Do Country and Industry Explain Macroeconomic Exposure?

Transcription

1 Do Country and Industry Explain Macroeconomic Exposure? Niclas Andrén Department of Business Administration, Lund University PO Box 7080, Lund, Sweden Phone: Fax: This draft: Abstract This paper studies the importance of the industry a firm competes in and the country where it is domiciled as cross-sectional determinants of the macroeconomic exposures of European listed, financial and non-financial, firms. Contrary to expectations, country is found to be an important determinant of variations in exposure, while industry is found to be rather weak after controlling for differences in industry specialization among countries and geographical concentration among industries. Further, it is shown that there are distinct differences in macroeconomic exposures between countries, while exposures are symmetric across industries. Keywords: Exchange rates, inflation, interest rates, macroeconomic exposure, exposure decomposition. JEL classification: F39, G12, G39 Copyright 2000, Niclas Andrén Research funding provided by the Swedish Network for European Studies in Economics and Business is gratefully acknowledged.

2

3 Do Country and Industry Explain Macroeconomic Exposure? Introduction A firm s performance and value can fluctuate due to unexpected macroeconomic changes macroeconomic risks. A firm s exposure to macroeconomic risk is unique to that firm in the sense that the exposure depends on the particulars of the firm s structure and operations. Exchange-, interest-, and inflation-rate risks expose the firm through, for example, its assets and liabilities, its financing, its choice of production function and location, and its investment practices, but also indirectly through the exposures of its suppliers, customers, and competitors. The firm s ability to adapt to exogenous change and its natural and contractual hedges-in-place also partake in determining the exposure. Given the firm specificity of macroeconomic exposure, should we expect to find any systematic similarities in exposures across firms? For two firms to be similarly exposed to real exchange-rate risk, the two firms must be trading, financing, and investing in the same (or correlated) currencies and have competitors with their costs in the same currencies unless the purchasing power parity holds ex post as well as face similarly intense competition and equally elastic demand (for input and output prices and sales volumes to fluctuate in similar ways). For two firms to be similarly exposed to real interest-rate risk, the two firms must have similar capital structures and be financing and investing in the same currencies and at the same maturities save the real interest-rate parity and the expectations hypothesis hold ex post as well as face equally

4 interest-elastic demand. For two firms to be similarly exposed to inflation risk, the two firms must have equally large net fixed assets and similar capital structures, have input and output prices that are indexed to the same extent and to the same measures of inflation, have competitors with their costs in the same currencies and similar production functions, and face similarly intense competition and equally price- and income-elastic demand. Two potential sources of similarities in macroeconomic exposure are focused here, namely the country where the firm is domiciled and the major industry the firm competes in. Can these two firm characteristics explain cross-sectional differences in macroeconomic exposure? If that were the case, we would find similarities in exposures within industries and countries, rather than across industry and country borders. Sweeney and Warga (1986) show differences among industries in interest-risk exposures, while Bodnar and Gentry (1993), Choi and Prasad (1995), Griffin and Stulz (1998), and He and Ng (1998) provide evidence of industry differences in exchange-risk exposures. Bodnar and Gentry (1993) and Rees and Unni (1998) provide partial evidence of country differences in exchange-risk exposures. In general, as macroeconomic exposure depends on the firm s operations in terms of locations of production, product mix, capital structure, demand elasticities, competitive structure, etc, which could be expected to be more industry than firm specific, it seems natural to expect differences in exposures between, but similarities within industries. On the other hand, if competing firms produce in different currencies the result would be diametrically opposing exposures to real exchange-rate and inflation risk, as well as exposure to inflation and interest-rate risks in different currencies. Economic and financial integration, which is the general rule in today s economic environment, suggest convergence in real exchange and interest rates, which, in turn, suggest leveling of competitive conditions through convergence of costs of capital, factor and production costs, and output prices in different currencies. Integration further allows optimization of capital structure, business portfolio, risk-management practices, etc globally instead of domestically. Leveling of competitive conditions can be expected to translate into convergence in risk 2

