1
The influence of the introduction of the
euro on exchange rate exposure
– The case of the Netherlands 1993-2007 –
2
ABSTRACT
In this paper, the effect of the introduction of the euro on exchange rate exposure is being examined. Using three timeframes („93-„97, „98-„02 and „03-„07), a sample of 68 Dutch non-financial firms has been tested on changes in volatility, and changes in direct exposure by regressing stock returns against a trade weighted currency index. Contrary to expectations, the results of the analyses suggest that immediately after the introduction of the euro, stock return volatility increased significantly, along with the number of firms that were significantly exposed to exchange rates. The same result is visible when examining industry portfolios and the aggregated market index. Moreover, the individual firms show an average change of the relationship coefficient from negative in the first period, to positive in the subsequent periods. Furthermore, the time lag that investors need to assess the influence of currency fluctuations on the value of a company has been significantly shorter for the period immediately after the introduction of the euro, than for the other two periods. The entire Dutch stock index has been significantly exposed to currency fluctuations for the first two periods, where the ‟98-‟02 period showed evidence of investors initially overreacting to changes in exchange rates.
3
TABLE OF CONTENTS
I INTRODUCTION ...5
II THEORY ...7
2.1 Exchange rate exposure in perfect markets ...7
2.2 Company value ...8
2.3 Predictors of exchange rate fluctuations ...8
2.4 Types of exchange rate exposure ...9
2.5 The joint determination of stock prices and exchange rates ... 10
III LITERATURE OVERVIEW ... 11
3.1 Earliest studies: little significance ... 11
3.2 Mispricing and sample selection biases ... 12
3.3 More convincing results ... 13
3.4 The effect of the introduction of the euro ... 15
3.5 Methodologies ... 16
IV METHODOLOGY ... 18
4.1 Research question ... 18
4.2 Research objective ... 19
4.3 Research design ... 19
4.4 Sample selection and data description ... 21
4.5 Equal weighted and value weighted portfolios ... 23
4.6 The trade weighted currency index ... 24
4.7 Data sources ... 26
V ANALYSIS ... 26
5.1 Stock return volatility – individual firm level ... 26
5.2 Stock return volatility – industry level ... 29
5.3 Stock return volatility – market level... 30
5.4.1 Actual exchange rate exposure – individual firm level ... 31
5.4.2 Coefficient of the relationship ... 33
5.4.3 Time lag... 35
5.5.1 Actual exchange rate exposure – industry level ... 36
5.5.2 Coefficient of the relationship ... 39
5.6 Actual exchange rate exposure – Market level... 40
VI CONCLUSION ... 41
VII LIMITATIONS AND RECOMMENDATIONS ... 43
4
APPENDIX A: SAMPLE FIRMS ... 51
APPENDIX B: NORMALITY OF STOCK RETURN VARIANCES ... 53
APPENDIX C: MANN-WHITNEY U-TEST ON DIFFERING VARIANCES... 55
5
I INTRODUCTION
The last few decades can be characterized by the explosion of a trend that has been prevalent in our society since before the beginning of the 20th century; globalization. The concept of globalization, however, has only been used by economists since the 1980s, and is often defined as; a phenomenon involving the integration of economies, cultures, governmental policies and political movements around the world (Encyclopedia Britannica). Although this development is not warmly welcomed by every group in society, the consequences of it are being experienced by everybody. When travelling to a variety of countries on various continents, one is very likely to see common brands, common styles of clothing and common habits. One might even speak the same language. The world is getting smaller, and it is doing so very rapidly. In Jules Vernes Le tour du monde en quatre-vingts jours (1873), Phileas Fogg has a very hard time travelling the world in eighty days. Nowadays, the same journey be done in less than a week.
This proverbial shrinking of the world has had a dramatic effect on the way that business has been conducted in recent years. In order to find new sources of competitive advantage to beat the competition, many companies have entered the global arena of trade by conducting (part of) their business abroad. Although many companies (and scholars alike) advocate the benefits of operating on a global basis, globalization has severely altered the rules of the game. Companies should, for example, take into account a variety of preferences of customers in different parts of the world, and competition from other continents may pose a threat to local entrepreneurs.
Another consequence of international trade is that a company will be exposed to foreign currencies. When a company signs contracts with either foreign buyers or suppliers, chances are high that the counterparty wants to be paid in a currency other than the companies home currency, which introduces risk as exchange rates may be volatile.
Typically, exchange rates are more than four times as volatile as interest rates, and more than ten times as volatile as inflation rates (Jorion, 1990). The effects of this can be exhibited by a small investigation among the annual accounts of some of the largest Dutch companies. Heineken, the Dutch beer producer, reported a negative 2007 result of €152 million due to exchange rate fluctuations, whereas electronics concern Philips made a loss of €112 million due to varying exchange rates in the same year. That these fluctuations can also turn out to be favorable to the results is proven by Shell, the oil giant, which reported a quarterly gain on exchange rate fluctuations of €123 million in its 3rd
6
The concept that defies the relationship between exchange rate fluctuations and the value of a certain asset is called exchange rate exposure. Dumas (1978), Adler and Dumas (1980) and Hodder (1982) use the following definition to describe this concept: The regression coefficient of the real value of a firm on the exchange rate across states of nature. Jorion (1990) adds that exchange rate exposure can be decomposed into two effects: the value of net monetary assets with fixed nominal payoffs, and the value of real assets held by the firm. That not only multinational firms, but also purely domestic firms experience the effects of the randomness of exchange rate fluctuations, is stipulated by the second effect. Because a fluctuating exchange rate can have an influence on the aggregate demand of, for example, the value of real estate, firms that operate purely in one country can still feel the consequences of a falling or rising value of a country‟s exchange rate.
In 2002, however, with the introduction of the euro many (European) currencies ceased to exist. Consequently, contracts that were signed between two euro countries were from then on denominated in the same currency. For many companies this might mean that the importance of exchange rate exposure declines, as trade might be conducted for a great part in the new common currency.
The purpose of this paper will therefore be to investigate whether exchange rate exposure has become less relevant with the introduction of the euro. The main question will be:
To what extent are Dutch firms influenced by exchange rate fluctuations, and how did this change after the introduction of the euro?
