The Importance of Cash Flow News for. Internationally Operating Firms

Transcription

1 The Importance of Cash Flow News for Internationally Operating Firms Alain Krapl and Carmelo Giaccotto Department of Finance, University of Connecticut 2100 Hillside Road Unit 1041, Storrs CT We thank Assaf Eisdorfer and Tom O Brien for their helpful comments. January 2011 Abstract Previous studies show that stock prices are primarily driven by discount rate news, indicating that changes in expected returns are more persistent than changes in expected cash flows. We argue that investors distinguish between foreign- and domestic based cash flow news in the sense that foreign based cash flow news is more permanent in nature than domestic based cash flow news. Potential explanations include additional frictions that internationally operating firms face, as well as the informational content of export news in the sequential internationalization process (Johanson and Vahlne (1977)). Applying Campbell s (1991) variance decomposition framework we find that cash flow news becomes more important for net importing and net exporting firms relative to firms with a low propensity of receiving foreign based cash flow news. Keywords: Variance Decomposition, Cash-Flow News, Foreign Currency Exposure, Internationalization Process EFM Classification Code: 330 1

2 1. Introduction This paper investigates if investors treat foreign based cash flow news of internationally operating firms differently from U.S. market based cash flow news of domestically operating firms. Using the log linear dividend ratio model of Campbell and Shiller (1988), Campbell (1991) uses a vector autoregressive (VAR) approach to decompose the variance of unexpected stock returns into the variance terms of cash flow news and discount rate news (expected rate of return news), and its covariance term; he shows that expected returns news are primarily responsible for moving market-level stock returns. Similar results are found by Campbell and Ammer (1993), who study the returns of bond and stock markets more broadly. Vuolteenaho (2002) shows that firm-level stock returns are mainly driven by cash flow news and then reconciles his findings with those of Campbell (1991) by demonstrating that cash flow news can be diversified within portfolios while discount rate news series are highly correlated across firms. Campbell and Vuolteenaho (2004) decompose the CAPM beta into a cash-flow-beta and discount-rate-beta, arguing that news about future cash flows should have a higher price of risk. Moreover, Eisdorfer (2007) finds that cash flow news become more dominant for financially distressed firms and that more bankruptcies occur following negative cash flow shocks than positive discount rate shocks. This paper contributes to the existing literature by demonstrating that foreign-based cash flow news is more important in the pricing process of internationally operating firms than U.S. market based cash flow news for the pricing of domestically operating firms. The increased relative importance of foreign based cash flow news can be explained by existing economic literature. First, internationally operating firms face substantial frictions such as entry and exit costs associated with exporting a product or operating production facilities abroad. Frictions can 2

3 make it more cumbersome and costly for a firm to react to a negative or positive cash flow shock therefore rendering the cash flow news for an internationally operating firm more permanent relative to the cash flow news of a purely domestically operating firm. Second, Johanson and Vahlne (1977) describe the internationalization process of firms as a sequence that starts by exporting a product and then moves to the establishment of a foreign sales subsidiary, to licensing agreements and similar contracts before actual investment in the form of foreign production facilities takes place. The individual firm moves from a relatively low risk export oriented stage to a higher risk foreign production stage that involves foreign direct investment (FDI). According to the Johanson and Vahlne s (1977) model, the firm s initial step in the internationalization process is product export to a foreign market whereby the firm acquires valuable knowledge of supply and demand conditions, business culture, risks (such as country risk), growth opportunities, competition, structure of foreign institutions etc. Based on the role of exports in the knowledge gathering process, one could infer an increased permanence of cash flow news (that is specifically related to news about exports) for net exporting firms (firms that have a net long foreign currency position). Invaluable knowledge that is gathered by firms during the export stage is used to make long term investment decisions. Assuming sufficiently efficient financial markets, investors would utilize this information and incorporate it into their valuation process of internationally operating firms, resulting in the increased importance of cash flow news. 3

