Understanding Brazilian companies' foreign exchange exposure

Emerging Markets Review 13 (2012) 352–365

Contents lists available at SciVerse ScienceDirect

Emerging Markets Review journal homepage: www.elsevier.com/locate/emr

Understanding Brazilian companies' foreign exchange exposure José Luiz Rossi Júnior ⁎ Insper institute of education and research, São Paulo, Brazil

a r t i c l e

i n f o

Article history: Received 18 May 2011 Received in revised form 7 March 2012 Accepted 13 March 2012 Available online 30 March 2012 JEL classification: G32 F23 F31 Keywords: Exchange rate exposure Linearities Exchange rate

a b s t r a c t The paper analyzes the exchange rate exposure of a sample of nonfinancial Brazilian companies from 1999 to 2009. The results confirm the importance of using nonlinear models to address companies' exchange rate exposure. The results indicate that when compared to the linear model commonly used in literature, the nonlinear model leads to an increase in the number of firms exposed to exchange rate fluctuations, which allows a more accurate analysis of the impact of exchange rate fluctuations on the value of firms. In addition, the paper shows that exporters and companies that hold foreign currency denominated debt are more likely to be exposed to exchange rate fluctuations and that the nonlinearity of companies' foreign exchange exposure is associated with the use of foreign currency derivatives. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Correctly measuring companies' foreign exchange exposure and identifying the main channels through which exchange rate fluctuations exert an impact on the value of companies are of great importance, not only for investors who wish to find the correct balance between risk and return of their portfolios, but also for company managers who want to assess the result of their past decisions and take further decisions about different aspects of their firm, such as forms of financing, risk management and choice of investment projects. The empirical literature on these subjects has been developing intensively since the early 1990s. On the measurement of companies' foreign exchange exposure, works such as Jorion (1990), Bartov and Bodnar (1994) and Amihud (1994) analyzed the exchange rate exposure of North American companies defining companies' exchange rate exposure as the elasticity of the firm returns with respect to changes on the exchange rate. The authors found that only a small fraction of companies in this country would be exposed to exchange rate fluctuations. Studies done for companies in other countries, using similar methodology to ⁎ Rua Quatá 300 sala 414, São Paulo, SP, Brazil 04546-042. Tel.: + 55 11 4504 2437; fax: + 55 11 4504 2350. E-mail address: [email protected]. 1566-0141/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.ememar.2012.03.007