5 Do country and industry explain macroeconomic exposure? 3 exposures; knowingly taking on a macroeconomic-exposure posture different from the main competitors is equivalent to the firm speculating on real exchange- and interest-rate fluctuations. Furthermore, integration can be expected to lead to specialization and agglomeration (Krugman, 1991), as well as other non-macroeconomic reasons for selecting sources of funding, locations of production, etc. Specialization and agglomeration will lead to converging exposures to the extent firms specialize in similar fashions and locate in the same places. Overall, economic and financial integration suggests that industry would be a more important determinant of macroeconomic exposure than country. In particular, differences in exposures across countries could be anticipated to be the result of differences in industry specialization. On the other hand, integration could also be an effective force of exposure divergence. Small firms and domestic firms could, for example, take advantage of integration by realizing that there are few benefits from moving operations from one currency to another, from one cost base to another, or from one financing source to another, and instead stick to the domestic currency, the domestic factor and product markets, or the domestic credit markets. In that case, there would rather be similarities in exposures across firms within than between countries. What determines macroeconomic exposure? A partial answer can be given by decomposing exposure into exposure due to country-ofdomicile, due to industry-belonging, and due to firm-specificity. The main question to be answered in this exploration accordingly is: to what extent is a firm s macroeconomic exposure due to country-of-domicile, industry-belonging, and firm-specificity? To accurately distinguish between these explanations of exposure, each effect on macroeconomic exposure must be separated. The empirical estimation strategy utilized to decompose exposure is two-way analysis of variance in a regression formulation (see Suits, 1984; Kennedy, 1986). This method has a rather recent history in financial research, where Heston and Rouwenhorst (1994) was the first study to use it to decompose cross-sectional variations in stock returns into country and industry effects. A few studies have followed suit, but the method has to my knowledge not been applied to decomposition of corporate risk exposures.

6 4 The second section of the paper, besides introducing the data sample, presents the exposures to global macroeconomic developments among 1,201 firms domiciled within the EU. A more detailed analysis of the macroeconomic exposures of these firms is found in Andrén (2000). Section 3 outlines the exposure-decomposition method employed. In Section 4, evidence of the presence of country-specific components of exposure variation, and not industrial-structure components, as an important determinant of the cross-sectional variations in exposures across firms is presented. However, it is also shown that macroeconomic exposure is primarily firm specific. In Section 5 the study is taken one step further by looking for patterns in macroeconomic exposures across industries and countries. Section 6 summarizes the main findings and provides concluding remarks. Assessment of macroeconomic exposure The sample includes 1,201 firms listed on the stock markets in 11 EU countries (see Table 1). The investigated period is March 1989 to December 1998 using monthly data. The sample contains all firms listed on a stock market in an EU country as of early Autumn 1999 with full sets of data in the databases of the Reuter Securities 2000 System. All stock prices are corrected for splits etc, but not for dividends. Each firm is assigned to the country where it is domiciled. The firms are also assigned to 26 industry categories, based on industry classifications provided by Wright Investors Service, and are further divided into seven industry sectors: Capital equipment (aerospace, electrical, electronics, and machinery and equipment) Construction (including building materials) Consumer goods (apparel and textiles, automotive, food and beverages, drugs, cosmetics, and health care, recreation, and retailers)

7 Do country and industry explain macroeconomic exposure? 5 Financial (banking, financial services, insurance, investments, and real estate) Materials (chemicals, energy, metal production and metal products, and paper) Multi (diversified and miscellaneous) Services (printing and publishing, services organizations, transportation, utilities, and wholesalers) Table 1, Panel A summarizes the performance of the stocks in the 11 countries and seven industries during the period 1989 to 1998 expressed in common currency (German mark). The average return of the firms in each country and industry is homogeneous, lying close to zero, while volatilities vary, more across countries than industries, averaging 9.2% per month. Close to 40% of the firms are domiciled in the UK, while between two and four percent are domiciled in Austria, Belgium, Finland, Ireland, and Sweden. In terms of market values, the distribution is slightly different, with Finland representing 14% of total market value (due to Nokia), and Denmark and France representing six percentage points smaller shares than when counting number of firms (due to many small firms in these countries). The industrial composition of countries, as well as the geographical distribution of industries is not uniform (see Panel B). For example, Finland has a large share of firms in the capital-equipment industry, while Austria has many firms in the construction industry and no firms in the multi industry. Belgium has many financial firms and few services firms, while the Netherlands has many services firms. Oxelheim and Wihlborg (1987; 1997) argue that macroeconomic exposure should be measured with multiple regression to derive marginal exposures to real exchange-rate, real interest-rate, and inflation risk: r b s c i d + u i, t = a + t + t + ðt t where r i,t is the unexpected relative change in firm i s real market value, s t is a set of exchange rates, i a set of interest rates, a set of inflation rates, and u t is a white-noise term capturing unexpected changes in the real market value orthogonal to the macro prices. An asterisk denotes deviations from expectation, r t = r t E t-1 [r t ] and denotes relative change. The coefficient vectors b, c, and d represent the sensitivity of the (1)

