The Introduction of the Euro and Foreign Exchange
Rate Risk Exposure, Dutch Evidence
Luuk Grosfeld, 10075232, Economics and Business – Finance and Organisation, Supervisor: J.E. Ligterink, 01-02-2015.
Abstract
This paper examines whether the introduction of the euro changed the exchange risk exposure of 47 Dutch firms. We test for exposure in the separate periods 1994-1997, 1998-2001, 2002-2005 and with a dummy variable in 1996-2001. We also examine whether lagged responses and asymmetric exposures can complement the models used to measure exchange risk exposure. In addition to exposure to 3 trade-weighted exchange rate indices we also test for exposure to different bilateral rates. We find that the number of Dutch firms that are significantly exposed to the indices has decreased after the introduction of the euro. This result is the same for both tests on separate periods, as for the test on one period using a dummy variable. We also find that asymmetric exposure and lagged exposure are of great explanatory power besides the current symmetric exposure. Furthermore we find that more firms are exposed to bilateral rates than to the indices. We also find that the number of bilateral rates that are included in the index does not significantly alter the exposure of firms to that index. Finally we find that, although overall exposure has decreased, the number of firms with asymmetric exposure has increased after the introduction of the euro.
Hierbij verklaar ik, Luuk Grosfeld, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.
Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.
1. Introduction
Advocates of the introduction of the euro as a common currency for the countries of the EMU have stated that it would reduce the corporate costs of managing exchange risk for European firms (Wildman, 1997). It is important to find out whether this claim was valid. To do this we need to know how firms are influenced by exchange rate changes and whether this has changed after the introduction of the euro. The most common way to do this is measuring the exposure of the returns on firm stock value to changes in exchange rates. This is called the exchange risk exposure (Adler and Dumas, 1984). There have been some studies on exchange risk exposure around the introduction of the euro. Bartram and Koralyi (2006) for example have measured the exchange risk exposure of 3921 EMU and non-EMU firms in the period 1992-2001. Furthermore Nguyen et al. (2007) have measured the exchange risk exposure of French companies in the periods 1996 and 2000. Where the former finds weak support for a change in exchange risk exposure of firms after the introduction of the euro the latter finds strong support.
This study contributes to past research in multiple ways. In the first place, we analyse the exchange risk exposure for a sample of 47 Dutch firms in the period around the introduction of the euro. The Dutch market is one of the most open economies in the world so foreign trade en hence foreign currencies are of major influence on the performance of companies (De Jong et al., 2006). In the second place, we test for exposure to different trade-weighted exchange rate indices, based on the euro as well as on the guilder, and with a different number of foreign currencies included. In the third place, we also test for the exposure to different bilateral rates to see whether this can complement the exposure to the trade-weighted exchange rate indices. In the fourth place, we also test for time-lagged and asymmetrical exposures around the introduction of the euro. Research from Dewenter et al. (2004) and Bartov and Bodnar (1994) has shown that these exposures can have major explanatory power in addition to current symmetric exposures. Finally, besides using a model with a post-euro dummy variable for the period 1996-2001, we also use a model with the separate sub-periods 1994-1997 and 1998-2001. This is done to see whether different methodologies yield different results. Furthermore we also include a 2002-2005 period to investigate whether the exposures have changed in the period 5 years after the introduction of the euro.
Using the standard methodology introduced by Jorion (1990) we test for the exposure of the monthly stock returns of 47 Dutch firms to the changes in 3 trade-weighted exchange rate indices in the periods 1994-1997 and 1998-2001. We find that in the period before the euro 21%-34% of the firms were exposed and in the period after the euro 13%-28%. Furthermore we find that the 1-month lagged indices are of equal explanatory power as the current indices. The exposure in the 2002-2005 period is not significantly different from the exposure in the 1998-2001 period. Testing for exposure to bilateral rates in these same periods we find that 35 firms are exposed to these rates where only 17 firms were exposed to the indices. Hence exposure to bilateral rates can complement the exposure on trade-weighted exchange rate indices. Using a dummy variable for the introduction of the euro in the period 1996-2001 we find again that the number of firms that show exposure has decreased after the introduction of the euro. However this difference is less profound than when we compared the periods separately. Furthermore we find that in all periods asymmetric exposure is of significant explanatory power in addition to symmetric exposure. We found an increase of 20%-30% in exposures when testing for asymmetric exposure besides symmetric exposure. We also find that this asymmetric exposure has increased with 54% after the introduction of the euro.
We conclude that overall the number of firms with exchange rate exposure has decreased after the introduction of the euro. However different models and different trade-weighted exchange rate indices used have varying results. We also conclude that both lagged and asymmetric exposures must be taken into account next to current symmetric exposure when testing for exchange risk exposure.
The remainder of this paper is organised as follows. Section 2 discusses the theory and evidence of exchange risk exposure. Section 3 describes the methodology and data. Section 4 analyses the results and robustness tests and section 6 concludes.
2. The theory and empirics of exchange risk exposure 2.1 Theoretical grounds for exchange risk exposure.
Economic analysis implies that exchange rate movements can affect the value of a firm via either the future cash flows of a firm’s operations or by affecting the discount rate used. The sensitivity of firm value to exchange rate fluctuations is called
exchange risk exposure (Adler and Dumas, 1984). A firm can be exposed to exchange risk through various channels. A firm can experience decreasing cash flows of their foreign sales, in terms of the domestic currency, when the applicable foreign currency depreciates and vice versa. However this firm can also experience positive economic results of the same appreciation of domestic currency due to cheaper import of raw materials it uses for its production. Obviously, the effect of changing exchange rates on cash flows depends on which currency is used to pay for the imports and exports. According to Muller and Verschoor (2006), the cash flow volatility resulting from these changing exchange rates enhances the likelihood that a firm will need to access capital markets, which will be reflected in firm value. Below follows an analytical partition of exchange rate exposure in different components.
An important source of exchange rate risk to firm value is the possible existence of assets and liabilities abroad. Foreign subsidiaries or foreign debt denominated in other currencies can have different values depending on the current exchange rate. As a result the balance sheet of a firm might change and this affects firm value. Sercu and Uppal (1995) have called this form of exposure translation exposure. Except from international trading or operating firms, even a domestic firm with no foreign activities can be exposed to exchange risk. This form of composure is called economic or competitive exposure and has a longer time horizon. Economic exposure is the affect that exchange rate variations have on the relative prices of goods in different countries. These relative price changes have influence on the foreign competition, the economic environment and the future growth and development options of a firm (Muller and Verschoor, 2006). Besides economic exposure and translation exposure firms can also experience transaction exposure. This happens when the a value of a future cash flow from a contract denominated in a foreign currency changes between the time when the firm commits to the contract and the time when the transaction is made.
Transaction exposure and translation exposure are forms of direct exposure. They have a short-term time horizon so effects are directly measurable. The management of a company can anticipate on these exposures by making use of Foreign Currency Derivatives to hedge against exchange rate risk. Hence these exposures can effectively be managed by well-structured hedging strategies, according to Nydahl (1999). Economic exposure on the other side is indirect. And because of the longer time horizon and the more complex relationship between
changes in exchange rates and competitiveness, harder to estimate. This is why Williamson (2001) states competitive exposure is the most difficult to effectively hedge against.