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

353

the work in the U.S. such as Bodnar and Gentry (1993) for Japan, the United States and Canada, showed similar results, corroborating the idea that exchange rate fluctuation would have an impact on only a small number of firms. Since then, many studies have questioned the methodology used to estimate the foreign exchange exposure of firms, indicating that previous results were not robust in the specification used. One possibility raised in the literature was that linear models are not appropriate for estimating the exchange rate exposure of firms. Muller and Reuer (1998), Bartram (2004), Tai (2005), Koutmos and Martin (2003a), Muller and Verschoor (2006a) and Aysun and Guldi (2011b) examined the possibility of using nonlinear models in the analysis of foreign exchange exposure of companies. Bartram and Bodnar (2012) show that the realized return to exposure is directly related to the size and sign of the exchange rate change. On the whole, these studies confirmed the importance of nonlinearity, showing differences between processes of appreciation and depreciation and the presence of different impacts according to the magnitude of the exchange rate fluctuation. In the Brazilian case, Rossi (2011) and Merlotto et al. (2008) estimate the foreign exchange exposure of Brazilian firms. None of them analyzed the non-linear hypothesis. The first objective of the paper is assess the exchange rate exposure of non-financial Brazilian companies from 1999 to 2009, not only establishing a linear relationship as shown in most studies, but also examining the possible existence of a nonlinear relationship between exchange rate movements and the value of firms. The Brazilian case makes the analysis of the existence of nonlinearity in foreign exchange exposure extremely interesting. After adopting a flexible exchange rate in 1999, the R$/US$ exchange rate not only had long periods of depreciation and appreciation, but also periods of high and low volatility; this means it is possible to identify the existence of nonlinearity in companies' foreign exchange exposure. Fig. 1 in the appendix shows the evolution of the exchange rate R$/US$. After a speculative attack in January 1999, Brazilian currency was allowed to float, and an inflation-targeting regime was adopted. The period from 1999 to 2002 was characterized by a relative stabilization of the exchange rate. In 2002, due to the possibility that a new president against current policies would be elected, a reversal of capital flows took place and the exchange rate depreciated more than 50% during the year with a consequent rise in inflation. After 2004, the exchange rate exhibited continuous appreciation, varying from 3.12 R$/US$ in May 2004 to 1.56 R$/US$ in July 2008. With the global financial crisis, the exchange rate depreciated over 50% in one semester, reaching a value of 2.37 R$/US$ in February 2009. As the crisis cooled, it started to increase in value. Initially, the linear model similar to the one used by Jorion (1990) is estimated. The results indicate that in a sample of 196 companies, approximately 20% of the firms have a statistically significant foreign exchange exposure. The results also show that depreciation of domestic currency is problematic for Brazilian companies. Everything else being equal, a 1% depreciation of the exchange rate leads, on average, to a 0.260% drop in return for the firms. Moreover, the number of firms which are negatively impacted by depreciation of the Brazilian Real is greater than the number of companies that benefit from these exchange rate movements. Next, the assumption of exchange rate exposure linearity is tested against nonlinear smooth transition (STAR) models. The results confirm the importance of adopting a nonlinear model for the analysis of foreign exchange exposure of firms. The assumption of linearity is rejected for 107 firms, or 54.6%, of the sample. Then, the foreign exchange exposure of each firm is estimated according to the best model indicated by the data. The results confirm the negative relationship between currency devaluations and returns of companies and indicate that nonlinearity in the exchange rate exposure exacerbate the problem. On average, exchange rate movements of greater magnitude have a negative impact on the foreign exchange exposure of firms. The paper goes one step ahead compared to the previous literature by analyzing the main determinants of companies' foreign exchange exposure with a special interest on the causes of the nonlinearity of companies' exposure. The results indicate that exporters and companies that hold foreign currency denominated debt are more likely to be exposed to fluctuations on the exchange rate and that the foreign currency denominated debt is the variable responsible for on average Brazilian companies' value to fall after a depreciation of the home currency. Finally, the paper indicates that the use of currency derivatives is the main determinant of nonlinearities on companies' exchange rate exposure especially the use of forwards, options and futures.

354

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

The work was ordered as follows: Section 2 discusses the estimation of nonlinear models of companies' exchange rate exposure. Section 3 presents the data used in the analysis. The results for the estimation of the models are reported in Section 4. Section 5 discusses the main determinants of companies' exchange rate exposure and the theoretical motivation for existence of nonlinearity in foreign exchange exposure. Section 6 shows the results for the estimation of the determinants of companies' exchange rate exposure. Section 7 presents the conclusions. 2. Estimating companies' exchange rate exposure Nonlinearities in companies' exchange rate exposure can be modeled in different ways. Several studies (Koutmos and Martin (2003a, 2003b), Koutmos and Knif (2004), Priestley and Odegaard (2004) among others) have modeled nonlinearity by differentiating between movements of currency appreciation and depreciation. However, different theoretical arguments indicate that movements of different magnitude are more likely to generate a nonlinear foreign exchange exposure. Thus, we chose to model exchange rate exposure nonlinearity using a STAR model (Smooth Transition Autoregressive Model). This model is more appropriate because it is expects that a firm's currency exposure will not change suddenly, but moves smoothly according to the magnitude of exchange rate variations. The estimated model can then be represented as follows:
 m S S0 r i;t ¼ α þ βi rm;t þ βi þ βi M i ðγ; c; zt−d Þ ΔSt þ εi;t

ð1Þ

Where ri,tindicates the return of firm i at time t. rm,trepresents the market portfolio return at time t and ΔStindicates the exchange rate variation at time t. The term (βiS + βiS' * Mi(γ,c,zt − d)) represents the firm's exchange rate exposure, which can be broken down into two terms. The first term, βiS, indicates the linear portion of the exchange rate exposure and the second term βiS' * Mi(γ,c,zt − d) represents the nonlinear part of the exchange rate exposure. Miis a function of continuous and smooth transition that will vary according to the firm in question, γis the velocity of transition between the regimes, cis the point where there is a change of regime and zt − dis the transition variable. The model therefore allows the exchange rate exposure to vary according to the distance of the exchange rate to a threshold value (threshold). The usual interpretation is that the model has two regimes associated with the end of the function of transition, where the transition takes place in a smooth way. The literature shows the possibility of different functions of transition. This work uses the two main functions found in the literature: exponential and logistic. The functions have the following equation: n h
io−1 i Exponential Function : M ðγ; c; :zt−d Þ ¼ 1 þ exp −γ zt −c ;γ > 0

