -16.43 19.02 3.467 4.516 3.581

Dividend dummy 730 0 1 0.93 0.25 1

Debt Ratio 724 0 7.928 0.768 0.793 0.559

Capex/Sales 716 0.0031 0.953 0.08 0.117 0.045

(11)

11 4. Analysis

4.1 Univariate analysis

The amount of currency derivative users has only been rising, be it for trading purposes or for hedging purposes, numbers have skyrocketed. In the study by Allayannis and Weston (2001), the percentage of firms with recorded FCD usage was 40% in 1995, whereas in this sample the percentage of firms with recorded FCD usage reached 89% in 2016. Despite the difference in geographical settings, given that Allayannis and Weston researched U.S. firms, the significant rise in overall FCD usage is evident. This means that it is not particularly useful to make a distinction between users and nonusers of FCDs, since in the sample of 146 firms with disclosed values there were only 14 recorded overall nonusers. It is therefore not statistically useful to compare 10 percent of the sample versus the other 90 percent,

especially since most of the nonusers are among the smallest firms in the sample. The main goal of this research is to find out if the usage of FCDs results in higher firm value, thus, the research should reflect this goal. To find out if this holds true, it is necessary to compare subsamples of types of FCD users. Allayannis and Weston (2001) compared nonusers to users, however with this sample this would result in unequal sample sizes (14 nonusers versus 132 users). Therefore, a different classification will be made, based on the

transformed hedging variable. The hedging variable in this study classifies firms on the basis of gross FCD usage as a percentage of total assets. In order to obtain relatively equal sample sizes and to retain the distinction between low FCD users and high FCD users, the sample will be divided into firm-year observations that have gross FCD usage below five percent of total assets and firm-year observations that have gross FCD usage above five percent of total assets. The resulting samples include 318 firm-year observations for the sample below five percent and 412 firm-year observations above five percent. Descriptive statistics on both subsamples can be found in table 3.

Observations Observations Mean Mean Median Median

FCD below 5% FCD above 5% FCD below 5% FCD above 5% FCD below 5% FCD above 5% General descriptives

(12)

12

Total Liabilities (millions) 315 412 17800 19900 7790 5100

Total Equity (millions) 315 412 8740 8940 3830 3430

Dividend Payment (millions) 318 412 449 542 167 157

Tobin's Q and FCD usage

Tobin´s Q 318 412 0.982 1.065 0.851 0.935

Gross FCD value (millions) 318 412 598 5890 118 2070

FCD to assets ratio 318 412 1.671 23.467 1.285 18.871 Control variables Foreign Sales% 300 396 59.47 65.66 63.07 69.18 Return on Assets 312 412 2.78 3.64 3.18 3.78 Dividend dummy 318 412 0.94 0.92 1 1 Debt Ratio 315 409 0.974 0.61 0.735 0.474 Capex/Sales 315 406 0.097 0.065 0.047 0.042

Table 3: Pairwise comparison of means and medians of the split panel. Subsamples in this table are created on the basis of the ratio of gross FCD usage to total assets, splitting the main sample in two at the 5% FCD to total assets mark.

There are a few interesting observations to be made regarding these subsamples. The main variable of interest is of course Tobin’s Q, which has a mean of 0.982 for firms with usage under five percent of total assets and a mean of 1.065 for firms above five percent, which is an indication that a hedging premium might exist. Further notable discrepancies are found within the debt ratio and investment growth. The investment growth proxy of capex over total sales yields a mean of 9.7% in the subsample below five percent and a mean of 6.5% within the subsample above five percent. This is not in line with existing theory, which states that hedgers have less variability, thus more certainty over future cash flows implying more investment opportunities.