4 Using Campbell s (1991) approach we decompose the returns of all firms on Nasdaq, Amex and the NYSE, using data from January 1973 to December 2009, and find that cash flow news is more persistent, i.e. dominant for firms that have an increased propensity of receiving foreign based cash flow news (either net importer or net exporter firms, as measured by their FX equity exposure). The paper proceeds as follows: Section two describes the variance decomposition framework as introduced by Campbell (1991), and the estimation process of a firm s propensity of receiving foreign based cash flow news as opposed to domestic based cash flow news. Section three describes the data, section four presents the empirical findings, and section five concludes. 2. Methodology 2.1 Variance Decomposition Framework Campbell (1991) and Campbell and Ammer (1993) use a log-linear dividend-ratio model that was introduced in Campbell and Shiller (1988) to decompose unexpected real stock returns into changes in the rational expectations of future dividend growth and future stock returns. Following Campbell (1991) the unexpected real stock returns follow: Where is the log real return on a stock from the end of period to the end of period, is the log real dividend paid during period, and is a constant of linearization, a number little smaller than one 1. Equation (1) signifies that for unexpected real stock returns to be 1 where is the sample mean of the log dividend price ratio. Campbell and Shiller (1988) discuss the approximation process whereas Campbell and Mei (1993) and Vuoleenaho (2002) find that equation (1) holds well for a wide range of possible 4

5 positive, future dividend growth has to increase, or expected future stock returns have to decrease, or both. Equation (1) can be rewritten as: Where: Following from equation (2), the variance of unexpected stock returns can be decomposed into the variance and covariance terms of cash flow and expected rate of return news: Further, Campbell (1991) shows that news about cash flows and news about future returns can be expressed as: Where:, being the matrix of vector autoregressive (VAR) parameters (the companion matrix of the VAR system) that result from the estimation of a first-order VAR system of the form: (where is the vector of the VAR variables and is the vector of the error terms with a corresponding variance/covariance matrix that is denoted by.). 5

6 denotes a vector whose first element is one and whose other elements are zero. Following (7) and (8), the variance and covariance terms of equation (6) can be expressed as: 2.2 Estimating a Firm s Propensity of Receiving Foreign Based Cash Flow News In order to test for the difference between foreign based cash flow news and domestic based cash flow news, it is necessary to estimate the level of international involvement or exposure of a firm, following the notion that a firm with a high level of international activity will experience more international cash flow innovations relative to the mostly domestic-operating firm that has neither cost nor revenue sources abroad. To avoid substantially limiting the data sample and to capture the whole spectrum of a firm s international exposure, this paper uses foreign currency equity exposure as a proxy of the level of international involvement. Jorion (1991), based on the work of Adler and Dumas (1984), suggested the following widely used model for the estimation of firm FX equity exposure: Where is stock s individual returns at time, is the return of the value-weighted U.S. market index at time, and is the return of foreign currency (usually a basket of currencies), measured in USD terms at time. Although equation (12) has been commonly used in the empirical literature, we modify the model to account for uncertainties in the timing 6

7 between changes in FX rates and equity returns 2 (see Amihud (1994)) by estimating a Dimsonversion of the equity FX exposure coefficient based on Dimson (1979). Where and are based on OLS-estimates of Equation (14): where is the CRSP holding period return of stock for month, is the monthly return of the value-weighted U.S. market index during month, is the continuously compounded rate of return of the major currency basket as measured by the USD for time period. During the first part of this study, the Dimson-Gamma, which is given by equation (13), estimates the firm s level of post-hedging FX equity exposure. We use this measure to form twenty-five size / FX equity exposure portfolios to demonstrate the importance of cash flow news for firms that have international exposure. This follows from the notion that firms with higher (in absolute value terms) FX equity exposure are more likely to experience international cash flow innovations relative to firms with little FX equity exposure. 3. Data As suggested by Eisdorfer (2007), we use three different VAR specifications. First, following Campbell (1991), we include the log of realized stock returns (including dividends), the dividend yield, and the relative bill rate as the predictive variables in the VAR system. Second, following Larrain and Yogo (2008), we replace the dividend yield with net payout, 2 The literature on FX exposure is vast. See Bartram and Bodnar (2007) for an overview of the FX exposure puzzle and issues such as the time-relationship between equity returns and changes in FX rates. 7