7
II THEORY
Before the analysis will be started, it is useful to examine how exchange rates are being set and how they can influence company results. In order to get a clear understanding of this system, the first step will be to investigate Modigliani and Millers irrelevance theorem (1958, 1961)
2.1 Exchange rate exposure in perfect markets
In order to be better able to understand exchange rate exposure, it is useful to start with the perfect market assumption. This situation has been described in Modigliani and Miller‟s irrelevance theorem (1958, 1961). According to this theory, a firm cannot increase its value by undertaking activities that investors can perform themselves. Although the primary concern of these authors is the capital structure of a firm, the theory has also been applied as a case against hedging (an often used method to avoid risk due to exchange rate fluctuations) by companies (Duffie, 1991). In a world with perfectly efficient markets, every individual will have the same set of information. Hence, if all parties in the financial system have equal information, the theorem states that the choice of hedging should be made by investors (the actual owners of the company), who can diversify the risk themselves by creating a portfolio according to their risk preference.
Furthermore, when one assumes perfect markets, the concept of exchange rate exposure would be non-relevant. This is due to two theories; the law of one price, and purchasing power parity. According to the law of one price, every good will have the same price in different countries (although this price may be quantified in a different currency) (Chambers & Lacey, 2004). Suppose that gold trades at $400.- per ounce in the US, and that the euro trades at $1,25. Because in a perfect market trade is not being distorted by for example costs of transport, the law of one price tells us that the gold price in the Eurozone will be 400/1,25= €320.-.
8
The importance of these theories for exchange rate exposure is obvious. When a company earns a certain amount of foreign currency, in a perfect market situation it does not matter whether the company decides to convert the money or not. The law of one price and purchasing power parity will make sure that price levels will be equal all over the world, and that the same goods can be bought for the same amount of money everywhere.
However, the perfect market assumption generally does not hold. First of all, in many cases markets can be characterized as being information asymmetric. Companies often possess more information than individual investors, who are therefore not in the best position to make hedging decisions. Secondly, empirical evidence supporting the existence of concepts like purchasing power parity and the law of one price is only very limited (Chambers & Lacey, 2004). It will thus be useful to investigate where exchange rate exposure comes from, and how it influences the value of a company.
2.2 Company value
According to basic financial theory, the value of any company is the sum of its future cash flows. Because a €100,000 profit in 2050 is not worth as much as a €100,000 profit tomorrow, these cash flows should be discounted against a rate based on e.g. inflation and risk. As we can see from the examples above, however, exchange rate fluctuations can have quite a large influence on a company‟s cash flows. This means that when an exchange rate rises or falls unfavorably for a company, it will prospect less earnings. Having said this, the link with financial theory is clear. A depreciating exchange rate means less profits, which in turn means lower (future) cash flows which should negatively affect the value of a company. If future cashflows fall, investors will sell their stocks, which will increase the supply and lower the demand, which will in turn accumulate in lower stock prices (Della Corte et al, 2008).
2.3 Predictors of exchange rate fluctuations
9
There is, however, a second model that explains fluctuations in exchange rates, namely the interest rate parity model.
The interest rate parity theorem states that the international interest rate differences will be equal to differences between current and forward exchange rates (Chambers & Lacey, 2004). Hence, the ratio between interest rates equals the ratio between forward and spot foreign currencies, or; 1+Domestic Yield1+Foreign Yield = Forward Exchange RateSpot Exchange Rate . The rationale behind this theory is that investors should be indifferent between holding two currencies. In other words; the difference between the exchange rates of two currencies should be the difference in interest rates, added to the expected appreciation of a currency. Imagine a situation in which the 12-month dollar interest rate is 5%, while the 12-month yen interest rate is 1%. In this case, US investors will be eager to borrow money in Japan, as the 12-month yen interest rate is very low compared to the dollar interest rate. Therefore, demand for the yen will rise, and so will the supply of dollars, making the yen/dollar exchange rate rise. According to the theory of interest rate parity, the yen/dollar rate will strengthen as much as the difference in interest rates, which is 5%-1%= 4% (Della Corte et al, 2008).
It should be noted that, although interest rates are very important in the valuation of currencies, the theories presented above are certainly not flawless, and scholars often fail to find convincing empirical evidence that supports them (Della Corte et al, 2008).
2.4 Types of exchange rate exposure
As mentioned several times before, one of the most prevalent features of international finance is that currencies fluctuate in value against each other. Therefore, companies with any international link in the value chain or in their competitive environments are by definition exposed to currency fluctuations (Lasserre, 2003). The way in which this process occurs, however, is not the same for each company. Lasserre (2003) acknowledges three different types of exchange rate exposure:
10
buying its supplies in the US, will have to pay more in order to get the same amount of materials. Assuming that the yen/euro rate remains constant, this results in a cost disadvantage for manufacturer A.
This type of exposure is also often called operational exposure (i.e. Yong-Cheol & McElreath, 2001), as it decreases or increases the costs of operating for a company.
Transaction exposure: Transaction exposure arises when a company has contractual obligations that are to be fulfilled in a foreign currency. Suppose a Dutch windmill manufacturer that signs a contract to build 20 windmills in the US, and will be paid in US dollars six months from now. When the dollars are received, these will be converted into euro‟s, as this is the common currency in the Netherlands. However, when the dollar depreciates unexpectedly within the next few months, the company will receive less euro‟s for its dollars when it converts them, and thus be less profitable than expected.
Translation exposure: This type of exposure is mainly an accounting affair. It reflects the effects that the change in currency values has on the financial statements of internationally operating firms. When a Spanish company operates subsidiaries in the US and in China, it will probably have cash flows in euro‟s, dollars and yuan. At the end of the year, however, the company will have to report a consolidated annual report in euro‟s. Although no currency transactions are being made, the company has to translate the figures from its subsidiaries abroad, to euro‟s. This means that differences may occur due to currency fluctuations, as compared to the year before.
2.5 The joint determination of stock prices and exchange rates
11
individual firm. It is very likely, however, that the entire market portfolio will be influenced by this second component. In order to be able to exclude this second component and measure the pure exchange rate exposure of individual firms, scholars usually add the market index to the regression analysis, to ensure that the results reflect the true exchange rate exposure of companies.
Adding the market index to the regression, however, is not a perfect method to prevent biases as a result of the joint determination of stock prices and exchange rates. The market index accounts for the aggregated exposure of the entire market to macroeconomic changes that also influence exchange rates. However, not every firm is equally related to the market. This is reflected in differing firm betas. Therefore, the true importance of macroeconomic shocks on a firm‟s stock value may be over- or underestimated when adding the market index to the regression, and results may still be marginally biased.
Now that the different theories underlying exchange rate exposure are known, this paper will continue with discussing the empirical evidence that has been found to prove the relationship between exchange rate fluctuations and firm value. The next section will therefore give an extensive overview of the literature written on the topic so far.
III LITERATURE OVERVIEW
3.1 Earliest studies: little significance
12
researchers focused on the impact of interest rates and inflation rates on the market value of companies (Jorion, 1990).