(13)

13

FCD below 5% (1) FCD above 5%(2) Difference(2)-(1) t-statistic

Mean Q 0.982 1.065 0.083 2.034**

FCD below 5% (1) FCD above 5% (2) Difference(2)-(1) Chi-squared statistic

Median Q 0.851 0.935 0.084 3.842**

Table 4: Equality test of mean and median q’s of the subsamples of firms with <5% and >5% of FCD usage of total assets. ** denotes significant at the 5% level

What the results in table 4 show is that, at the five percent level, the means and medians of the subsamples are significantly different. Practically, this can show evidence of the

existence of a hedging premium of around 0.083 with regard to Tobin’s Q, which is robust when looking at the median difference, which is almost the same at 0.084.

4.2 Multivariate analysis

In order to control for any other exogenous variables that could potentially influence the Q ratio, multivariate tests have to be conducted. These tests will use the model as described in equation (2), with any additional dummies that are needed to control for subsamples. For every separate regression, pooled OLS will be performed, as well as time and entity fixed-effects equations, in order to control across periods and across entities. The raw dependent variable Tobin’s Q is skewed, therefore from here on, the natural log of Tobin’s Q will be taken as the dependent variable, as this improved the distribution and it now approaches normality. Furthermore, in these tests, the main independent variable FCD usage is proxied by the percentage of usage when compared to total assets. Controls are put into place for size, leverage, access to financial markets, percentage of foreign sales, investment growth and return on assets. Three separate regressions will be performed; firstly the main

(14)

14

Secondly, the sample will also be tested on an industry basis and a country basis. Similar to the previous model, these regressions will also run with the regression equation as outlined in equation (2), with the addition of dummies. For the industry analysis, given that the majority of firms are found to be operating in the manufacturing industry (87 out of 146 sampled firms), there will be a manufacturing dummy put into place which takes on a 0 if the firm does not operate in the manufacturing industry and takes on a 1 if a firm does operate in the manufacturing industry. Time and entity fixed-effects will also be performed on this regression.

Finally, a model will also be made for the country sample. Again, this model follows equation (2) with the addition of dummy variables. It was found that the largest representative

countries in the sample were France (57 out of 146 firms) and Germany (42 out of 146 firms), thus there will be two dummies put into place. Firstly, a dummy which takes on a 0 if the firm is not headquartered in France and 1 if the firm is headquartered in France, and secondly a dummy which takes on a 0 if the firm is not headquartered in Germany and 1 if the firm is headquartered in Germany. For this regression, as usual a time fixed-effects test will be performed, however for this specific modeling, an entity random-effects test will be performed, based on the Hausman test, which came out as not significant with a Chi-squared statistic of 3.661 and an associated p-value of 0.812.

Tables 5, 6 and 7 will present respectively the model with the FCD percentage dummy, the model with the manufacturing and the model with the country dummies, as well as their associated fixed- and random effects tests and adjusted R-squared. Economic interpretation of the results is provided after each model.

Panel: All firm-year observations with disclosed notional FCD values

Dependent variable: ln Tobin's Q

Observations: 676 Pooled OLS Time fixed-effects Firm fixed-effects

Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic

FCD percentage -0.003515 -3.43*** -0.003661 -3.659*** 0.003425 2.405**

Size (log assets) -0.090842 -7.329*** -0.09318 -7.7*** -0.022255 -0.46

Debt ratio 0.037053 1.839* 0.042835 2.175 0.033358 1.555

(15)

15

Foreign sales/Total Sales 0.000176 0.312 0.000167 0.305 0.000606 2.052**

Investment Growth -0.167027 -1.108 -0.122268 -0.83 -0.031657 -0.371

Return on Assets 0.050982 14.548*** 0.050981 14.916*** 0.012868 5.167***

Percentage dummy 0.11225 2.979*** 0.109093 2.968*** 0.08072 2.514**

Adj. R-squared 0.353 0.385 0.857

Table 5: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression on the dependent variable Tobin’s Q, which is calculated as market value of the firm’s equity, plus the book value of its preferred stock and long term debt, divided by the book value of total assets. FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Investment growth is proxied by capital expenditures/total sales and Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