8 which is dividend plus equity repurchase net of newly issued equity 3. As suggested by Vuolteenaho (2002), the third VAR specification uses return on equity (ROE) as the predictive cash flow variable. All data span the period from January 1973 to December 2009 and include all firms listed on Nasdaq, Amex and the NYSE. Monthly stock return data were taken from CRSP, the monthly major currency index (MCI) data come from the Federal Reserve s H10 tables, the monthly risk free rate was retrieved from Kenneth French s website, and firm net income data was obtained from the quarterly Compustat files. Table I presents the summary statistics, Pearson correlation matrix, and tests of stationarity for: 1) The monthly excess return of the U.S. value-weighted market index, 2) the relative bill rate which is the difference between the one-month U.S. treasury bill rate and its twelve-month backward moving average, 3) the FX equity exposure measure, or Dimson- Gamma, which follows the approach suggested by Jorion (1991) with a modification suggested by Dimson (1979), 4) the monthly dividend yield that was calculated by using with- and without dividend stock returns data from CRSP, following the approach by Fama and French (1988), 5) net payout which is dividends and stock repurchases net of new equity issued, following Larrain and Yogo (2008) and Eisdorfer (2007), 6) returns on equity (ROE) which is the sum of four quarters of net income, divided by current firm market capitalization (Vuolteenaho (2002) and 3 Following Boudoukh, Michaely, Richardson, and Roberts (2007), the net payout for firm in month is calculated by: Where is the number of shares outstanding, is the cumulative factor to adjust shares, is the month-end share price, and is the cumulative factor to adjust price. All data is available from CRSP. 8

9 Eisdorfer (2007)), and 7) the trade weighted continuously compounded monthly return of a basket of foreign currencies of major U.S. trading partners, expressed in USD terms 4. All variables with the exception of Dimson-Gamma and dividend yield are stationary at the 5 percent confidence level. [Insert Table I approximately here] 4. Empirical Findings In the first part of this section we present the results of the variance decomposition using three different specifications of the VAR model to ensure robustness of our results with respect to different cash flow proxy variables. In the second part of this section we expand our analysis by estimating a series of multivariate regression models to further illustrate our main findings. Vuolteenaho (2002) shows that cash flow news can be diversified by forming portfolios while expected rate of return news is highly correlated across firms. This means that portfoliolevel variance decomposition results can be affected by the level of diversification within portfolios, therefore we choose to decompose individual firm stock returns first, and then subsequently form our different portfolios of interest. This approach rules out any diversification effects on the variance decomposition results. In order to study the difference between domestic and foreign cash flow news for a firm, it is necessary to first, sort firms by size (equity market capitalization) and then second, by a proxy for the firm s propensity of receiving foreign cash flow news as opposed to domestic cash flow news (Dimson-Gamma which is a measure of firm FX equity exposure). It is necessary to 4 This monthly rate of return is calculated by first-differencing the log of the reciprocal of the major currencies index (MCI) as reported by the Federal Reserve Bank of the United States, H10 Tables. 9

10 control for firm size due to size effects on both, the firm s variance decomposition results, as well as its effect on FX equity exposure. Vuolteenaho (2002) finds that the variance of cash flow news, the variance of expected rate of return news, and the covariance terms of the two news components decrease approximately monotonically with firm size (as measured by equity market capitalization). Moreover, Vuolteenaho (2002) finds that cash flow news becomes more important with firm size. There is also empirical evidence that firm size affects FX equity exposure measures mostly supporting the hypothesis that small firms have higher FX exposure measures than bigger firms (see Dukas et al and Chow and Chen 1998). Table II shows the VAR decomposition results using realized stock returns, dividend yield and the relative bill rate as the predictive variables. Individual firm variance decomposition results are trimmed by excluding 1% of the outliers in each tail, and sorted into twenty-five size/fx equity exposure double sorted portfolios. Sorting the portfolios based on FX equity exposure (Dimson-Gamma) should separate firms with a high post-hedging FX equity exposure from firms that have little exposure to changes in FX rates. It follows that firms with a high (in absolute value terms) FX equity exposure have a higher propensity of receiving foreign-based cash flow news relative to firms with little FX equity exposure. Size 1 denotes the portfolio that contains the smallest size firms whereas the Size 5 portfolio contains the largest firms (as measured by average equity market capitalization). Firms with the most negative FX equity exposure are assigned to FX 1. In this study this portfolio contains firms whose stock price declines with a strengthening of the foreign currency basket, i.e. a weakening of the USD. Such firms have a net short FX position or could be essentially called net importers from an FX perspective. FX 3 contains firms that have little-to-no FX equity exposure, which are likely to be 10

