“Exchange Rate Exposure at the Firm and Industry Level: Impact on Company Value and the Role of Hedging, Evidence from Greece”

(1)

MASTER THESIS:

MSc in International Business and Management – International Financial Management

(Double Joined Degree)

“Exchange Rate Exposure at the Firm and

Industry Level: Impact on Company Value and

the Role of Hedging,

Evidence from Greece”

AUTHOR: Mylonas Spyridon

SUPERVISOR: Prof. Dr. C.L.M. Niels Hermes

MAY 29

th

, 2009

university of groningen

faculty of economics and business

University of Uppsala

(2)

May 29, 2009 Author: Spyridon Mylonas Student-number: 1752324 e-mail:S.Mylonas@student.rug.nl Spyridon.Mylonas@gmail.com

Supervisor: Prof. Dr. C.L.M. Niels Hermes

University of Groningen Faculty of economics and business Msc International Business & Management Specialization International Financial Management

(3)

Acknowledgments

This work would not have been completed without help and support of several individuals. I would like to express my gratitude to everyone who has participated in any way to the completion of this paper. My cordial appreciation goes particularly to: Dr. Niels Hermes for providing me an opportunity to conduct my master’s research under him and for his guidance and valuable comments over the course of it. A number of anonymous individuals for helping me to get in contact with companies and gather the required data. My friends for making my stay in Uppsala and in Groningen easy and for sharing with me plenty of memorable moments. My professors who gave me the intellect to achieve my goal.

(4)

“Do not follow where the path might lead. Go instead where there is no path and leave a trail”

Ralph Waldo Emerson.

(5)

Abstract

Within international financial management, one of the most challenging issues is the identification and modeling of the relationship between exchange rate fluctuations and firm value. In this paper I study 81 firms listed in the Athens Stock Exchange over the period 2004-2007. Through the use of a questionnaire I collect unique data about hedging and the foreign operations of each examined firm. I construct firm specific exchange rates and I document significant levels (42%) of negative contemporaneous foreign exchange exposure. The evidence shows also that the impact of the lagged exchange rate changes is quite significant, indicating that the Greek market suffers by somewhat lagged response. The analysis at the industry level reveals that the aggregation of firms into portfolios might underestimate the exposure significance. Moreover, the openness of the economy, and the internationalization of the firm’s operations increase the exposure, while the opposite holds for the use of operating hedging techniques. Financial hedging tools do not seem to affect the relationship. Furthermore, results corroborate the assumption that even firms with no foreign operations might be indirectly exposed to foreign exchange variation, thus, even though firm-specific exchange rates yield more significant and reliable results, trade-weighted indices must be always used as a complement. Finally, accounting for multicollinearity and endogeneity is of crucial importance in measuring foreign exchange exposure.

(6)

Introduction

In the early 1970s the Bretton Woods fixed-parity system collapsed bringing about instability in exchange rates and insecurity in the international financial markets. This breakdown and the subsequent dominance of floating rates systems led to increasing foreign exchange rates volatility. At the same time the rapid expansion of the international trade increased the need to analyze, understand, and if necessary, manage exchange rate exposure.

Financial and economic theory suggests that exchange rate movements affect in various ways both the cash flows of a firm’s foreign operations and the discount rate employed to price these cash flows 1. Since the value of the firm denotes the current value of its future cash flows it becomes evident that changes in the exchange rates could play a role in the determination of company value. Obviously, the importance of managing efficiently foreign exchange exposure becomes of crucial importance for companies, especially multinationals.

Starting from the work of Shapiro (1974) who tried to formulate the conceptual framework of the exchange rate exposure and firm value relationship, and the papers by Adler and Dumas (1984) and Jorion (1990) that developed the empirical examination approach, a substantial number of related studies have been conducted. However, the empirical results do not meet adequately the theoretical expectation and appear rather conflicting and mixed.

In an effort to explain and deal with these inconsistencies, Bartov and Bodnar (1994), Bodnar and Wong (2003), and Dominguez and Tesar (2001a), among else, have extensively studied the specifications of the empirical methodologies. A big portion of the criticism refers to the choice of the exchange rate. The majority of the researchers use trade-weighted indices provided by financial institutions. However, the utilization of firm-specific indices is highly suggested, despite the difficulty in constructing them. The focus on the US market of the early studies is also a potential reason that could explain the failure of the empirical studies to meet the researchers’ expectations. US are a rather

(8)

closed economy and this might have an impact on the results. Indeed, some evidence from researchers such as De Jong et al (2002) who studies Netherland, and Nydahl (2001) who studies Sweden which are smaller but more open economies confirm this assumption. The fact that the firms are aware of their risk and thus they are hedging their positions is a much debated issue. Allayanis and Ofek (2001) provide related evidence showing that the use of derivatives reduces exposure. Moreover, Gao (2000) and Doukas et al (2003), among others, have shown that because movements in exchange rates and overall market behavior might be attributed to common macroeconomic factors, which in turn, influence the firm value as well, the problems of multicollinearity and endogeneity have serious impact on the results. Finally, the inclusion of control variables in the model of Jorion (1990) has revealed that such factors have to be taken into consideration. In general, the examined relationship appears to be exceptionally multifaceted and complicated, being dependent on an immense number of macro- and micro-economic parameters that are many times interrelated2.

All these questions and inconsistencies motivated me to further investigate the topic. In this paper I aim to shed more light on the issue and to add to the existing literature by identifying:

(1) The impact of foreign exchange rate fluctuations on firm value, (2) Through which channels these fluctuation affect firm value, and

(3) Which is the role of hedging in determining the exchange rate and firm value relationship?

I apply my research on Greek listed in the Athens Stock Exchange firms for a period which spans from 2004 to 2007. The case of the Greek market is interesting and challenging at the same time. Since the middle 90s Greek companies were not internationally oriented to a large extent. However, during the last decade Greek companies began expanding internationally. According to the National Bank of Greece, the major trading partners of Greece (without taking EU countries into account) are the USA, and the neighboring economies, with Romania, Bulgaria, Turkey, Cyprus, as well

(9)

as FYROM being the most important3. This makes the case even more interesting since the currencies of these countries have been rather volatile. Moreover, the Greek market is quite small but relatively open. Data from the WDI database of World Bank referring to countries’ import and export activities in percentage of GDP suggest that for 2003, for example, the respective numbers for Greece are 29% and 20%, for USA 14% and 10%, for Japan 10% and 12%, for Sweden 36% and 43%, and for Netherlands 56% and 63%. In addition, the Greek market, to the best of my knowledge, has never been studied before in terms of exchange rate issues.