ð2Þ


2  i Logistic Function : Mðγ; c; zt−d Þ ¼ 1− exp −γ zt −c

ð3Þ

The exponential function has values between 0 and 1. It is interesting to note that this is symmetric around the threshold. When γ → − ∞, the exponential function tends to 0 and whenγ → + ∞, the function tends to 1; in both cases, the exchange rate exposure of the firm becomes linear. With the exponential function, the function's symmetry means the magnitude of the distance between the transition variable and the threshold is important, whether negative or positive. The logistic function also has the property of varying between 0 and 1, which means no matter which function is used, the exchange rate exposure of firm varies between βiSand βiS + βiS'. If γ = 0the model becomes linear, and if γ → ∞, the model becomes an autoregressive threshold (ART). Unlike the exponential function, for the logistic function, both the position of the transition variable (above or below the threshold) and the distance between the transition variable and the threshold are important for determining the function value. That is, the model allows differences between currency depreciation and appreciation. The methodology used to estimate the Eq. (1) is consistent with that proposed by Teräsvirta (1998) and Dijk et al. (2002). Initially, the most appropriate linear model is determined for the data. In this work,

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

355

the standard model for estimating exchange rate exposure is adopted as proposed by Jorion (1990). Afterwards, the hypothesis of nonlinearity is tested against the exponential and logistic models using the test proposed by Luukkonen et al. (1988). If the hypothesis of linearity is rejected, the model most appropriate for the data – exponential or logistic – is tested. Therefore, the model indicated by the tests is estimated by the method of maximum likelihood. Finally, diagnostic tests are carried out to check possible specification problems.

3. Data Share price data was collected from the website Economática. The study used all the weekly data of 196 non-financial Bovespa-listed companies from February 1999 to March 2009. The analysis began in 1999 in order to coincide with the period of flexible exchange rate in the country. Some parameters were used to construct the sample. The data cover all active firms at the end of the sample period. Only data from firms with at least 24 months of trading were used to ensure a minimum of information for estimating exchange rate exposure. Additionally, the sample only included companies that traded for at least 80% of the weeks and that went no more than 3 weeks without trading. It is worth noting that the sample of 196 companies represents more than half of all non-financial companies in the country (360), according to Economática. Unlike Muller and Verschoor (2006a), the analysis was developed for all companies, not only the multinationals or exporters because there are other reasons besides foreign currency income that are important components of foreign exchange exposure for Brazilian companies. One example is the importance of debt acquired in foreign currency. The share price data were extracted from the weekly closing prices, always using the value of the last trade that occurred in the week. The log-return prices were then calculated to determine the company's share return. The use of weekly, rather than daily, data was chosen because sometimes the market takes a while to understand and realize the effects of exchange rate on various asset prices, so in the short term it is harder to properly monitor a company's foreign exchange exposure. In fact, many works, such as Rossi (2011), Dominguez and Tesar (2006), show an increase in the number of companies exposed to the exchange rate when weekly or monthly data are adopted instead of daily data. Table 1 shows the distribution of firms by year. The data in Table 1 indicate that 74 companies remained in the sample during the whole period. It is worth noting that there was a significant increase in the number of businesses after 2005 due to the large number of IPOs that occurred in that period. On average, the sample gives 301 observations per company, or approximately 6 years of returns per share. The data presented in Table 2 indicate that firms in the sample are distributed throughout various sectors of activity, and that there is no dominance of any sector. Table 2 shows that construction, electrical energy and the mining and steel industry are the sectors with the highest number of firms in the sample. For exchange rate data, weekly closing values of the same days were used in the stock price analysis. The R$/US$ rate was used, i.e., positive values of the exchange rate variation are indicative of depreciation of the Real. Unlike other works, an exchange rate weighted by the country's international trade was not used since Muller and Verschoor (2006a) showed that the use of such a rate reduces the significance of the exposure. As most Brazilian international trade is expressed in US dollars and almost the entire amount of foreign currency denominated debt is expressed in dollars, the R$/US$ rate was chosen.