11 firms that have a lower propensity of receiving foreign cash flow news. FX 5 contains net exporter firms, or firms that have a net long FX position. Panel A in Table II reports the average relative importance of cash flow news for firms that are assigned to each portfolio. The relative importance of cash flow news is defined as the variance of cash flow news as a percentage of the variance of unexpected stock returns. Consistent with Vuolteenaho (2002), cash flow news dominates expected rate of return news on individual firm level. More importantly, Panel A shows that within each of the size portfolios we can observe that the average relative importance of cash flow news is higher for the FX 1 and FX 5 portfolios, and the lowest for the FX 3 portfolio. This U- or V- shaped pattern of the relative importance of cash flow news confirms our hypothesis that cash flow news is more important for firms with high FX equity exposure, i.e. firms that have a higher propensity of receiving foreign-based cash flow news relative to firms that have little FX equity exposure, or a much lower propensity of receiving foreign-based cash flow news. This observation would be consistent with the notion that investors perceive foreign based cash flow news to be more persistent than domestic based cash flow news. Panel C in Table II demonstrates that this U- or V- shaped pattern is also present in the average covariance terms, indicating that internationally operating firms experience a higher level of interaction between cash flow and discount rate news, adding to the total unexpected return volatility; this observation is similar to Vuolteenaho s (2002) finding that the covariance terms of cash flow and expected rate of return news are of a higher magnitude for smaller firms. The results in Panel A indicate that cash flow news is, on average, more important for net exporting firms within the Size 1, Size 2, Size 3, and Size 4 portfolios (105.7%, 113.1%, 11

12 114.8%, and 111.2% for net exporters, compared to 102.8%, 109%, 109.1%, and 107.4% for net importers). There does not seem to be a substantial difference in the importance of cash flow for net importers and net exporters within the largest firm size portfolios. [Insert Table II approximately here] Figure 1 demonstrates graphically, the increasing importance of cash flow news for firms that are exposed to the risks of international operations. The relative importance of cash flow news is defined as the variance of cash flow news as a percentage of the variance of unexpected returns which is depicted on the vertical axis of the graphs. The firm-level variance decomposition used realized stock returns, dividend yield and the relative bill rate as the predictive variables. Firms where then sorted into 25 portfolios based on average firm size and FX equity exposure. The horizontal axis show the various FX portfolios where 1 contains net importing firms, 3 has firms with little to no FX equity exposure, and 5 contains net exporting firms. As hypothesized most firms display predominantly U- or V- shaped patterns of the relative importance of cash flow news, confirming the more permanent nature of foreign based cash flow news. The U-shape pattern is least pronounced for small size firms and becomes more substantial with the size of the firm (Size 4 and Size 5 portfolios depict the most pronounced U- shapes). [Insert Figure 1 approximately here] The different treatment of cash flow news by investors may be due to two main differences: First, firms who operate internationally, i.e. firms that have either production facilities abroad or export their products to foreign markets, experience substantially higher frictions. These additional frictions are the result of numerous factors that could be summarized 12

13 under the umbrella of country risk. Exporting firms face barriers to entry and often barriers to exit which affects their ability to react to negative demand shocks for their product in foreign based markets 5. Similarly, internationally operating firms might find it more cumbersome and costly to adjust their cost structure when their production facilities are located in countries with economies that do not conducive to rapid cost cutting (countries with large proportions of organized labor, or complex labor laws etc.). This can lead to foreign based cash flow news being more permanent than domestic based cash flow news. Second, as described by Johanson and Vahlne (1977), exports play a key role in the internationalization process of firms. Based on their model, a firm s exporting activities are used as an important step in their knowledge development which is a key factor in the internationalization process. Based on this notion, foreign based cash flow news, especially for a net exporting firm would have more permanent effects on the unexpected returns for the firm than comparable news for a non-exporting firm. Investors would be aware that firms use international cash flow news to make key long-run decisions which would result in an increased persistence of cash flow news. Variance decomposition results are affected by the choice of predictive variables in the vector autoregressive system. Following Eisdorfer (2007), we use two alternative specifications of the vector autoregressive model. As suggested by Larrain and Yogo (2008), net payout, which is dividend payments plus net repurchases (repurchases minus newly issued equity) might be better cash flow measure for the VAR system. Table III reports the variance decomposition results that use realized stock returns, net payout and the relative bill rate as predictive variables. 5 One of the consequences of market frictions and additional uncertainties in international trade is hysteresis. See Dixit (1989), Baldwin (1988), Baldwin and Krugman (1986) for readings on this topic. 13