8 6 Table 1 Summary information for European sample firms A: Average returns and standard deviations Country Number of firms Average real return Average Industry standard deviation Number of firms Average real return Average standard deviation Austria Capital Belgium Construction Denmark Consumer Finland Finance France Material Germany Multi Ireland Service Italy Netherlands Sweden UK All firms 1, B: Number of firms by country and industry Capital Constructiosumer Con- Finance Material Multi Service Austria Belgium Denmark Finland France Germany Ireland Italy Netherlands Sweden UK real return of firm i to a one-percent unexpected change in each macro price, after controlling for movements in the other macro prices. Here, macro-price variables representing the dominant economic regions in the world, Germany (representing the EU), Japan, and the US, were chosen. Focusing on these countries would capture the majority of global trade and capital flows, and the macroeconomic developments in these countries could be assumed to proxy for global macroeconomic developments. The model used to assess macroeconomic exposure is

9 Do country and industry explain macroeconomic exposure? 7 r i, t = β β i β s 1 JPST, t β π USIN, t DEM / USD, t + β i 6 + β 11 DELT, t π + β s JPIN, t 2 + β i 7 + β DEM / JPY, t USLT, t r 12 DCm, t + β i + β i 8 + u t 3 DEST, t JPLT, t + β i + β π 9 4 USST, t DEIN, t where all variables are measured as unexpected changes and all changes are measured as continuous returns. Nominal domestic-currency denominated stock returns were converted to real returns by deflating with the domestic consumer-price index. DEM/USD and DEM/JPY are the monthly returns in the German mark vs the US dollar and the Japanese yen; DEST, JPST, and USST are the monthly changes in the short-term German, Japanese, and US interest rates; 1 DELT, JPLT, and USLT are the monthly changes in the yields on long-term German, Japanese, and US government bonds; DEIN, JPIN, and USIN are the monthly changes in the German, Japanese, and US consumer-price indices; and r DCm,t is the domestic stock-market return orthogonal to the other macro-price variables, which controls for country-specific left-out 2 variables and extraneous events. The conservative random-walk assumption was employed for all variables, thus making all changes in the variables unexpected. 3 The macro-price data were taken from IMF s International Financial Statistics and the OECD Statistical Compendium. Unit-root tests were performed with augmented Dickey-Fuller tests, showing that all variables are stationary in returns. 4 Tests for multicollinearity were undertaken. There are some indications of multicollinearity, as shown in the bivariate-correlation matrix in Table 2, but they are not considered serious enough to motivate any changes in the model specification. + + (2) The short-term rates used are 3-month T-bill rates in the case of the US and callmoney rates in the cases of Germany and Japan. The stock-market indices used are the ATX50 (Austria), Brussels General (Belgium), Copenhagen General (Denmark), HEX (Finland), CAC40 (France), DAX (Germany), Irish General (Ireland), MIB General (Italy), AEX (the Netherlands), Stockholm General (Sweden), and the FTSE All Share (the UK). A series of time-series forecasts for the macro-price variables were evaluated, showing that AR (1) processes and random walks provide equally good forecasts in terms of unbiasedness and forecast accuracy. The results are available upon request from the author. The results of the tests are available upon request from the author.

10 8 Table 2 Bivariate-correlation matrix for macro-price risks, DEM USD DEM JPY DEST USST JPST DELT USLT JPLT DEIN USIN JPIN DEMUSD DEMJPY DEST USST JPST DELT USLT JPLT DEIN USIN.329 JPIN The results of the estimations of macroeconomic exposures are presented in Table 3 (for the entire sample and for industries), Table 4 (for countries), and Table 5 (shares of significant firms). Altogether, 80% of the regressions are significant, with an average explanatory value of 32%. The most important explanatory variable is the domestic market index, which has an average partial explanatory value of 22%. Most firms (70%) are significantly exposed to at least one macro-price risk. A fifth of the firms are significantly exposed to the DEM/USD with a large majority being affected positively by depreciations of the DEM. A quarter of the firms are significantly exposed to the German bond yield with a large majority being affected negatively by interest-rate increases. Between 10% and 16% of the firms are significantly exposed to the German and US short-term, and the US long-term interest rates and to German and US inflation. For almost all macro prices, there is homogeneity across industries in the average signs and sizes of the exposures. However, some industries stand out in terms of shares of significant variables deviating from the overall average. The capital-equipment industry has a small share of significant exposures to the German long-term interest rate, while the construction industry is relatively more exposed to the yen and to the German bond yield and inflation. The consumer-goods, multi, and services industries have larger-than-average shares of significant exposures to German inflation, while the materials industry is overly