Al these different components of exchange risk exposure indicate that the sensitivity of firm value to the changes in exchange rates depends on a large number of parameters. Muller and Verschoor (2006) name the nature of activities the firm is involved in, its import and export structure, the amount of foreign operations, the currency denomination of its competitors and the competition in faces in its input and output markets, besides many others. There is still no consensus in theoretical literature on which parameters are the most important. However the amount of theoretical arguments explaining the origins of exchange risk exposure is considerable.
2.2 Measuring exchange risk exposure.
Contrasting the theoretical predictions, empirical investigations have only found limited support for a significant coefficient indicating a relationship between changes in firm value and changes in exchange rate. A possible explanation for the lack of strong empirical evidence might be that firms successfully manage to eliminate their foreign currency risk by hedging as Bartov and Bodnar (1994) suggest. However it is doubtful that firms are able to hedge against their indirect long-term economic exchange risk exposure so this might not be the whole explanation. Therefore researchers are trying to develop ever more comprehending and specified estimation models to find empirical evidence for the exchange risk expected by theory.
To measure the exposure of firm value to exchange rate Adler et al. (1984) suggest using stock returns and exchange rate changes to construct this simple regression model:
Rit = αi + φiθt + εit
Here Rit designates the total return on the stock of firm i in period t, θt is the exchange
rate change in period t, αi the constant term, φi the coefficient of firm i’s stock returns
to changes in exchange rate and εit the white noise error term. Hence the exchange
rate exposure is designated by the coefficient φi. Firms should have a positive φi
coefficient to exchange rate changes if they are net-exporters and a negative if they are net-importers when the exchange rate is defined as units of domestic currency per unit of foreign currency.
A problem with the model developed by Adler et al. (1984) is that is measures the total correlation of stock i’s returns with exchange rate movements. It is very well possible that other macroeconomic variables have covariation with both exchange rate changes and stock i’s returns. These macroeconomic variables would influence the market as a whole and thus also stock i’s return. If these variables are omitted the estimate of exchange rate exposure of the individual firm i will be exaggerated. Therefore Jorion (1990) came up with an addition to the model to measure firm i’s specific exchange rate sensitivity by controlling for market effects:
Rit = αi + βiRmt + φiθt + εit
Here Rmt is the overall stock market benchmark return in period t and βi is firm i’s
return sensitivity to market risk. Hence the firm specific exchange rate exposure is the residual between the firm’s total exposure and the market’s exposure adjusted by the firm’s market beta βi. Jorion’s model became the standard model used by subsequent
research.
Contrasting this capital market model Walsh (1994) proposed a current cash flow approach. Where the capital market model is essentially forward looking by incorporating all future cash flows in its stock value variable, Walsh’s approach is orientated on current cash flows. By looking at current cash flows and their exposure to exchange rate risk it is possible to differentiate between transaction and economic exposure. This decomposition of exchange rate exposure is an advantage of the current cash flow model. However it is necessary to measure cash flow exposure of cash flows net of hedging. That data is harder to obtain so this makes it more difficult to detect. Furthermore, as Marin and Mauer (2005) point out the model does not measure the total impact of currency movements on firm value because it does not include expectations about the future cash flows.
Jorion’s capital market model also has some weaknesses. If for instance all firms in the market benchmark are affected in the same way by a currency depreciation/appreciation, the model shows no significant residual exchange rate exposure. This is because the firm specific exchange rate exposure is the residual of the market risk. So when in reality the every firm’s stock value in the market is affected, the model does not specify this, as Glaum et al. (2000) point out. Besides this careful consideration is needed when choosing the appropriate benchmark for the market risk factor. Different specifications can have immediate implications for the sign, magnitude and significance of the estimation for exchange rate exposure. A
value-weighted portfolio will have an over-representation of larger firms that are more likely to have significant exchange rate exposure (Bodnar and Wong, 2003). Because of this positive bias in the exposure coefficient due to the weight given to larger and thus more export-orientated firms, Bodnar and Wong suggest to use equal-weighted benchmark portfolios instead.
Just as the appropriate benchmark index, the variable for the exchange rate needs consideration as well. Most of the empirical research makes use of a single proxy for all the exchange rate fluctuations that are likely to affect the values of the firms studied. In this proxy multiple bilateral exchange rates are brought together in a single trade-weighted exchange rate index. Studies that employ this method are for example Jorion (1990), Bodnar and Gentry, (1993), Bartov and Bodnar (1994), He and Ng, (1998), Nydahl (1999) and Dominquez and Tesar (2001). Furthermore some have constructed company specific trade-weighted exchanger rate indices (De Jong et al., 2006). Miller and Reuer (1998) explain that the use of this trade-weighted exchange rate indices neglects negative correlation between the movements of exchange rates. Therefore divergent exchange rate changes will cancel each other’s effect and the exposure of a firm is underestimated. Furthermore exchange rate exposure will be underestimated when the firm is exposed to only a few currencies from the trade-weighted index. A possible solution for this defect is using multiple bilateral exchange rate variables for the most important currencies for the firm instead. Dominquez and Tesar (2001) for example show that in their sample a lot of firms that are exposed to one or more bilateral exchange rates are not exposed to the trade-weighted index. However Bartram (2002) contradicts this when he shows that trade-weighted indices do not understate the exposure of firms in comparison to bilateral exchange rates. Furthermore Fraser and Pantzalis (2004) show that the more currencies are included in the trade-weighted exchange rate index, the more firms of their sample show significant exchange risk exposure. This can be seen as indirect evidence for the notion that companies are exposed to exchange rate fluctuations of currencies from countries where they are not directly operating.
Because of information asymmetries it might be possible that investors are not instantly able to evaluate the relationship between a firm’s future cash flows and an unexpected exchange rate shock. Hence they will revalue the firm’s stock only after they learned how the performance of the firm was influenced by the shock. This can lead to a lagged relationship between stock returns and exchange rate changes.
Therefore Bartov and Bodnar (1994) suggest including a lagged exchange rate change variable next to the current exchange rate change variable in the capital market model. They find that the lagged exchange rate does have more explanatory power on the stock returns than the current exchange rate has in their sample of US firms. However He and Ng (1998) for Japanese multinationals and Nydahl (1999) for Swedish multinationals also test for a lagged response but find contradicting results.
Another source of possible errors in measuring exchange risk exposure is the fact that the economic environment, the competition, and the operational structure of a firm are constantly changing. Hence a firm’s exchange risk exposure must be time varying to. Jorion (1990) among others separates his time series in multiple sub-periods. The time varying nature of exchange risk exposure is thus integrated when the periods are small enough. Next to the periods the observation frequency is also an area of debate. Most authors use monthly data but due to market inefficiencies authors like Chamberlain et al (1997) suggest using daily data, where others like Dominguez and Tesar (2001) suggest using intervals longer than 1 month. There is no consensus however which observation frequency works best but the market efficiency hypothesis suggests that exposure should not depend on which time horizon is used.
