!
Exchange Rate Exposure during
Recession and Growth
Jos Pol
University of Groningen
Uppsala University
Programme: MSc International Financial Management
Student nr.:
1463659
Date:
April 18, 2009
Supervisor:
Prof. dr. Niels Hermes
!
Exchange Rate Exposure during
Recession and Growth
Jos Pol
University of Groningen
Uppsala University
Abstract
This study estimates the exchange rate exposure of 188 Japanese nonfinancial companies for the period between July 1997 and December 2007 and focuses on the differences in exposure between recessions and periods of growth. In the research period, two recessions and two periods of growth are identified and the outcomes of these periods are compared. Results show differences in company exposure between an economic downturn and a period of growth. Not only is a much larger share of the sample companies significantly exposed to the exchange rate during a recession, the exposure is stronger too. The outcomes are robust for the choice of important variables of the regression model. Subsequently, the determinants of exposure are examined to find out the causes of the differences between recession and growth periods. As it turns out, the changes in the values of the significant determinants company size, foreign sales ratio and debt ratio have a measurable impact on the exposure differences between the states of the economy. No evidence is found for the importance of the current ratio, dividend payout ratio, book-to-market value of equity, keiretsu affiliation or industry characteristics in the determination of exchange rate exposure.
JEL classification: F31, G32, C32
Table of Contents
1. Introduction...3
2. Theoretical framework...5
2.1. Economic foundation ... 5 2.2. Empirical research... 7 2.2.1. Company exposure ... 8 2.2.2. Industry exposure ... 9 2.3. Time variation ... 10 2.4. Methodological issues ... 12 2.5. Determinants of exposure... 13 2.5.1. Hedging ... 143. Methodology and Data ...17
3.1. Regression model ... 18 3.2. Sample ... 19 3.3. Financial data ... 20 3.4. Robustness tests... 24 3.5. Determinants of exposure... 25 3.5.1. Company size ... 25
3.5.2. Foreign sales ratio ... 26
3.5.3. Hedging activities... 26
3.5.4. Keiretsu firms... 27
3.5.5. Industry... 28
3.5.6. Regression analysis of determinants ... 29
4. Results and Interpretation ...32
4.1. Exposure of individual sample firms ... 32
4.2. Robustness tests... 34
4.3. Determinants of exposure... 39
5. Conclusion ...44
List of References...46
Appendix I – Lagged exposure ...49
Appendix II – Regression model of exposure determinants ...51
Appendix III – Differences amongst recession and growth periods...53
1. Introduction
Many papers have been written on the subject of exchange rate exposure. As the exchange rate is a serious factor of uncertainty, a number of academics have built theoretical models that predict a relationship between exchange rates and company performance. Following these predictions, a vast range of empirical studies on the impact of exchange rates on company performance, measured by stock returns, are performed. However, results of these studies are no more than moderately successful, as often only little evidence of exchange rate exposure is found.
In this research, not only a new attempt is made to establish the relationship between exchange rates and stock prices, but also argumentation is given that the exchange rate exposure of a company is dependent on the state of the economy. The main point of this argument is that 1) company’s characteristics that define exchange rate exposure differ between recession periods and growth periods and 2) investors may assess different factor to be determinants of exposure between recessions and growth periods. This argument is tested for a sample of 188 Japanese listed companies. In order to determine exposure, stock price fluctuations are explained by the independent variables of the trade-weighted exchange rate of the yen and the returns of the broad Japanese Topix index for the period between July 1997 and December 2007. Based on quarterly GDP movements, two recession periods and two growth periods are distinguished in this research period. The two recessions are the periods from July 1997 to September 1999, subsequent to the Asian Financial Crisis, and April 2001 to March 2003, following the burst of the Internet bubble.
The results of this research prove that exchange rate exposure is much more severe during recession periods than during growth periods. The number of companies with significant exposure increases heavily during recessions and the average exposure coefficient is higher during recessions. Besides, statistical tests show that the coefficients of all sample companies are significantly different between the recession and growth periods. All these results are robust to the use of different variables in the regression models.
or more keiretsu financial institutions, a typical phenomenon in the Japanese economic environment. The risk sharing associated with keiretsu has no impact on exchange rate exposure in this study. Moreover, a company’s industry does not influence exposure coefficients. The most striking result of this analysis, however, is that no difference exists between the outcomes of the tests between recessions and growth periods. Therefore, the vast differences in exposure that is proven in this study cannot be explained by the fact that different factors determine exposure during a recession. As the determinants remain the same, the exposure differences are caused by changes in the value of the determinants, i.e. the increase in foreign sales ratios or the lower debt ratio for example.
Although it has been proven in previous literature that company exposure changes over time and depends on the general direction of the exchange rate, the link between exposure and the state of the economy has not been made. This study aims to prove this relationship and its underlying arguments by analysing the exchange rate exposure of Japanese listed companies over a period that exceeds ten years while distinguishing between periods of recession and economic growth.
2. Theoretical framework
In this section the economic foundation of exchange rate exposure is explained and an overview of the most important empirical research is provided. Problems arising from this empirical research, as well as specific characteristics that can have an influence on the outcome of the subsequent analysis presented in this study, are discussed.
2.1. Economic foundation
Economic reasoning argues that fluctuations in exchange rates have a direct impact on the performance and market valuation of companies. Movements in exchange rates influence the value of international, cross-currency cash flows. According to accepted theory, the valuation of a company is based on the present value of its future cash flows; hence a relationship between exchange rates and stock prices should exist. However, the relationship is not that straightforward that a certain decrease in the exchange rate leads to a same extent decrease in home currency cash flow values. Many different variables impact this relationship, as is proven by numerous theoretical and empirical studies, of which several are described below.
The impact of currency fluctuations on company value is generally categorised into three different types of exposure (see e.g. Stulz & Williamson, 2000). Firstly, translation exposure refers to the value changes of balance sheet items in a straightforward relation with the exchange rate. The values of balance sheet items are easily observable and can be hedged using financial products, so that the actual exposure is limited. Secondly, transaction
exposure considers previously booked transactions which values are affected by exchange
rate movements. This kind of exposure can be measured through accounting techniques and can be managed using financial hedges.
on the price elasticity in the foreign market. If cost changes cannot be passed through, the company’s profit will be harmed.