Table 5 shows the results of a pooled OLS regression, together with time and firm fixed effects tests on regression equation (2) with the addition of a percentage dummy. This percentage dummy has been added to capture the effect of hedging. This dummy variable takes on 0 when a firm has underlying FCD positions worth below 5% of total assets and takes on 1 when a firm has underlying FCD positions worth above 5% of total assets. In the pooled OLS regression, most variables show significance, with the exception of the foreign sales percentage and the investment growth, which is conflicting with earlier studies. Allayannis and Weston (2001) also found no significance with regard to investment growth, whereas their foreign sales percentage was significant at the 1% level, while Jin and Jorion (2006) and Jin and Jorion (2007) found significance of investment growth at the 5% level in both studies, with the caveat being that these were industry-specific studies, namely on oil and gas producers and on the gold mining industry respectively.

In this model, the main variable of interest is the percentage dummy, which has a coefficient of 0.11225 and is significant at the 1% level. The implication of this is that on average, firms that have an underlying position of foreign currency derivatives worth more than five percent of their total assets are valued 11.23% higher on the market than firms with

(16)

16

hedging premium found in Allayannis and Weston (2001), who found an average hedging premium of 4.87%. However, this hedging premium was specifically for the distinction between hedgers and hedgers, whereas in this case, the distinction between non-hedgers and non-hedgers is less clear due to the small amount of non-non-hedgers. However it is clear that a higher level of hedging will often lead to higher firm value.

As mentioned before, the foreign sales and investment growth control variables are not significant; however the remaining control variables are all significant, usually corresponding to their hypothesized directions. The coefficient sign for size, proxied by natural log of firm assets, is negative, which is in line with previous results found by Lang and Stulz (1994) and Allayannis and Weston (2001). The profitability proxy ROA is also significant at the 1% level, with the hypothesized positive direction, in line with Allayannis and Weston (2001) and Jin and Jorion (2007). The leverage proxy debt ratio is only significant at the 10% level. Across studies, the effect of leverage has been ambiguous; Allayannis and Weston (2001) found significance at the 1% level, but a very small coefficient, whereas Jin and Jorion (2007) found no significance.

The variable that stands out is the dividend dummy, which took on a 0 if a firm did not pay dividends that year and a 1 if a firm did pay dividends that year. Theory suggests that

dividend payment will be negatively correlated to firm value due to the decrease in financial constraints allowing firms to undertake more risk, which is also what was found in Allayannis and Weston (2001), which also included a dividend dummy, but there the coefficient was negative and significant at the 1% level. One reason could be the dividend signaling theory, which states that the announcement of dividend payments sends out positive signals to investors, who in turn reward the firm with higher valuation.

(17)

17

control variables in the time-fixed effects model are consistent with the pooled OLS, with the exception of the debt ratio not being significant at the 10% level anymore. The control variables in the firm fixed-effect model are less consistent, debt ratio is also not significant anymore at the 10% level, but also size is not significant at all anymore, while the foreign sales percentage variable is now significant at the 5% level.

The next model will include an additional control variable based on industry, the results of which are presented in table 6.

FCD percentage -0.003517 -3.429*** -0.003662 -3.658*** 0.003426 2.404**

Size (log assets) -0.091034 -7.337*** -0.09337 -7.704*** -0.021722 -0.448

Debt ratio 0.036693 1.819* 0.042467 2.153** 0.032915 1.532

Dividend dummy 0.226687 3.692*** 0.219648 3.661*** 0.223017 3.807***

Foreign sales/Total Sales 0.000176 0.312 0.000167 0.304 0.000603 2.042**

Return on Assets 0.050892 14.496*** 0.05089 14.863*** 0.012857 5.159***

Manufacturing dummy 0.01489 0.504 0.014931 0.519 0.007317 0.464

Adj. R-squared 0.353 0.385 0.857

Table 6: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression on the dependent variable Tobin’s Q, which is calculated as market value of the firm’s equity, plus the book value of its preferred stock and long term debt, divided by the book value of total assets. FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Investment growth is proxied by capital expenditures/total sales and Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. An additional dummy is constructed for the industry, taking 0 if a firm does not operate in the manufacturing industry and 1 if a firm does operate in the manufacturing industry. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