The market value of a firm might react asymmetrically to positive and negative exchange rate changes. This is called sign asymmetry. Furthermore large and small exchange rate shocks can have different asymmetrical impacts. This is called size symmetry. Dewenter et al. (2004) explain it is hard to imagine a firm’s stock price reacting linear and symmetrical to exchange rate fluctuations. Therefore Di Iorio and Faff (2000) use dummy variables in their model to make partitions in the sign and size of exchange rate movements. They however find weak support for the explanatory power of asymmetries. On the other side Muller and Verschoor (2004) show that including asymmetries does increase the significance of exchange risk exposures in samples. They state that especially the size asymmetries have important explanatory power in measuring exchange risk exposure.
2.3 Empirical evidence for exchange risk exposure.
Research on exchange rate exposure has come up with different results depending on the time period, sample of companies and models used. The table below gives a summary of some of the relevant findings in empirical literature.
Table 1: Empirical findings in literature
Author Sample Period Observation
Frequency Method Results Jorion (1990) US multinationals
1971-1987 Monthly TWI 15 out of 287 companies** Bodnar and Gentry (1993) 2-digit industry portfolios: US, JPN, CA 1978-1993 Monthly TWI 9/39 industry portfolios** He and Ng (1998) Japanese 10% export-ratio multinationals
1978-1993 Monthly TWI 26% positive from 171 companies** Nydahl (2001) Swedish companies
1992-1997 Weekly TWI & single FX’s 26% from 47 companies** Dominquez and Tesar (2001) 2000 firms of 8 industrial non-US countries
1980-1999 Weekly TWI & single FX’s Significant amounts all countries** De Jong et al. (2006) Dutch companies 1994-1998 Bi-weekly Specific TWI & single FX’s Over 50% of 47 companies**
The results are the number and percentage of firms that show significant (5%) exchange rate exposure. TWI means that a trade-weighted index is used as proxy for the relevant exchange rates.
In regard to which factors are determinants for the companies that show exchange risk exposure, empirical research has identified different determinants. Jorion (1990) shows that the degree to which companies are involved in foreign activities is an important explanatory variable. De Jong et al. (2006) have linked Foreign Currency Derivative usage to exchange risk exposure. Others are industry (Shin and Soenen, 1999), liquidity and cash dividends (Chow and Chen, 1998), industry structure (Marston, 2001) and firm size (Chow et al., 1997).
2.4 Literature on exchange risk exposure and the introduction of the euro.
The introduction of the euro in January 1999 is interesting phenomenon for research on exchange risk exposure. Bartram and Karolyi (2006) measure the stock market volatility, stock return variances, market Beta’s and exchange risk exposure for 3921 companies from 1992-2001. Their sample includes firms from Japan, the US and 18 European Eurozone and non-Eurozone countries. They use Jorion’s capital market model with dummies for post-euro market Beta’s and post-euro exchange risk exposure. The weekly trade-weighted exchange rate index per country is from the database of the Bank of England (with IMF weights). For the Eurozone countries this index is denominated in the pre-euro currency, so they use the Dutch Guilder index for the Netherlands.
Their main finding is that for the majority of companies the market Beta’s have decreased after the introduction of the euro. Just a small fraction of the firms show significant exchange risk exposure in the whole period (between 1.6% and 5.2% for positive exposure and between 5.7% and 7.3% for negative exposure). Hence it seems that overall firm value decreases when the home currency depreciates. After the introduction of the euro they find that only 0.8%-3.3% of the positive exposed firms show significant changes and only 2.1%-8.1% of the negative exposed firms. These changes in exposure are of the opposite sign as the exposure beforehand for most EMU firms. This means overall exposure of EMU firms decreases slightly after the introduction of the euro. However most firms do not show significant exchange risk exposure or a significant change in exchange risk exposure at all. Because of their finding that market Beta’s have significant decreased after the introduction of the euro they suggest that an important part of exchange risk exposure lies in change in sensitivity to the market index.
Another paper on exchange risk exposure around the introduction of the euro from Nguyen et al (2007) focuses on French companies. From their sample of 99 firms 32% shows significant exposure in 1996 against 11% in 2000. They do not make use of a dummy variable for pre- and post-euro data but simply run a separate regression for the periods 1996 and 2000. Hutson and Driscoll (2010) also perform research on Eurozone and non-Eurozone companies around the introduction of the euro. They find a significant larger amount of companies exposed in Eurozone companies in comparison with European non-Eurozone companies. They use trade-weighted exchange rate indices from the IMF based on the euro for the different
Eurozone countries. Hutson and Driscoll perform separate regressions for the periods 1990-1998 and 1999-2008 after which they find exchange risk exposure has significantly increased after the introduction of the euro. They also find a decrease in market-level exchange risk exposure. Therefore they suggest a shift from systematic to firm-specific exchange risk exposure has occurred after the introduction of the euro.
2.5 Contribution of this research.
This research contributes in five ways to past research done on exchange risk exposure of companies around the introduction of the euro. In the first place, we specifically chose to examine the exchange rate exposure of firms form the Dutch market. The Dutch market is as De Jong et al. (2006) say one of the most open economies in the world so foreign trade and hence foreign currencies are of major influence on the performance of companies. The Netherlands does not only trade with most of his fellow EMU countries but also with the rest of the world so it should be possible to expose the differences made by the introduction of the euro clearly. Moreover examining one particular market instead of multiple will avoid the generalisation errors made by applying European-level parameters and findings to specific markets.
In the second place, past research made use of different trade-weighted exchange rate indices while not specifically account for it. Batram and Karolyi (2006) for example use a Dutch Guilder-based index for the Netherlands, where Hutson and Driscoll (2010) use a euro-based index. This difference does matter because the Dutch Guilder-based index includes trade weights from the Netherlands with EMU countries but ends after December 2001. This is for the simple reason that the Guilder was brought out of circulation after 2001. In contrast the Euro-based index includes data from before the introduction of the euro. However the exchange rates applied for the construction of this index before the euro exists are based on a proxy for a ‘theoretical euro’ (Buldorini et al., 2002). This proxy uses weights for the exchange rates of different pre-euro currencies based on their share in the trade of the Eurozone as whole vis-à-vis the rest of the world. As a consequence the intra-Eurozone trade is neglected in modelling this proxy. Hence there is a major difference in trade weights between Guilder-based and euro-based trade-weighted indices that must be accounted for in using them. Therefore this research uses both indices for separate regressions
for the period 1996-2001 to see whether these differences significantly matter. Besides these two indices we also use a third broad index to investigate whether the inclusion of additional bilateral exchange rates has a significant effect. Fraser and Pantzalis (2004) for example found that the more currencies were included in the trade-weighted exchange rate index, the more firms show significant exchange risk exposure.
In the third place, we also test for the exchange risk exposure of Dutch companies to single bilateral exchange rates of major trading partners. Exchange risk exposure to pre-euro currencies of major Eurozone can complement the exposure to the trade-weighted index. This is done because, as stated above, divergent bilateral exchange rates might cancel each other out when brought together in indices. Other research on exchange risk exposure around the introduction of the euro failed to include this.