Moreover, other factors that define economic exposure are the structure of the markets in which the company sells its products and the structure of the markets in which the company (and its competitors) purchase their inputs, as explained in Flood & Lessard (1986). The market structure is defined by the origin of the producers and consumers that dominate the market (i.e. if a market is primarily composed of U.S.-based producers and consumers, exchange rates will have little impact on the U.S. dollar price) and the demand and supply elasticities. However, Hodder (1982) showed that even a company without any international operations could still be hurt by exchange rate changes, as he recognises that surges in ‘foreign competition’ and domestic commodity price booms are frequently related to fluctuations in the exchange rate. A purely domestic company can be hurt by strategic decisions of international competitors (e.g. market entry, expansion of operations), after an exchange rate fluctuation has made it more attractive for this international competitor to do so. Apart from that, the customer base of the domestic company may include importing or exporting companies whose activities are influenced by the exchange rate. Due to the changes in market environment, the purely domestic company may be forced to adjust its prices and may incur a lower profit.
the domestic currency, the exporting firm has to pass through his cost increase to the market, whereas the purely foreign firm keeps its price constant, as its marginal costs are constant. When perfect substitutability between products is assumed, demand will shift to the products of the foreign firm, which captures a share of the profit of the exporting firm.
2.2. Empirical research
Much of the research performed in this field is based on a model proposed by Adler & Dumas (1984), in a later stage adjusted by Jorion (1990). Adler & Dumas (1984) argue that the factors influencing the impact of exchange rates on company cash flows are too widespread to capture separately. Instead, they develop a technique to measue the exposure of company value to the exchange rate. Since all consequences of exchange rate changes are bound to impact the company’s future cash flows, Adler & Dumas suggest using changes in the company’s stock price as an approximate for exchange rate exposure. Efficient market theory states that all available information is included in the stock prices, which are based on investors’ expectations of future cash flows. The model laid out by Adler & Dumas (1984) is a regression model in which stock prices as the dependent variable are explained by exchange rates as the independent variable, in formula:
t = 1, , T, (1)
in which Rit is the rate of return on the ith company’s common stock and Rst is the rate of
change in a trade-weighted exchange rate, measured as the home currency price of the foreign currency (a positive value for Rst indicates home currency depreciation). Jorion (1990) points
out that several macroeconomic factors exist that have a simultaneous influence on both stock prices and the exchange rate, such as inflation and the risk-free interest rate. Therefore, Jorion suggests adding a market variable to the model in order to capture the macroeconomic factors and to isolate the influences of exchange rates on stock prices. This leads to the following regression formula, in which Rmt represents the market variable.
t = 1, , T, (2)
2.2.1. Company exposure
Jorion (1990) regressed monthly stock prices against monthly trade-weighted exchange rate for a sample of 287 U.S. multinational corporations between 1971 and 1987. Contrary to prior expectations, based on theoretical models regarding the impact of exchange rates on company performance, only 15 (5.22%) companies of his total sample showed significant exposure to the exchange rate. However, Jorion did identify differences in exposure across firms according to their involvement in international trade. Amihud (1994) uses this observation as a feature of sample selection. Amihud selected a group of sample firms that are expected to have large exposure, as his 32 sample companies are all named in Fortune’s ‘50 Leading Exporters’ between 1982 and 1988. Moreover, he constructs a portfolio of the eight largest U.S. exporters. In his study, there are no significant stock price responses to contemporaneous exchange rate changes for both the firm level and the constructed portfolio. Following these unexpected results, other scholars have performed studies on the U.S. Chow, Lee & Solt (1997) find only 2.3% of their sample of 323 multinational firms to be significantly exposed within the one-month period. Bodnar & Wong (2003) increase their sample size to 910 companies by adding smaller firms. Smaller firms may have less available resources for hedging currency exposure and their international operations may be limited to fewer countries. Due to the lack of (natural) hedging smaller companies are expected to have a higher exposure. Moreover, Bodnar & Wond focus on a longer research period, from 1977 to 1996. These differences resulted in 14.6% of sample companies with a statistically significant exchange rate exposure.
sensitivity of Australian companies to changes in the trade-weighted index value of the Australian dollar for the period of 1984 to 1989, which are the first six years of the current floating exchange rate regime. Of the 141 sample companies, 15 companies (10.6%) are exposed to the exchange rate.
Dominguez & Tesar (2006) focused their research on eight developed countries, on a country-by-country basis. With regards to the monthly interval, the following shares of the sample consist of significantly exposed companies: Chile (14%), France (19%), Germany (21%), Italy (26%), Japan (31%), the Netherlands (26%), Thailand (21%) and the United Kingdom (19%). Muller & Verschoor (2006) conduct a comprehensive analysis on European multinational firms and test their sample of 817 companies for being exposed to the U.S. dollar, the Japanese yen and the U.K. pound between 1988 and 2002. They do not differentiate across countries but pool all the European companies into one sample. As it turns out, the U.K. pound has the most impact on European companies, influencing 22% of the sample. The U.S. dollar is of importance to 13% of the sample and the Japanese yen to 14%. Doidge, Griffin & Williamson (2006) create a sample of 17.929 non-financial companies from 18 different countries, with the median country contributing 299 firms to the sample. Of this large sample, 4.2% of all companies are positively exposed to the exchange rate and 4.0% are negatively exposed, leading to a worldwide total exposure of 8.2%.
2.2.2. Industry exposure
non-traded products: mining and ‘other retail’ (consisting mainly of department stores). This observation does support the statements of Bodnar & Gentry (2003). Choi & Prasad point out that companies in the same industry are not necessarily homogenous in their operational activities or financial strategies. Whether or not a company operates internationally, or whether a company uses hedging instruments is not defined by its industry; these decisions are company-specific. By aggregating companies into industries, these company-specific characteristics averaged out and may therefore lead to distorted results.
Furthermore, several researchers focus on multinational companies that are part of the same (single) industry, such as Williamson (2001), who measures exposure and its determinants in the automotive industry. This industry is characterised by high export sales and foreign competition and is thus likely to be sensitive to foreign exchange rates. He finds that the U.S. and Japanese automotive industry react differently to the tested exchange rates (U.S. dollar, Japanese yen, Deutschmark) based on the degree of foreign competition and the structure of the companies’ operations. U.S.-based automotive companies are positively exposed to the Deutschmark due to the production facilities these companies have in European countries. U.S. automotives have negative exposure to the yen, is all their cars for the Japanese market are manufactured outside Japan. The same argument holds for the Japanese automotive companies, who manufacture all the cars they sell worldwide in their home country. This leads to negative exposure to both the U.S. dollar and the Deutschmark. Another comparison between U.S. and Japan is made by Chamberlain et al (1997), who study the exposure of companies in the banking industry in both countries. Financial firms have a different attitude towards currency risk, as dealing foreign currency is part of their operations. Consequently, these firms are usually excluded from research. Of the U.S. bank holding companies, nine out of thirty (30%) show significant exposure. However, only about 10% of the Japanese banks are sensitive to the exchange rate. Chamberlain et al attribute this dissimilarity to fundamental differences in operations and regulatory conditions of the companies in the two countries. The exposure of U.S. banks is analysed and can be explained for roughly 25 to 40 percent by a bank’s net foreign assets position and hedging through off-balance sheet activities in foreign contracts.