It wasn‟t until the early 1990‟s that the body of literature written on the topic of exchange rate exposure began to grow. Jorion (1990) is often quoted as being the key person in the development of this field of research (e.g. De Jong, Ligterink & Macrae, 2006). According to him, the lack of research on the topic that existed until then, could be due to the general absence of data sources in international finance, and the fact that the absolute size of currency exposure was generally small relative to the measurement error. By focusing on multinational firms in the US he is the first to investigate the topic extensively. By regressing a Trade Weighted Currency Index (TWC, which will be addressed later in this paper) against cross-sectional portfolios of US firms, he is able to find a weak correlation in which 15 out of the 287 investigated firms show a significant relation with exchange rate fluctuations. With 5,2% of the total amount of firms investigated, however, this is little more than random chance. Secondly, he finds evidence that the rate of exposure might differ between industries by studying and comparing portfolios with selections of companies in different sectors.
The lack of significance was a surprising outcome, as most scholars believed that a relationship should exist. A series of studies following up on these results, however, still proved unable to find convincing significant evidence (e.g. Amihud, 1994; Choi & Prasad, 1995). Still focusing on US multinationals, many authors failed to find distinguishing evidence both for individual firms, as for sets of portfolios (e.g. Dukas et al., 1996).
3.2 Mispricing and sample selection biases
This lack of evidence of what scholars still assumed to be existent led Bartov and Bodnar (1995) to make two revolutionary propositions. According to them, there was a situation in which mispricing was prevalent, and scholars were focusing on the wrong set of companies. To support this, the authors developed two hypotheses; the sample selection hypothesis and the lagged response hypothesis.
13
According to the sample selection hypothesis, it is therefore important that only firms that have strong international linkages are selected in the sample of analysis. Another interpretation of the sample selection hypothesis is that, until then, the vast majority of studies on the topic had been focusing on the US. Because the US has an enormous home market, however, many companies only earn a relatively small part of their turnover abroad. Realizing this, it is not more than logical that statistical evidence had been only very limited until then. When companies do not operate abroad, the influence of exchange rates will be considerably less than when a company does have international linkages in its organizational design (that companies that solely operate in one market are still influenced by exchange rate fluctuations has been discussed in the introduction). Therefore, according to Bartov and Bodnar (1995), researchers should only incorporate those companies that have strong international linkages in their research design.
Secondly, Bartov and Bodnar (1995) introduce the effect of lagged responses into the field of research. Until then, scholars had been focusing on the contemporaneous effect of currency fluctuations. It is likely, however, that investors will need some time when characterizing the relationship between firm value and changes in the investigated currency. There are three reasons for this:
It is complex to identify possible asymmetries in the impact of appreciations and depreciations of a currency on firm value
It is complex to determine whether a change in the value of a currency is permanent or temporary
It is complex to determine the impact that a change in the value of a currency will have on the economic performance of a company
These reasons all contribute to the inability of investors to make an immediate assessment of the consequences of a depreciating or appreciating currency. Therefore, investors will need time in order to make a buy or sell decision, resulting in a lagged response.
3.3 More convincing results
14
even when incorporating these propositions many authors still failed to find the results they were hoping for, namely convincing statistical evidence for exchange rate exposure (e.g. Donnelly & Sheehy, 1996; Walsh, 1994). The growing body of literature written on the topic, and the failure to find the expected results led to the introduction of the term “exchange rate exposure puzzle”. Authors still believed that exchange rate fluctuations influenced firm value, but contributed their failure to find statistical evidence to the use of the wrong methodologies (Bartram & Bodnar, 2005).
Until the beginning of the 21st century, the vast majority of researchers still focused on US multinationals, although some exceptions were being made for other major economic powers as Germany and Japan. This is a remarkable situation, as it contradicts Bartov and Bodnar‟s (1994) propositions. As mentioned above, they argue that researchers should focus their attention on firms that are heavily exposed to currency fluctuations. By focusing mainly on firms in the US, this is exactly what researches had not been doing. The US exports only a little more than 11% of its total production (www.ppionline.org), and therefore scholars might have been focusing on the wrong market.
It wasn‟t until recently that academics have acknowledged this critique (Forslöf & Nilsson, 2007; De Jong, Ligterink & Macrae, 2006). From the beginning of this century, researchers shifted their focus from measuring predominantly US multinationals, to measuring firms in (emerging) markets (e.g. Chue & Cook, 2008; Dominguez & Tesar, 2006). Research on small, industrialized countries that as a result rely heavily on their exports and therewith might be influenced heavily by exchange rate fluctuations, however, is still very scarce. Secondly, although recently authors have shifted their attention beyond the US, the timeframe studied is typically during the 1990s. For the few European countries that have been examined, this means that these researches have been done in the pre-euro era and that the situation might have changed drastically as a result of the introduction of the common currency.
15 3.4 The effect of the introduction of the euro
To date, several authors have empirically studied the impact of the introduction of the euro on a variety of economic dimensions. Evidence suggests that the event has had a mixed effect on stock market volatility. In Germany, the volatility of the stock market increased, whereas in Spain and Italy it decreased (Billio & Pelizzon, 2002; Morana & Beltratti, 2002). Flam and Nordström (2003) and Barr et al (2003) have shown that the introduction of the euro has given rise to a significant increase in trade between euro countries, as well as with outside countries.
On a more aggregated level Danthine et al (2001) concludes that the impact of the euro has been positively when looking at the European capital markets. Adjaouté and Danthine (2003), however, argue that the effect of the introduction of the euro on equity markets has been only very limited.
The influence of the introduction of the euro on the level of exchange rate exposure, however, remains a question that has largely been unaddressed. One of the very few exceptions on this are Bartram and Karolyi (2006). These authors base their research method partly on a paper written a decade earlier by Bartov et al (1996). By studying the period of the breakdown of the Bretton Woods system, Bartov et al (1996) discuss the effect of an increase in exchange rate variability on the riskiness of US multinationals. Bartram and Karolyi (2006) argue that the introduction of the euro is the exact opposite of the abandonment of the Bretton Woods system. Where the breakdown of the Bretton Woods system led to an increase in exchange rate variability, the introduction of the euro is expected to lead to a decrease in exchange rate variability. Therefore, Bartram and Karolyi (2006) use the same approach, but with opposite hypotheses. For the purpose of this paper, only the first and third hypotheses of these authors are relevant. The first hypothesis focuses on stock return volatility;
H1: Nonfinancial firms with sales/assets in euro area countries should exhibit a larger reduction in stock return volatility than firms with no sales/assets in the euro area.
16
Secondly, Bartram and Karolyi (2006) focus on the actual change in exchange rate exposure;
H3: Nonfinancial firms with sales/assets in euro area countries should exhibit a larger reduction in incremental foreign exchange rate risk exposure than firms with no foreign sales/assets in the Euro area.