11 Do country and industry explain macroeconomic exposure? 9 Table 3 Distribution of exposure coefficients across industries All firms Capital Construction Consumer Finance Material Multi Service β (.01) (.02) (.02) (.01) (.01) (.01) (.01) (.01) DEMUSD (.34) (.41) (.37) (.34) (.29) (.32) (.33) (.35) DEMJPY (.30) (.37) (.33) (.30) (.26) (.29) (.30) (.31) DEST (.24) (.30) (.26) (.24) (.21) (.23) (.24) (.25) USST (.23) (.29) (.25) (.23) (.20) (.22) (.23) (.24) JPST (.09) (.11) (.10) (.09) (.08) (.09) (.09) (.10) DELT (.30) (.37) (.33) (.30) (.26) (.29) (.30) (.31) USLT (.29) (.35) (.31) (.29) (.25) (.27) (.28) (.30) JPLT (.08) (.10) (.09) (.08) (.07) (.08) (.08) (.09) DEIN (2.18) (2.68) (2.38) (2.18) (1.90) (2.09) (2.15) (2.28) USIN (4.34) (5.32) (4.73) (4.32) (3.77) (4.15) (4.28) (4.53) JPIN (1.79) (2.19) (1.95) (1.78) (1.55) (1.71) (1.76) (1.87) Market (.17) (.20) (.18) (.17) (.14) (.16) (.16) (.18) R F DW Standard errors in parentheses. exposed to German and US interest-rate fluctuations. The financial industry is relatively more exposed to the yen, to German and US shortterm, and German long-term interest rates, and to US inflation. There are differences in exposures across countries, with heterogeneity in average signs, sizes, and shares of significant exposures. In particular, Austria, but also Belgium, stand out with small exposures to almost all macro-price risks and with low explanatory values overall. Finland and the UK stand out with the opposite pattern: relatively large

12 10 shares of significant exposures to many variables. About half of the Finnish firms are significantly exposed to German and US long-term interest rates and German inflation. More than half of the Swedish firms are significantly exposed to the dollar. The same is true for two-fifths of the German and Dutch firms. A quarter of the Swedish firms are significantly exposed to the Japanese short-term interest rate, while a quarter of the Finnish firms are significantly exposed to the Japanese long-term interest rate. Decomposition of exposures by country and industry In order to separate country, industry, and firm-specific exposures, I advance the following model for the exposure of firm i that belongs to industry k in country l: β = α + γ + δ + ε i k l i where β i is firm i s exposure to a macro-price variable, α is the equallyweighted average exposure across all firms in the sample (equal to the average exposures presented in the first column of Table 3), γ k is the industry effect, δ l is the country effect, and ε i is a white-noise term capturing the firm-specific component of exposure. The model allows separate influences of industry and country, but rules out any interaction between the two. Defining an industry dummy I i,k that is equal to one if firm i belongs to industry k and zero otherwise, and a country dummy C i,l that is equal to one if firm i belongs to country l and zero otherwise, (3) can be written as a cross-sectional regression model: β α γ! γ δ δ + ε (4) i = + 1 Ii, Ii,7 + 1Ci,1 +! + 11Ci, 11 When the coefficients of the dummy variables are ignored, (4) shows the grand-mean exposure, α. The coefficients of the dummy variables then show the extent to which exposures in the respective countries (averaged over all industries) and in the respective industries (averaged over all countries) differ on average from the grand mean. i (3)