2.3. Time variation
and increasing internationalisation of the competitive landscape and (iii) 1989-95, with volatile yen exchange rate and financial market concerns in Europe and Japan. Williamson (2001) found that the levels of exposure of U.S. and Japanese automotive companies vary significantly with the different time periods, as the industry evolves. This is generally referred to as time variation. Time variation is also proven in Choi, Hiraki & Takezawa (1998), Choi & Prasad (1995) and Bodnar & Wong (2003). In these three studies research period is divided into multiple subperiods based on the general trend of the exchange rate, i.e. in periods of rising and falling exchange rates. All studies demonstrate that the more companies are exposed and exposure coefficients are larger during weak dollar periods. Given the reluctance to decrease prices following a decrease in the exchange rate, currency depreciations and appreciations do not bring about symmetric changes in company value.
Time variation is an important issue in this paper, as the focus of the research is on the differences in company exposure between recession periods and periods of economic growth. The studies mentioned above report that exposure differs following changes in the market circumstances and that the trend in the exchange rate has a certain impact on exposure, but ignore the argument that the state of the economy may influence exchange rate exposure. This is due to the changes companies make in their marketing and financial strategies during recessions.
Shama (1993) used a questionnaire to find out the changes in marketing strategy of a sample of 180 companies during the 1990-1991 recession in the U.S. In response to an economic downturn, one of the most taken measures is the (international) broadening of target markets. These results are confirmed by Pearce & Michael (1997) who test the impact of changes in marketing strategies on company’s resistance to the 1990-1991 recession. Using questionnaires, they find that among the 114 sample firms a significant part extended their internationalisation strategy during the recession, along with among others the improvement of marketing efficiency and the streamlining of the value chain. That the internationalisation was not a one-time phenomenon is proven by the earlier study of Rao et al (1990), who prove that U.S. exporters intensified their exports during the 1980-1982 recession. A consequence of the increased internationalisation in economic downturns is the increase in international cash flows, thereby increasing the economic exposure of a company.
company, the increased focus on short-term liquidity increases exchange rate exposure as more capital is exposed to the exchange rate. Retrenchment increases the translation and transaction exposure of a company.
As financial and marketing strategies differ between recessions and growth periods, the argument is made that exchange rate exposure differs as well. How the exposure differs with the business cycle is measured in this study, and an attempt is made to look at the underlying reasons of the differences in exposure.
2.4. Methodological issues
In general, the percentages of companies or industries that are significantly affected by exchange rate movements are lower than researchers’ priors predict. As a consequence, several scholars have proposed changes in the methodology. Bartov & Bodnar (1994) addressed two factors that could possibly explain the ‘disappointing’ results of previous studies. Firstly, they question the sample selection procedures of previous studies. They argue that the impact of a certain exchange rate fluctuation is determined by whether a company has a long or short economic position in the involved currencies. Therefore, consists their sample only of companies that have reported impacts from changes in the U.S. dollar in their financial statements. By selecting their sample firms on this aspect they hope to strengthen their results. However, they fail to present significant results supporting the sample selection argument. Secondly, due to the complexity of the relation between exchange rates and firm performance, assets and liabilities, they argue that the market response is delayed until corporate information is disseminated. In other words, contemporaneous stock price responses are subject to mispricing and therefore lagged stock returns are a better approximate for exposure. Corporate announcements of the consequences of exchange rate fluctuations are included in lagged stock returns, so that investors have a better picture in order to value these consequences. This argument is supported in their analysis. More studies prove that exchange rate exposure gets more significant as the time horizon lengthens (see e.g. Chow, Lee & Solt, 1997, Chow & Chen, 1998 and Bodnar & Wong, 2003). In this research, the impact of lagged returns is tested and reported in Appendix I.
variable leads to a different economic interpretation of the exposure coefficient. The ‘new’ exposure coefficient measures the difference between a company’s total exposure and the market’s total exposure, adjusted by the company’s beta. Bodnar & Wong refer to this statistic as the ‘residual exposure’, where a zero residual exposure implies that a firm has the same exposure as the market portfolio. Since it is unlikely that the exposure of the market portfolio used to control for macroeconomic effects will be zero, the choice of the market portfolio directly influences company exposure. Bodnar & Wong examine the importance of the choice for the market variable by using three different portfolios: the Center for Research in Security Prices (CRSP) U.S. value-weighted index, the CRSP equal-weighted index and the Morgan Stanley Capital International world market portfolio. They show that the choice of the market variable has a substantial impact on the results of company exposure, as the use of the three indices in their research led to significantly different results. The robustness of the research results of this study is tested with two different market variables, a broad national index (Topix) and a global index (Dow Jones STOXX Global 1800).
Furthermore, the interpretation of the exchange rate variable has been subject to research. Often, researchers choose to use trade-weighted multi-lateral exchange rates. Bartram (2004) proves, however, for a sample of 447 German nonfinancial companies that using a bilateral exchange rate (in this case the US Dollar) does not lead to substantially different results. Generally, actual exchange rate fluctuations are used to approximate the exposure of companies. As firm performance is only subject to unexpected changes in the exchange rate, Choi & Prasad (1995) regressed stock returns to abnormal changes of the exchange rate, where abnormal changes are defined as the differences between actual rates and previous forward rates. Results of this analysis did not lead to substantially different outcomes. Fraser & Pantzalis (2004) notice that all studies apply a common exchange rate index to the analysis, whilst all companies in the sample have different international operations. They determine a firm-specific currency basket to overcome this problem. For every sample company they weigh the exchange rate variable by the company’s distribution of foreign subsidiaries. Although they prove this adjustment in the analysis leads to a higher number of firms with significant exposure than when a common index is used, their results are not radically different from the results found in previous research. In this study, the choice for a trade-weighted index is examined by additional tests performed with the bilateral yen-dollar relationship as the explanatory variable.
2.5. Determinants of exposure
is employed to find the root causes of any changes in exposure that occur between recession and growth periods. Commonly, the regression coefficient for the exchange rate variable is set as the dependent variable and a set of company characteristics is included as explanatory variables in an ordinary least squares regression. Formula (3) indicates the regression formula used in such analyses.