Because the euro will reduce the amount of foreign cashflow of European companies or companies that operate in Europe drastically, one would expect exchange rate exposure to decrease more for these firms than for firms not operating in Europe.
The results of Bartram and Karolyi (2006) are surprising. Contrary to expectations, the introduction of the euro has led to an increase in stock return volatility for both their samples, although the non-euro companies showed an increase that was significantly larger than euro companies. The amount of firms that show statistically significant exchange rate exposure, however, is low. Only around 6% of the firms in their samples is significantly affected by changes in exchange rates.
Although this research acknowledges the importance of the introduction of the euro on exchange rate exposure, the researched timeframe after the introduction of the euro is only slightly more than 3,5 years (from January 1st 1998 to August 2001). As the authors also acknowledge, the results are therefore probably biased because one would expect a transition period immediately after the introduction that may influence the results. During this period, companies will assess the influence of the euro introduction on corporate strategy and future contracts, and might for example decide to switch to suppliers inside the euro area instead of overseas. The duration and magnitude of this period, however, is not clear, although it is a good indication that more research on this topic is needed.
3.5 Methodologies
17
𝑅𝑖𝑡 = 𝛽0𝑡 + 𝛽1𝑡𝑅𝑠𝑡 + 𝛽2𝑡𝑅𝑚𝑡 + 𝜖𝑖𝑡, 𝑡 = 1
In this, 𝑅𝑖𝑡 is the stock return, 𝑅𝑚𝑡 the return on the market index and 𝑅𝑠𝑡 the change in the
exchange rate (or in this case, the TWC). According to Jorion (1990), combining all relevant exchange rates into one multilateral exchange rate results in a simple but effective representation that is convenient to use, as the change of only one number (the TWC) represents the change in many foreign currencies.
As can be seen from the formula, the return on the market index is included. The advantages of this procedure are described in the theory section above. Although the author failed to find significance, the same technique was used by other scholars, also without significant results (f.e. Amihud, 1994; Godnar & Gentry, 1993).
Because day to day trading is often characterized by buy and sell decisions based upon incomplete information, rumors or other irrational motives, usually the interval period of one month is taken. By doing so, authors try to avoid the problem of having biased results due to this so-called noise trading (f.e. Parzalis et al, 2003; Doidge et al, 2006). Some authors deliberately choose a shorter bi-weekly time interval (f.e. De Jong, Ligerink & Macrae, 2006) or one week (f.e. Dominguez, 1998; Griffin & Stulz, 2001) in order to overcome the problem of having only a limited amount of measuring intervals when studying a relatively short timeframe. A shorter interval, however, does not seem to significantly affect the outcomes. Furthermore, the time period studied varies from study to study, with some studies exhibiting a period of less than 5 years (f.e. Allayannis & Ofek, 2001), and others up to 25 years (Doidge et al, 2006). Here also, outcomes between different time periods studied tend to be similar, with no period showing exceptionally higher significance. The last dimension in which studies differentiate themselves is when looking at the determinants of exchange rate exposure. Examples of this are a significantly higher exposure of firms that are also listed at at least one other foreign stock market (Booth & Rotenberg, 1990), differences in leverage (He & Ng, 1998), and the foreign / total assets ratio (Miller & Reuer, 1998a). For a complete overview of studies on the determinants of exchange rate exposure, Bartram and Bodnar (2005) are suggested.
18
questionnaires to determine per company which exchange rate influenced company value most. Based on this, they created a firm specific TWC. Furthermore, they introduce on- and offbalance hedging and assess this based on the questionnaires. Of their sample, an exceptional amount of 50% shows a significant relationship with the TWC. Secondly, Dewenter et al (2005) use a radically different approach. In their research they use an event study method to avoid the difficulties mentioned by other authors. The results, however, are slightly disappointing with only approximately 15% of the firms showing significance.
As can be deducted from the overview given above, research on the topic of exchange rate exposure is still developing but so far non-conclusive. Moreover, studies focusing on the effect of the introduction of the euro on exchange rate exposure are still very rare. The remainder of this paper will therefore focus on the methods that are used to attempt to fill this gap, and the results that this approach has yielded
IV METHODOLOGY
4.1 Research question
As mentioned above, the primary concern of this paper will be to investigate whether the introduction of the euro has had a significantly altering effect on the exchange rate exposure of firms. The main research question that will be answered will therefore be:
To what extent are Dutch firms influenced by exchange rate fluctuations, and how did this change after the introduction of the euro?
The research will focus on the Netherlands for several reasons. First of all, in order to measure the effect of the introduction of the euro on exchange rate exposure, it is logical to focus on a country that has indeed introduced the euro. Furthermore, the Netherlands adopted the euro as early as in 2002 as their official national currency. Therefore, the timeframe that can be studied will be sufficiently large to draw solid conclusions.
19
Germany had a higher absolute amount of exports in 2008. Moreover, the Netherlands have the second highest trading surplus in Europe, again following Germany (www.cbs.nl). By focusing on the Netherlands, therefore, it has been tried to include as many international linkages as possible.
Thirdly, data availability for the Netherlands is relatively high, which increases the possibility of significant results.
4.2 Research objective
The objective of this paper will be twofold. As shown in the literature review above, many authors have dedicated time and effort in order to solve the exposure puzzle. Most researches, however, have focused their attention on US firms. The amount of researches that focus on firms in smaller countries, that as a results rely more heavily on exports, is still limited. By focusing on the Netherlands, this is the first contribution of this paper.
Secondly, and more importantly, although the introduction of the euro may have had a significant effect on the exchange rate exposure of multinationals, research on this topic is still barely existent. The second contribution of this paper to the existing body of literature is therefore to fill this gap, and to make recommendations for further research.
4.3 Research design
In order to be able to answer the main question, as stated above, several hypotheses have been proposed. The first hypothesis is based on the research conducted by Bartov et al (1996). These authors state that when stock prices are indeed influenced by exchange rate variability, a reduction in exchange rate variability should lead to a reduction in stock return volatility. By studying the breakdown of the Bretton Woods system (a situation in which exchange rate variability increased due to the abandonment of a system with fixed rates), they find evidence that this relationship is indeed significant.