Table 1 Sample distribution Table 1 shows the distribution of the number of firms during the analyzed period. Year

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Number of firms

74

81

82

87

90

102

109

134

193

196

196

356

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

Table 2 Distribution of sample firms per activity sector Table 2 shows the distribution of firms in different activity sectors. The sector classification of the website Economática is used. Sector

Number of firms

Agriculture and fishing Food and beverage Commerce Construction Electronics Electrical energy Mining and steel Machines, vehicles and parts Paper and cellulose Petrochemicals Telecommunications Textiles Transport and services Others Total

4 12 10 24 6 21 19 12 4 15 14 9 13 33 196

The Ibovespa index was used as the market portfolio. Both the exchange rate and the Ibovespa index were collected directly from the Central Bank website. Table 3 presents the descriptive statistics of variables used in the study. The paper also analyzes the main determinants of companies' exchange rate exposure. Data about foreign sales and imports, the currency composition of the debt and the use of currency derivatives were collected directly from companies' annual reports. Information on derivatives and foreign-currencydenominated debt is available in the explanatory notes on firms' yearly balance sheets. The foreigncurrency-denominated debt is located under the item “loans and financing, and the information on derivatives is located under “financial instruments”. Table 4 shows that that are multiple possible sources for companies to be exposed to exchange rate fluctuations. More than one-third of the companies in the sample export their products and around onefourth are importers. The results in Table 4 indicate the importance of foreign currency denominated debt for Brazilian firms. 122 out of 196 companies in the sample hold liabilities expressed in foreign currency. One advantage of the paper is that we are able to establish a direct link between companies' exchange rate exposure and the use of derivatives since firm-level data on the use of derivatives is available for Brazilian firms under the explanatory notes in their annual reports. Aysun and Guldi (2011a) found that the use of derivatives reduce companies' foreign exposure using an indirect proxy for derivatives usage by firms. Table 4 also shows that the use of currency derivatives is relatively common among Brazilian firms. Almost 50% of the firms in the sample use currency derivatives. Confirming Rossi (2009), swap is the most preferred derivative among Brazilian companies.

Table 3 Descriptive statistics Table 3 presents the descriptive statistics of variables used in the analysis. ΔSrepresents the change in the R$/US$ rate. IBOVESPA represents the return of the IBOVESPA index. The data show weekly periodicity. The analysis period is from February 1999 to March 2009.

Mean Median Standard deviation Maximum Minimum Observations

ΔS

IBOVESPA

0.020% − 0.156% 2.411% 11.72% − 9.40% 531

0.318% 0.631% 4.44% 16.84% − 22.33% 531

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

357

Table 4 The main determinants of companies' exchange rate exposure. Table 4 displays the number and the proportion of companies' in the sample that are exporters, importers, foreign currency denominated debtors, users of currency derivatives divided on companies' that use swaps and other currency derivatives including options, forwards and futures. Foreign Operations stands for companies that have foreign subsidiaries. Number of companies (%) Exporter Importer Foreign currency debtors Users of derivatives Swap Other derivatives Foreign Operations Number of companies

68 (34.7%) 47 (23.9%) 122 (62.2%) 96 (48.9%) 69 (35.2%) 60 (30.6%) 24 (12.2%) 196

Finally, Table 4 present that there are only a small percentage of the companies in the sample (12.2%) that have foreign subsidiaries. 4. Results for the estimation of companies' exchange rate exposure As discussed above, in the first stage, the results of the linear model are considered. Afterwards, the results of testing and estimation of the nonlinear model are presented. 4.1. Linear model Using the model estimated by Jorion (1990), the following linear specification was estimated. The results are in Table 4. m S r i;t ¼ α þ βi rm;t þ βi ΔSt þ εi;t

ð4Þ

The results presented in the table indicate that the market risk is lower than 1 for a considerable percentage of the sample firms, which indicates the possible fact that Brazilian firms are diversified to the point of reducing market risk. Table 5 shows the pattern of exchange rate exposure of Brazilian companies. Considering a significance level of 5%, 34 companies or 17.3% of firms have a statistically significant foreign exchange exposure; this proportion is higher than that shown by Jorion (1990) and Muller and Verschoor (2006a) for North American companies. This number rises to 56 companies or 28.6% of the sample with a 10% level of significance. Therefore, regardless of the presence of nonlinearity, it can be seen that the exchange rate fluctuations affect a reasonable proportion of Brazilian companies. Table 5 Results of estimation of the linear model Table 5 presents the results of the equation's estimate (4). ri,t represents the return of firm i at time t. rm,tindicates the market return (Ibovespa). ΔStis the variation of the R$/US$ exchange rate in the period. The estimation was carried out by ordinary least squares + + with robust covariance matrix for the presence of heteroskedasticity. N5% and N10% indicate the number of firms with positive and − − statistically significant exposure, respectively, at 5% and 10% level of significance. N5% and N10% indicate a negative and statistically significant exposure. N indicates the total number of firms used in the analysis. Coefficient