14 Consistent with the findings of Larrain and Yogo (2008) and Eisdorfer (2007), the average importance of cash flow news for firms is higher when using the net payout as the cash flow proxy. Similar to the findings in Table II, cash flow news play a more important role for larger firms. More importantly, cash flow news on average is more persistent for net importer and net exporter firms, confirming our hypothesis. As observed in the results in Table II, cash flow news seem to be slightly more permanent for net exporter firms relative to net importers with the exception of size quintile five (containing the largest firms). The news covariance terms are also more substantial for internationally active firms (confirming our findings in Table II). Figure 2 is a graphical representation of the results in Panel A of Table II and emphasizes the U- and V- shaped patterns of the relative importance of cash flow news. [Insert Table III approximately here] [Insert Figure 2 approximately here] The second alternative cash flow proxy, as used by Vuolteenaho (2002), is return on equity (ROE). Due to the availability of financial statement data, the frequency changes from monthly to quarterly, significantly reducing the number of observations. In addition, the availability of financial data from Compustat decreases the number of firms in our sample substantially (from 11,968 to 6,305 firms). As suggested by Eisdorfer (2007), in order to avoid any seasonality in firm earnings, we measure ROE as the sum of the past four quarters of net income divided by the current market capitalization of the firm. Table IV, and Figure 3 show that, albeit with more noise (likely due to the loss of observations and the noise generated by using financial statement data), our previous findings are robust to the choice of predictive variables in the VAR system. 14

15 [Insert Table IV approximately here] [Insert Figure 3 approximately here] In the following part of this section we include an analysis of regression results that use the relative importance of cash flow news (measured as the following ratio: ) as the dependent variable and the firm s propensity of being exposed to foreign based cash flow news as the explanatory variable while controlling for firm size (measured as the log of average firm equity market capitalization). The VAR system used to estimate the relative importance of cash flow news uses log realized returns, dividend yield and the relative bill rate as predictive variables. Table V presents the regression results of equation (15): Where is the average of firm s variance of cash flow news as a percentage of the variance of unexpected returns ( ), is the absolute value of the Dimson-Gamma measure of firm, and is the natural log of the average market capitalization of firm. Based on the hypothesis that cash flow news become more persistent or dominant with the increasing propensity of the firm of obtaining foreign-based cash flow news, should be positive in sign. Panel A presents the estimates of equation (15) using the entire sample (11,968 firms), whereas Panel B breaks the sample into net importers (6,673 firms) and net exporters (5,295 firms), Panel C presents the results of the Chow test where There is no statistical difference between the estimates of equation (15) between net importers and net exporters. 15

16 The results in Panel A show that, on average, for each increase in the absolute value of the Dimson-Gamma measure of a firm, the relative importance of cash flow news increases by 4.5 percent, controlling for firm size. This result confirms my initial findings in Table II and provides further support for my hypothesis. The slope coefficient on the log of average firm size indicates that the importance of a firm s cash flow news increases with firm size. This finding is consistent with Vuolteenaho s (2002) observation that discount rate news is relatively more important for small firms. Both results are statistically significant at the 99-percent level. Separating the sample into net importers and net exporters allows testing for differences between net importers and net exporters. The Chow test (in Panel C) rejects the null hypothesis of no difference between the two groups on the 99-percent level. The results in Panel B suggest that, on average, the positive relationship between the importance of cash flow news and the firm s level of internationalization is stronger for net exporting firms ( is for net exporters compared to for net importers). [Insert Table V approximately here] 5. Conclusions This study tests the hypothesis that cash flow news for internationally operating firms is more permanent in nature than cash flow news for firms that mostly operate domestically, which leads to increased persistence of international cash flow news. Using three different VAR specifications, we decompose firm level stock returns of all firms on Nasdaq, Amex and NYSE and find empirical evidence that cash flow news is more important for net importing and net exporting firms relative to firms that have little to no post-hedging FX equity exposure. This 16

17 indicates that investor s value firms that have foreign-based cash flows in a different manner than they do firms that mostly have domestic based cash flow innovations. Furthermore, our regression results suggest that the dominance of cash flow news is more substantial for net exporting firms than net importing firms. We also confirm the findings of Vuolteenaho (2002) who finds that cash flow news is relatively more important for large firms and that on individual firm-level, cash flow news dominates discount rate news. The increased permanence of cash flow news could be attributed to the existence of additional frictions for internationally operating firms and the sequential process of internationalization that is described by Johanson and Vahlne (1977). 17