(3)
In equation (3), !fxi represents the coefficient of the exchange rate variable in the
determination of exchange rate exposure (similar to models (1) and (2)). Xi represents the
variables that are tested for being determinants of exchange rate exposure in this analysis; n corresponds to the number of variables included. The ordinary least squares results are indicated by the intercept " and the coefficients #i to #n. Variables that are often included in
these analyses are company size and foreign sales ratios. The argument that company size influences exchange rate exposure is based on the observation that larger firms generally have more extensive foreign operations, e.g. production facilities in different countries. These foreign operations ensure more international cash flows and a larger transaction exposure, as is proven in Chow & Chen (1998), Bodnar & Wong (2003), He & Ng (1998) and Doidge, Griffin & Williamson (2006). Foreign sales ratios indicate the dependence of companies on their international sales activities and are therefore positively related to exchange rate exposure, i.e. the higher the foreign sales ratio of a company the higher its exposure to the exchange rate. This variable is a significant determinant of exchange rate exposure in a.o. Jorion (1990), He & Ng (1998), Allayanis & Ofek (2001), Bartram (2004) and Doidge, Griffin & Williamson (2006). Other variables included in analyses of exposure determinants are mainly hedging variables. A description on the use of hedging is described below, and the variables used to approximate for hedging in this study are discussed in paragraph 3.5.3.
2.5.1. Hedging
with regards to tax and financial distress. Hedging reduces the variability of pre-tax firm values, which reduces the expected corporate tax liability and, thus, increases the after-tax firm value. Furthermore, as hedging reduces the variability of the future value of the firm, hedging lowers the probability of incurring bankruptcy costs. Thirdly, as Froot, Scharfstein & Stein (1993) point out, without hedging, companies may be forced by its high variability in cash flows to underinvest in some states of the world because it is more costly to raise external finance. Finally, DeMarzo & Duffie (1995) drop MM’s assumption of the absence of asymmetric information. Instead, the more realistic assumption that managers are better informed about the sources and magnitudes of the risks faced by companies then shareholders are. This information asymmetry puts managers in a better position to hedge these risks, and gives an explanation for corporate hedging.
Nance, Smith & Smithson (1993) examined a sample of 169 firms, of which 104 used financial hedging products as swaps, futures and options. It is found that firms reduce their expected tax liabilities, lower transaction costs and control agency problems through hedging. Also, the hedging firms face more convex tax functions, have less coverage of fixed claims, are larger and have more growth opportunities. As a consequence, hedging was empirically proven to be value increasing for the firm. Allayannis & Weston (2001) use Tobin’s Q as a measure to test whether the use of currency derivatives is associated with higher firm market value for a sample of 720 nonfinancial U.S. companies. They show that for companies with foreign sales, hedging activities consistently adds value, both during a currency’s period of appreciation and depreciation. In Allayanis & Ofek (2001) it is demonstrated that the use of currency derivatives leads to lower exchange rate exposure. Therefore, it is proven that currency derivatives’ accompanying increase in market value does not result from speculative activities, but from hedging use.
The use of financial hedges can be measured empirically for U.S. companies as they are obliged to file their use of currency derivatives on the SEC 10-K reports, whereas foreign debt data is usually available from annual reports. Outside the U.S., the regulations regarding the registration of hedging products is not as strict. As a consequence, data collection on hedging is more complicated for these countries as there is no central database of hedging use. Consequently, researchers (e.g. He & Ng, 1998) approximate hedging behaviour based on the incentives that companies have for hedging.
3. Methodology and Data
In this study, the exchange rate exposure of 188 Japanese companies is examined between July 1997 and December 2007. In order to find out the differences in exchange rate exposure between recession periods and periods of economic growth, a subdivision is made in the research period. Japan is chosen as the country of research because of its dependence on international trade, the fact that it has had large swings in the economy in recent years, and because its unique governance structure offers additional research opportunities regarding the impact of keiretsu on exchange rate exposure.
Figure 3.1 below is a graphical representation of the quarterly GDP movements of Japan during the period of study. During the full period, two recessions took place: one following the Asian Financial Crisis, starting in July 1997 and ending in September 1999, and one following the burst of the internet bubble, starting in April 2001 and ending in March 2003. Between these two periods and between April 2003 and December 2007 the GDP of Japan has been increasing continuously. There is no reason to assume different behaviour of exchange rate exposure amongst the two periods of recession (or the two periods of growth). Thus, two subsamples are created, one in which all months in the two recession periods are aggregated and one in which all the months of the periods of growth are aggregated. In Appendix III, following a discussion of the differences and similarities between the two recession and the two growth periods, it is proven that the aggregation of the two regression and two growth periods is valid. All analyses are performed for the full research period and the two subsamples; the differences in results are discussed.
Figure 3.1 - Quarterly GDP of Japan and trade-weighted yen values
0! 50! 100! 150! 200! 250! 300! 350! 400! 470! 480! 490! 500! 510! 520! 530! 540! 1997/06 ! 1998/03 ! 1998/12 ! 1999/09 ! 2000/06 ! 2001/03 ! 2001/12 ! 2002/09 ! 2003/06 ! 2004/03 ! 2004/12 ! 2005/09 ! 2006/06 ! 2007/03 ! 2007/12 ! TW Y en ! GDP (Billions) ! GDP! TW Yen!
Previous research on time variation has mainly focused on differences of a rising and falling trend in the exchange rate, see e.g. Williamson (2001) and Choi, Hiraki and Takazewa (1998) This study focuses on rising and falling GDP figures. One could argue that these macroeconomic variables are interrelated and that therefore a study towards the differ
ences in exposure between periods of economic growth and decline is not needed. Howevcr, by taking a closer look at figure 3.1 it is evident that the exchange rate and the GDP of Japan do not follow the same pattern. During the obvious upswings and downturns in GDP, as indicated by the striped vertical lines, the exchange rate seemed to have followed a random walk. The correlation between GDP and the trade-weighted yen over the full research periods is -0.19, thereby indicating that the relationship is not very strong.
In this chapter, the details of the methodology in estimating exposure are outlined and the motivations of the choices regarding research period and sample companies are presented. Moreover, a first glance at the data is provided. The robustness of the results of this research is tested regarding the use multiple variables. The methodology of these sensitivity analyses is discussed in this section. Finally, the way in which determinants of exposure are estimated is explained.
3.1. Regression model
This study uses the technique of estimating exchange rate exposure that is the most common in this field. As discussed in paragraph 2.2, Adler & Dumas (1984) define exchange rate exposure as the impact of exchange rates on the present value of future cash flows of the firm. Since stock prices represent investors’ estimations of future cash flows of companies, this is the best measure to use for exposure. Therefore, Adler & Dumas laid out a regression model in which stock returns are regressed against exchange rates (see equation (1), paragraph 2.2). Jorion (1990) argued for the inclusion of a market variable to capture macroeconomic factors that influence both the stock price and the exchange rate. Scholars generally accept this argument and, as a consequence, the model outlined in equation (4) below became the usual way of measuring exchange rate exposure, with variables being constructed at a monthly interval.