Examining a sample of 81 Greek companies from 13 industries, I aim to deal with the key issues mentioned above. I employ a rather unique methodology by integrating and adjusting important elements of previous approaches. That is, following Nydahl (2001) and mainly De Jong et al (2002), I use questionnaires and I gather unique data about hedging and foreign activities of each firm. By combining some methodological suggestions by Gao (2000), Nydahl (1999), and Doukas et al (2003), I employ a 3-stage analysis methodology which allows diminishing multicollinearity and endogeneity problems. This methodology, together with the decomposition of the exposure coefficient into its (potential) determinants allows dealing also with the time-varying character of the foreign exchange exposure. Moreover, by adding in the model the possible determinants of exposure, I can identify potential channels, e.g. foreign sales and/or expenses, through which exchange rate fluctuations affect firm value. In this way, I can discover also the role that hedging plays in identifying exposure. I also gather firm-specific information, such assets, liabilities etc, on monthly and/or quarterly basis so as to incorporate control variables for the model. Finally, I conduct my research by estimating the most common used models. This approach, which has not been followed before, will allow me to compare empirical results derived from the same dataset and draw conclusions about the efficiency of different methodologies in estimating exposures.

The remainder of the paper is organized as follows: In the first part I develop the theoretical framework upon which the empirical investigation will be based. The theoretical foundations of foreign exposure and its relationship with firm value are

(10)

(11)

A. THEORETICAL BACKGROUND

1. Foreign Exchange Exposure

1.1 Theoretical Considerations

Adler and Dumas (1984) defined foreign exchange exposure as the sensitivity of changes in the real domestic-currency value of assets or liabilities to changes in exchange rates. There are many different theoretical approaches to portray how exchange rate fluctuations result in changes in the firm value. The main argument suggests that exposure affects firm value via its impact on the firm cash flows. Yet, the challenge is to identify through which channels this impact is realized.

To understand the nature of the relationship between exchange rates and firm value variability, literature4 has decomposed foreign exchange exposure into three types: transaction, translation, and economic/competitive exposures. Transaction exposure refers to the contractual transactions (i.e. accounts receivable, accounts payable, repatriated dividends) of a firm in foreign currencies. An example: if an exporting company denominates its exports in a foreign currency, a 10% decline in that currency’s value will result in a 10% reduced value of the receivables. The impact of the exchange rate on firm value is obvious. Likewise, the value of assets and liabilities of foreign subsidiaries, denominated in foreign currency, can influence the corporation’s overall value when translated, remitted and integrated to the consolidated financial statements in the parent country/currency. This accounting-based effect is marked as translation exposure. Finally, economic exposure reflects, like transaction exposure, the possibility a firm’s present value of operating cash flows to be affected by changes in exchange rates. However, it is a broader notion which actually includes transaction exposure. The basic idea behind is that as exchange rate variations affect the relative prices of goods sold and of raw materials in different countries/markets of the world, not only the transactional accounts of a firm are affected. The purchasing power of consumers and the opportunities

(12)

for suppliers and competitors change as well. Such changes influence, evidently, a firm’s competitive position, export opportunities5, and thus, indirectly its future cash flaws and value. Hence, even a firm with no foreign operations can be exposed to foreign exchange exposure via changes in the overall market conditions. Evidence that proves this assumption is provided by early study of Hodder (1982), and confirmed by subsequent papers such as Doukas et al (2003)6.

As mentioned above, the main and most important challenge for the researchers is to identify through which specific channels foreign exchange exposure impacts on company value. Shapiro (1975) is the first who attempted to model the theoretical effects of the exchange rate movements on firm value. He built a two-country model and, in essence, focused on the direction of the foreign operations of an internationally oriented company. The theory derived from his work and the subsequent literature, postulates that while exporting firms should benefit when a depreciation of the local currency occurs due to the fact that the products become more affordable to foreign consumers, firms whose production depends on imported raw and other intermediate materials/products may see their profit margins shrinking due to the increasing production costs.

A stream of the literature studies how the degree of the internationalization of a firm, industry, or a market is connected to the level of the estimated exposure. Elaborating on the assumption of Shapiro (1975), Gao (2000), Allayanis and Ihrig (2001), Marston (2001), Doukas et al (2003), and Fraser and Pantzalis (2004), and others, have shown that there is a strong connection between measurements of international activities and the degree of the exposure of a firm. In general, the higher the level of cash flows, e.g. sales and expenses, denominated in foreign currencies the higher the level of the exposure. The same holds for the operational network of a company; a positive and significant relationship between a firm’s exposure and its proportion of foreign to total assets, its number of foreign subsidiaries, its number of countries where it operates etc. has been documented. Likewise, Bodnar and Gentry (1993) and De Jong et al (2002) hypothesize and confirm that not only the firm level but also the total economy’s degree of openness constitutes a factor of increasing/decreasing exposure. That is, firms that

5

Booth and Rotenberg, 1990

(13)

belong to open economies, i.e. economies engaging in intense exporting/importing activities, they face higher possibilities to be exposed to foreign exchange rate risk.

Even though there is still a lot of discussion related to the relevance of managing foreign exchange risk, Nydahl (2001), De Jong et al (2002) and Fraser and Pantzalis (2004) provided evidence to support the assumption that hedging activity is a means of diminishing exposure. Companies that recognize and act so as to insulate themselves against exposure to foreign exchange risk, they can prevent, to some extent, their firm’s value from the effects of exchange rate movements.

Among others, Doukas et al (2003), show that to the degree that the larger a company is the higher the possibility is to engage into foreign operations, it is expected the firm size to be a mean through which exchange rate exposure affects company value. On the other hand, Dominguez and Tesar (2001) found that the size of a company is connected to higher degree of hedging instruments usage, and thus, lower exposure.

Overall, size, industry affiliation, hedging activities, degree of internationalization, and other factors7, constitute channels through which firms/industries can be influenced by exchange rates fluctuations. However, as Dominguez and Tesar (2001) state, the precise linkage among (some of) those factors and the direction of the exposure is still quite unclear. Further empirical research on the issue can provide more insights related to the theoretical assumptions surrounding the foreign exposure and firm value relationship.