In the fourth place, we test for lagged response to exchange rate changes around the introduction of the euro. As Bartov and Bodnar (1994) have found, the lagged response can be as important or even more important than the current response. Besides the lagged response we also test for an asymmetric response to exchange rate changes. More specific, we test whether the returns of the firms respond differently to a depreciation of the home currency than to an appreciation of the home currency and whether this changed after the introduction of the euro. Koutmos and Martin (2003), for example found that 40% of the exposure that their sample of firms from the UK, USA, Germany and Japan showed was asymmetric. Both the lagged response and the asymmetric response were not included in previous research on exchange rate exposure and the introduction of the euro (Batram and Karolyi, 2006; Nguyen et al., 2007; Hutson and Driscoll, 2010)
Finally, besides using a model with a post-euro dummy variable for the period 1996-2001 we also use a model with sub-periods. So we do different regressions for the sub-periods 1994-1997 and 1998-2001 that are shorter in length than the 1996-2001 period. This is done to control for the time-varying nature of exchange risk exposure. Furthermore we include a 2002-2005 period to investigate whether the exchange risk exposure of companies might have taken longer to adjust to the introduction of the euro.
3. Research design and data 3.1 Research design.
We follow previous research in this area by estimating exchange risk exposure using the following model:
Rit = β0i + β1iRmt + β2iFXt + εit for t = 1, … T (1)
Where Rit designates the total return on the stock of firm i in period t, Rmt is the return
on the equal-weighted market portfolio for period t and FXt is the return on the
trade-weighted exchange rate index or bilateral exchange rate measured in units of foreign currency per guilder/euro in period t. The regression coefficient β2i measures the
exchange risk exposure of firm i.
We use this model to estimate the exchange risk exposure of firms in the 1994-1997, 1998-2001 and 2002-2005 periods. The first and second of these periods are chosen to measure the exchange risk exposure of companies in a 4-year period just before the introduction of the euro and a 4-year period just after the introduction of the euro. The period is 4 years and not more because of the possible time-varying nature of exchange risk exposure (Jorion, 1990). We follow Batram and Karolyi (2006) in using 1 January 1998 as the effective event date of the introduction of the euro. This way we take in the fact that investors might have anticipated the introduction of the euro. The last 4-year period of 2002-2005 is chosen to measure whether there is a different in exposure of companies in de period directly after the introduction of the euro and 4 years after this period. This is done because it might be the case that investors perceived that companies needed a longer time to adjust to the introduction of the euro.
For the FXt variable we use the guilder-based trade-weighted exchange rate
index, the euro-based trade-weighted exchange rate index and the bilateral exchange rate of the guilder versus the 5 largest trading partners of the Netherlands in separate regressions. For the estimation of the exchange risk exposure in period 1996-2001 we use the following model:
Where Deurot is a dummy variable that equals 1 after the introduction of the euro and 0
before the introduction of the euro. This period is 6 years long to include the 3 years just before the theoretical introduction of the euro (1998) that showed the most exchange rate changes as well as 3 years after this introduction. A period of longer than 6 years would decrease the quality of the results due to the aforementioned time-varying nature of exchange risk exposure.
We also test for a lagged relation between stock returns and exchange rate changes as Bartov and Bodnar (1994) did. To do this we complement model (1) and model (2) with a 1-month time lag and a 2-month time lag coefficient:
Rit = β0i + β1iRmt + β2iFXt + β3iFXt-1 + β4iFXt-2 + εit for t = 1, … T (3)
Here the coefficients β3 and β4 measure the 1-month lag and the 2-month lag
respectively. Model (2) is complemented in the same way.
To test for a sign asymmetric response to exchange rate changes we follow Koutmos and Martin (2003) in complementing model (1) with a dummy variable:
Rit = β0i + β1iRmt + β2iFXt + β3iFXtDnegative,t + εit for t = 1, … T (4)
Here the dummy variable Dnegative equals 1 if the change in exchange rate is negative
(a depreciation of the home currency) and 0 if the change in exchange rate is positive (an appreciation of the home currency). Hence if there is an appreciation, β2 measures
the exposure of the firm. If there is a depreciation of the home currency, β3 measures
the exposure of the firm to this depreciation that is not already measured by β2. This
means that if β3 is significant, the firm is asymmetric exposed regardless of the sign of
the coefficient. We test for sign asymmetric exposure in the periods 1994-1997, 1998-2001 and 2002-2005. Unfortunately we were not able to test for size asymmetry due to data constraints.
To test to for robustness we run all regressions with bilateral rates again with weekly data. Due to data constraints it is not possible to do additional weekly data robustness tests for the trade-weighted exchange rate indices. Hence we could only test the robustness of the regressions with the bilateral rates. However these results should be a good indicator for the robustness of the regressions with trade-weighted
exchange rate indices. Because of the possibility of autocorrelation and heteroskedasticity we make use robust standard errors.
3.2 Data sources.
The focus of this research is on the exchange rate exposure of Dutch firms listed on Euronext Amsterdam. Datastream provides monthly and weekly data on the stock-price of these 75 companies. We follow previous research in this field by excluding all financial firms due to their abnormal activities using foreign currencies. Firms that were not listed during the whole of the 1994-2005 period are also excluded. This leaves us with a sample of 47 firms.
For the benchmark we use the Datastream provided equal-weighted non-financial Dutch market portfolio. We have chosen an equal-weighted benchmark instead of a value-weighted benchmark because of the overrepresentation of large firms in the last one. This could bias results due to the likeliness that large firms have foreign operations and subsidiaries, which will influence their exchange risk exposure.
The bilateral exchange rates are from Thomson Reuters and provided by Datastream. They are measured in units of home currency per unit of foreign currency. Hence an increase in the exchange rate will mean an appreciation of the home currency. We use the bilateral rates of the 5 biggest trading partners of the Netherlands. In the pre-euro period these are Belgium, Germany, France, the United Kingdom and the United States. In the post-euro period the countries that have switched to the euro are replaced with China, Japan and Russia. The trade figures are taken from EUROSTAT.
We use 3 different effective exchange rate indices. The first one is calculated by the Bank of England using IMF data and is based on the Dutch Guilder versus 21 other currencies. This index stops after 2001 because then the Guilder was officially taken out of circulation. The other 2 indices are both calculated by the Bank of International Settlements and are based on the euro and Dutch trade figures. For the pre-euro period these indices make use of a ‘theoretical euro’ based on the currencies and trade figures of the euro area countries, as is explained by Buldorini et al. (2002). The first one uses a narrow group of other currencies (27) and the last one uses a broad group of other currencies (52). These effective exchange rate indices assign
different weights to bilateral rates based on the trade of the Netherlands with the relevant country. Hence they are trade weighted exchange rate indices.
We choose to use nominal effective exchange rates (NEER) indices and bilateral rates instead of real effective exchange rate indices and bilateral rates as is done by most other research in this area. As Bodnar and Gentry (1993) state this is not likely to bias the result due to the high correlation between real and nominal rates in low inflation countries.