Table 6 presents an extension from the regression equation used in table 5, with the

(18)

18

effect that may exist as an influence on Tobin’s Q. The inserted dummy is called manufacturing dummy, as it most of all represents the effect that operating in the

manufacturing industry has. Firms that operate in the manufacturing industry make up half the sample, therefore the dummy that was constructed takes on 0 when a firm does not operate in the manufacturing industry and takes on 1 when a firm does operate in the manufacturing industry, as delimited by US primary SIC codes.

Comparing coefficients and their significance across tables 5 and 6, there are virtually no differences; the hedging premium in the pooled OLS still remains 11.25% and no

significances have been altered with the addition of this manufacturing dummy. The new dummy variable is not significant, thus implying that on average it does not matter in what major sector a firm operates, as it should not have a statistically significant influence on their market value.

The final multivariate model will include two additional control variables with regard to country effects, the results of which can be found in table 7.

Observations: 676 Pooled OLS Time fixed-effects Firm random-effects

FCD percentage -0.003964 -3.821*** -0.004098 -4.047*** 0.000236 0.212

Size (log assets) -0.086357 -6.831*** -0.088803 -7.189*** -0.080049 -3.869***

Debt ratio 0.028832 1.396 0.034838 1.725* 0.017719 0.973

Dividend dummy 0.237486 3.871*** 0.230392 3.842*** 0.264968 5.037***

Foreign sales/Total Sales 0.0000483 0.086 0.0000416 0.076 0.000542 1.853*

(19)

19

Table 7: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression on the dependent variable Tobin’s Q, which is calculated as market value of the firm’s equity, plus the book value of its preferred stock and long term debt, divided by the book value of total assets. FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Investment growth is proxied by capital expenditures/total sales and Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. Additional dummies are constructed for country controls. France dummy equals 0 if the firm is not headquartered in France and 1 if it is, while Germany dummy equals 0 if the is not headquartered in Germany and 1 if it is. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

Table 7 presents an extension of the regression equation from table 5 by adding country dummies in order to isolate country effects on Tobin’s Q. The two largest representative countries in the sample were France and Germany, thus there have been two additional dummies created, namely France dummy and Germany dummy. France dummy takes on 0 if a firm is not headquartered in France and 1 if it is, while Germany dummy takes on 0 if a firm is not headquartered in Germany and 1 if it is.

As was the case with the industry dummy, the addition of country dummies does not seriously change the variables’ coefficients or their significances, except for the debt ratio, which always has around a 10% significance level. However, what is notable is that the France dummy has a negative value and is significant at the 5% level. This implies that, on average, a firm headquartered in France has a lower market value than a firm

headquartered somewhere else in the Eurozone, 7.15% less market value to be exact.

5. Robustness analysis

(20)

20

Dependent variable: ln Simple Tobin's Q

Observations: 676 Pooled OLS Time fixed-effects Entity fixed-effects

FCD percentage -0.003339 -3.122*** -0.003312 -3.088*** -0.003688 -3.188***

Size (log assets) -0.085544 -5.424*** -0.085406 -5.4*** 0.009804 0.315

Debt ratio 0.017748 0.702 0.017941 0.707 -0.014141 -0.513

Dividend dummy 0.310199 3.978*** 0.310745 3.975*** 0.361257 4.191***

Foreign sales/Total Sales -0.000129 -0.181 -0.000257 -0.352 -0.00005 -0.065

Return on Assets 0.055781 12.519*** 0.055845 12.476*** 0.056414 11.942***

Percentage dummy 0.007917 0.209 0.006672 0.175 0.096326 1.208

Adj. R-squared 0.298 0.287 0.337

Table 8: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression on the dependent variable Tobin’s Q, which is calculated as (enterprise value/book value of total assets). FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Investment growth is proxied by capital expenditures/total sales and Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

Table 8 shows the coefficients for the variables and their significances with a new dependent variable, namely the simple version of Tobin’s Q. Almost all variables retained their

coefficients and significances. However the largest, and only, differences can be found with regard to the percentage dummy, which is also the most important variable in this equation. The coefficient has dropped to near zero, while there is no significance to be found, as compared to a coefficient of 0.112 and significance at the 1% level when using the extended calculation of Tobin’s Q. This test implies that level of hedging does not matter for the size of Tobin’s Q, therefore initial results are not robust to a simple calculation of Tobin’s Q.