Although Bartov et al (1996) focus on a situation in which the amount of exchange rate fluctuations increased, the relevance of their research for this paper is obvious. The euro has decreased the amount of foreign currencies (the exact opposite effect of the breakdown of Bretton Woods). Therefore, aggregate exchange rate variability in Europe has decreased and one would expect that stock return volatility has also decreased. In order to verify this, the first hypothesis will be:
H1. The introduction of the euro has led to a decreased level of stock return volatility of stock
20
In order to be able to measure volatility, market, industry and individual stock returns will be calculated on a bi-weekly basis. Based on this, the return variances will be calculated for three timeframes. The reasoning for using three timeframes is given in paragraph 4.4 below. Moreover, statistical significance will be tested by using a student t-test in the case that the collected data is normally distributed, or a Mann-Whitney U-test in case of non-normally distributed variances. When doing so, the following hypothesis will be tested:
𝐻0 = Stock return variance of Dutch non-financial firms will be the same for the 1993-1997 (1998-2002) period and the 1998-2002 (2003-2007) period.
𝐻1 = Stock return variance of Dutch non-financial firms will differ between the 1993-1997 (1998-2002) period and the 1998-2002 (2003-2007) period.
Both the paired samples t-test, and the Mann-Whitney U-test have been chosen because only two groups of return variances need to be compared to answer the hypotheses.
As mentioned, the analyses will be conducted on a market level, an industry level and an individual firm level to obtain an as rich as possible understanding of the influence of the introduction of the euro on Dutch listed firms.
Having measured the change in exchange rate exposure of Dutch firms by focusing on the stock return volatility, the second part of the analysis will then measure the phenomenon directly by running a regression analysis. This analysis will also be conducted on an individual firm level, an industry level and the entire market level. As mentioned several times above, the abandonment of many currencies within the European Union will probably have led to a reduction in exchange rate exposure. The second hypothesis will therefore be:
H2. The introduction of the euro has led to a decrease in exchange rate exposure of
nonfinancial firms in the Netherlands.
21
𝑅𝑖𝑡 = 𝛽𝑖0+ 𝛽𝑚𝑖 𝑅𝑚𝑡 + 𝛽𝑥𝑖 𝑅𝑥𝑡 + 𝛽𝑥𝑗 𝑅𝑥𝑡−1+ 𝛽𝑥𝑘 𝑅𝑥𝑡−2+ 𝜖𝑖𝑡 t = 1, .…, T [1]
In this, 𝛽𝑚𝑖 and 𝑅𝑚𝑡 are the coefficient and the return on the market index, 𝛽𝑥𝑖, 𝛽𝑥𝑗 and 𝛽𝑥𝑘 are the coefficient of the relationship of the currency index with the stock returns, whereas 𝑅𝑥𝑡, 𝑅𝑥𝑡−1 and 𝑅𝑥𝑡−2 are changes in the TWC (which will further be addressed in section 4.6)
for time period t. In this, stock return is defined as [log 𝑃𝑡− log 𝑃𝑡−1]. Log values are used in the regression in order to assure that the data will be normally distributed. This is in line with for example De Jong et al (2006).
Because it is not exactly known how much time investors need to assess the influence of exchange rate fluctuations on the value of a company, and therewith their buy and sell decisions, three different time frames are included in the formula. 𝑅𝑥𝑡 equals the return of the currency index x on time t. This will represent the contemporaneous effect of changes in the exchange rates. Moreover, 𝑅𝑥𝑡−1 and 𝑅𝑥𝑡−2 measure the return of the TWC at time t-1 (a time lag of 2 weeks), and t-2 (a time lag of 4 weeks).
Furthermore, it is very likely that heteroskedasticity will be occurring in the regression. When conducting a regression analysis, the assumption is made that the variances of the error terms are constant. Because the sample used in this research consists of 68 different firms, this assumption is likely to be violated. In order to prevent the results being biased due to this heteroskedasticity of the error terms, the Newey-West (1987, 1994) syntax will be used. 4.4 Sample selection and data description
22
period of analysis will run until December 31st, 1997. January 1st, 1998 will thus be used as starting date for measuring the exchange rate exposure after the introduction of the euro. This is a few months before the actual decision on which countries would join the final phase of the EMU, and in line with Bris et al (2007).
The second timeframe will run from January 1st, 1998 to December 31st, 2002. This timeframe will represent the situation immediately after the introduction of the euro. There is, however, a major problem with doing this. As mentioned above, Bartram and Karolyi (2006) suggest that a certain transition period, in which companies might change their trading patterns to inside the future euro area in order to avoid exchange rate exposure, is very likely. Although the entire length of this period is not known, it is very likely that this will be somewhere close to the introduction of the euro in May 1998, or in the following years. The period of analysis that starts from January 1999, might therefore be biased because some companies might have started their transition to the euro later than others. Because the transitions are very likely to be finished after the second five-year period, it has been chosen to examine a third period that runs from January 1st, 2003 to December 31st, 2007. For this period, the assumption is made that the transition of companies has been completed, and that companies are not further changing their trading patterns to reap benefits that arise as a result of the introduction of the euro.
The first hypothesis will then be tested by calculating the variance of bi-weekly stock returns, or volatility, for each timeframe. This is done on a bi-weekly basis, in order to exclude any biases due to noise and day trading. This is in line with De Jong, Ligerink and Macrae (2006). In order to make the analysis as extensive as possible, it will be performed on three different levels. First of all, the individual firm level is analyzed to be able to assess the influence that the introduction of the euro has had on the variability of stock returns of individual firms listed in the Netherlands. Having done this, an analysis will be done on an industry level. Based on this, a judgment can be made on whether the introduction of the euro has had different effects on different industries. Lastly, the return volatility of the entire Dutch stock market will be analyzed in order to draw conclusions on the aggregated effect of the introduction of the euro on stock price volatility in the Netherlands.
23
of a change in exchange rates on the value of a company. In order to increase the likelihood of significant results, the contemporaneous effect is added in the regression, as well as a two-week and a four-two-week time lag. This is in line with De Jong, Ligerink and Macrae (2006). In order to be able to compare the results of the first hypothesis with the second, the same time frames will be used, and the same levels of analysis. This means that the second hypothesis will also be tested on a general market level, on an industry level and on a firm specific level. In order to show the hypothesis and the levels of analysis graphically, a conceptual model is presented in figure 1 below.
Figure 1: Conceptual model
Furthermore, the firms that will be analyzed have been selected on a variety of criteria. In order to make a fair adjustment, only firms are included that are listed on the Dutch stock market for the entire 1993-2007 period. Secondly, only nonfinancial firms are included, as financials have considerably different activities with regard to their international transactions, their imports and their exports than firms in other segments (Martinez-Solano, 2000). Thirdly, the listing of the firm should be active (meaning that the stock is frequently traded), and be traded in either guilders or euro‟s. In total, 68 firms meet these criteria. They are presented graphically in appendix A.
4.5 Equal weighted and value weighted portfolios
24
that each firm accounts for an even share of the portfolio, regardless of market capitalization and stock price.