1.2 Empirical Evidence

The empirical investigation has met the theoretical expectations with limited -at least- success. Researchers, in general, have failed to document significant levels of association between exchange rate fluctuations and firm value.

The basic methodological approach followed in order to estimate currency exposure is a linear regression of stock returns on an exchange rate variable. This

(14)

approach traces back to the work of Adler and Dumas (1984). Their model suggests the use of stock returns as the best representation of the current value of the future cash flows of the company8and the change of the exchange rate as the variable to be regressed on. The coefficient of the regression demonstrates a firm’s stock returns elasticity to exchange rate changes. Despite the fact that the exchange rate exposure is by definition the amount of the firm’s value which is at risk, still this elasticity is what the literature has named as foreign exchange exposure:

R

it

= α

i

+ φ

i

θ

t

+ ε

it

where

R

it denotes the total return of firm

i

in period

t

,

θ

t the exchange rate change in

period

t

,

ε

it represents the white noise error term, while

φ

i describes the sensitivity of

firm

i

’s stock returns to changes in exchange rates.

As Adler, Dumas and Simon (1986) in a later paper highlight, the above model estimates the part of a firm’s stock returns variation that is correlated to currency movements, and thus, referred to as total exposure. However, other macroeconomic conditions might concurrently covary with exchange rates affecting stock returns. Hence, the above simple model results in biased results in terms of the proportion of the stock movements that are attributable to exchange rates. Taking these disadvantages into account, Jorion (1990) augments the above model including a market variable so as to isolate the effects of the economy’s behavior:

R

it

= α

i

+ β

i

R

mt

+ γ

i

θ

t

+ ε

it

where

R

mt designates the overall stock market return, measured by some market index.

This model measures exposure as the residual between the firm’s total exposure and the market’s total exposure and served as a roadmap for numerous of following researchers9.

Using his augmented model, Jorion (1990) examined 287 US MNCs for the years 1971-1987. Despite selecting this sample under the condition that the firms have

8

If the market is efficient the stock prices represent the evaluation of the investors of the value of the company (Adler and Dumas, 1984).

(15)

significant international operations, only 5.2% of the sample exhibits a significant level of exposure. The sign of the exposure coefficient is negative implying that a depreciation of the USD causes losses to the US firms that are significantly affected. Amihud (1994) examines a sample of 32 leading exporter US companies between 1982 and 1988 and also fails to find any significant exposure, though he finds higher levels of exposure when using lagged exchange rate values. Bartov and Bodnar (1994) study contemporaneous and lagged effects of exchange rate exposure for 208 US firms and like Jorion (1990) and Amihud (1994), they found significant results for the lagged values but not for the contemporaneous. Choi and Prasad (1995) examine a dataset comprising 409 MNCs over a period of 12 years both in firm and industry level. They document just 14% and 10% of the companies to be exposed respectively at the 10% level. On the contrary to Jorion (1990, 1991) they identified positive exposure coefficient.

While the first wave of research was mainly focusing on US companies, after mid 90s the focus shifted also to other, non-US markets. Since US is a quite closed economy, it is expected that smaller and more open economies might be more sensitive to exchange rate changes10. However, despite more significant, the results still do not meet adequately the theoretical expectations. Bodnar and Gentry (1993) conduct the first multi-country study examining US, Canadian, and Japanese companies, and find that stock returns are rather insensitive to exchange rate changes reporting 23%, 21%, and 25% of the companies respectively being significantly exposed. They hypothesize that their failure can be attributed partly to the use of hedging. He and Ng (1998) examine 171 Japanese multinationals for the period 1979-1993, and identified a percentage of exposure equal to 26% while they do not document any significance when they use lagged values. Nydahl (1999) studies 47 Swedish stock returns between 1992 and 1997. When he employs the Jorion (1990) model, he documents a significant level of exposure for the 26% of the sample firms in the 10% significance level. The work of Doidge, Griffin, and Williamson (2002) is the most comprehensive to date. They form a dataset of more than 17000 stocks from 18 markets, but they find that exchange rate changes impact on firm value mainly during periods of exchange rate shocks. Particularly, they document that firms with high international sales outperform those with no international sales only during periods of

(16)

large currency depreciations and the opposite holds during periods of large currency appreciations. Priestley and Ødegaard (2005) examine 7 Norwegian industries against USD and ECU and find that none of them are significantly exposed. Kurt and Yücel (2003) study 152 Turkish companies and their results reveal that only a 12% of the examined firms are significantly exposed.

A number of researchers investigate the issue using industry portfolio returns. That is, the sample firms are divided and aggregated into industries and then the regression equations are estimated for each industry rather than for each individual firm. Khoo (1994) however suggests that this analysis can cause bias if the firms that constitute the portfolio are not homogeneous in terms of exposure sign, size, foreign countries where they operate, and so on. In particular, it is expected that the exposure coefficients will be underestimated, because exchange rate fluctuations may result in offsetting affects on the importing and exporting firms within the industry. His results confirm these expectations and are in line with those of Choi and Prasad (1995), Muller and Verschoor (2004a) and Allayanis (1996) who corroborate that the level of statistical significance of the currency exposure is lower when calculated at industry comparing to the firm level.

(17)

subsidiaries. Corroborating results were presented by Fraser and Pantzalis (2003) in a paper where they build firm-specific index following the same as Ihrig’s (2001) approach. Nevertheless, even though the results are somewhat more significant when they use firm-specific indices, the sensitivity of the stock returns to the exchange rate movements remains low (8,5% to 5%). De Jong et al (2002) study the Dutch market. They employ firm-specific bilateral exchange rates based on information gathered via questionnaires, and they found that 50% of the 47 sampled firms are significantly exposed. Nevertheless, although the results do not change much when they use a trade-weighted exchange rate index, the firms that are found to be exposed are different. Thus, they conclude that the methodologies are complementary.

