Authors: Craig Doidge, John Griffin, Rohan Williamson Journal of Empirical Finance 13 (2006) 550–576 Buuruljin Enkhbold Jens Dahl Haagensen International Finance JEM044 Institute of Economic Studies IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Measuring the Economic Importance of Exchange Rate Exposure

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Outline 1.Introduction (Literature review, Paper’s approach) 2.Data (Source of data, Summary statistics) 3.Methodology: Linear Regression (Firm level regressions, Determinants of exchange rate exposure, Pooled Regressions) 4.Methodology: Portfolio Analysis (High minus low portfolio returns, Portfolio formed on FX betas, Controlling for BE/ME and size, Time series regressions) 5.Other Issues: Robustness Tests (Lagged exposure, Foreign income and exposure, Cash flow forecasts) 6. Conclusion and Limitations

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests What is Exchange Rate Exposure? The risk a company is facing due to fluctuations in the exchange rate A high exchange rate exposure, means that a change in the exchange rate, will have a “big” effect on a firm’s value, future cash flow etc. and vice versa The exposure comes from: Foreign sales Foreign future cash flow etc

Shapiro (1975): Multinational firm with export sales and foreign competition should exhibit exchange rate exposure. Exposure related to the proportion of export sales, foreign competition, substitutability between local and imported factors of production. Levi (1994): Exposure through value of foreign sales Marston (2001) Exposure in the net foreign revenues, but can be the form of competition, demand elasticities etc can reduce it. Who pays? Costumer or firm? Allayannis and Ihrig (2001) Mark-up and competition play roles in the exposure. Low mark- up, higher exposure. IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Theoretical exchange rate exposure literature

Past 30 years, once national industries have become more global  large increase in international activity Large real exchange rate changes after Bretton Woods These deviations away from PPP have an average half-life of 4 or 5 years, leads to large movements in price mark-ups and profit margins = Exchange rate should have a measurable effect on firm value Given all the theory we have, the empirical evidence is limited, and mostly concentrated on the US. IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

P1:The theory suggest a complex relation of exposure, while the data is limited. S1: Data gathered from a consistent source across firms and countries P2: Exposure at an industry level can be problematic, as movements in in ex-rate may lead to offsetting affects within the industry S2: Large sample of individual firm P3: Traditional regression framework, assumes ex-rate linear and constant impact firm value S3: Portfolios! Make portfolios based on int. sales. Then compute avg return return in times of ex-rate movements => no need assuming linear and constant exposure relation IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Problems and solutions

P4: Use of derivatives S4: No solution, but large firms more likely to use derivatives, indirectly account for that in the analysis, because examine the relation between firm size and exposure. IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Problems and solutions (cont.)

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests 1.Introduction (Literature review, Paper’s approach) 2.Data (Source of data, Summary statistics) 3.Methodology: Linear Regression (Firm level regressions, Determinants of exchange rate exposure, Pooled Regressions) 4.Methodology: Portfolio Analysis (High minus low portfolio returns, Portfolio formed on FX betas, Controlling for BE/ME and size, Time series regressions) 5.Other Issues: Robustness Tests (Lagged exposure, Foreign income and exposure, Cash flow forecasts) 6. Conclusion and Limitations

Stock return and market capitalization are from the Datastream International database Foreign sales, export sales, total sales, foreign assets, total assets, foreign income and total income, from Worldscope database For each country, uses value-weighted stock market index, constructed by Datastream, as proxy for market Uses mostly BoE trade-weighted ex-rate, but for robustness, the country’s bilateral cross-rate with the predominate regional rate Sample period: Jan Jul 1999 IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Data

Foreign sales – sales revenue from goods produced and sold abroad Export sales – sales revenue from goods produced domestically and sold abroad International sales – Foreign sales + Export sales IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Data (cont.)

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests 1.Introduction (Literature review, Paper’s approach) 2.Data (Source of data, Summary statistics) 3.Methodology: Linear Regression (Firm level regressions, Determinants of exchange rate exposure, Pooled Regressions) 4.Methodology: Portfolio Analysis (High minus low portfolio returns, Portfolio formed on FX betas, Controlling for BE/ME and size, Time series regressions) 5.Other Issues: Robustness Tests (Lagged exposure, Foreign income and exposure, Cash flow forecasts) 6. Conclusion and Limitations

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Firm-level exposure regressions – just a quick check R_i =monthly stock return R_m = Country specific value weighted market return R_fx = % change in FC/HC D_i = measures exposure elasticity of the firm

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests The determinants of exchange rate exposure

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Pooled regressions Begin by pooling firm-level across all countries Dependant vaiable in each regressions is d_i from (2) 5 year interval Include independent variables from Worldscope Sample period, 1990 to 1999, divided two sup-periods Using the different determinants of ex-rate exposure

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests 1.Introduction (Literature review, Paper’s approach) 2.Data (Source of data, Summary statistics) 3. Methodology: Linear Regression (Firm level regressions, Determinants of exchange rate exposure, Pooled Regressions) 4.Methodology: Portfolio Analysis (High minus low portfolio returns, Portfolio formed on FX betas, Controlling for BE/ME and size, Time series regressions) 5.Other Issues: Robustness Tests (Lagged exposure, Foreign income and exposure, Cash flow forecasts) 6. Conclusion and Limitations

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Why Portfolio Approach? Stock returns may have many sources of variation (idiosyncratic movements). Then, what are the common co-variation that affect groups of stocks? Authors aggregated stocks into portfolios that should exhibit high and low exposure. Firms are regrouped annually in every June based on previous year’s international sales. It allows for variation and possible non-linearity (Linear regression assumes that exposure is constant or linear throughout the 5-year period).