13 Do country and industry explain macroeconomic exposure? 11 Table 4 Distribution of exposure coefficients across countries AT BE DK FI FR DE IE IT NL SE UK β (.02) (.01) (.01) (.02) (.01) (.01) (.01) (.02) (.01) (.02) (.01) DEMUSD (.42) (.31) (.32) (.47) (.33) (.29) (.34) (.43) (.30) (.40) (.32) DEMJPY (.38) (.28) (.28) (.42) (.30) (.26) (.30) (.38) (.27) (.36) (.29) DEST (.30) (.22) (.23) (.33) (.24) (.21) (.24) (.31) (.22) (.29) (.23) USST (.29) (.21) (.22) (.32) (.23) (.20) (.24) (.29) (.21) (.28) (.22) JPST (.12) (.08) (.09) (.13) (.09) (.08) (.09) (.12) (.08) (.11) (.09) DELT (.38) (.28) (.28) (.42) (.30) (.26) (.30) (.38) (.27) (.36) (.29) USLT (.36) (.26) (.27) (.40) (.28) (.24) (.29) (.36) (.26) (.34) (.27) JPLT (.11) (.08) (.08) (.12) (.08) (.07) (.08) (.11) (.08) (.10) (.08) DEIN (2.74) (2.00) (2.04) (3.02) (2.15) (1.86) (2.21) (2.76) (1.96) (2.62) (2.08) USIN (5.44) (3.98) (4.06) (6.00) (4.27) (3.69) (4.39) (5.49) (3.90) (5.21) (4.13) JPIN (2.24) (1.64) (1.67) (2.47) (1.76) (1.52) (1.81) (2.26) (1.61) (2.15) (1.70) Market (.11) (.19) (.18) (.16) (.14) (.12) (.16) (.12) (.16) (.16) (.20) R F DW Standard errors in parentheses. When assigning dummy variables, the traditional approach is to use reference coding. Unfortunately, reference coding does not allow uniquely identifying country and industry effects; it is only possible to measure cross-sectional differences across industries and across countries, which means that a benchmark must be chosen. An alternative is to use effects coding, however, when there are unequal cell numbers, as is the case here, the resultant α will not be the grand mean; it will be the average of the cell means. The alternative used here is to use reference

14 12 Table 5 Shares of significant exposures DEM USD DEM JPY DEST USST JPST DELT USLT JPLT DEIN USIN JPIN AT BE DK FI FR DE IE IT NL SE UK Capital Construction Consumer Finance Material Multi Service coding with the grand mean as the reference. To implement this measure, the following restrictions are imposed, 7 k k = 1 n γ 11 l l= 1 l k = 0 n δ = 0 where n k and n l are the number of firms in industry k and country l. In practical terms, (4) is estimated with the constraints γ 7 = 0 and δ 11 = 0. The required estimates are produced by adding a constant g to γ 1 to γ 7 and a constant h to δ 1 to δ 11, and then subtracting (g + h) from α, where g is determined by (5) and h by (6) as, g = 6 k= 1 γ k n k n (5) (6) (7) h = 10 l= 1 nl δl n (8)

15 Do country and industry explain macroeconomic exposure? 13 where n is the total number of firms. The related standard errors are derived from the covariance matrix with the shares of firms in the calculations of g and h as weights. 5 By construction, the firm-specific components of exposure, ε, are orthogonal to all industry and country dummies, and are equal to zero in each individual industry and country as well as overall. Since the sum of the industry- and country-specific components of exposure is zero by definition, the least-squares estimate of α is equal to the grand mean. αˆ + γˆ k is an estimate of the pure exposure of an equally weighted, geographically diversified (same country composition as the grand mean) portfolio of firms in industry k. This estimate is accordingly free of country-specific components of exposure. Equivalently, αˆ + δ ˆ l is the pure exposure of the equally weighted, industrially diversified country portfolio l. Empirical results F-tests from two-way ANOVAs show that country is a significant divisor of cross-sectional variations in exposures for all variables, 5 The derivation of the converted coefficients and standard errors for Austria and the UK (the left-out variable) for the DEMUSD model: h = = b AT = =.22 and σ AT X where UK 10 = 10 i= 1 j= 1 i X j σ i, j X AT = 1 + ; X BE = ;!; X SE = ; = UK i= 1 j= 1 b =.11 and σ X i X j σ i, j where X AT = ;!; X SE =

16 14 Table 6 Industry components of macroeconomic exposures α Capital Construction Consumer Finance Material Multi Service R 2 DEMUSD (.05) (.09) (.08) (.07) (.06) (.06) (.05) (.03) DEMJPY (.04) (.07) (.06) (.05) (.04) (.04) (.03) (.02) DEST (.03) (.06) (.05) (.04) (.04) (.04) (.03) (.02) USST (.04) (.07) (.05) (.05) (.04) (.04) (.03) (.02) JPST (.01) (.02) (.02) (.02) (.02) (.01) (.01) (.01) DELT (.04) (.08) (.07) (.06) (.05) (.05) (.04) (.02) USLT (.04) (.08) (.06) (.05) (.05) (.05) (.04) (.02) JPLT (.01) (.02) (.02) (.02) (.02) (.01) (.01) (.01) DEIN (.33) (.59) (.49) (.42) (.39) (.37) (.31) (.17) USIN (.65) (1.16) (.95) (.83) (.77) (.73) (.61) (.34) JPIN (.26) (.47) (.39) (.34) (.31) (.30) (.25) (.14) Standard errors in parentheses. Underlined values are significant at the 5% level. whereas industry is significant for all variables except DEM/USD, the long-term Japanese interest rate, and US inflation. The decomposition of exposures into industry and country components is presented in Table 6 (for industries) and Table 7 (for countries). The overall R 2 s vary between macro prices, from.02 for the Japanese long-term interest rate to.20 for the US long-term interest rate. All regressions are significant on the 1% level. The partial r 2 s are consistently higher for countries than for industries; the average partial r 2 for the addition of country dummies to a regression only containing industry dummies is 10.1%, while the average partial r 2 for the addition of industry dummies to a regression only containing country dummies is 1.9%. On average, then, country is five times as important as industry but the firm-specific component is by far the most important.