A number of researchers had focused on issues related to problems of

multicollinearity and/or endogeneity between market and exchange rate risk factors, as

same factors may drive both market returns and exchange rate fluctuations11. The majority of these studies reports higher levels of significant foreign exchange exposures comparing to previous research that does not account for the aforementioned problems. Choi and Prasad (1995) orthogonalize the exchange risk factor by running a first stage regression of exchange rate changes on the market risk factor. The residual from this regression expresses the part of the exchange rate variations that are not explained by the market conditions. Then they use this term instead of the actual exchange rate changes in the Jorion’s regression model12. Gao (2000) extends this approach. In a first step he runs a side regression of exchange rate changes on their determinants, i.e. interest rate, money supply, level of industrial production, net exports, rate of inflation; then he uses the residuals from this orthogonalization procedure to a market model. However, in the latter equation, he also replaces market risk with macroeconomic variables that might influence simultaneously exchange rate variations and market returns leading to overestimated exposure values: news to unemployment rate, to producer price index, to the money supply, to an energy price index, to an aggregate wage index, and to an industry-specific wage index. He reports statistically significant results, despite the fact that none of the macroeconomic variables individually exhibit significant impact on the equation. Finally,

11

Muller and Verschoor, 2005.

(18)

Doukas et al (1999), in their three stages analysis of 1079 Japanese MNCs, choose to orthogonalize in side regressions both exchange rate variations and market returns on the same macroeconomic variables that are expected to be their common determinants: the industrial production, the unexpected inflation, the term structure, the money supply, the US-Japan interest rate spread, the trade balance series, and the Fama and French (1996) value minus growth and small minus large capitalization return spreads. Then they run a regression of the company returns on the six macroeconomic variables and the unanticipated risk factors (i.e. the residuals from the first two equations) they previously estimated. They report that the Japanese MNCs are significantly influenced by contemporaneous foreign exchange rate exposure.

A number of researchers attribute partially the failure of the related literature to document significant relationship between exchange rate fluctuations and stock returns to the fact that investors’ systematic mispricing of the exchange rate effects. The fact that the exposure effects are a result of a wide range of factors interrelated in a complex way, as well as the fact that company information associated with such factors are disclosed with some delay, may lead exchange rate effects to be reflected in stock prices with a time lag. Moreover, is difficult for investors to distinguish between temporary and permanent exchange rate changes, and this enhances the problem of mispricing. Researchers dealt with the problem by using lagged exchange rate values. Their results, reveal some statistical significance comparing to the use of contemporaneous exchange rate values, but in general are rather inconsistent. Jorion (1990), Amihud (1994), Bartov and Bodnar (1994), study US companies and they provide supporting evidence. On the contrary, when Nydahl (1999) adds lagged exchange rate changes in his model the percentage of the significantly exposed Swedish firms drops from 26% to 18%. Likewise, He and Ng (1998) investigating the Japanese market find that less than 4% of their sample is exposed when lagged values are used (26% were found exposed in the contemporaneous exchange rate changes).

A critical point of discussion within the exposure literature concerns the

(19)

firm’s operating structure change, and so must exposure13. Indeed, the majority of the papers that consider sub-periods in their analysis do prove that the exposure is unstable over time14. An interesting question, as Levi (1994) and Bartram and Bodnar (2005) note, is whether this variation is driven by economic factors or estimation error. In this context, Allayanis (1997) models the exchange rate exposure of 137 4-digit SIC manufacturing US companies, as a function of economic factors. He documents that even though the explanatory power of the time variation in the exposure is not high enough to explain the insignificant impact of exchange rate fluctuation on the stock returns, the exposure of the companies is significantly correlated to the export and import ratios of the industry. Likewise, Williamson (2001) studies the automotive industries of US and Japan and relates exposure time-variation to the changes in the competitive structure within the sector (i.e. market shares).

Yet, the low significance of the empirically estimated exchange rate exposures has been largely attributed to the use of hedging instruments by the companies. As Levi (1994) suggests, since hedging tools are designed in order to diminish several types of risk, the efficient implementation of hedging strategies can have direct negative impact on a firm’s exposure. However, even though a large number of authors have recognized hedging as a factor with important impact on the exchange rate and company value relationship, still only a few have tried to incorporate it in their models15. Nydahl (1999) was among the first who used primary data to control for hedging. A questionnaire-based survey by the Sveriges Riksbank provides him with data on the use of financial and operating hedging instruments by Swedish firms. His results suggest that the use of currency derivatives diminishes the estimated exposure. Proxying operational hedging with the operating network of firms Fraser and Pantzalis (2001) find that US multinationals with wide operational network have lower exchange rate exposure. De Jong et al (2002) test the impact of the on- and off-balance sheet hedging on an individual firms’ exposure. They study Dutch multinationals and they collect hedging data with the use of questionnaires and their findings suggest that the use of derivatives has no

13_{Bartov and Bodnar, 1994.} 14

Amihud (1994), Choi and Prasad (1995), Glaum et al (2000), He and Ng (1998), and Doukas et al (2003).

(20)

significant impact while the use of the on-balance sheet techniques reduce significantly the firm exposure. Allayanis and Ofek (2001) provide evidence that derivatives usage have a significant negative relationship with exposure for a sample of 378 US multinationals. Muller and Verschoor (2004b) report that the use of financial derivatives reduces exposure but the reduction of the coefficient is only weak. They justify their findings by the fact that companies hedge their exposure partially and selectively.

The above literature review refers to the most debated parameters that seem to determine the relationship between exchange rate changes and firm value16, and reveals the complexity of this relationship. Evidently, the more of these issues an empirical model can efficiently address the more reliable the results will be and a clearer picture of the nature of the examined relationship could be drawn. This paper constitutes such an effort. The following sections of the paper describe the employed methodology which in essence deals with all of the above mentioned problems: a non-US, small and open market is studied (i.e. Greece), both firm and industry level analysis are conducted, firm-specific exchange rates are constructed, a three stages analysis is applied to control for multicollinearity and endogeneity, the time-varying nature of the foreign exposure and the hypothesis of the lagged market response to exchange rate changes are taken into account, and unique data on hedging activities are collected.

16

(21)

B. EMPIRICAL RESEARCH

1. Methodology

The discussion of the empirical specifications of the previous literature indicates the difficulty that researchers face when aiming to model empirically the relationship between exchange rate changes and the company value. This relationship is driven by numerous parameters, which are in many ways interrelated. In order to document the relationship thus researchers have to implement sophisticated statistical models in an effort to accommodate as many as possible of the factors that might impact on the results. In this study I try to address many of the parameters that affect the exchange rate variations and firm value relationship by forming an econometric model that incorporates elements from several previous attempts. At the same time, by applying different methodologies on my sample, I aim to compare the efficiency of these approaches.