It is very likely, however, that the portfolios will consist partly of lower priced stocks, and partly of higher priced stocks. In order to make a fair distinction between these differences in stock prices, a value weighted portfolio will be constructed as well to test significant relationships with the TWC. In these value weighted portfolios, the proportion of the firms will be relative to their stock prices. Suppose a two-stock portfolio consisting of stock A, selling for €5,- and stock B, selling for 10,-. In this portfolio, the weight of the changes in stock prices will be relative to their stock prices, and thus weigh twice as much for stock B, than for stock A.
4.6 The trade weighted currency index
In order to be able to test hypothesis two, a trade weighted currency index (TWC) will be created that represents changes in the currencies of the most important trading partners of the Netherlands, where trade is defined as the combined value of imports and exports. The approach taken here will differ from Bartram and Karolyi (2006). Bartram and Karolyi (2006) use the Bank of England‟s New Sterling Exchange Rate Index, which is based on imports, exports and competitiveness of England. Obviously, a TWC based on the Pound Sterling would be inappropriate for this research. Moreover, a second party TWC (f.e. from JPMorgan, as is used by de Jong et al; 2006) is not available in the databases that are accessible by the author. Therefore, it has been chosen to create an independent TWC based on trading figures that have been collected from the IMF website (www.imf.org).
25
The second approach is to only include those currencies that still exist after the introduction of the euro. This will increase comparability between the periods, as the TWC will be created with the same currencies in the pre-euro situation as compared with the post-euro situation. As a result, however, the focus of this approach will be on changing trading patterns due to the introduction of the common currency. Companies may, for example, decide that having suppliers inside the euro area gives them a strategic advantage, as it reduces exchange rate exposure. They might therefore abandon their suppliers outside the euro area for ones that operate in Euroland. A framework with a TWC that is calculated out of the same currencies pre- and post-euro will grasp this phenomenon, as the amount of trade in these currencies will be reduced (due to companies decisions to find suppliers and buyers inside the euro-area), and the number of significant relationships will thus decrease.
Because of the increased comparability between the pre-euro and post-euro periods, and the ability to detect a reduced amount of foreign currency exposure as a result of changes in trading patterns due to the introduction of the euro, the second approach will be taken in this research.
Because the TWC will be regressed against each of the 68 individual companies, including relatively small trading partners in the TWC might distort the results as it is likely that only a very small proportion of the sample operates (and thus may be exposed to currency fluctuations) in these countries. Therefore, only the five major trading partners are included. To identify these trading partners, average trading volumes (imports + exports) over the entire 1993-2007 period are calculated. The results of this are represented in figure 2 below.
26
Combined 112385,228 100% 65399,021 100% 97781,984 100% 174001,680 100% World 503926,200 304580,000 420595,800 786602,800
Figure 2: Top 5 trading partners of the Netherlands. Export + Import, average per period in millions of dollars, TWC equals the relative weight of each currency in the TWC per period. Data collected from www.imf.org.
When this has been done, the value of the currencies of each country are multiplied by their relative weight, and added so that the trade weighted currency index is created. This will be done on a bi-weekly basis, in order to exclude day trading and other minor distortions. This is in line with De Jong, Ligerink and Macrae (2006).
4.7 Data sources
In order to be able to grasp the information that is needed to perform this research, several data sources will be used. First of all, the information on stock returns will come from Thomson Financial Datastream. Here, biweekly Wednesday closing rates will be found. Wednesdays have been chosen in order to avoid beginning-of-the-week and end-of-the-week effects (de Jong et al, 2006). Secondly, the data on exchange rate fluctuations will be taken from Reuters Ecowin. The data that will be gathered here will be biweekly averages of daily end-of-day spot rates. Because direct guilder quotations are not available for every spot rate that is needed in the analysis, the guilder quotations will be calculated from the dollar rates (that are fully available). By doing so, a non-arbitrage situation will be automatically assumed. Furthermore, a guilder/euro conversion rate is assumed of ƒ2,20371 / €1-
V ANALYSIS
5.1 Stock return volatility – individual firm level
As mentioned in the methodology section above, the first part of the analysis will focus on the effect of the introduction of the euro on the stock return variability of Dutch non-financial companies. The following hypothesis will be tested:
The introduction of the euro has led to a decreased level of volatility of stock returns of nonfinancial firms denominated on the Dutch stock market.
27
Having done this, the variance of stock returns of each individual firm will be compared over the three periods, in order to be able to draw conclusions of a change in stock return variances over time. The results of this analysis are shown in figure 3 and 4 below.
Average variance 1993-1997 0,2119% 1998-2002 0,3880% 2003-2007 0,2470%
Figure 3: Average bi-weekly variance of stock returns per period, calculated on a bi-weekly basis.
Direction of change Number of firms Percentage of firms
uu 13 19,0%
du 6 8,8%
ud 44 64,7%
dd 5 7,4%
Total 68 100%
Figure 4: Changing variances of stock returns of Dutch non-financial firms. Direction shows the direction of change for the period directly after the introduction of the euro (1998-2002) as compared with the pre-euro period (1993-1997), followed by the direction of change of the 2002-2007 period as compared to the 1997-2002 period. u=up, d=down, hence ud means an increase in variance directly after the introduction of the euro (1998-2002), but a decrease in variance in the 2002-2007 period.
As can been seen in the figures above, the variances differ for each of the three periods. Therefore, the next questions that will be addressed is whether the change in variances is significant. A histogram of the distributions of these variances is shown in appendix B. As can be seen from this appendix, the results are only partly normally distributed. It should be noted that because the associated data are stock return variances, no negative values are shown. Hence, only part of the left tail of the distribution is visible. Because this equals only slightly more than half of the normal distribution, this result is not convincing enough to proceed with doing a paired sample t-test. To strengthen this conclusion, the P-P plots and Shapiro-Wilks tests are also included in appendix B, indicating that a t-test is not suitable in this situation and that this research should proceed by doing a nonparametric Mann-Whitney U-test.
The Mann-Whitney U-test will be conducted to test two sets of hypothesis. Firstly, the test will validate whether the observed difference in stock return variance around the introduction of the euro is statistically significant. Hence;
𝐻0 = Stock return variance of Dutch non-financial firms will be the same for the 1993-1997 (pre-euro) period and the 1998-2002 (post-euro) period.
28
The results of this analysis are shown in appendix C. Based on this analysis, the null hypothesis can easily be rejected at the 1% confidence level.
The second hypothesis that will be tested using a Mann-Whitney U-test, is whether the 1998-2002 period is significantly different from the 2003-2007 period. The following hypothesis are set up:
𝐻0 = Stock return variance of Dutch non-financial firms will be the same for the 1998-2002 period and the 2003-2007 period.
