Imports, Exports, Dollar Exposures, and Stock Returns

Imports, Exports, Dollar Exposures, and Stock Returns Suparna Chakrabortya a, Yi Tang b, Liuren Wu a a Baruch College and b Fordham University April 20, 2012 The Fifth Annual Triple Crown Finance Conference Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 1 / 19

Overview We explore three related questions: 1 Are stock returns exposed to exchange rate risk? How do the exposures differ for different companies? 2 Can we trace the exchange rate exposure to firm fundamentals? What firm fundamental characteristics explain the cross-sectional difference in exchange rate exposures? 3 Does the exchange rate exposure induce a risk premium for stocks? Can the linkage with firm fundamental characteristics help enhance the identification of the risk premium? by examining the exposures of U.S. industries on a dollar index. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 2 / 19

1. Are stock returns exposed to exchange rate risk? The literature: Jorion (1990): 287 U.S. multinationals 15% w. significant exposure. Amihud (1994): 32 large U.S. exporting firms not much. Allayabbis and Ihirg (2001): The exposures on 4 of 18 U.S. industries are significant. Our thinking: Identifying the percentage of companies with significant exposure is not that meaningful. With dollar appreciates, some companies benefit, some companies suffer, and some others do not care. Instead of examining how many companies have significant exposure estimates, we examine how the exposure estimates differ across different companies. Are these different exposure estimates purely driven by random errors, or can they be linked systematically to firm fundamental characteristics? Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 3 / 19

2. Can we trace the exposure estimates back to firm fundamental characteristics? The literature: Firms with more international trades should have more significant exposures. Jorion (1990) uses the share of foreign sales in total sales as a measure of openness. Dominguez and Tesar (2006) uses the aggregate bilateral trade flows with the U.S. as a measure of openness. Our thinking: Openness is only half the story. We agree that a company with no international trade shall have little currency exposure. But a company with balanced imports and exports activities may not have much exposure, either. Dollar appreciation may hurt exports, but can benefit imports. It is not the aggregate openness (imports + exports) that matters, but rather it is the imbalance (imports - exports) that generates the currency exposure,... when the imbalance is left unhedged. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 4 / 19

3. Does the dollar exposure induce a risk premium? The literature considers two main approaches: 1 The ICAMP approach: Time-series regression of returns on covariance estimates The identification relies on the time-variation of the covariance term. Issue 1: Individual companies can have strong currency exposures even if the market portfolio does not. Issue 2: The business of a company (and hence its currency exposure) should not vary that much from month to month, even though the business/exposure can be quite different from company to company. 2 The Fama-MacBeth approach: cross-sectional regression of stock returns on exposure (beta) estimates. The good: The regression relies on the cross-sectional difference in exposures, which is larger than the time-series variation. The bad: Rolling-window beta estimates tend to be very, very noisy. Our approach: We start with the FM approach, but enhance the identification of the dollar exposures by exploiting its link to firm characteristics. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 5 / 19

Results overview 1 Regress the stock returns for each of 404 U.S. industries against returns on a broadly defined dollar index, and other risk factors... The dollar exposure estimates vary widely from industry to industry. 2 Regress cross-sectionally dollar exposure estimates against imports and exports of each industry. The estimate on total trade (imports+exports) is not significant. The estimate is highly positive on imports, and negative on exports. Import-driven companies benefit from dollar appreciation whereas export-driven companies suffer from it. 3 Regress cross-sectionally future stock returns on dollar exposure estimates. The average risk premium on the dollar exposure is negative. The estimate is not significant when the dollar exposure is estimated purely from stock return regressions, but becomes significant when we enhance the dollar exposure identification with information from imports and exports. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 6 / 19

Data collection Our analysis involves data from several sources: 1 The dollar index: Federal Reserve Statistical Release 2 Stock returns: CRSP 3 Stock market risk factors: Kenneth French online data library 4 Imports and exports: Annual U.S. imports and exports data by 4-digit SIC coded industries. 1972-1988: Feenstra, Center for International data at UC, Davis 1989-2007: United States International Trade Commission. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 7 / 19