(21)

21

FCD percentage -0.00301 -1.981** -0.0031 -2.035** -0.002467 -0.554

Size (log assets) -0.092715 -5.044*** -0.094413 -5.121*** -0.06319 -0.417

Debt ratio 0.023223 0.777 0.025749 0.858 0.14639 2.181**

Dividend dummy 0.129583 1.425 0.120118 1.316 0.220812 1.207

Investment Growth 0.206406 0.924 0.224896 1.003 0.173421 0.651

Return on Assets -0.006544 -1.258 -0.006514 -1.251 -0.004096 -0.526

Percentage dummy 0.057713 1.033 0.057238 1.023 -0.019972 -0.199

Manufacturing dummy -0.008031 -0.184 -0.00836 -0.191 -0.021957 -0.446

Adj. R-squared 0.042 0.04 0.059

Table 9: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression on the dependent variable Tobin’s Q, which is calculated as (enterprise value/book value of total assets). FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Investment growth is proxied by capital expenditures/total sales and Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. An additional dummy is constructed for the industry, taking 0 if a firm does not operate in the manufacturing industry and 1 if a firm does operate in the manufacturing industry. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

Table 9 captures industry effects with the new dependent variable, a simple calculation of Tobin’s Q. The initial outcome, operating industry has no statistically significant effect on Tobin’s Q is robust to a different calculation of Tobin’s Q. However, many variables lost their significance, with the exception of FCD percentage and size, implying that the overall results were not robust.

(22)

22

Observations: 676 Pooled OLS Time fixed-effects Firm random-effects

FCD percentage -0.002962 -1.920* -0.003042 -1.968** -0.002925 -1.83*

Size (log assets) -0.96694 -5.143*** -0.0986 -5.228*** -0.097003 -4.93***

Debt ratio 0.029873 0.973 0.032768 1.063 0.309 0.982

Dividend dummy 0.124632 1.366 0.114682 1.253 0.126926 1.351

Investment Growth 0.213264 0.953 0.231799 1.032 0.208983 0.93 Return on Assets -0.006456 -1.236 -0.006402 -1.225 -0.006394 -1.202 Percentage dummy 0.062386 1.113 0.062012 1.105 0.059533 1.034 France dummy 0.047209 0.876 0.049831 0.923 0.047329 0.839 Germany dummy 0.052521 0.934 0.052984 0.941 0.053174 0,902 Adj. R-squared 0.042 0.041 0.039

Table 10: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression on the dependent variable Tobin’s Q, which is calculated as (enterprise value/book value of total assets). FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Investment growth is proxied by capital expenditures/total sales and Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. Additional dummies are constructed for country controls. France dummy equals 0 if the firm is not headquartered in France and 1 if it is, while Germany dummy equals 0 if the is not headquartered in Germany and 1 if it is. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

(23)

23

The final robustness test concerns the omission of non-significant variables. Only the main hypothesis will be tested, which means a redo of the regression equation of table 5, with the omission of the foreign sales variable and the investment growth variable, the results of which can be found in table 11.