𝐻1 = Stock return variance of Dutch non-financial firms will differ between the 1998-2002 and the 2003-2007 period.
The results of this analysis, graphically represented appendix C, are again sufficiently convincing to reject the null hypothesis at the 1% confidence level.
According to these statistical tests, the introduction of the euro was followed by a significant increase of stock return volatility, as 57 companies (or 85,3%) in the sample show an initial increase in volatility. This is remarkable, as it contradicts the hypothesis that a decrease in the amount of relevant foreign currencies should result in lower stock return variability. Furthermore, the risen volatility was only temporary, as more than 72% of the companies that were examined show a decrease in volatility in the 2002-2007 period.
The initial increase in stock return volatility is in line with Bartram and Karolyi (2006), who report the same result for their sample of 218 European non-financial firms. However, the results presented above give more insight in this phenomenon. Bartram and Karolyi (2006) studied the 1990-2001 period, whereas the timeframe studied above continues for six more years. Bartram and Karolyi‟s (2006) conclusion may therefore turn out to be a bit premature, as variances significantly decreased for the 2002-2007 period.
On the other hand, the results are conflicting with Bartov et al (1996), who studied the exact opposite situation. Their study on the effect of the breakdown of the Bretton Woods system on stock return volatility of US firms concluded that this did indeed lead to an increase of stock return volatility. Turning this logic around would have been in line with the hypotheses stated in the methodology section.
29
stock market volatility to increase temporarily. Furthermore, the terrorist attacks of September 11th, 2001, might also have resulted in higher variances, biasing the 1998-2002 results upwards.
Future analysis should uncover whether the 1998-2002 increase in stock return volatility was indeed only temporarily, and whether the decrease that seems to have started from 2003 will eventually result in a lower level of stock market volatility as compared with the pre-euro situation (as would be expected according to the hypotheses presented earlier in this research). 5.2 Stock return volatility – industry level
The variance analysis has also been conducted on an industry level. The composition of each of the twelve portfolios is given in appendix A. The companies are mostly characterized according to the Euronext Industry Classification Benchmark, released by Euronext on November 11th, 2005. However, because some portfolios then consisted of only one or two firms, they are combined if the characteristics of both industries allowed to do so. Because of their distinct industrial characteristics, both Crown van Gelder and Heineken did not seem to fit in any of the portfolios. Therefore, they are being ignored in the analysis.
The results of the analysis are presented in figure 5 below:
1993-1997 1998-2002 2003-2007
Chemicals 0,03% 0,05% 0,05%
Construction & Materials 0,04% 0,05% 0,06% Electronic & Electrical Equipment 0,06% 0,26% 0,08%
Food Producers 0,03% 0,09% 0,04%
Household Goods & Home Construction 0,10% 0,16% 0,09%
ICT 0,23% 0,67% 0,19%
Industrials 0,05% 0,10% 0,06%
Media 0,04% 0,15% 0,05%
Oil 0,05% 0,13% 0,07%
Real Estate Investment 0,25% 0,08% 0,05%
Retailers 0,06% 0,23% 0,03%
Support Services 0,04% 0,06% 0,05%
Figure 5: portfolio variance. The left column shows the industry represented by the portfolio. Variance is calculated over relative weekly stock returns.
30
that due to the relatively low number of companies in some of the portfolios, it is not possible to test for statistical significance. Worth noticing is that especially the ICT and the electronic & electrical equipment portfolios exhibit an abnormally high increase in variance. That these two portfolios are among the portfolios that show the highest increase in variability confirms the proposition made in the individual firm level analysis above, that the increased variability may (partly) be due to the internet bubble. Furthermore, retailers and food producers show a high increase in variability. The explanatory factor(s) behind this pattern for these two portfolios is not clear.
5.3 Stock return volatility – market level
Lastly, the variance analysis will be performed on the aggregated market level. For this, the variance of the returns on the AEX index (main Dutch stock index) for the 1993-1997, 1998-2002 and 2003-2007 periods has been calculated. The results of this are given in figure 6 below.
1993-1997 1998-2002 2003-2007 AEX variance 0,07% 0,26% 0,10%
Figure 6: Average bi-weekly AEX return variance.
Knowing the results of the previous analyses, it will not be very surprising to discover that the average bi-weekly returns of the AEX index have become more volatile for the 1998-2002 period, as compared with the 1993-1997 period. Neither will it be a surprise that for the 2003-2007 period the variance declined again. All evidence suggests that during the period immediately after the introduction of the euro, stock return variance increased quite drastically. Although this contradicts the hypothesis stated in the methodology section, the expected decline is still visible with a lagged effect, as the 2003-2007 has been a period of relatively low variance. The logical explanation that is also supported by the industry level analysis presented above, is that this is due to the explosion of the dotcom bubble. Another likely explanation is that the terrorist attacks of September 11th 2001 caused an increase in stock return volatility. However, the nature of this research does not provide evidence to support this statement.
31
5.4.1 Actual exchange rate exposure – individual firm level
Realizing that the variance analysis remains relatively inconclusive about the effect that the introduction of the euro has had on stock return volatility, the second part of the analysis will focus on measuring the exchange rate exposure of Dutch non-financial companies directly. As mentioned in the methodology section, the accompanying hypothesis states:
The introduction of the euro has led to a decrease in exchange rate exposure of nonfinancial firms in the Netherlands.
Again, the underlying mechanism will be that firms have changed their trading patterns to countries inside the euro-area, in order to benefit from the reduced influence of currency fluctuations. This will in turn reduce the influence that exchange rate fluctuations have on company performances, and thus reduce exchange rate exposure. In order to verify whether this hypothesis is valid or not, the following regression has been conducted:
𝑅𝑖𝑡 = 𝛽𝑖0+ 𝛽𝑚𝑖 𝑅𝑚𝑡 + 𝛽𝑥𝑖 𝑅𝑥𝑡 + 𝛽𝑥𝑗 𝑅𝑥𝑡−1+ 𝛽𝑥𝑘 𝑅𝑥𝑡−2+ 𝜖𝑖𝑡 t = 1, .…, T [1]
In this, 𝛽𝑚𝑖 and 𝑅𝑚𝑡 are the coefficient and the return on the market index, 𝛽𝑥𝑖, 𝛽𝑥𝑗 and 𝛽𝑥𝑘
are the relation coefficients of the TWC with the individual stocks, whereas 𝑅𝑥𝑡, 𝑅𝑥𝑡 −1 and 𝑅𝑥𝑡−2 are the returns of the TWC for time period t. In this, return is defined as [log 𝑃𝑡− log 𝑃𝑡−1]. Again, log values are being use in order to normalize the data. This is in line with recent studies as De Jong et al (2006).