The monthly returns are calculated from the last workday of the month. We are aware of possible end-of-the-month effects as in Williamson (2001), but the data on the effective exchange rate indices was only available on an end of the month basis. The weekly returns are calculated from Wednesday to Wednesday to prevent possible end-of-the-week effects. Table 2 shows the summary statistics of the returns/changes of all the data.
Table 2: Summary statistics of data N. FIRMS GLD NEER EURO NEER NAR. EURO NEER BR. GLD/ USD GLD/ GBP GLD/ F.FR. GLD/ B.FR. GLD/ DEM EUR/ USD EUR/ GBP EUR/ JPY EUR/ CNY EUR/ RUB Minimum -.2830 -.0135 -.0170 -.0158 -.0596 -.0499 -.0131 -.0030 -.0023 -.0435 -.0342 -.0639 -.0483 -.0343 Median .0437 -.0011 -.0005 .0002 .0047 -.0004 -.0012 0 -.0003 .0037 .0048 .0053 .0022 .0046 Maximum .3444 .0205 .0263 .0318 .0604 .0413 .0146 .0023 .0021 .0607 .0499 .0551 .0664 .0676 Mean .0099 -.0005 .0000 .0009 -.0004 -.0026 -.0004 -.0001 -.0001 .0062 .0026 .0043 .0064 .0050 Standard Deviation .0956 .0067 .0072 .0090 .0255 .0214 .0058 .0009 .0009 .0261 .0173 .0241 .0283 .0246
This table shows the summary statistics of the data used in this research. The minimum, median, maximum, mean and standard deviation of all the changes in the indices and bilateral rates are displayed. For the sample of 47 firms the averages of the returns for each statistic are given. The period for the sample of firms and for the euro-indices is 1994-2005, for the guilder-index 1994-2001, for the guilder bilateral rates 1994-1997 and for the euro bilateral rates 2001-2005.
4. Results
In this section we discuss the results of our data analysis on exchange risk exposure of Dutch firms around the introduction of the euro. We begin by examining the exposure
of our sample of 47 Dutch firms to the 3 different NEER-indices in the periods 1994-1997 and 1998-2001. Then the exposure against the 2 euro based NEER-indices in the 2002-2005 period is examined. We also add the results of the exposure against the time-lagged NEER-indices to test for robustness. By comparing the results of the different periods we can see whether the exposure has changed after the introduction of the euro. After this we examine the exposure of the sample of Dutch firms against 5 bilateral rates both in the guilder period (1994-1997) and in the euro period (2002-2005) to see whether they can complement the previous results of the NEER-indices. Here we add the time-lagged results as well. Then we examine the exposure of our sample of 47 Dutch firms to the 3 NEER-indices in the period 1996-2001 using a dummy variable for the introduction of the euro to see whether these results affirm the findings of the regressions done on the separate periods. After that we test for sign asymmetric exposure to the broad euro NEER-index in the periods 1994-1997, 1998-2001, 2002-2005. At last we test for the robustness of the results of the exposure to bilateral rates using weekly instead of monthly data.
4.1 Exposure to Nominal Effective Exchange Rate indices.
Table 3 presents the result of the monthly stock returns of 47 Dutch firms against the guilder-based, narrow euro-based and broad euro-based NEER indices in the period 1994-1997 and the results against the 1-month and 2-month time lag of these indices.
Table 3 shows that only 15% of the Dutch firms show significant exposure at a 10%-level to the guilder-based NEER-index and only 13% show significant exposure to both narrow and broad euro-based indices in the 1994-1997 period. The mean exposure coefficient is negative for all three indices varying from -0.240 for the guilder-based index to -1.029 and -0.823 for the euro-based indices. This means that if the guilder or ‘theoretical euro’ appreciates with 1% against the basket of currencies used in the NEER-indices the mean stock value of the firms decrease with respectively 0.24%, 1.03% or 0.82%.
Another important result is that for the both the narrow and broad euro-based indices significantly more firms are exposed to the 1-month time lagged index than to the current index (11 against 6 in both cases). However firms are more exposed to the current guilder-based index than to the 1-month time lagged index. The significant exposed firms in the guilder-based index are significantly more positive (5) than
negative (2) exposed. For both broad and narrow euro-based indices the significant exposed firms are equally spread between positive and negative exposure.
Table 3: NEER-indices 1994-1997 GUILDER-NEER GUILDER-NEER LAG-1 GUILDER-NEER LAG-2 EURO-NEER NARROW EURO-NEER NARROW LAG-1 EURO-NEER NARROW LAG-2 EURO-NEER BROAD EURO-NEER BROAD LAG-1 EURO-NEER BROAD LAG-2 Minimum -5.439 -4.977 -4.256 -7.774 -7.548 -2.561 -5.396 -6.071 -2.171 Median -1.045 -1.362 -0.477 -1.231 -2.294 1.272 -1.140 -1.923 1.579 Maximum 3.068 2.623 9.262 2.651 4.718 6.529 2.831 1.833 5.106 Mean -0.240 -1.209 0.279 -1.029 -1.985 1.083 -0.823 -1.861 0.984 Standard deviation 2.214 1.857 2.998 2.414 2.967 2.280 2.010 2.135 2.125 Number of firms with significant exposure*:
Significant at 1% 1 0 1 2 2 1 2 2 1 Significant at 5% 2 2 4 2 6 2 3 8 3 Significant at 10% 7 4 6 6 11 3 6 11 4 Positive at 10% 5 0 4 3 2 2 3 2 3 Negative at 10% 2 4 2 3 9 1 3 9 1
This table represents the results of the regression analysis, using model (1) and (3). The minimum, median, maximum, mean and standard deviation of the β2’s of the different NEER-indices on the stock
returns of the 47 firms are displayed at the top half of the table. The bottom half of the table represents the number of firms that showed exposure and at which level of significance and whether it was positive or negative exposure. (*): The number of firms with exposure to both the current and the 1-month time lagged index are for the guilder-based 1, euro-based narrow 1, euro-based broad 2.
Table 4 presents the result of the monthly stock returns of 47 Dutch firms against the guilder-based, narrow euro-based and broad euro-based NEER indices in the period 1998-2001 and the results against the 1-month and 2-month time lag of these indices.
Table 4 shows that only 2% of the Dutch firms show significant exposure at a 10%-level to the guilder-based NEER-index and that 15% show significant exposure to the narrow based index and 19% show significant exposure to the broad euro-based index the 1998-2001 period. The mean exposure coefficient is slightly positive for the guilder-based (0.135) and narrow euro-based index (0.083) and slightly
negative for the broad euro-based index (-0.137). This means that an appreciation or depreciation of the euro against the basket of currencies in the indices has no significant effect for the mean stock value of all the firms in the sample. However the minima and maxima show that some of the firm’s stock values do change significant when the indices do. Furthermore firms are again more exposed to the 1-month time lagged NEER indices in the case of the guilder-based and narrow euro-based indices. However more firms are exposed to the current euro-based broad index than to the 1-month time lagged. The explanatory power of the 2-1-month time lag is low in comparison to the other two, as was the case in the 1994-1997 period.