The dollar index The dollar index 130 120 110 100 90 80 70 60 50 40 The dollar index Monthly returns on the dollar index, % Monthly returns on the dollar index 5 4 3 2 1 0 1 2 3 4 30 71 76 82 87 93 98 04 09 5 71 76 82 87 93 98 04 09 The dollar index is a weighted average of foreign exchange rate values of the dollar for a large group of major U.S. trading partners. Industry specific dollar index (Goldberg) Statistical driven dollar index We use the dollar index to measure the strength of the dollar against the basket of other currencies. The index shows two sustained periods of dollar appreciation, followed by two periods of depreciation. Monthly returns on the dollar can be very volatile (±4%). Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 8 / 19

Cross-sectional v. time-series variation ln(im) ln(ex) ln( IM ME ) EX ln( ME ) ER A. Cross-sectional statistics of time-series averages Mean 5.2 4.9-0.7-0.9 1.1 Std 1.8 1.8 2.2 1.8 0.8 B. Time-series statistics of cross-sectional averages Mean 5.5 5.2-1.0-1.1 0.9 Std 1.0 1.0 0.4 0.4 5.8 Imports and exports activities show much larger cross-sectional (from industry to industry) variation than time-series (from year to year) variation. Firm fundamental variations are better captured cross-sectionally than over time. The excess stock returns show much larger time-series variation than cross-sectional variation. Large randomness in stock return realization can overwhelm the cross-sectional differences in risk premiums. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 9 / 19

Risk exposure estimates ERt i = β i0 + βi fx ERt fx + βi mkt ERt mkt + βi smb SMB t + βi hml HML t + et, i β fx β mkt β smb β hml R 2 Mean 0.046 0.970 0.971 0.333 0.391 Std 0.667 0.277 0.536 0.472 0.158 Minimum -3.557-0.337-0.362-1.629 0.043 Maximum 3.095 2.134 3.352 2.350 0.876 The cross-sectional average of the dollar exposure estimates is small, But the estimates show large cross-sectional variations across 404 industries. The cross-sectional variation of the dollar exposure is larger than the cross-sectional variation of market, size, book-to-market exposures. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 10 / 19

The cross-sectional variation of dollar-exposure estimates Dollar exposure estimates t-statisics 250 120 200 100 Number of observations 150 100 Number of observations 80 60 40 50 20 0 4 3 2 1 0 1 2 3 4 Dollar exposure estimates, β fx 0 4 3 2 1 0 1 2 3 4 Dollar exposure t statistics There is definite cross-sectional variation in the dollar exposure estimates... Negative for steel investment foundaries, space vehicle equipment Positive for men s underwear, electronic resistors, refrigerator Are these variations driven purely by random noise... or do they contain some information about the actual fundamental differences across these industries? Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 11 / 19

Linking dollar exposures to imports and exports Cross-sectional regression of full-sample dollar exposure estimates against average imports and exports of each industry. Intercept ln( IM+EX IM ME ) ln( ME ) EX ln( ME ) IM ln( EX ) Adj. R2 A. 0.042 0.065-0.056 1.49% (1.29) (2.84) (-2.11) B. 0.042 0.020 0.07% (1.29) (1.13) C 0.042 0.048 0.69% (1.29) (1.94) The dollar exposure is positively related to imports and negatively related to exports. The relations are highly significant. The exposure does not have a strong link with the total trade, but has a significant link to import/export imbalance. The low R-squared suggests that the return-regression generated dollar exposure estimates are very noisy. The significant link to imports/exports suggests that they are not all noise. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 12 / 19

Time-variation in the cross-sectional linkage Regress 10-yr rolling dollar exposure estimates against 10-yr rolling average of imports and exports: Dollar exposure dependence on imports and exports 0.1 0.08 0.06 0.04 0.02 0 0.02 0.04 0.06 0.08 ln(im/me) ln(ex/me) 84 87 90 93 95 98 01 04 06 The link with imports is positive and the link with exports is negative. The relation is reasonably stable over time. Import-driven companies benefit from dollar appreciation whereas export-driven companies suffer from it. The result makes intuitive economic sense, and suggests that firms do not fully hedge their trade-induced currency exposures. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 13 / 19