The statistical model used in this paper is mostly based on the papers by Doukas et al (1999, 2003), Gao (2000), and Nydahl (1999). The authors of these papers address many of the problems that the previous research has identified, e.g. autocorrelation, multicollinearity, time-variation, endogeneity. At the same time, following De Jong et al (2002) I gathered primary data concerning foreign operating and hedging activities by sending questionnaires. These questionnaires also helped me to construct firm-specific exchange rates, as it will be explained later. The choice of the control and instrumental variables is guided by several previous papers since the literature has not been conclusive in terms of the set of variables that determine the examined relationship.

1.1. Model Development

The starting point for the discussion of the development of the empirical model is the augmented regression equation of Jorion (1990):

(22)

where

R

it is the rate of return of the ith firm’s stock;

R

Mt is the rate of return of the

market index;

FX

it is the rate of change of the selected exchange rate; and

ε

it is the error

term of the regression estimation. We are interested in the

γ

it coefficient which is the

sensitivity of the firm’s stock returns to the changes of the exchange rate, i.e. the firm’s exposure, in excess of the overall market behavior against exchange rate fluctuations. If the exchange rate is quoted as EUR per one unit of foreign currency, then a positive and significant

γ

means that a depreciation of the EUR corresponds to an increase in the value of the firm

i

17

. However, the interpretation of the coefficient leads to a conceptual point of caution. In this model, one insignificant

γ

does not mean that there is no exchange rate exposure; rather the market in total might have fully reacted to the exchange rate change and thus, the market factor has fully capture the exchange rate effect.

Gao (2000) and Doukas et al (1999, 2003) in their discussion on the equation 1 model, note that the relationship between endogenous variables such as exchange rates and the market returns may just disclose the contemporaneous effects of same macroeconomic factors on both of these variables. That is, the changes in exchange rate and stock market values might be partially determined by common variables. Insofar as that market participants form their expectations based on publicly available information, exchange rates will adapt to economic conditions and the subsequent exchange risk perceptions. Currency exposure investigation should thus rely on unexpected currency and market movements that are orthogonal to each other 18 and orthogonal to the factors that shape investors’ expectations, e.g. macroeconomic conditions.

In this study, following Doukas et al (1999, 2003), I deal with the aforementioned multicollinearity and endogeneity problems by applying a three-stage analysis. The model comprises the estimation of two side-regressions in order to orthogonalize19 the exchange rate and the market factors on some instrumental variables. The unexplained

17

Glaum, Brunner, and Holger, 2000. 18_{Doukas, Hall, and Lang, 2003.}

(23)

terms of these two stages are then used in the main equation instead of the actual values of the independent variables.

FX

jt

= φ

0

+

∑

φ

1j

IV

FXjt-1

+ ε

FXjt

(2)

RMKT

t

= θ

0

+

∑

θ

1i

IV

MKTt-1

+ θ

2

ε

FXjt

+ ε

MKTt

(3)

where

FX

jt is the rate of change of the exchange rates;

RMKT

t is the rate of return of the

market index;

IV

FXjt-1 represents a set of instrumental variables that determine exchange

rate changes;

ε

FXjt describes the unanticipated exchange rate changes;

IV

MKTt-1represents

a set of instrumental variables that determine the market behavior; and

ε

MKTt is the error term of the second equation reflecting the unexplained market movements.

So, the equation 1 can now be reformulated as follows:

R

it

= α

i

+ β

it

ε

MKTt

+ γ

it

ε

FXit

+

∑

δ

it

CV

it

+ ε

it

(4)

where

ε

MKTt and

ε

FXit are the residual terms of the equations 2 and 3, while

CV

it is a set

of control variables that are expected to influence the individual firms’ stock returns. The examination up to this point tests the contemporaneous relationship between exchange rate changes and company value. Research evidence, however, has shown that the market is not perfectly efficient and investors need some time to assess the information they get before they react. As Doukas et al (2003) summarize, the systematic mispricing of the exchange rate risk of the stock returns on the part of the investors might be attributed to the relatively small history of the floating exchange rate regime, the incomplete financial information disclosure, the time they need to analyze the complex market relationships. Thus, Amihud (1994), Bartov and Bodnar (1994) among others, found empirical support via the use of exchange rate lagged values. The regression model changes in order to check the lagged response hypothesis in the following way:

R

it

= α

i

+ β

it

ε

MKTt

+ γ

it

ε

FXit

+ γ*

it

ε

FXit-1

+

∑

δ

it

CV

it

+ µ

it

(5)

(24)

One of the objectives of the paper is to further examine the determinants of the firm exposure to exchange rate changes in order to better understand the nature of the examined relationship. This part of the analysis aims (1) to reveal evidence about channels through which exchange rate fluctuations affect firm value, (2) to partially deal with the time-varying nature of the exposure, and (3) to reveal the impact that hedging has on the estimated exposures. The last point is of utmost importance, since Bartov and Bodnar (1994), Chow and Chen (1998) and a large number of other researchers attribute the low level of documented significant exposure to the fact that companies actively hedge their foreign exchange risk.

The standard methodology used to identify the determinants of foreign exchange exposure is a two step regression analysis. The first step is the previous equation that aims to measure exposure. Then the estimated exposures are regressed on the set of the variables that are expected to determine the extent of the exposure, hedging being one of them:

γ

i

= χ

i

+

∑

λ

i

F

i

+ η

it

(6)

where

γ

it is the exposure estimation as obtained from the equation 4 and

F

i is a set of

factors that are expected to have an influence on the extent of the firm exchange rate exposure, i.e. the percentage of sales and expenses that are denominated in foreign currency, the financial and operating hedging activity of each firm, and the size measured as the logarithm of total assets. However, Jorion (1990) argues that the two-step estimation with OLS leads to dependence between

γ

iacross the equations, because the coefficients are estimated over the same sample period. As a result the estimation of equation (6) violates the OLS assumption that error terms are not correlated. Thus, I adopt the one step approach of Nydahl (1999) and Gao (2000) and combining equations 4 and 6 I estimate the following model:

(25)

factors change constantly over time, if there are significant determinants of exposure, then the model efficiently controls for the time-variability of exposure.