18 References Adler, M. and B. Dumas, 1984, Exposure to Currency Risk: Definition and Measurement, Financial Management 13, Amihud, Y., 1994, Exchange Rates and the Valuation of Equity Shares, Exchange Rates and Corporate Performance, Irwin, New York, NY, Baldwin, R., 1988, Hysteresis in Import Prices: The Beachhead Effect, American Economic Review 78, Baldwin, R. and P. Krugman, 1986, Persistent Trade Effects of Large Exchange Rate Shocks, NBER Working Paper No Bartram, S. M., and G.M. Bodnar, 2007, The Exchange Rate Exposure Puzzle, Managerial Finance 33, Boudhouk, J., R. Michaely, M.P. Richardson, and M.R. Roberts, 2007, On the Importance of Measuring Payout Yield: Implications for Empirical Asset Pricing, Journal of Finance 62, Campbell, J.Y., 1991, A Variance Decomposition for Stock Returns, The Economic Journal 101, Campbell, J.Y. and J. Ammer, 1993, What Moves the Stock and Bond Markets? A Variance Decomposition for Long-Term Asset Returns, Journal of Finance 48, Campbell, J.Y. and J. Mei, 1993, Where Do Betas come from? Asset Price Dynamics and the Sources of Systematic Risk, Review of Financial Studies 6, Campbell, J.Y. and R.J. Shiller, 1988, The Dividend Price Ratio and Expectations of Future Dividends and Discount Factors, Review of Financial Studies 1,

19 Campbell, J.Y. and T. Vuolteenaho, 2004, Bad Beta, Good Beta, American Economic Review 94, Chow, E.H. and H.L. Chen, 1998, The Determinants of Foreign Exchange Rate Exposure: Evidence on Japanese Firms, Pacific-Basin Finance Journal 6, Dimson, E., 1979, Risk Measurement when Shares are Subject to Infrequent Trading, Journal of Financial Economics 5, Dixit, A., 1989, Entry and Exit Decisions under Uncertainty, Journal of Political Economy 97, Dukas, S.P., A.M. Fatemi, and A. Tavakkol, 1996, Foreign Exchange Rate Exposure and the Pricing of Exchange Rate Risk, Global Finance Journal 7, Fama, E.F. and K.R. French, 1988, Dividend Yields and Expected Stock Returns, Journal of Financial Economics 22, Jorion, P., 1991, The Pricing of Exchange Rate Risk in the Stock Market, Journal of Financial and Quantitative Analysis 26, Johanson, J., and J.E. Vahlne, 1977, The Internationalization Process of the Firm A Model of Knowledge Development and Increasing Foreign Market Commitments, Journal of International Business Studies 8, Larrain, B. and M. Yogo, 2008, Does Firm Value Move Too Much to be Justified by Subsequent Changes in Cash Flow?, Journal of Financial Economics 87, Priestley, R., 2001, Time-Varying Persistence in Expected Returns, Journal of Banking and Finance 25, Vuolteenaho, T., 2002, What Drives Firm-Level Stock Returns? Journal of Finance 57,

20 Table I Descriptive Statistics, Stationarity Tests and Correlation Matrix Table I presents descriptive statistics, stationarity tests and a Pearson correlation matrix for the following variables: The market excess return is the difference between the monthly returns of the value-weighted U.S. market index including dividends and the one-month treasury bill rate. The relative bill rate (RREF) is the difference between the one-month treasury bill rate and its one-year backward moving average. where and are OLS estimates of: therefore Dimson-Gamma denotes the FX equity exposure coefficient that is estimated by using a Dimson (1979) version of the approach suggested by Jorion (1991). The monthly dividend yield is computed using the returns with and without dividends from CRSP, using the approach developed by Fama and French (1988). Net Payout is computed following the methodology of Boudoukh, Michaely, Richardson, and Roberts (2007) and is equals dividends plus stock repurchases minus new equity issued, divided by market equity. Return on equity (ROE) is calculated by adding the past four quarters of net income, divided by market equity (Eisdorfer (2007), Vuolteenaho (2002)). The major currency index denotes the monthly log returns of a trade-weighted basket of currencies of major U.S. trading partners as defined by the Federal Reserve Bank of the United States (This is equal to 1/MCI Index). P25, P75 indicate the 25th and 75th percentiles of each variable, and the ADF statistic denotes the augmented Dickey-Fuller unit root test. All results are based on monthly data on all firms listed on the NYSE, Amex, and Nasdaq during the period of January 1973 to December Descriptive Statistics Tests of Stationarity Mean Stdev. P25 Median P75 ADF Tau P-Value Market Excess Return RREF Dimson-Gamma Dividend Yield Net Payout ROE Major Currency Index Correlations Mkt. Return RREF Dimson- Gamma Dividend Yield Net Payout ROE MCI Index Market Excess Return RREF Dimson-Gamma Dividend Yield Net Payout ROE Major Currency Index