Table 4: NEER-indices 1998-2001 GUILDER-NEER GUILDER-NEER LAG-1 GUILDER-NEER LAG-2 EURO-NEER NARROW EURO-NEER NARROW LAG-1 EURO-NEER NARROW LAG-2 EURO-NEER BROAD EURO-NEER BROAD LAG-1 EURO-NEER BROAD LAG-2 Minimum -4.351 -5.082 -5.092 -3.426 -4.092 -6.982 -3.323 -3.117 -6.366 Median -0.291 -0.625 0.056 -0.231 -0.462 0.060 -0.343 -0.394 0.025 Maximum 3.068 9.339 6.107 3.225 7.928 7.093 2.437 7.591 5.229 Mean 0.135 -0.011 -0.069 0.083 -0.114 -0.028 -0.137 -0.035 -0.156 Standard deviation 2.491 2.541 2.280 1.759 2.339 2.260 1.598 1.984 1.791 Number of firms with significant exposure*:
Significant at 1% 0 1 0 1 1 0 0 1 0 Significant at 5% 0 3 0 3 3 1 4 4 1 Significant at 10% 1 5 2 7 8 3 9 5 3 Positive at 10% 0 3 1 4 2 1 5 1 2 Negative at 10% 1 2 1 3 6 2 4 4 1
This table represents the results of the regression analysis, using model (1) and (3). The minimum, median, maximum, mean and standard deviation of the β2’s of the different NEER-indices on the stock
returns of the 47 firms are displayed at the top half of the table. The bottom half of the table represents the number of firms that showed exposure and at which level of significance and whether it was positive or negative exposure. (*): The number of firms with exposure to both the current and the 1-month time lagged index are for the guilder-based 0, euro-based narrow 2, euro-based broad 2
Table 5 presents the result of the monthly stock returns of 47 Dutch firms against the narrow euro-based and broad euro-based NEER indices in the period 2002-2005 and the results against the 1-month and 2-month time lag of these indices. Obviously the guilder-based index is omitted because the guilder was taken out of circulation at the end of 2001.
Table 5: NEER-indices 2002-2005
This table represents the results of the regression analysis, using model (1) and (3). The minimum, median, maximum, mean and standard deviation of the β2’s of the different NEER-indices on the stock
returns of the 47 firms are displayed at the top half of the table. The bottom half of the table represents the number of firms that showed exposure and at which level of significance and whether it was positive or negative exposure. (*): The number of firms with exposure to both the current and the 1-month time lagged index are for the euro-based narrow 2 and for euro-based broad 2.
Table 5 shows again that firms are exposed to both the current and the 1-month time lagged indices. Another important result is that the mean exposure coefficient for both the narrow and broad euro-based, current and lagged indices are positive which means that an appreciation of the euro means that the stock value of the firms will on average appreciate to.
So what can we learn when we compare table 3, table 4 and table 5? One of the main results is that for all 3 indices an amount of firms is exposed to the current
EURO-NEER NARROW EURO-NEER NARROW LAG-1* EURO-NEER NARROW LAG-2 EURO-NEER BROAD EURO-NEER BROAD LAG-1* EURO-NEER BROAD LAG-2 Minimum -5.162 -4.290 -4.924 -4.380 -3.408 -4.314 Median -0.955 1.253 0.581 -0.817 1.019 0.503 Maximum 7.328 8.357 7.646 5.205 5.810 4.771 Mean 0.599 1.859 -0.203 0.589 1.395 -0.358 Standard deviation 2.576 2.661 2.199 1.924 1.952 1.649 Significant at 1% 1 3 0 1 2 0 Significant at 5% 7 6 0 5 6 0 Significant at 10% 7 8 3 8 7 2 Positive at 10% 0 7 2 0 6 1 Negative at 10% 7 1 1 8 1 1
index and an almost equal amount is exposed to the 1-month time lagged index. Furthermore only very few of the exposed firms are exposed to both the current and the 1-month time lagged index as is shown after the (*). This leads to the conclusion that both the current and the 1-month time lagged results should be taken into consideration in the remainder of this research. Hence we do not follow Bartov and Bodnar (1994) in saying that the time lagged exposure is more important than the current exposure but we find that both can complement each other. This contradicts the findings of He and Ng (1998) and Nydahl (1999) as they find no evidence for a lagged exposure of firms in their samples.
If we then compare the number of firms that are significantly exposed to both the current and the lagged indices while we control for the ones that are exposed to both we gain the following results: In the 1994-1997 period 10 firms (21%) are exposed to the guilder-based index, 16 firms (34%) to the narrow euro-based index and 15 (32%) to the broad euro-based index. In the 1998-2001 period 6 firms (13%) are exposed to the guilder-based index, 13 (28%) to the narrow euro-based index and 12 (26%) to the broad euro-based index. In the 2002-2005 period 13 firms (28%) are exposed to the narrow euro-based index and 13 firms (28%) are exposed to the broad euro based index. Hence we see that the number of firms that show significant exposure had decreased after the introduction of the euro but this difference is more substantive with the guilder-based index than with the euro-base indices. The results of the 2002-2005 period show that the exposure of firms has not significantly changed in comparison to the period directly after the introduction of the euro. Hence we state that investors did not need a longer period to adjust for changes that were made by the introduction of the euro.
If we compare these results with the results from previous research on this subject we find some similarities and some differences. In their sample of Dutch firms in the period 1994-1998 De Jong et al. (2006) find that 50% are exposed to exchange risk. They use bi-weekly data and exposure to the current index. In the 1994-1995 period we find only 13%-15% of the firms exposed to the current index and 21%-32% exposed to the current or the lagged index. This is a significant difference. However the results are not entirely comparable because De Jong et al. use bi-weekly instead of monthly data. They find a mean exposure coefficient of 1.55% which is a bit higher than our mean exposure coefficient of -0.24% to -1.03% (the sign is the same as they use a home/foreign currency instead of a foreign/home currency index). Both results
indicate that a depreciation of the home currency leads to increased returns on firm stock value.
Bartram and Karolyi (2006) in find that in the 1992-2001 period with weekly data for their sample of European and non-European firms that 0.8%-3.3% of the positive exchange firms changed in exposure and 2.1%-8.1% of the negative exposed firms changed in exposure after the introduction of the euro. We find larger changes in the exposure to the guilder-based index, from 15% before the euro to 2% after the euro (-87% change). The changes in exposure to our euro-based indices and with the inclusion of the lagged exposures are also bigger, but less profound than to the guilder-based index. Overall Bartram and Karolyi find a slight decrease in the amount of firms that show significant exposure after the introduction of the euro. This result is in line with our findings.
Nguyen et al (2007) find that in their sample of French firms with monthly data 32% are exposed in the period 1996 and 11% are exposed in the period 2000. This is a larger decrease in exposures than we have found, but their periods have a 3-year gap in between which makes the results harder to compare. Their result, although in another country, are however in line with our results. Hutsion and Driscoll (2010) find in their comparison of the periods 1990-1998 and 1999-2008 with monthly data a decrease from 17.6% exposure to 10.3% exposure of Dutch firms. Theses changes are comparable with our results although the periods of investigation are significantly different. In their sample there are more firms that show negative exposure than positive exposure (3 positive and 9 negative before the euro and 3 positive and 4 negative after the euro). We find 3-5 positive and 2-3 negative exposed before the euro and 0-5 positive and 1-4 negative exposed after the euro. However the results are different if we include the lagged exposures.