1.2. Variables Rationalization

As Muller and Verschoor (2005) point out, it seems that the relationship between exchange rates and company value is so complicated and is driven by numerous parameters that it is unlikely that researchers will manage to account for all the relevant issues. Therefore, the choice of control variables is bound to be somewhat arbitrary and based upon the previous literature and the intuition and stance of the author.

1.2.1. Exchange Rates

The failure of the previous literature to document significant levels of foreign exchange exposure has been attributed to a great extent to the choice of the exchange rate factor. The majority of the researchers up till now employ either trade-weighted exchanges rate indices as constructed by various financial organizations or single currency exchange rates. The utilization of such exchange rate variables assumes that all the companies of the market bear uniform operational characteristics. Nevertheless, such an assumption can clearly bias the results. In contrast, one of the contributions of this paper entails the construction of firm specific exchange rate indices. However, I also use trade-weighted exchange rate index as given by the Bank of International Settlements in order to compare the results and draw conclusions about the efficiency of each methodology.

(26)

weighted for the construction of the firm-specific index. Note that the exchange rate index that applies to each firm can differ from year to year including the currencies as these are indicated by the questionnaire responses.

Apart from the firm-specific indices, I construct also industry indices for the industry level analysis. These comprise all the currencies that the firms of the respective industry have reported and are weighted according to how many times they have been indicated as exposure factors. For example, for the Chemical industry and for the year 2004, 5 out of the 6 companies have reported potential exposure to USD; so the USD weights 5/6 in the index. Similarly to the firm-specific, each industry’s exchange rate can include different currencies for each year.

Finally, I also use the trade-weighted exchange rate index of the Bank of International Settlements. The BIS constructs exchange rate indices for 58 countries. The real exchange rates indices are calculated as geometric weighted averages of bilateral exchange rates adjusted by relative consumer prices. The weighting occurs on the basis of the trading activities of each country, and the indices are reviewed and released mid-month on a mid-monthly basis. It is calculated as EURO per unit of foreign currency, thus, an increase in the index represents a devaluation of the EURO.

1.2.2. Orthogonalization of the Exchange Rate factor

The aim of equation 2 is to estimate the unanticipated exchange rate changes. There is a vast body of literature that aims to spot the dynamics of the exchange rate determination, but there is no clear-cut deduction. In principle, exchange rates change in response to the changes in the demand and supply of a currency. Hence, factors that may alter these two aspects can be proxied and used as macroeconomic instrumental variables in equation 2. Note that the EURO currency is common in 16 European countries. Thus, the macroeconomic variables that are going to be employed on the part of the EURO currency refer to EU totals.

(27)

level both determine the purchasing power of the consumers in a country, the unemployment growth rate (UMPL) and the wages level growth rate (WL) are another two variables to be used. The trade of balance changes is another indicator of exchange rate changes as is the national productivity level. Thus, the logarithmic difference between exports and imports (trade of balance - TB) and the industrial production growth rate (IP) are the next two instrumental variables in the equation. Since central Banks have the ability to respond or alter the demand and supply of their respective national currencies via the supply of money, the monthly change of money supply (MS) is employed as an additional variable. Finally the rate of change of the crude oil prices (OIL) is the last variable to be used. With the all growing dependence of the economies on energy consumption and the prevalence of oil as main energy source to fuel production, oil prices are considered as variable that should be included in any market relationships’ investigation20,21. An important point to highlight is that the use of firm-specific exchange rates allows me to use a unique approach. Because the exchange rate between two currencies changes in response to changes of factors in either of the two respective countries I consider macro variables from both countries in the model. This comes in contrast to Gao (2000) and Doukas et al (2003) who employ only home country macro variables; since they use only trade-weighted exchange rate indices which include currencies of a large number of countries, it would be rather impossible for them to follow the same as mine approach. Nonetheless, in order to orthogonalize the BIS index, which includes currencies from 52 countries, I also use only the macroeconomic variables of the home country (i.e. the EU area).

1.2.3. Orthogonalization of the Market factor

Relating theoretical reasoning and its empirical investigation have revealed the macroeconomic determinants of the exchange rates to be in a great extent the same as those of the aggregate market return. In essence, capital asset pricing models considering

20

Chen, Roll, and Ross, 1986.

(28)

both the market factor and the exchange rates to be endogenous explanatory variables of the individual firm stock returns. As such, common determinants affect the movement of both of these factors. Thus, the orthogonalization procedure of the market factor is conducted by using the same as previous macroeconomic variables as well as the changes in exchange rates as represented by the residuals of the first orthogonalization equation (

ε

FX). Note here, that in contrast to the previous stage, the values of the macro variables used in this equation refer only to the numbers of the Greek economy.

1.2.4. The Actual Equation

(29)

to stock price, since bigger companies are related to low-risk investments (size effect). Finally, the leverage of a firm is calculated as debt to equity, and expresses the extent to which a firm is financing its operations with external credit. It is expected that the relationship between leverage and stock price is negative, since the more leveraged a company is the higher are the chances for the firm to be unable to pay its due.

1.2.5. Determinants of Exposure

With this equation I aim to identify the forces that affect exchange rate exposure of individual firms. The choice of the variables is based on previous literature and basically refers to firm-specific factors related to the operating network of the firms, the extent of the international operations, and the degree of the use of hedging strategies. Namely: the foreign to total sales ratio (FSAL), the foreign to total expenses ratio (FEXP) are both indicators of the foreign operations of a company and they capture the percentage of the firm’s cash flow as that are denominated into foreign currency. They are expected to have positive relationship to exposure since the higher the proportion of the firm cash flows that are generated abroad the higher should be the exposure to the respective exchange rate fluctuations. The size calculated as the natural logarithm of the firm’s total assets (ASIZE) and is expected to have a positive impact on exposure. This is because large organizations are expected to have bigger operational network including foreign countries. Thus, it is expected to be more exposed to exchange rate movements. Finally, the operating hedging variable (OPHEDG), and the financial hedging variable FINHEDG) are of utmost interest for this paper. The former refers to the structuring of the company’s operations in such a way that the mix of production inputs and outputs will minimize the net exposure to exchange rate changes22. The latter refers to financial hedging tools comprising OTC and market-traded derivatives, e.g. forward, future, option, swap, and other contracts. Through questionnaires, I gather related information from primary sources and I aim to identify if hedging indeed drives previous research findings.

22

(30)

It is expected that both variables should have a negative sign, since the higher the hedging activity of a firm the lower the degree of its exposure should be.