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Portfolio Analysis Portfolios HIGH exposure stock: if a firm has more than 25% international sales LOW exposure stock: if a firm has 0 international sales. 4 exchange rate ‘regimes’ (The standard deviation of the exchange rate is computed for each country over the sample period) 1. Large depretiation: exchange rate depreciation that is greater than one standard deviation from zero. 2. Small depreciation: exchange rate depreciation that is less than one standard deviation from zero. 3. Small appreciation: exchange rate appreciation that is less than one standard deviation from zero. 4. Large appreciation: exchange rate appreciation that is greater than one standard deviation from zero.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests High minus No international sales portfolio returns

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Portfolio formed on FX betas Portfolios are formed according to firms’ past estimated foreign exchange exposures Estimate market model regressions over 5 year windows. Created value-weighted portfolios: long firms in the bottom 15% (within a country) of standardized exposure betas and short those in the top 15%. The portfolio is referred as Low Minus High (LMH).

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Portfolio formed on FX betas For large home currency appreciations, the magnitude of exposure is even greater than that os using IS.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Controlling for BE/ME and Size 1.Controlling for Book to Market Ratio Double-sorting procedure: First, sort firms into high and no international sales portfolios. Second, split each into high and low book to market ratio (BE/ME). 2.Controlling for Size Split portfolios based on firm size (large and small) after sorting them into high and no international sales portfolios. To address whether the results are driven by book to market ratio and size.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Controlling for BE/ME The difference between high and low international sales firms is evident in firms with both high and low book-to-market equity.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Controlling for Firm Size Even though the results indicate that the there is less exposure in small stocks, one should be hesitant to conclude that because 1.Some of the portfolios have very small number of firms due to additional data requirements for data sorting. 2.Smaller firms in some countries might have IS figures which are noisier and less stable than those of large firms.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Time Series Regressions Using the Portfolio Approach On a country-by-country basis, estimated regressions of the high and no international sales portfolio returns on contemporaneous exchange rate movements. These regressions do not impose a constant relation between a particular firm and exchange rates across long periods of time but do impose one for portfolio. Portfolio regressions results across countries indicate a more important role for exchange rates than those at the firm level. Expectations: Firms with no IS - either have no foreign activity or could be net importers - therefore exposure coefficient on the “no IS” portfolio is >0 Firms with high IS - are more likely to be exporters - should have a negative exposure coefficient.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Time Series Regressions Using the Portfolio Approach 1% appreciation in home currency leads to 0.21% loss in firm value for firms with high international sales as compared to firms with no international sales. ggg Firms with high levels of IS outperform those with no IS during periods of large currency depreciations and underperform during currency appreciations.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests 1.Introduction (Literature review, Paper’s approach) 2.Data (Source of data, Summary statistics) 3. Methodology: Linear Regression (Firm level regressions, Determinants of exchange rate exposure, Pooled Regressions) 4.Methodology: Portfolio Analysis (High minus low portfolio returns, Portfolio formed on FX betas, Controlling for BE/ME and size, Time series regressions) 5.Other Issues: Robustness Tests (Lagged exposure, Foreign income and exposure, Cash flow forecasts) 6. Conclusion and Limitations

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Other Issues – Robustness Tests 1.Lagged exposure The lagged exchange rate effect is insignificant for most of the countries except for the US. 2. Foreign income and exposure There are fewer firms that report foreign income The results of the regressions using foreign income were similar to the previous results with IS. 3.Cash flow forecasts Whether the cash flow exposure is greater or less than stock price exposure? Authors find almost no evidence that contemporaneous or lagged exchange rate movements are related to changes in cash flow estimates. It is possible that analysts do not update their earnings estimates frequently in response to small changes in cash flows.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests 1.Introduction (Literature review, Paper’s approach) 2.Data (Source of data, Summary statistics) 3. Methodology: Linear Regression (Firm level regressions, Determinants of exchange rate exposure, Pooled Regressions) 4.Methodology: Portfolio Analysis (High minus low portfolio returns, Portfolio formed on FX betas, Controlling for BE/ME and size, Time series regressions) 5.Other Issues: Robustness Tests (Lagged exposure, Foreign income and exposure, Cash flow forecasts) 6. Conclusion and Limitations

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Conclusion More firms are exposed to exchange rate movements than can be attributed to chance; however exchange rates do not explain a large portion of the variation in individual firm stock returns. Cross sectional regressions of exhchange rate betas on determinants of exposure find that firm size, the level of international sales, foreign income, and foreign assets are negatively related to exposure. Firms with high international sales outperform (0.72% per month) those with no international sales in periods of large currency depreciations but underperform (1.1% per month) during periods of large currency appreciations according to portfolio analysis. While exposure may only explain a small proportion of the variation in stock returns for a particular firm, this effect is common across firms with international activities; therefore, exchange rates are important for explaining cross-sectional differences in stock returns. Overall, the results are consistent with theory.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Limitations Did not consider derivatives’ usage due to the lack of data, however, authors believe that not using derivatives in the analysis could only underestimate the magnitude of the exposure if the firms hedge effectively. Some portfolios did not have enough data when they were split based on additional requirement such as their international sales, size, book-to-market ratio etc. For example, the results indicate that small sized firms have less exposure which contradicts with the intuition that the large firms have less exposure because they hedge more. Authors explain that this is due to the small number of firms that the portfolios have after data sorting. Some results are statistically insignificant.

IntroductionDataLinear RegressionPortfolio AnalysisConclusion and LimitationsRobustness Tests Thank You!