17 Do country and industry explain macroeconomic exposure? 15 Table 7 Country components of macroeconomic exposures AT BE DK FI FR DE IE IT NL SE UK DEMUSD (.07) (.06) (.04) (.07) (.03) (.04) (.08) (.03) (.04) (.06) (.01) DEMJPY (.05) (.05) (.03) (.06) (.02) (.03) (.06) (.03) (.03) (.05) (.01) DEST (.04) (.04) (.02) (.05) (.02) (.03) (.05) (.02) (.02) (.04) (.01) USST (.05) (.04) (.03) (.05) (.02) (.03) (.06) (.02) (.03) (.04) (.01) JPST (.02) (.01) (.01) (.02) (.01) (.01) (.02) (.01) (.01) (.01) (.00) DELT (.06) (.05) (.03) (.06) (.02) (.03) (.07) (.03) (.03) (.05) (.01) USLT (.06) (.05) (.03) (.06) (.02) (.03) (.07) (.03) (.03) (.05) (.01) JPLT (.02) (.01) (.01) (.02) (.01) (.01) (.02) (.01) (.01) (.01) (.00) DEIN (.44) (.37) (.25) (.48) (.18) (.25) (.51) (.21) (.23) (.36) (.09) USIN (.86) (.72) (.49) (.94) (.35) (.50) (1.00) (.42) (.45) (.71) (.18) JPIN (.35) (.29) (.20) (.38) (.14) (.20) (.41) (.17) (.18) (.29) (.07) Standard errors in parentheses. Underlined values are significant at the 5% level. Ordinary t-tests on the coefficients test for differences between the individual country and industry effects and the grand mean. Beginning with industries, it can be noted that few industries are significantly different (5% level) from the grand mean. The construction industry is, on average, significantly more negatively exposed than the grand mean, and the materials industry significantly less negatively exposed, to the German long-term interest rate. The multi industry is significantly less negatively exposed to the US short-term interest rate, and the services industry is significantly more negatively exposed to the DEM/JPY and significantly more positively exposed to the Japanese short-term interest rate and German inflation. Overall, though, it seems as if industry only provides limited explanation of cross-sectional differences in exposures. Instead turning to country differences, it is found that there are many cases of significant differences between country exposures and the grand

18 16 Table 8 Decomposition of macroeconomic exposure Grand mean Industry exposure Country exposure Firm-specific exposure DEMUSD 558% 51% 166% 436% DEMJPY 69% 108% 372% 368% DEST 208% 49% 101% 205% USST 324% 104% 262% 456% JPST 103% 53% 185% 201% DELT 4,207% 554% 703% 4,185% USLT 393% 102% 193% 319% JPLT 244% 23% 71% 282% DEIN 454% 195% 180% 316% USIN 377% 44% 134% 345% JPIN 97% 86% 137% 191% means. For all countries except Germany (with four significant), Ireland (two), and the Netherlands (three), there are significant differences between the pure country exposures and the grand mean for at least six of the eleven macro prices; Denmark deviates significantly on ten variables. For all variables except the long-term Japanese interest rate (with two significant) between five and eight countries deviate significantly from the grand mean. Overall, the results indicate that the country where a firm is domiciled is an important determinant of cross-sectional differences in macroeconomic exposure, whereas the major industry that the firm competes in is not. The levels of the R 2 s indicate, though, that the firmspecific component of exposure is by far the most important. Another way of comparing the relative importance of each of the four components of exposure the grand mean, the industry effect, the country effect, and the firm-specific component is to relate the size of each to each firm s total exposure. Since exposures are both positive and negative, an evaluation of the size of each component must be performed on absolute values, by relating the absolute value of each component to the absolute value of the total exposure. The results are presented in Table 8. The average share of exposure explained by each component varies extensively between variables and components. Most of the average shares are above 100%, which just shows that different components in many cases have opposite signs and that the variances of total exposures