FCD percentage -0.003462 -3.52*** -0.0036 -3.754*** 0.003107 2.214**

Size (log assets) -0.087113 -7.297*** -0.089727 -7.703*** -0.037357 -0.811

Debt ratio 0.040134 2.061** 0.045488 2.393** 0.038592 1.834*

Dividend dummy 0.229329 3.815*** 0.223 3.803*** 0.204216 3.496***

Return on Assets 0.05130 14.983*** 0.051663 15.358*** 0.012914 5.249***

Adj. R-squared 0.350 0.384 0.854

Table 11: Pooled OLS, time fixed effects and firm fixed-effect tests performed on the multivariate regression equation as outlined in equation (1). The dependent variable in this regression is the natural log of Tobin’s Q, measured as ((MV of equity + BV of debt) / BV of total assets). FCD percentage represents the notional underlying amount a firm possesses as a percentage of its total assets. Debt ratio is calculated as ((Long-term debt + short-term debt + current portion of long-term debt) / Common equity) *100. The dividend dummy takes on 0 if a firm did not pay dividends in a year and 1 if it did pay dividends in a year. Return on Assets is defined as the year-end net income of a firm divided by the book value of its assets. This particular regression includes a percentage dummy for FCD usage which takes on 0 if a firm reported FCD usage below five percent of total assets in a given year and takes on 1 if a firm reported FCD usage above five percent of total assets in a given year. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.

(24)

24 6. Conclusion

This study examined the usage of foreign currency derivatives in the Eurozone through a sample of 146 large nonfinancial firms headquartered within the Eurozone. The main goal of the research was to find out if hedging using foreign currency derivatives increases firm market value, as proxied by Tobin’s Q. It adds to existing literature by extending the geographical scope of the framework designed by Allayannis and Weston (2001), who focused on US firms, as did many of their successors; this study is one of the first of its kind to be conducted on Eurozone firms.

(25)

25 References

Allayannis, George, and James P. Weston. "The use of foreign currency derivatives and firm market value." Review of financial studies 14.1 (2001): 243-276.

Allayannis, George, Ugur Lel, and P. D. Miller. "The use of foreign currency derivatives, corporate governance, and firm value around the world." Journal of International

Economics (2011).

Allayannis, George, and Eli Ofek. "Exchange rate exposure, hedging, and the use of foreign currency derivatives." Journal of international money and finance 20.2 (2001): 273-296. Belghitar, Yacine, Ephraim Clark, and Salma Mefteh. "Foreign currency derivative use and shareholder value." International Review of Financial Analysis 29 (2013): 283-293.

Carter, David A., Daniel A. Rogers, and Betty J. Simkins. "Does hedging affect firm value? Evidence from the US airline industry." Financial management 35.1 (2006): 53-86.

Clark, Ephraim, and Amrit Judge. "Foreign currency derivatives versus foreign currency debt and the hedging premium." European Financial Management 15.3 (2009): 606-642.

Froot, Kenneth A., David S. Scharfstein, and Jeremy C. Stein. "Risk management: Coordinating corporate investment and financing policies." the Journal of Finance 48.5 (1993): 1629-1658.

Géczy, Christopher, Bernadette A. Minton, and Catherine Schrand. "Why firms use currency derivatives." the Journal of Finance 52.4 (1997): 1323-1354.

Hutson, Elaine, and Anthony O’Driscoll. "Firm-level exchange rate exposure in the Eurozone." International Business Review 19.5 (2010): 468-478.

Jin, Yanbo, and Philippe Jorion. "Does hedging increase firm value? Evidence from the gold mining industry." Working Paper, California State University-Northridge and University of California-Irvine. 2007.

(26)

26

Leland, Hayne E. "Agency costs, risk management, and capital structure." The Journal of

Finance 53.4 (1998): 1213-1243.

Mayers, David, and Clifford W. Smith Jr. "On the corporate demand for insurance."

Foundations of Insurance Economics. Springer Netherlands, 1982. 190-205.

Nguyen, Hoa, Robert Faff, and Andrew Marshall. "Exchange rate exposure, foreign currency derivatives and the introduction of the euro: French evidence." International Review of

Economics & Finance 16.4 (2007): 563-577.

Smith, Clifford W., and Rene M. Stulz. "The determinants of firms' hedging policies." Journal