(4)
In the formula above, Rit is the rate of return on the ith company’s common stock, Ret is the
rate of change in a trade-weighted exchange rate, Rmt represents the market variable. $i, !ni are
the ordinary least squares coefficients and %i represents the company-specific error term. As
company in the sample. The regression model above is run for the full research period and the recession and growth periods separately, and the results in exchange rate betas are discussed.
With regards to the variables, the monthly nominal trade-weighted index of the yen is used, as supplied on the website of the Bank of Japan (www.boj.or.jp, last visited January 8, 2009). The trade-weighted index is used since the sample companies conduct business in a vast array of countries. Therefore, use of a bilateral exchange rate variable (e.g. yen vs. U.S. dollar) theoretically leaves out measurable impact on stock prices. The market index variable used is the Topix-index, a broad index of all first section firms listed on the Tokyo Stock Exchange (approximately 1700 index members). This diversified index is a good measure of the macroeconomic factors influencing Japanese stock prices. Following the choice for these variables, the main regression model used in this research is described below.
(5)
Rit, RFXtw,t and RTOPIX,t depict the returns on the stock of the sample company, the
trade-weighted yen and the Topix index, respectively. $i, !ni are the company-specific ordinary
least squares coefficients and %i represents the error term. The exchange rate variable is
measured as the trade-weighted Japanese yen price of the foreign currency. Hence, an appreciating yen causes an increase in the exchange rate variable. For each sample company, a t-test of being significantly different from zero of !1i is performed. Companies with a
significantly positive coefficient are positively exposed (appreciating yen leads to increase in stock price); the opposite holds for companies with a significantly positive coefficient. Shares and percentages of the sample companies with positive and negative exposure are reported. This research covers the period between July 1997 and June 2008. Existing studies on the exchange rate exposure of Japanese companies focus on time periods preceding the time span subject to research in this study. He & Ng (1998) focus on the period between 1978 and 1993, whereas Chow & Chen (1998) study exposure between 1975 and 1992. The recent period subject to research in this study ensures renewed results on the impact of the value of the yen on Japanese stock prices.
3.2. Sample
The Bloomberg Professional database (“Bloomberg”) provides data on historical index members, going back to the year 2001. The earliest available list was retrieved from this database. Of these companies, monthly stock prices are downloaded from Bloomberg. As not all Nikkei-firms were listed over the whole research period, due to e.g. mergers or de-listings, another selection is made. Only the companies with at least 70 observations of monthly stock returns (out of a possible 126) are included in the sample, to ensure statistically significant tests. Following this selection, 11 companies1 are removed from the sample, leaving a final
set of 188 companies2.
Since Japanese companies do not make use of the commonly used industry codes as SIC, NACE or NAICS, a large share of the sample firms has no data available on this matter. Bloomberg uses its own industry qualification system for all its firms in the database. Thus, for reasons of data availability, the Bloomberg level I classification (Industry Group) is used to divide the sample by industries. The number of sample firms per industry is presented in table 3.1.
3.3. Financial data
This study requires two types of financial data: exchange rate data and stock market data. In this section, the choice for different data, its sources and its descriptive characteristics are discussed.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 The excluded companies are: AC Real Estate Corp, Aoki Corp, Japan Energy Electronic Mate, JFE Steel Corp,
Mitsukoshi Ltd/Old, Niigata Engineering Co Ltd, Nippon Paper Industries Co, Nissho Iwai Corporation, NKK
Corporation, Panasonic Mobile Communication and Sato Kogyo Co Ltd.
!
2 A list of all sample companies is provided in Appendix IV
!
Table 3.1 - Sample firms per industry
Industry Amount Basic Materials 25 Communications 4 Consumer, Cyclical 46 Consumer, Non-cyclical 32 Energy 3 Industrial 67 Technology 6 Utilities 5 Sample: 188
Excluded: Data Availability 11
Excluded: Financial 26
A nominal trade-weighted index of the yen against a set of 15 major currencies3 is used as
input for the exchange rate variable. The data is constructed by the Bank of Japan on a a monthly basis and published on its website. The trade-weighted index is used because of its frequent use in previous literature (e.g. Jorion, 1990; Choi & Prasad, 1995; He & Ng, 1998). The trade-weighted index captures the value of the yen relatively to its use in international trade. As the sample consists of Japan’s largest companies, many sample members perform trade with a variety of countries and are thus subject to a range of exchange rates. Although not company-specific, this range of exchange rates is best approximated by the trade-weighted exchange rate as this variable relates to a number of currencies. Besides, the choice has been made to use nominal exchange rates rather than real exchange rates, as is common in researching exchange rate exposure (Bodnar & Wong, 2003). The first reason underlying this choice is practicality; the nominal and real exchange rates are virtually identical on a monthly basis. The second reason is that real exchange rates are generally not directly priced in the stock market. Real exchange rates are defined by the nominal exchange rates corrected for inflation. As inflation is not instantly observable, stock prices react to the nominal exchange rate rather than the real exchange rate. Stock prices only adjust to inflation (and thus real exchange rates) when the inflation statistics are announced.
A graphical overview of the trade-weighted value of the yen is displayed in figure 3.2. The yen was at its weakest early in the investigated time period, and remained relatively low (under 300) during the Asian Financial Crisis, which had its onset in July 1997 and the consequences lasted roughly to early 1999. The yen then gained until 2000 before gradually declining in value until the end of the research period.
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3 US dollar / Chinese yuan / EMU Euro / Korean won / New Taiwan dollar / Hong Kong dollar / Thai baht /
Singapore dollar / Pound sterling / Malaysian ringgit / Australian dollar / Indonesian rupiah / Philippine peso / Canadian dollar / Mexican peso. Currencies are weighted on their share in Japanese exports, adjusted on a yearly
Monthly stock price data of the 188 sample companies is taken from Bloomberg. In the analysis, closing rates on the last trading day of each month are used. Data of the monthly Topix index is retrieved from www.econstats.com. Again, closing rates on the last trading day of each month are used for the analysis. The Topix index is chosen, as it is a very diversified index of small and large Japanese listed companies. The role of the market variable is to correct for macroeconomic movements that influence both stock prices and exchange rates. However, when the correlation between index members and the index itself is too high, the index itself is obviously exposed to the exchange rate results. Due to its diversification, the Topix index has a slightly lower correlation with the sample companies than its main alternative, the Nikkei-225 index, which is the index consisting of Japan’s 225 largest listed companies (67 and 74 sample companies with correlation over 0.5, respectively). As a test for robustness, a sensitivity analysis will be performed in which a global index variable will be used as an alternative market variable. Yen fluctuation should not influence a global stock index, although the downside is that Japan-specific macroeconomic circumstances are not captured in a global variable.