It must be clear that in all this previous research the NEER-indices used have not been accounted for and the lagged responses to exposure have not been included.
Our result that there is no significant difference between the exposure to the broad (52 currencies) and the narrow (27 currencies) euro-based indices is in contradiction with the findings of Fraser and Pantzalis (2004). They find that the more currencies are included in the trade-weighted exchange rate index, the more firms show significant exposure.
4.2 Exposure to bilateral rates.
The results stated above have not shown whether the use of these indices to measure the exposure of firms is a good proxy for the exposure that firms actually have to the different bilateral exchange rates that are applicable. Furthermore if the firms show drastically different exposure to these bilateral rates than to the indices it might be possible that these indices understate the actual exposure as for example Miller and Reuer (1998) have stated. In measuring the exposure to the bilateral rates we include both the current and the 1-month time lagged rates to be in harmony with the previous results.
Table 6: Bilateral rates Dutch Guilder 1994-1997 USD USD LAG-1 GBP GBP LAG-1 F.FRANC F.FRANC LAG-1 B.FRANC B.FRANC LAG-1 G.MARK G.MARK LAG-1 Minimum -4.706 -1.740 -1.418 -5.193 -5.099 -6.449 -32.144 -40.557 -38.061 -78.920 Median -0.219 -0.034 0.250 -0.061 0.359 -1.154 2.931 -0.208 -2.984 -1.156 Maximum 0.707 1.312 3.495 1.862 6.571 4.614 31.267 35.973 70.833 20.787 Mean -0.384 -0.055 0.275 -0.014 0.607 -1.214 3.693 0.484 -1.780 -4.745 Standard deviation 0.922 0.600 0.971 1.003 1.872 2.025 13.068 13.247 18.994 16.410 Total number of firms with at least one significant exposure*: 35
Significant at 1% 3 2 0 0 0 0 1 0 1 1 Significant at 5% 5 4 4 3 2 3 3 1 5 3 Significant at 10% 10 4 8 7 3 6 6 4 13 7 Positive at 10% 0 1 4 6 2 0 6 3 6 2 Negative at 10% 10 3 4 1 1 6 0 1 7 5
This table represents the results of the regression analysis, using model (1) and (3). The minimum, median, maximum, mean and standard deviation of the β2’s of the different bilateral rates on the stock
returns of the 47 firms are displayed at the top half of the table. The bottom half of the table represents the number of firms that showed exposure and at which level of significance and whether it was positive or negative exposure. (*): The number of firms with exposure to both the current and the 1-month time lagged rates are 0 for the USD, GBP and F.FRANC, 1 for the B.FRANC and 2 for de G.MARK
Table 6 shows that a significant number of firms is exposed to the bilateral rates of the 5 biggest trading partners of the Netherlands in the period 1994-1997.
Ranging from 9 (19%) to the French Franc to 18 (38%) to the German Mark. Firms are mostly negative exposed to the US Dollar (13), while they are mostly positive exposed to the Belgian Franc (9). The total number of exposures is 65. Overall the mean exposure coefficient of the sample varies from -1.780 to -4.745 for the German Mark and from 3.693 to 0.484 for the Belgian Franc. Furthermore there are more firms negatively exposed (37) than positively exposed (28). This indicates that Dutch firms are largely export orientated because their stock value depreciates when the guilder appreciates against most currencies. Overall more firms are exposed to the bilateral rates than that they are exposed to the indices, although not a very significant amount. This indicates that the indices do understate the exposure that is experienced in the 1994-1997 period. These findings are in line with the results that De Jong et al. (2006) have found in their research.
Table 7: Bilateral rates Euro 2001-2005 USD USD LAG-1 GBP GBP LAG-1 JPY JPY LAG-1 CNY CNY LAG-1 RUB RUB LAG-1 Minimum -3.786 -6.045 -4.455 -3.669 -0.803 -1.314 -3.003 -5.368 -6.149 -5.053 Median 0.215 0.432 0.091 -0.350 -0.129 -0.136 0.039 0.192 -0.494 -0.629 Maximum 4.766 8.835 6.003 1.298 1.151 2.227 4.803 5.797 2.927 2.372 Mean 0.402 0.808 -0.060 -0.487 -0.025 -0.093 0.178 -0.055 -0.685 -0.477 Standard deviation 1.683 2.412 1.496 0.917 0.487 0.696 1.428 1.865 1.717 1.408 Total number of firms with at least one significant exposure*: 33
Significant at 1% 2 0 0 1 0 1 2 3 2 2 Significant at 5% 3 3 5 6 1 3 3 6 5 6 Significant at 10% 4 4 8 8 1 5 6 8 7 9 Positive at 10% 3 4 4 0 1 1 2 5 1 1 Negative at 10% 1 0 4 8 0 4 4 3 6 8
This table represents the results of the regression analysis, using model (1) and (3). The minimum, median, maximum, mean and standard deviation of the β2’s of the different bilateral rates on the stock
returns of the 47 firms are displayed at the top half of the table. The bottom half of the table represents the number of firms that showed exposure and at which level of significance and whether it was positive or negative exposure. (*): The number of firms with exposure to both the current and the 1-month time lagged rates are 0 for the USD, 2 for the GBP, 1 for the YEN, 1 for the CNY and 1 for the RUB.
Table 7 shows that a significant number of firms are exposed to the bilateral rates of the 5 biggest non-euro trading partners of the Netherlands in the period 2002-2005. Ranging from 15 (32%) in case of the RUB to 5 (11%) in case of the JPY. Firms are mostly negative exposed to the GBP (12) and to the RUB (14) and mostly positive to the USD. Total number of exposures is 55. The mean exposure coefficients of the different bilateral rates are all between -0.685 and 0.808.
If we compare the results from table 6 and table 7 we see that in both periods more firms show exposure to the bilateral rates than to the different NEER-indices. The decline in the number of exposures after the introduction of the euro is also less profound (from 65 to 55) than the decline in exposures that the NEER-indices have shown. However these results are hard to compare because it was not possible to do a regression for the 1998-2001 due to the euro being introduced in the middle of this period. Furthermore the bilateral rates shown in both tables differ because some of the rates from the first table obviously seized to exist after the introduction of the euro.
Hence these results cannot be used to compare the number of exposures in the different periods before and after the introduction of the euro. However they can be used to see whether the results from the NEER-indices in the periods 1994-1997 and 2001-2005 are a good proxy for the actual exposure that is experienced by the firms. It is clear that in both periods more firms were exposed to at least one bilateral rate than to the different NEER-indices (35 > 17/17/11 and 33 > 15/15) so, as mentioned above, the indices understate the actual exposure. These results are in line with Miller and Reuer (1998) say about the underestimation of exposure by indices. Dominquez and Tesar (2001) find the same results in their sample of multinational firms. The results are in contradiction with the results of Bartram (2002). He finds that trade-weighted indices do not understate the exposure of firms in comparison to bilateral rates.