Identifying the dollar exposure and dollar risk premium SFM. ERt+1 i = ηt 0 + η fx EFM. ERt+1 i = ηt 0 + ηt fx (control) = ηt mkt βit mkt ( it + control + ei t+1, β fx it + λ IM ln( IM ME ) it + λ EX ln( EX ME ) it) + control + e i t+1 t β fx + ηt smb βit smb + ηt hml βit hml Fama-MacBeth η 0 η fx η mkt η smb η hml Standard 0.484-0.009 0.379-0.085 0.016 (1.38) (-0.09) (1.14) (-0.37) (0.07) Enhanced 0.485-0.066 0.351-0.062 0.038 (1.36) (-2.02) (1.04) (-0.27) (0.15) The dollar exposure induces a negative risk premium on average. The estimate is not significant when the dollar exposure is estimated purely from return regressions, potentially because the estimates are too noisy. Adding imports and exports to enhance the exposure identification leads to a significantly negative risk premium estimate on the dollar exposure. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 14 / 19

Time-varying dollar risk premium Exponential-weighted moving average of the monthly risk premium estimates (η fx t ) 0.01 0.02 0.03 Currency risk premium 0.04 0.05 0.06 0.07 0.08 0.09 0.1 82 84 87 90 93 95 98 01 04 06 09 The dollar risk premium becomes increasingly negative over time, potentially because increasing aggregate currency exposure and/or increasing investor awareness of the risk inherent in the currency exposure. The risk premium also becomes more negative during the two recessions (early 90s and early 2000). Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 15 / 19

The economic implications of a negative risk premium on the dollar exposure Stocks for import-driven companies tend to co-move positively with the dollar index, whereas export-driven companies tend to negatively co-move with the dollar index. Positive dollar exposure generates a negative risk premium. Import-driven companies tend to generate lower average returns than export-driven companies. Dollar appreciation is regarded as a bad state for the overall economy. Companies that positively co-move with dollar can serve as a hedge of the bad state and hence investors ask for a lower return for such companies. The absolute magnitude of the currency risk premium increases during recessions, either due to increased risk, or due to increased risk aversion, or both. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 16 / 19

The practical implications on risk exposure and risk premium identification The Fama-MacBeth approach is good in exploiting the information in the cross-sectional variation of risk exposures. The issue is that the risk exposure (beta) estimates from return regressions tend to be very noisy. The regression suffers from severe errors-in-variable problems. Textbooks propose various remedies to reduce the noise in market beta: Averaging within industries, bottom-up methods, etc. Practitioners (e.g. BARRA) move away from beta estimates to firm characteristics, as they can be measured more accurately and rely less on time-series regressions. Academic studies (Daniel & Titman (1997), Ang, Hodrick, Xing, Zhang (2009)) also find that it is the characteristics (e.g. size and book to market ratio), not the beta estimates on the SMB and HML portfolios, that predict stock returns. We propose to combine the information in both... Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 17 / 19

Concluding remarks Different U.S. firms have different exposures to the dollar movement. The exposure difference can be traced back to differences in firm fundamental characteristics. Stock returns on import-driven firms move with the dollar. Stock returns on export-driven firms move against the dollar. These companies do not fully hedge their business risk exposure. Investors ask for different returns for different dollar exposures. Investors ask for lower expected returns for companies with positive dollar exposures. Export-driven companies tend to generate higher returns than import-driven companies. The risk premium becomes more negative during recessions. Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 18 / 19

What is next? Combining return regression beta estimates with firm characteristics: Given the highly noisy nature of beta estimation, should we follow the practitioners lead in switching to firm characteristics from beta estimation? (How) can the idea of combining beta estimates with firm characteristics be applied to other risk factors? Why is dollar appreciation bad for the U.S. economy? Is it specific to the U.S. economy? Is it a proxy for something more fundamental? Why export-driven companies generate higher average returns? Liuren Wu (Baruch) Imports, Exports, Dollar Exposures, and Stock Returns 4/20/2012 19 / 19