During the period of research, large fluctuations in the Topix have taken place. Starting at a rather high level, the Asian Financial Crisis caused sharply falling stock prices. Ahead of the economy, the index started recovering in the final months of 1998 and reached a peak in early 2000. The technology boom ended around this time, which caused stock prices to decline
Figure 3.2 - Monthly nominal trade-weighted yen values
again until mid-2002. A healthy recovery followed until October 2007, at which the highest point in the dataset is reached (1774). This moment corresponds with the start of the credit crisis in the U.S., upon which stock prices reacted negatively. To exclude the possible consequences of this crisis, the research period ends at 31 December 2007. A graphic overview is provided in figure 3.3.
Descriptive characteristics and the distributions of the three variables (stock prices, Topix and exchange rate) included in the regression model are shown in table 3.2. The Jarque-Bera test statistics displayed in the table indicate that all three variables are normally distributed. Therefore, an ordinary least squares regression method is justified for analysing the relationships between the variables.
Figure 3.3 - Monthly Topix values
3.4. Robustness tests
The results of the analysis are subject to several choices regarding the use of variables. In order to ensure that the results are reliable and do not differ significantly depending on the choices made, the exchange rate variable and the market variable are subject to robustness tests.
Table 3.2 – Descriptive statistics of stock price data of sample firms
Stock returns of 188 sample companies
Observations 23688 Average 0.0047 Standard Deviation 0.1084 Minimum -0.5776 25th Percentile -0.0582 Median 0.0018 75th Percentile 0.0656 Maximum 1.9640 Jarque-Bera (sign.) 39185.66 (0.00) Topix index Observations 126 Average -0.0001 Standard Deviation 0.0488 Minimum -0.1233 25th Percentile -0.0343 Median -0.0021 75th Percentile 0.0352 Maximum 0.1314 Jarque-Bera (sign.) 31.77 (0.00)
Trade-weighted exchange rate
Bartram (2004) argues that the trade-weighted exchange rate is determined on a country level, whereas the research on exposure is done on a company level. Consequently, the sample companies’ distributions of international operations do not match the distribution in the exchange rate variable. Bartram chooses to use the bilateral exchange rate with the U.S. dollar instead, to which most companies are exposed. His findings show that the research outcomes using the trade-weighted and bilateral exchange rate variable are similar for his sample of 373 German companies. Through changing the exchange rate variable, the following model is used in this robustness test:
(6)
In addition, the choice of the market variable is subject to a robustness test. Due to its high correlation with the sample firms, the Topix is likely to be exposed to the exchange rate. Consequently, the actual influence of the exchange rate on stock prices can be misrepresented as part of the impact is captured in the market variable. This is what Bodnar & Wong (2003) refer to as residual exposure; only the exposure different from the market exposure is captured in a regression model with an included market variable. The use of an index not solely made up of Japanese companies limits this problem, although the macroeconomic factors affecting only Japan will not be included in the market variable. The Dow Jones STOXX Global 1800 index is used to approximate for world stock price movements, an index made up of the 600 largest companies of Europe, the Americas and Asia-Pacific, respectively. Equation (7) represents the model with the alternative market variable
(7)
3.5. Determinants of exposure
As was proven in different empirical studies, exchange rate exposure is influenced by several company characteristics and by the industry in which a company operates. In this study, the changes in the company characteristics that occur between recessions and periods of growth are tested for having an impact on the changes in exchange rate exposure. A regression, similar to (3) in paragraph 2.5, of company exposure to a set of company characteristics (firm size, foreign sales ratio, hedging incentives and keiretsu affiliation) is performed to find out which factors determine exchange rate exposure.
3.5.1. Company size
exchange rate movements. To measure the relationship between company size and exchange rate exposure, total assets data is retrieved from Bloomberg and its natural logarithm is used as the variable that determines company size. The natural logarithm is taken because of the sheer size differences in the value of this variable compared to the other variables, which may distort results. By taking the natural logarithm, this problem is avoided.
3.5.2. Foreign sales ratio
A different variable that is often examined in research concerning exchange rate exposure is the foreign sales ratio. The higher the foreign sales, the more a company is exposed to the direct transaction exposure and the more the exchange rate can influence future cash flows through the operating exposure. This argument is used in different studies as Jorion (1990), Choi & Prasad (1995) and Allayannis & Ofek (2001). Foreign sales data is available for 137 of 198 sample firms in the Japan Company Handbook (Spring 1999 until Autumn 2008), which defines a company’s foreign sales ratio by its international sales divided by its total sales.
3.5.3. Hedging activities
is due to a firm’s dependence on external financing and access to costly external financing. The underinvestment cost hypothesis suggests a negative relationship between growth opportunities and costly external financing. The book-to-market ratio is applied to proxy for a company’s growth opportunities. The lower the BM, the greater the incentive to use currency derivatives to hedge, in order to reduce underinvestment costs. As a result, the exchange rate exposure decreases. The BM is calculated as the ratio of a company’s book value of equity to its market value of equity.
In this study, He & Ng’s approximation of hedging incentives is followed. All data is taken from the Japan Company Handbook (Editions Spring 1998 to Spring 2008) to estimate the variables in the same manner as He & Ng (1998). However, as inventory data is unavailable, the quick ratio cannot be used. Instead, liquidity is approximated by the current ratio, calculated as the ratio of current assets to current liabilities. Data regarding the determination of these financial ratios is available for 170 out of 188 sample firms.
3.5.4. Keiretsu firms
In Japan, many enterprises are part of a complex network of customers and supplies, with interlinked financing and trading ties. This is contrary to Anglo-American use, where investors are viewed as outsiders to the firms: shareholders hold only equity and creditors only debt (Berglöf & Perrotti, 1994). The most well known examples of such networks are the Japanese keiretsu, an informal grouping of Japanese firms grouped around several financial institutions (Koch & Saporoschenko, 2001). The main feature of the keiretsu is the reciprocal shareholdings amongst member firms. Cross-holdings by group financial institutions are important in all keiretsu, with the main bank holding an average 2 to 5 percent of every member firm. When indirect holdings are included, group financial institutions account for more than half of intragroup shareholdings. As a result, banks own stock in the institutions that they provide with credit, while borrowers own stock in the banks. (Koch & Saporoschenko, 2001) In the more cohesive keiretsu (Mitsubishi, Mitsui and Sumitomo) cross-holdings by non-financial firms are much more developed than in the less cohesive keiretsu (DKB, Fuyo and Sanwa) (Berglöf & Perrotti, 1994).