21 Table II Variance Decomposition when the Predictive Variables are Realized Return, Dividend Yield, and the Relative Bill Rate Table II reports the variance decomposition results of unexpected returns of 25 double-sorted equalvalued portfolios, using holding period stock returns including dividends, dividend yield (calculated using Fama and French (1988) approach), and the relative bill rate, as the predictive variables. The variance components are estimated as follows:,, and, where, is a vector whose first element is one and whose other elements are zero, and denote expected return news and cash flow news, is the coefficient matrix of the following first order VAR system:, and is the average ratio of the market price to the sum of the market price and the dividend, in this sample. Size 1 is the portfolio that contains the smallest firms, Size 5 the largest firms (as measured by average market capitalization). FX 1 contains firms with low Dimson-Gamma measures, i.e. firms that have a net short foreign currency position, such as net importing firms, FX 3 contains firms with little equity FX exposure, while FX 5 contains firms that have a net long foreign currency position, such as exporting firms. Panel A, B and C report the portfolio averages of individual firm variance decompositions. Panel A reports the variance of cash flow news as a percentage of the variance of unexpected stock returns, Panel B shows the variance of expected return news as a percentage of the variance of unexpected returns, and Panel C reports the covariance between and multiplied by a factor of -2. Panel A: Variance of Cash Flow News (Percentage Weight) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX Panel B: Variance of Discount Rate News (Percentage Weight) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX Panel C: Covariance Term - (Multiplied by Factor -2) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX

22 Table III Variance Decomposition when the Predictive Variables are Realized Return, Net Payout, and the Relative Bill Rate Table III reports the variance decomposition results of unexpected returns of 25 double-sorted equalvalued portfolios, using holding period stock returns including dividends, net payout (calculated using Boudoukh et al (2007) approach), and the relative bill rate, as the predictive variables. The variance components are estimated as follows:,, and, where, is a vector whose first element is one and whose other elements are zero, and denote expected return news and cash flow news, is the coefficient matrix of the following first order VAR system:, and is the average ratio of the market price to the sum of the market price and the dividend, in this sample. Size 1 is the portfolio that contains the smallest firms, Size 5 the largest firms (as measured by average market capitalization). FX 1 contains firms with low Dimson-Gamma measures, i.e. firms that have a net short foreign currency position, such as net importing firms, FX 3 contains firms with little equity FX exposure, while FX 5 contains firms that have a net long foreign currency position, such as exporting firms. Panel A, B and C report the portfolio averages of individual firm variance decompositions. Panel A reports the variance of cash flow news as a percentage of the variance of unexpected stock returns, Panel B shows the variance of expected return news as a percentage of the variance of unexpected returns, and Panel C reports the covariance between and multiplied by a factor of -2. Panel A: Variance of Cash Flow News (Percentage Weight) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX Panel B: Variance of Discount Rate News (Percentage Weight) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX Panel C: Covariance Term - (Multiplied by Factor -2) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX

23 Table IV Variance Decomposition when the Predictive Variables are Realized Return, Return on Equity (ROE), and the Relative Bill Rate Table IV reports the variance decomposition results of unexpected returns of 25 double-sorted equalvalued portfolios, using holding period stock returns including dividends, ROE (calculated using Eisdorfer (2007) approach), and the relative bill rate, as the predictive variables. The variance components are estimated as follows:,, and, where, is a vector whose first element is one and whose other elements are zero, and denote expected return news and cash flow news, is the coefficient matrix of the following first order VAR system:, and is the average ratio of the market price to the sum of the market price and the dividend, in this sample. Size 1 is the portfolio that contains the smallest firms, Size 5 the largest firms (as measured by average market capitalization). FX 1 contains firms with low Dimson-Gamma measures, i.e. firms that have a net short foreign currency position, such as net importing firms, FX 3 contains firms with little equity FX exposure, while FX 5 contains firms that have a net long foreign currency position, such as exporting firms. Panel A, B and C report the portfolio averages of individual firm variance decompositions. Panel A reports the variance of cash flow news as a percentage of the variance of unexpected stock returns, Panel B shows the variance of expected return news as a percentage of the variance of unexpected returns, and Panel C reports the covariance between and multiplied by a factor of -2. Panel A: Variance of Cash Flow News (Percentage Weight) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX Panel B: Variance of Discount Rate News (Percentage Weight) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX Panel C: Covariance Term - (Multiplied by Factor -2) Size 1 Size 2 Size 3 Size 4 Size 5 FX FX FX FX FX

24 Table V Regression Results Variance Decomposition based on Dividend Yield Table V reports the regression analysis results of the relative importance of cash flow news for internationally operating firms using variance decomposition results that included the holding period return, the dividend yield, and the relative bill rate as predictive variables. Panel A reports the estimates of regression 1: where the dependent variable is the variance of cash flow news of firm as the percentage of the variance of unexpected stock returns of firm, is the absolute value of the dimson gamma foreign currency exposure measure of firm, and is the natural log of the average equity capitalization of firm. Panel B reports the estimation results of regression 1 with the exception of only including Net importer firms, i.e. firms that have a natural short FX position. Panel C repeats the analysis for Net exporting firms. Panel C reports the results of the Chow test where states that there is no statistically significant difference between Net importing and Net exporting firms. Panel A All Firms Observations F-value p-value < Variable Parameter Estimate t-statistic p-value Intercept <.0001 Dimson Gamma <.0001 Log Firm Size <.0001 Panel B Net Importing Firms Observations F-value p-value < Variable Parameter Estimate t-statistic p-value Intercept <.0001 Dimson-Gamma <.0001 Log Firm Size <.0001 Net Exporting Firms Observations F-value p-value < Variable Parameter Estimate t-statistic p-value Intercept <.0001 Dimson-Gamma <.0001 Log Firm Size Panel C Chow Test Degrees of Freedom Chow Test Statistic p-value <

25 Figure 1: Variance Decomposition using Returns, Dividend Yield and the Relative Bill Rate as the Predictive Variables Figure 1 shows the average portfolio values of the relative importance of cash flow news for individual firms that are sorted into 25 double sorted equal valued portfolios. The variance decomposition used stock returns, dividend yield and the relative bill rate as predictive variables. The relative importance of cash flow news is defined as the variance of cash flow news as a percentage of the variance of unexpected returns and depicted on the vertical axis. The horizontal axis show the five FX portfolio categories where 1 contains net importing firms, 3 low FX equity exposure firms, and 5 net exporting firms. Size 1 Size Size 3 Size Size

26 Figure 2: Variance Decomposition using Returns, Net Payout and the Relative Bill Rate as the Predictive Variables Figure 2 shows the average portfolio values of the relative importance of cash flow news for individual firms that are sorted into 25 double sorted equal valued portfolios. The variance decomposition used stock returns, net payout and the relative bill rate as predictive variables. The relative importance of cash flow news is defined as the variance of cash flow news as a percentage of the variance of unexpected returns and depicted on the vertical axis. The horizontal axis show the five FX portfolio categories where 1 contains net importing firms, 3 low FX equity exposure firms, and 5 net exporting firms Size Size Size Size Size

27 Figure 3: Variance Decomposition using Returns, ROE and the Relative Bill Rate as the Predictive Variables Figure 3 shows the average portfolio values of the relative importance of cash flow news for individual firms that are sorted into 25 double sorted equal valued portfolios. The variance decomposition used stock returns, ROE and the relative bill rate as predictive variables. The relative importance of cash flow news is defined as the variance of cash flow news as a percentage of the variance of unexpected returns and depicted on the vertical axis. The horizontal axis show the five FX portfolio categories where 1 contains net importing firms, 3 low FX equity exposure firms, and 5 net exporting firms Size Size Size Size Size