2. Data

2.1. Selection and Sources

To estimate the above mentioned models, I gather monthly data over the period 2004-2007, for non-financial companies that are listed in the Athens Stock Exchange for at least 2 full years. The exclusion of the financial companies is justified by the fact that this type of firms has different operations and cash flow structure and their inclusion might bias the sample23. The 4 years horizon has been chosen for several reasons: (1) it is long enough to capture exchange rate exposures, (2) it is not too long, and thus, alleviates the effects of the time-varying nature of the exposure, (3) it is a time horizon that will not discourage the recipients of the questionnaires to reply. At the same time, the monthly observations horizon choice it is mostly suggested by the literature since: (1) it is long enough so as not to be affected by daily volatility noise and long enough to capture the long swings that currencies experience, thus, revealing the more fundamental relationship between exchnage rate fluctuations and company value24, (2) it is short enough so as not to be affected by the time-varying nature of the exposure, and (3) it is a horizon that allows to gather also control variables data (it would be impossible to find available macro- and micro-economic data on weekly, for example, basis).

By the end of 2007 there were 257 non-financial firms listed in ASE and were classified into 15 different industries. During October 2008 I mailed a questionnaire to the Financial Departments25 of all of these companies. The questionnaire aimed to provide information about the foreign operations of the company, the hedging activity, and the currencies that the company might be exposed to. A 2nd reminder was sent during February 2009. A total number of 51 questionnaires were returned. During March 2009, I

23

Dufey and Giddy, 1992.

24_{Muller and Verschoor (2006), Dominguez and Tesar (2001).}

(31)

have contacted by phone a number of companies in an effort to collect more filled questionnaires and extend the sample. The final sample that was formed numbers 81 companies from 13 industries.

To build up the monthly firm and industry exchange rate indices I use real exchange rates collected from the DataStream database. In total, 26 different currencies have been reported in the questionnaires as potentially having impact on firm value. I also use the EURO trade-weighted exchange rate index for Greece as provided by the Bank for International Settlements. The BIS indices include 51 different currencies weighted according to each country’s trading facilities. I collect monthly stock returns from Wednesday to Wednesday so as to avoid end-of-week effects, from the DataStream database. The market returns are those of the FTSE/ASE 140 and the returns are taken from the DataStream26.

To gather monthly macroeconomic data for the 26 countries, Greece and the EU area in total, I mainly used the IFS database provided by IMF. However, a big number of needed data were not available or were available in quarterly basis. To complement the gaps I used a number of other well-known and reliable web-online sources such as the OECD Statistics Portal, EBRD Statistics, Eurostat, CIA World Factbook, Trading Economics Statistics, World Bank and European Bank Statistics, EconStats, and many official national statistical organizations. I find at least quarterly data for all the variables and for all the countries. Particularly, from the (28 x 7 + OIL=) 197 constructed macroeconomic variables, 138 (70.4%) are on monthy basis. For the occasions that I not quarterly only data I consider that these remain stable over the quarter. The crude oil monthly prices were obtained from the EIA Statistics of the US Government.

To estimate the main equation regression I needed firm-specific financial data. Because the exchange rate and stock returns observations are on a monthly basis it is essential the firm-specific control variables to refer to the same horizon. In previous literature, researchers either avoid to use control variables, or they consider their values as annually constant, e.g. Gao (2000). On the contrary, I obtained quarterly data from the quarter financial statements of the companies as provided within the firms’ web-sites

(32)

and/or in the ASE database. However, the latter maintains statements only for the three latest years (i.e. 2004 was missing) and the same is valid for the majority of the firms’ websites. Thus, I had to contact companies’ financial departments to ask for the quarterly statements. During these phone/e-mail contacts I managed to collect also monthly data for 32 (40%) companies. For the rest of the companies I consider that the values remain stable over the quarter. The assumption should not bring serious bias in the results since items like assets and liabilities that I used do not change considerably from month to month. One drawback of the monthly data I collected is that they are not audited.

2.2. Analysis

Table 1 presents a statistical description of the dataset. First of all, I do not apply any selection criteria related to the level of foreign activity of a firm. Bartov and Bodnar (1994) attribute part of the failure of the empirical research to document significant levels of exposure to poor sample selection. That is, they hypothesize that the inclusion of purely domestic firms which (probably) have very limited exposures might bias the results, underestimating the total exposure. However, I do include such firms because (1) I want to test the abovr hypothesis (2) I want to examine the possibility firms without direct linkage to the international financial markets to be indirectly exposed to exchange rate movements.

(33)

papers. However, such an effort is also limited by data availability. Accordingly, the Greek firms rarely report segmental information or detailed information about their foreign activities. Thus, it would be impossible to construct firm specific exchange rate indices and/or variables like the foreign to total sales, which are marked as exposure determinants, only by using annual and other published reports27. The questionnaire can be found in the Appendices. There are four questions that ask the recipient to indicate the percentage of the exchange rate risk that is hedged through financial and operating hedging, and the percentage of their foreign sales and expenses. A fifth question asks the recipient to list the currencies that are mostly used by the company and could have an impact on the firm’s value, in order of importance. To ensure some degree of validity, I also compare the given answers to information provided in annual reports in case any palpable inconsistencies can be identified.

I examine 81 companies that were listed in the ASE for at least 2 full years28 within the 2004-2007 period. The sample covers 13 out of the 15 non-financial industries of the ASE while the firms are almost equally divided in terms of their size; therefore, the sample can be considered as satisfactorily representative of the total market. In 6 out of the 13 covered industries more than 33% of the listed firms were included in the sample, and in 10 more than 23%. Travel & Leisure, Telecommunications, and Media industries are not well represented in the sample, but still, these are industries that are mainly nationally oriented while they do not face foreign competition29. To further validate the representativeness of the dataset, I conduct mean comparison between the firms that responded to the questionnaire and the rest of the market, in terms of size and age. Results are presented in Table 1b. For both criteria, the null hypothesis that the means do not statistically differ cannot be rejected at the 10% level, implying that the dataset does not suffer from sample bias, or in the current case, non-response bias. The US directory names as multinational a corporation with minimum 5% of foreign cash flows or assets.

27

Apart from the case that the methodology introduced by Fraser and Pantzalis (2003) is used, i.e. the construction of firm-specific exchange rates based absolutely on the number of the foreign subsidiaries of a firm and the number of countries where these subsidiaries are located.