19 Do country and industry explain macroeconomic exposure? 17 Table 9 Pure country and industry exposures DEM USD DEM JPY DEST USST JPST DELT USLT JPLT DEIN USIN JPIN AT BE DK FI FR DE IE IT NL SE UK Capital Construction Consumer Finance Material Multi Service around the grand means are high. For example, the grand mean for DEMUSD is on average four times larger than each firm s total exposure to the DEMUSD, while the industry component counts for 51%, the country component for 166%, and the firm-specific component is on average three times larger than the total exposure. As can be seen, the industry component of exposure tends to be the smallest component and the country component the second smallest. For most variables, the firmspecific component is the largest in terms of absolute size. Patterns in exposures among countries and industries Table 9 presents the estimated pure industry ( αˆ + γˆ k ) and pure country ( αˆ + δ ˆ l ) exposures. The entries in the table can be directly compared with the total exposures in Table 3 and Table 4 to evaluate the importance of industry composition for country exposures and country

20 18 composition for industry exposures. The differences between the pure and total exposures are equal to the average pure-industry (pure-country) exposure of the firms in the country (industry). If industry composition were important for explaining cross-country differences in exposure, the pure country exposures would look more homogeneous after correction. Similarly, if country were important, pure industry exposures would be more homogeneous than total industry exposures. However, what is most striking from a comparison are the small differences between the total and pure industry and country exposures. Differences in exposures across countries are apparently not caused by differences in industrial specialization. Similarly, differences in exposures between industries are not due to geographical concentration. Are there any patterns in pure exposures across industries and countries, that is to say, do countries and industries tend to cluster together in terms of exposures? This can be evaluated by testing for significant differences between industries and countries. The problem with this type of ex-post testing is that a large number of tests have to be performed (605 tests for countries and 231 tests for industries), which means that the risk of type-i error occurring becomes large. In fact, with a 5% significance level, on average one in every 20 significant tests could be expected to be wrong, giving a number of anticipated faulty significances of Table 10 summarizes the major patterns found among country and industry exposures by showing the number of insignificant tests for each country and industry. The average number of insignificant bivariate tests of industry differences is large, 7.5, which shows that two industries on average have statistically similar exposures to 7.5 out of eleven macro variables. The capital, multi, and services industries seem to be even closer connected than the average, with similar exposures to all macro prices (with the exception of the German short-term interest rate for the 6 The type-i error is binomially distributed with a 5% probability of occurring. Using a normal approximation, the probability of finding 42 type-i errors is P Z > 45.6% A 95% confidence interval gives a range of 30 < number of type-i errors < 54.

21 Do country and industry explain macroeconomic exposure? 19 Table 10 Patterns in macroeconomic exposures A: Number of insignificant country tests AT BE DK FI FR DE IE IT NL SE UK AT BE DK FI FR DE IE IT NL 5 5 SE 4 B: Number of insignificant industry tests Capital Construction Consumer Finance Material Multi Service Capital Construction Consumer Finance Material 8 7 Multi 11 capital and multi industries). A somewhat weaker, but still bloc of industries are the capital, consumer, material, and multi industries, which each have between eight and ten pair wise statistically similar exposures. The construction and financial industries are the two industries with fewest relations in terms of exposures. The average number of insignificant bivariate country tests is 5. Germany and the Netherlands have similar exposures to all variables, as do Ireland and the UK. Germany and the Netherlands further seem to make up a bloc together with Ireland, with between eight and eleven statistically similar exposures. Germany and the Netherlands also seem to group together with Austria on the one hand and France on the other, with between seven and eleven insignificant differences in exposure. Italy and the UK are the countries with the fewest statistically insignificant bivariate tests, indicating that these countries deviate the most from the rest of the EU countries. The Nordic countries, Sweden, Denmark, and