Another important results of the comparison of table 6 and table 7 is that the mean exposure coefficients in the 2002-2005 period (ranging from -0.685 and 0.808) are smaller in both the positive and the negative sign than the mean exposure coefficients in the period 1994-1997 (ranging from -4.745 to 3.693). This indicates that if exposure is experienced in the 2002-2005 period, it is less severe than in the 1994-1997 period.
4.3 Exposure before and after the introduction of the Euro
We have seen that if the periods are taken separately, the sample of firms are slightly less exposed in the period after the introduction of the euro (1998-2001) than in the period before the introduction of the euro (1994-1997). However to make a good assessment of the actual influence of the introduction of the euro it is useful to make use of dummy variables as in model (3). Therefore we use the 3 NEER-indices again to test if the exposure of firms changed in the period 1996-2001 by using a dummy variable to distinguish the periods before and after the introduction of the euro. To be in harmony with the previous results we include the current and the 1-month time lagged NEER-indices.
Table 8 shows that the introduction of the euro has mixed effects on the number of firms that show exposure to the indices. For the guilder-based index: 7 firms (13%) are exposed before and after the introduction of the euro, 2 firms (4%) only before and 2 firms (4%) only after. For the narrow euro-based index: 9 firms (19%) are exposed before and after the introduction of the euro, 6 firms (13%) only before and 4 firms (9%) only after. For the broad euro-based index: 6 firms (13%) are exposed before and after the introduction of the euro, 6 firms (13%) only before, and 3 firms (6%) only after. As indicated at the (*), these numbers are controlled for the double counting of firms that are exposed to both the current and the 1-month time lagged index.
These results show that in the case of the euro-based indices, the introduction of the euro has had an effect on the number of firms that are exposed. Some firms are only exposed after the introduction of the euro and some only before. However the number of firms that were exposed only before the introduction of the euro is larger (6>4 and 6>3) than the number of firms that were exposed only after the introduction of the euro. Furthermore some firms (9 and 6) were exposed both before and after the introduction of the euro. The guilder-based index does not show a change in the number of firms that are exposed before and after the introduction of the euro. Hence 2 of the 3 indices show that the introduction of the euro reduced the amount of firms with significant exchange risk exposure, however slightly. These results contradict the earlier results of tables 3 and 4 in that the changes in firms with exposure to the guilder-based index in the periods before and after the introduction of the euro were greater than the changes in exposure to the euro-based indices in these periods.
Table 8: Exposure to NEER-indices in 1996-2001 including dummy GUILDER -NEER GUILDER -NEER LAG-1 GUILDER -NEER AFTER EURO GUILDER -NEER AFTER EURO LAG-1 EURO-NEER NARR. EURO-NEER NARR. LAG-1 EURO-NEER NARR. AFTER EURO EURO-NEER NARR. AFTER EURO LAG-1 EURO-NEER BRO. EURO-NEER BRO. LAG-1 EURO-NEER BRO. AFTER EURO EURO-NEER BRO. AFTER EURO LAG-1 Minimum -5.718 -22.036 -7.701 -7.281 -9.080 -16.13 -15.731 -5.121 -5.007 -13.791 -13.495 -5.882 Median 0.053 -0.518 -0.619 0.203 0.321 -2.176 0.093 0.801 -0.335 1.849 -0.435 0.971 Maximum 10.817 3.060 10.969 28.795 14.774 3.473 7.178 22.063 11.994 4.298 4.332 21.638 Mean 0.106 -0.518 -0.025 0.750 0.262 -2.582 -0.205 2.435 0.193 -2.197 -0.454 2.140 Standard deviation 3.085 3.575 3.926 5.144 3.451 3.851 3.641 5.090 2.676 3.379 2.933 4.536 Number of firms with significant exposure*:
Significant at 1% 2 0 1 1 0 1 1 1 1 2 1 1 Significant at 5% 4 2 3 2 3 8 4 2 2 8 3 4 Significant at 10% 6 4 6 4 6 11 7 7 4 11 6 5 Positive at 10% 1 2 5 2 4 1 3 6 2 1 2 4 Negative at 10% 5 2 1 2 2 10 4 1 2 10 4 1 Sign. pre- and post Euro 5 2 5 2 4 6 4 6 2 5 2 5 Sign. pre Euro 1 2 - - 2 5 - - 2 6 - - Sign. post Euro - - 1 2 - - 3 1 - - 4 0 Different sign - - 5 1 - - 4 6 - - 2 5
This table represents the results of the regression analysis, using model (2) and (3). The minimum, median, maximum, mean and standard deviation of the β2’s of the different NEER-indices on the stock
returns of the 47 firms are displayed at the top half of the table. The bottom half of the table represents the number of firms that showed exposure and at which level of significance and whether it was positive or negative exposure. Furthermore the number of firms that show significant exposure in only the pre euro, only the post euro or both the pre- and the post euro period are given. In the last row the number of firms that have a different coefficient exposure sign (+) or (-) before and after the euro are given. This obviously can only be said of firms that showed exposure in both the pre- and the post euro period. (*): The number of firms with exposure in both the current and 1-month time lag are: 1 GUILD, 1 for DGUILD, 2 EURO(N), 1 DEURO(N) (1 of them in pre- and post euro), 2 EURO(B), 2 DEURO(B) (1 of them in pre- and post euro).
However both the results in tables 3 and 4 as the results in table 8 show an overall reduction in the number of firms with significant exposure after the introduction of the euro.
Since these results show a less profound decrease in the number of firms with significant exposure after the introduction of the euro they are more in line with Bartram and Karolyi (2006) than the results of our tests in separate periods. Hence it is clear that the dummy variable model, which is also used by Bartram and Karolyi, results in smaller differences found in exposure before and after the euro than when the periods are compared separately. Hutsion and Driscoll (2010) and Nguyen et al (2007) have also compared the periods before and after the introduction of the euro separately (1996 and 2000 in case of the former and 1990-1998 and 1999-2008 in case of the latter). As is explained above, they indeed find larger difference in exposure before and after the euro than we find in table 8.
Another important result shown in table 8 is that for almost all firms that are exposed in both the pre- and the post euro period the sign of the exposure (+ or -) has changed. This is true for 23 of the 24 firms that show exposure in the entire period. The mean exposure coefficients of the dummy variable of both the current and the lagged indices are also in each case of a different sign than the mean exposure coefficient of the non-dummy variable. We have checked the data for possible errors or changed definitions but found nothing. Interestingly, Bartram and Karolyi (2003) also find that for most of their companies the change in exposure after the introduction of the euro was of opposite sign. Furthermore Bartov et al. (1995) also find a change in exposure of opposite sign in their study of exchange rate exposure before and after the breakdown of the Bretton Woods system. Both studies make use of the same dummy variable model. There is no theoretical or empirical explanation for these results that we know of.
4.4 Asymmetrical exposure.
We test for sign asymmetric exposure using model (4). Since our previous results have shown that it makes no significant difference whether the broad or the narrow based index are used, we only test for asymmetric exposure to the broad euro-based index. The guilder-euro-based index is not included because it has no data for the 2002-2005 period.