The argument that the keiretsu firms have a different exposure to the exchange rate than the non-keiretsu firms follows from the difference in hedging incentives. As keiretsu firms are less likely to face bankruptcy, since the group financial institutions will intervene, they have a lower probability of financial distress and a lesser need to maintain high short-term liquidity (He & Ng, 1998). Consequently, keiretsu firms are expected to demonstrate a higher individual exposure to the exchange rate. This is empirically confirmed in He & Ng (1998). All sample companies are qualified as either a keiretsu or non-keiretsu firm, according to the data provided in Osono (1995) and Ibata-Arens (2005). The assumption is made that since 2005 no changes in the ownership of companies has taken place. Accordingly, a dummy variable is constructed that takes the value of 1 in case of a keiretsu firm and a value of 0 in case of a non-keiretsu firm.
3.5.5. Industry
Following Bodnar & Gentry (1993), who argued that the products that are traded in different industries are important in the determination of exchange rate exposure, many studies have added an industry-analysis to their research, e.g. Choi & Prasad (1995), Griffin & Stulz (2001) and Aquino (2005). Several reasons for industry differences in exchange rate exposure exist. Bodnar & Gentry claim that industries of internationally traded goods are more exposed to currency risk than non-traded goods, i.e. goods for which high transportation costs prevent international trade, and prove this point with a regression analysis. Marston (2001) argues that the competitive environment of an industry largely determines exchange rate exposure, and shows that monopoly and duopoly firms react differently towards exchange rate fluctuations.
Williamson (2001) focused his research on the transport industry and found significant exposure for automotive producers in the U.S. and Japan with regards to the U.S. dollar and the Deutschmark, proving that the competition in Europe with U.S. firms partly determines the companies’ exchange rate exposures. In the extensive research that He & Ng (1998) performed on Japanese companies, they found the transport industry to be negatively exposed to the U.S. dollar too. Other industries that showed results of negative exposure in their research are chemicals, iron and steel, machinery and precision equipment. In the research of Bodnar & Gentry (1993), the Japanese construction and oil industries show positive exposure and the chemicals, electrics and precision instruments show negative exposure at the 5% significance level. Special attention is given to these industries in the following analysis, to see whether the research results correspond with previous research.
goods, industrial), due to the sample size. Of each of these four industries a dummy variable is constructed that takes the value of 1 when a company is part of that specific industry, and zero if otherwise. This method is similar to the analysis of He & Ng (1998).
3.5.6. Regression analysis of determinants
The variables described in paragraph 3.5.2 to 3.5.5 are tested for having a significant impact on the exchange rate exposure of a company and, especially, on the differences in exposure between periods of economic growth and decline. This is done in a way similar to e.g. Jorion (1990), He & Ng (1998). Chow & Chen (1998) and Doidge, Griffin & Williamson (2006). An ordinary least squares regression is constructed with the exchange rate coefficients (!1i) of
regression model (5) as dependent variable and the above named factors as independent variables. The regression formula is as follows
(8)
Besides, a second regression is performed in which the impact of industry characteristics is determined. Four dummy variables of the four industries that make up a substantial part of the sample (basic materials, cyclical consumer goods, non-cyclical consumer goods and industrial) are added to the regression model. As a consequence, the following model is tested:
(9)
In the equations above, TAi represents a sample company’s total assets, FSi its foreign sales
ratio, DEi its debt ratio, CRi its current ratio, DIVi its dividend payout ratio and BMi its
book-to-market value of equity. KEi is a dummy variable that takes the value of one if the firm
In order for least squares regressions to provide reliable results, it is important to determine the independence amongst variables. When independent variables are highly correlated with each other, multicollinearity becomes a problem. Multicollinearity leads to distorted variable coefficients, although the overall power of the tests is not inflicted. Table 3.4 below displays the correlation matrix of the variables used in this regression model. The correlation between debt ratio and current ratio is significant and the regression is therefore run without each of these two variables to ensure the reliability of the results.
As it turns out, the correlation between debt ratio and current ratio has a substantial impact on test results. The impact of the current ratio is overstated when both variables are included in the model. Appendix II provides the test results of the regressions in its original form and when only the current ratio is included. It shows that the significance of the current ratio is due to the correlation with the debt ratio. The test results as described in the following chapter
Table 3.3 - Summary statistics of regression variables
Total Assets Foreign Sales Debt Ratio Current Ratio Dividend Ratio Book-to-Market
N 186 133 162 162 162 162 Average 2,189,900 0.29 0.69 1.82 0.49 2.21 Standard Dev. 2,736,624 0.19 0.22 1.24 0.39 3.12 Minimum 59,876 0 0.05 0.11 0.00 0.00 25th percentile 495,439 0.11 0.46 0.97 0.24 0.89 Median 893,553 0.26 0.65 1.59 0.43 1.55 75th percentile 1,706,243 0.42 0.78 2.01 0.73 2.95 Maximum 19,415,760 0.79 0.93 14.60 6.00 15.43
Total assets is displayed in million yen, the natural logarithm of this variable is used in the regression. Of the Keiretsu dummy variable, 80 out of 188 companies are affiliate of at least one keiretsu.
Table 3.4 - Correlation Matrix
Total Assets For. Sales Debt Ratio Cur. Ratio Div Ratio B-t-M ratio Keiretsu
are the outcomes of the adjusted regression models with the debt ratio included and the current ratio excluded in the equation, as displayed below.
(8a)
4. Results and Interpretation
In the following paragraphs the results of the tests as described in the methodology section are discussed. The aim is to find out whether the value of the yen has a significant impact on the stock prices of Japanese companies and what the impact of the state of the economy (i.e. growth or recession) is on this relationship. Moreover, underlying reasons for differences in exchange rate exposure between recessions and growth periods are examined by analysing the determinants of exchange rate exposure.
4.1. Exposure of individual sample firms
For each sample company, monthly stock returns are regressed against contemporary exchange rate movements, as explained in equation (5). Exchange rate exposure is proven when the coefficient of the exchange rate variable is significantly different from zero. Table 4.1 provides an overview of the test results of the exposure coefficients.