The results of this analysis are graphically represented in figure 7 and figure 8, below.
Significance 1993-1997
N Nositive Negative % of sample
1% 6 3 3 8,82%
5% 8 5 2 11,76%
10% 13 4 11 19,12%
11% 2 1 1 2,94%
Significance 1998-2002
N Positive Negative % of sample
1% 2 2 0 2,94%
5% 7 5 2 10,29%
10% 19 13 6 27,94%
32
Significance 2003-2007
N Positive Negative % of sample
1% 3 1 2 4,41%
5% 8 1 5 11,76%
10% 9 4 5 13,24%
11% 4 4 0 5,88%
Figure 7: significance of firms in the periods measured. The first column in each table represents the level of significance. N = the number of firms at the particular significant level, positive/negative represent the number of firms with either a positive regression coefficient, or a negative regression coefficient. % of sample represents the percentage of firms at a particular significance level, measured over the entire 68 firm sample. The categories are non-cumulative, the 5% category thus only includes firms with a confidence level of between 1,5 and 5,5%. Note: no distinction is made between time lags. This will be discussed in section 5.4.3.
Total significant firms (11% level)
N Positive Negative % of sample
1993-1997 23 9 14 33,82%
1998-2002 29 19 10 42,65%
2003-2007 22 10 12 32,35%
Figure 8: Total number of significant relationships of firms to the exchange rate fluctuations variable, per period. N = the total number of firms per period, positive/negative represents whether the coefficient of the relationship is positive or negative, % represents the total percentage of significant firms per period, on an 11% significance level. Note: a limited amount of companies showed significance at more than one time lag, these firms are represented only once.
The first thing that stands out from the figures above is the relatively high amount of firms that have their stock returns related to changes in the exchange rates (as shown in figure 8). This seems to justify the decision to study the Netherlands because of the open nature of its economy. Where many studies on US companies fail to find convincing evidence on exchange rate exposure, the results above show a significant relationship for almost 43% of the companies studied for the 1998-2002 period. It should be noted, however, that there seems to be a discrepancy between the amount of firms in figure 7 and figure 8. This is due to the fact that a small amount of firms showed significant exchange rate exposure at more than one time interval. In order to give a complete description of the data, all these time intervals are included in figure 6. However, in order to give a fair representation of the amount of firms that are significantly related to the changes in exchange rates, these firms are included only once in figure 8.
33
As mentioned in the methodology section, a fall in exchange rate exposure was expected after the introduction of the euro as a common currency in many European countries. The results presented above seem to confirm this statement at least partly. In the 5 years prior to the introduction of the euro, 33,82% of the sample firms show a significant exchange rate exposure with the trade weighted currency index. Directly after the introduction of the euro (the 1998-2002 period), however, an increase in exposure is visible. Compared with the pre-euro situation, six more firms (or about 10% of the sample) show a significant relationship. The third period then shows the expected decline of exchange rate exposure from 42,65% in the 1998-2002 period to 32,35% in the 2003-2007 period. Although initially these results seem counterintuitive, they are confirmed by the variance analysis presented above. Here also, the expected downturn was only visible after an initial increase in stock return variances. It is not clear what is the cause of this result. As said before, Bartram and Karolyi find the same result but also fail to find convincing reasons to explain this.
5.4.2 Coefficient of the relationship
As mentioned above, the following regression formula has been used in order to make an assessment of the exchange rate exposure of the individual firms:
𝑅𝑖𝑡 = 𝛽𝑖0+ 𝛽𝑚𝑖 𝑅𝑚𝑡 + 𝛽𝑥𝑖 𝑅𝑥𝑡 + 𝛽𝑥𝑗 𝑅𝑥𝑡−1+ 𝛽𝑥𝑘 𝑅𝑥𝑡−2+ 𝜖𝑖𝑡 t = 1, .…, T [1]
In this formula, 𝛽𝑥𝑖, 𝛽𝑥𝑗 and 𝛽𝑥𝑘 represent the coefficient of the relationship of the individual
firms at three different intervals. Moreover, a positive beta indicates that an appreciation of the TWC (or a depreciation of either the guilder or the euro) resulted in an increase in firm value. A negative beta will result in the exact opposite effect. Knowing that a decline in exchange rate exposure has indeed been realized after the introduction of the euro, it is then interesting to investigate whether a positive or a negative relationship exists between the change in exchange rates and the stock returns of a firm, and whether this beta has remained the same after the introduction of the common currency. A histogram of the firm betas is given in figure 9 below.
34
Figure 9: Histograms of coefficients of the regression formulas with a significant relationship. The right handed section represents the mean, minimum and maximum coefficient of each period.
To interpret the effect of outliers on the analysis, z-scores (or standard scores) are calculated. A z-score equals the number of standard deviations that an observation differs from the mean of the sample. Although some authors argue that data with z-scores of 3 (Sincich, 1986) or even 4 (Younger, 1979) should be identified as outliers, for this analysis a stricter z-score of 2,5 will be used. Although a small amount of the data is near 2, none of them comes close to 2,5. Moreover, the boxplots of the distributions of the coefficients in each timeframe are
35
shown in appendix D. As can be seen from the boxplot of the 1993-1997 period, only one observation can possibly be identified as being an outlier. However, because this observation has a z-score of only 2,2, and the number of observations is relatively high, it will be unlikely that this event distorts the analysis. Therefore, it has been chosen to continue the analysis with the full sample.
The results presented in figure 9 above are interesting in two ways. First of all, the nature of the coefficient has on average changed from being negative in the 1993-1998 period, to being positive in the two periods after the introduction of the euro. This means that in the first period, an appreciating TWC will on average have had a negative effect on stock values, whereas in the two succeeding periods the opposite is true. One possible explanation for this might be that Dutch companies before 1998 were mainly reliant on foreign countries for their suppliers. A negative relationship would then be expected, as increasing currencies will increase prices of materials bought abroad. From 1998, however, it is likely that Dutch companies sought alternatives for their foreign currency suppliers when they realized that the euro was being introduced as a common currency for Europe. In order to avoid exchange rate exposure they might have switched to euro-area suppliers, keeping their customers outside the euro area. In this case, a positive coefficient is expected as an appreciating TWC will increase translated earnings and therewith firm value.
Secondly, a trend is clearly visible in which the average exposure of Dutch non-financial firms is getting closer to zero. This is partly due to the fact that the 2003-2007 period shows a much more evenly distributed sample than the other two periods, which makes the average exposure of the significantly exposed firms go down.
5.4.3 Time lag