28

Actually, only 3 firms were not listed during the whole examined period, with two being listed three years, and the third one for slightly more than two years.

(34)

In my sample, 60 companies (74%) have reported more than 10% of their cash flows (sales/expenses) being denominated in foreign currencies, during the whole examined period. 14 companies can be marked as purely domestic, since they have 0-1% cash flows denominated in foreign currency. Reported foreign sales percentages range between 0% and 94% while foreign expenses between 0% and 72%. The average firm of the sample receives 20% of its inflows (sales) and 22,7% of its outflows (expenses) into foreign currency. This indicates that in general Greek companies rely a lot on imported production inputs. Chemicals and Basic Resources are the most internationally oriented industries since 100% of the sample firms appear as multinational organizationsIn terms of hedging, financial hedging instruments have been used by 38 firms (47%) for at least one year of the examined period and by 27 (33%) during the whole period. Operating hedging on the other side seems to be more popular for the Greek companies; 56 (70%) firms applied such techniques for at least one year, and 43 (53%) for the whole period. Overall, 61 (75%) companies used ether financial or operating (or both) hedging tools for at least on year of the examined period, and 51 (63%) during the whole period.

Concerning the exchange rates that the Greek companies use, 26 currencies have been named as potential sources of exchange rate exposure. Table 2 shows details about the number of companies that are exposed to each of the currencies per year. The USD ranks first with 60 (74%) companies in average indicating it as an influential currency for their firm’s value. 28 (35%) firms are exposed to Romanian Lei, 17 (21%) to GBP, 14 (18%) to Bulgarian Leva, and 12 (15%) to Cypriot Pound.

3. Empirical Results and Analysis

3.1 First-Stage Regressions

(35)

FX

jt

= φ

0

+

∑

φ

1j

IV

FXjt-1

+ ε

FXjt

(2)

RMKT

t

= θ

0

+

∑

θ1i

IV

MKTt-1

+ θ

2 εFXjt

+ ε

MKTt

(3)

The estimations are calculated under the OLS methodology. Table 3 reports the results for the 26 exchange rates regressions on their macroeconomic determinants. In general, the crude oil prices are significant determinant in 14 exchange rates, while the interest rate differential in 11. Doukas et al (2003) have found only the interest rate differential to be significant, without including in their instrument variables set any kind of energy variable. The GDP variations seem to play a role in the determination of the exchange rates as well. The average significant level for the regressed equations, as given by the F-statistic is 30%, which is relatively low, but still indicates that, to some extent, the exchange rate fluctuations have in the long-run a component economically significant. Moreover, the R-sq values are in average 37%. This level is quite low but much higher than the results of Doukas et al (2003) who report R-sq values as low as 11-13%. This result reveals that in the short-run there is a fraction of exchange rate changes that can be predicted. However, it also shows that in the short-run, the biggest fraction of exchange rate movements is unanticipated. Moreover, when comparing to Doukas et al (2003) the much higher R-sq in my specification, which includes macroeconomic factors from both countries that constitute each exchange rate, attests to the quality and the efficiency of the instruments used in this paper.

(36)

Equation 3 orthogonalizes the market factor on its macroeconomic determinants. Only the wages level in Greece, and the Turkish Lira, Swiss Franc, Indonesian Rupee, Romanian Lei, and GBP exchange rate changes have significant coefficients at either of the 10%, 5%, or 1% levels. The signs of the exchange rate residuals coefficients vary, with 14/26 being negative. A negative sign indicates that a devaluation of the EURO against this currency would cause the market return to lower. Even thought the majority of the variables are insignificant in explaining alone the variations of the market return, the joint effect of the coefficients on the dependent variable is significant at the 12% level with an R-sq of 88%. This indicates that the overall market behaviour is affected by the general macroeconomic conditions in the world market (i.e. the joined effect of the variables rather than the individual). The use of the BIS residual as the exchange rate instrumental variable results in more coefficients being significant. The wages level, the inflation rate, the trade of balance and the BIS residual are significant determinants of the market return. However the overall models explanatory power is substantially lower than when the bilateral exchange rates are used with R-sq 58% and significance level of 22%.

3.2 Contemporaneous Firm-Specific Exchange Rate Exposure

This part of the analysis aims to reveal the extent to which Greek firms are affected by unexpected exchange rate fluctuations. In order to identify potential reasons of previous literature failure to document significant levels of exposure, I estimate exposure not only via the suggested by this paper model, but via other specifications that have been largely employed in the past as well. The analysis is conducted in two stages according to the exchange rate factor that is used, i.e. the BIS index and the firm-specific indices that I constructed.

3.2.1. Trade-Weighted Exchange Rate Index

(37)

R

it

= α

i

+ β

it

R

Mt

+ γ

it

FX

it

+ ε

it

(1)

where the exchange rate exposure is based on the BIS index and it is estimated using GLS method so as to mitigate possible omitted variables bias. In a second stage I add the firm-specific control variables and the equation becomes as follows:

R

it

= α

i

+ β

it

R

Mt

+ γ

it

FX

it

+

∑

δ

it

CV

it

+ ε

it

(1b)

This specification is estimated using OLS method. Finally, I apply the two-stage analysis as explained in the methodology part:

R

it

= α

i

+ β

it

ε

MKTt

+ γ

it

ε

FXit

+

∑

δ

it

CV

it

+ ε

it

(4)

OLS method is used in this model specification as well. All the regressions are estimated using monthly data under the Newey-West standard errors methodology in order to control for potential autocorrelation and heteroscedasticity30. The trade-weighted index of the BIS is calculated as EURO per unit of foreign currency; therefore, an increase in the index implies depreciation of the EURO against the foreign currency, i.e. the index. In that case a positive exposure coefficient implies that devaluation of the EURO increases the stock-prices of the firms. Table 5- Panel A exhibits the overall results of the regressions.

For the first regression, following Jorion (1990) and the majority of the subsequent papers, I regress the monthly firm stock returns on the monthly changes of the BIS exchange rate index and the monthly returns of the market index. The exposure coefficients range between -4,63 and 5,17 with mean 2,18. This means that the market

“Exchange Rate Exposure at the Firm and Industry Level: Impact on Company Value and the Role of Hedging, Evidence from Greece”