As can be derived from the table above, with regards to the whole sample period, 33 companies have a significant exposure to the exchange rate, which represent 17.5% of the sample. This percentage corresponds with previous research, in a way that Japanese companies have a higher exchange rate exposure than commonly found for companies from different industrialised countries. In respect of the United States, with exposure shares ranging from 2.3% (Chow, Lee & Solt, 1997) to 14.6% (Bodnar & Wong, 2003), Germany (7.5%; Bartram, 2004), Australia (10.3%; Loudon, 1993) and the Eurozone (13%; Muller & Verschoor, 2006), the percentage of significantly exposed Japanese companies found in this research is high. However, previous studies on Japanese companies often show a higher exposure. In the sample of He & Ng (1998), the exchange rate has a measurable impact on
Table 4.1 - Exchange rate exposure coefficients (trade-weighted yen movements)
Quartiles
Sample Period N min quartile 1 median quartile 3 max N+
N-Full Period 188 -1.3539 -0.2848 0.0636 0.4509 1.4979 20 13
Recession 187 -1.5885 -0.2953 0.2217 0.7091 2.1871 22 16
Growth 187 -2.0007 -0.4476 -0.0807 0.2845 1.9478 6 9
stock prices of 25% of the sample companies; in Chow & Chen (1998) this percentage is almost 30%.
With regards to the subsamples, there is an overwhelming difference in the amount of companies with significant exchange rate exposure between recession periods and growth periods. During the recession months 38 companies (20%) were exposed to the exchange rate, whereas during the months of the growth periods only 15 companies (8%) were significantly exposed. Thus, the argumentation that companies are more sensitive to the exchange rate in an economic downturn is fully supported by the results of this analysis. Both positive and negative exposure increases during a recession, thereby implying that the stock prices of both net importers and net exporters are more impacted by the exchange rate. A closer look is taken at the differences between results of the recession and the growth samples below. Table 4.2 provides and overview of the differences in exposure coefficients.
The fact that only three companies are consistently significantly exposed to the exchange rate indicates that the state of the economy has a substantial impact on test results. During a period of recession, different companies are subject to the exchange rate. Of the 38 companies with significant exposure during a recession period, 33 (87%) are not exposed at all during economic growth and two companies have changed sign. The paired t-test of equality, used for testing whether the sample exchange rate betas of the two subsamples are likely to come from the same underlying population, points out that at the 5% significance level the results of the recession period are statistically different from the growth period.
Table 4.2- Differences between growth and recession (trade-weighted yen)
Significant in both states (same sign) 3
Significant in both states (opposite sign): 2
Significant during recession, but not during growth: 33
Not significant during recession, but significant during growth: 10
Paired t-test of equality (N=186)
Full Sample Recession Growth
Full Sample x
Recession 2.4747 x
Growth 3.0984 4.0280 x
An overview of the coefficient of significantly exposed companies, separated by positive and negative exposure, is provided in table 4.3 below. The average coefficients indicate that for significantly negative exposed companies a one percent decrease in the exchange rate (ceteris
paribus) leads to an average stock price decline of 0.8761% over the whole sample period.
During the recession period this elasticity is considerably larger (±1.6%) than during the growth period (±1.0%) and the full sample period (±0.8%). Thus, it can be concluded that not only leads the recession period to different companies being exposed to the exchange rate, the companies that face exposure are influenced by the exchange rate to a higher extent.
Before continuing to the analysis of the underlying reasons for these evident differences in exposure between periods of recession and economic growth, the robustness of the results to the use of different variables is tested in the subsequent paragraph.
4.2. Robustness tests
As outlined in the previous chapter, the results are subject to choices in the research methodology. The power of the results increases when the outcomes of this research are not dependent on the choices made. Consequently, the choices regarding the use of a multilateral rather than a bilateral exchange rate variable and a national market index rather than a global market index are inspected. These tests are purely performed as a check for the robustness of the research results. Table 4.4 below gives an overview of the results when the yen-dollar relationship is used instead of the trade-weighted yen index, thus using the following regression model:
(6)
Table 4.3 - Coefficients of significantly exposed companies
N Average St.Dev Min Max
Full sample period
The results of this alternative regression are rather similar to the outcomes of the original regression model. Over the entire sample period, slightly more sample companies show significant exposure (35 versus 33). Again, the number of significantly exposed companies is much bigger in the recession period than in the period of growth. However, the difference has declined slightly, due to the fact that fewer companies are exposed during the recessions (38 versus 33) and more companies during times of growth (15 versus 21). Still, during a recession the number of companies with significant exposure in this sample increases with more than 50%.
Table 4.4 - Exchange rate exposure coefficients (yen - U.S. dollar movements)
Quartiles
Sample Period N min quartile 1 median quartile 3 max N+
N-Full Period 188 -1.2747 -0.1924 0.0584 0.2804 1.4937 15 20
Recession 187 -1.6504 -0.2745 0.0863 0.3641 1.2004 13 20
Growth 187 -1.6564 -0.2699 0.0371 0.3423 1.5204 9 12
N represents sample size, N+ and N- report the number of firms with positive and negative significant exchange rate exposure, coefficients at the 5% level, respectively.
Table 4.5 - Differences between recession and growth periods (yen-dollar)
Significant in both states (same sign) 6
Significant in both states (opposite sign): 1
Significant during recession, but not during growth: 26
Not significant during recession, but significant during growth: 14
Paired t-test of equality (N=186)
Full Sample Recession Growth
Full Sample x
Recession 2.6309 x
Growth 1.9896 2.0238 x
The differences between results of the two subsamples in this robustness test are presented in table 4.5. Results are consistent with the original analysis as a very large share of companies is significant in one of the subsamples, but not in the other. Moreover, the equality tests shows that the beta coefficients of the exchange rate variable are statistically different between the two subsamples, parallel to the original analysis. The differences and similarities between the outcomes of the original regression with the trade-weighted yen variable and the alternative variable with the yen-dollar exchange rate variable for the full sample are outlined in table 4.6. Eleven companies are consistently positively or negatively affected by the exchange rate regardless the specification of the currency variable. In both samples, approximately half of the exposed companies have significant exchange rate coefficients with the used currency variable, but not with the alternative specification. Interestingly, six firms are significantly exposed in both regressions, but with opposite signs between the two analyses. This may be an indication that different currencies are priced separately in the stock market and that these six firms have extensive ties with both U.S. and non-U.S. firms. A logical explanation would be that these companies import raw materials from the U.S., thereby gaining from a stronger yen, and subsequently export the finished goods to different markets, for which a stronger yen denotes a disadvantage.
Table 4.6 - Differences between sample results (TW vs yen-dollar)
Significant in both samples (same sign): 11
Significant in both samples (opposite sign): 6
Significant with TW yen, but not with yen-dollar: 16
Not significant with TW yen, but significant with yen-dollar: 18
Paired t-test of equality (N=186)
TW: Full Sample TW: Recession TW: Growth
Yen-dollar: Full Sample 1.1692
Yen-dollar: Recession 2.5140
Yen-dollar: Growth 1.3018