22 20 Finland also deviate significantly from a majority of the investigated countries for most variables. Belgium and Italy, as well as Italy and the UK have significantly different exposures to all variables except the German long-term interest rate (Belgium and Italy) and Japanese inflation (Italy and the UK). Furthermore, Belgium and the UK are significantly different on eight variables (all except the dollar and the Japanese short- and long-term interest rates). Denmark, Italy, and the UK make up a second group of countries with few similarities in exposures (between one and three insignificant differences), as do Belgium, France, and the UK (with three insignificant differences per pair wise combination). Discussion and concluding remarks In the introduction to the paper, it was hypothesized that the industry that a firm competes in would be a better determinant of cross-sectional differences in macroeconomic exposures than the country where the firm is domiciled. After all, it is the setup of the firm s operations that should determine exposure, not the location of its headquarters. Still, it is found that industry on average functions rather poorly as a determinant of differences in macroeconomic exposure, whereas country turns out to be significant. It is further found that there are differences in exposures among industries corrected for geographical concentration, as well as among countries corrected for industrial specialization. These results suggest the conclusions that (i) cross-country differences in exposures are largely the consequence of distinct differences in exposures across countries, rather than the result of industrial specialization, and (ii) cross-industry differences in exposures are largely the result of distinct differences in exposures across industries, rather than the result of geographical concentration. The conclusion on industry differences was expected; different industries put different requirements on firms. The conclusion on country differences is more striking: the European Union is highly

23 Do country and industry explain macroeconomic exposure? 21 economically and financially integrated, yet diverse in industrial specialization. So, there are distinct differences in exposures across countries and country is a significant determinant of cross-sectional variations in exposure. What could explain these findings? One potential explanation is that country-of-domicile proxies for other firm-specific explanations of exposure unrelated to industry. The difficulty with this explanation is to specify what these firm-specific explanations could be. A second explanation is that the EU countries perhaps not are perfectly integrated. The 1990s was a decade characterized by low inflation rates within the EU. Still, exchange and interest rates fluctuated extensively. This suggests that real exchange and interest rates did fluctuate, at least ex post, and that firms operating choices, by pure chance or as an effect of superior forecasting abilities, did in fact lead to differing exposures to macroeconomic risks. A third potential explanation is that stock markets might be segmented. The exposure of the firms market values was estimated. The estimated exposure accordingly is a function of the particulars of the stock market where the firm is traded. In segmented stock markets, there could well be differences in which macroeconomic risks that are considered important to corporate performance, and to what extent. The finding of the importance of country of domicile is intriguing and suggests the need for further inquiry into what lies behind the crosscountry differences in exposures. In particular, it could be fruitful to investigate cross-border differences in relevant corporate structures to evaluate if they could explain the differences in exposures, as well as the importance of country as an exposure determinant. References Andrén, N. (2000), "Macreconomic Exposure of European Firms", paper to be presented at the annual meeting of the Financial Management Association International, Seattle, USA, October 25-28, 2000.

24 Bodnar, G.M. and W.M. Gentry (1993), Exchange Rate Exposure and Industry Characteristics: Evidence from Canada, Japan, and the USA, Journal of International Money and Finance 12, Choi, J.J. and A.M. Prasad (1995), Exchange Risk Sensitivity and Its Determinants: A Firm and Industry Analysis of US Multinationals, Financial Management, Autumn, Griffin, J.M. and R.M. Stulz (1998), International Competition and Exchange Rate Shocks: A Cross-Country Industry Analysis of Stock Returns, Working Paper 98-7, The Dice Center for Research in Financial Economics, Ohio State University, Ohio, USA. He, J. and L.K. Ng (1998), The Foreign Exchange Exposure of Japanese Multinational Corporations, Journal of Finance 51, Heston, S.L. and K.G. Rouwenhorst (1994), Does Industrial Structure Explain the Benefits of International Diversification?, Journal of Financial Economics 36, Kennedy, P. (1986), Interpreting Dummy Variables, Review of Economics and Statistics 68, Krugman, P. (1991), Geography and Trade, Cambridge, MA: MIT Press. Oxelheim, L. and C. Wihlborg (1987), Macroeconomic Uncertainty. International Risks and Opportunities for the Corporation, Chichester, UK: Wiley. Oxelheim, L. and C. Wihlborg (1997), Managing in a Turbulent World Economy. Corporate Performance and Risk Exposure, Chichester, UK: Wiley. Rees, W. And S. Unni (1999), Exchange Rate Exposure Amongst European Firms: Evidence From France, Germany and the UK, Working Paper 99:8, Department of Accounting and Finance, University of Glasgow, UK. Suits, D.B. (1984), Dummy Variables: Mechanics v. Interpretation, Review of Economics and Statistics 66, Sweeney, R.J. and A.D. Warga (1986), The Pricing of Interest-Rate Risk: Evidence from the Stock Market, Journal of Finance 41,