Mean

Median

Standard deviation

Maximum

Minimum

Constant Market return (βim) Exchange rate exposure (βiS)

− 0.003 0.564 − 0.260

− 0.004 0.544 − 0.197

0.0091 0.294 0.503

0.039 1.555 1.714

− 0.038 − 0.557 − 3.100

+ N5% 9

− N5% 25

+ N10% 12

− N10% 44

N 196

358

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

Another characteristic of the currency exposure of Brazilian companies presented in Table 5 is that the mean (−0.260) and median (−0.197) are negative. This means that domestic currency depreciation against the US dollar leads to a fall in return for Brazilian companies. In addition, not only the mean, but also the distribution of foreign exchange exposure of firms, indicates a negative impact from currency devaluation. These results show that the number of firms with negative and statistically significant exposure is greater than the number of firms with positive and significant exposure. According to Rossi (2011) this result can be explained by the fact that Brazilian companies have significant foreign currency debt, which creates a discrepancy in the monetary composition of their assets and liabilities, so devaluation has a negative impact on the firm's value. The author also shows that since the adoption of flexible exchange rates, the problem has decreased; however, according to the results presented in Table 5, it is still relevant for the Brazilian economy. This fact will be analyzed later in the text. 4.2. Nonlinear models The model suggested by Teräsvirta (1998) and Dijk et al. (2002) is followed. Initially, the assumption of linearity was tested against the nonlinear STAR models. The test suggested by Luukkonen et al. (1988) was chosen. The test is based on a third-order approximation of the logistic function around the point γ = 0 (null hypothesis of linearity). The test was used; however, although it was designed to test the logistic model (LSTAR), Teräsvirta (1994) argues that the test has power against exponential type linearity (ETL). Due to the sample size and number of parameters to be estimated, this study used the F version of the LM test of nullity of parameters (Teräsvirta, 1998). The squared deviation of the exchange rate was defined as the model's transition variable. This variable is appropriate as it allows the analysis of the dependence of the exchange rate exposure with respect to the magnitude of the change in the exchange rate and has the statistical property of stationarity, needed for smooth transition models. The threshold used will then be the mean during the sample period of variable transition. The choice between the exponential and logistic model is made using the procedure described by Van Dijk et al. (2002). It is a common F test of exclusion of variables in the equation – which helps to test the linearity model. One can see which hypothesis of the parameters is most strongly rejected and determine which specification is the most appropriate. Finally, the model is estimated by ordinary least squares. Table 6 presents the results. The results in Table 6 indicate that the null hypothesis of linearity of the model was rejected for 107 companies, or 54.6%, of the sample. This is a first indication of the importance of considering nonlinearity when estimating exchange rate exposure of Brazilian companies. Among the 107 companies that rejected the linear model, 35 rejected it in favor of the logistic model, but the majority (72 companies) rejected linearity in the direction of adopting a specification that uses the exponential model. Thus it is evident that although the sign function (depreciation or appreciation) of the exchange rate change is important, the magnitude of the movement is more important in determining a firm's exchange rate exposure. This result contradicts the arguments about the importance of nonlinearity that emphasize the dependence of the exchange rate exposure with respect to the magnitude of movements in the exchange rate. Panel B, presented in Table 6, shows the results of the estimation of exchange rate exposure of firms. It is important to stress that if the test does not reject the null hypothesis of the adequacy of using the linear model, it is used for estimation. So, for 89 companies, the linear model was considered the most appropriate. For the other 107 companies, Eq. (1) was estimated by considering a function of exponential or logistic transition according to the results presented by the tests. The results contained in Table 6 confirm some results presented for the linear model with emphasis on the fact that the mean exchange rate exposure of Brazilian companies considering only the linear part is negative (−0.222), which indicates that domestic currency devaluations cause a fall in return for Brazilian firms. The results presented in panel B also confirm the importance of using nonlinear models when analyzing foreign exchange exposure of firms. Considering a significance level of 5%, the results show that 55 firms (28.1% of the sample) have a statistically significant term corresponding to the nonlinear portion of foreign exchange exposure. This number grows to 74 companies (37.7% of the sample) when one considers a 10%

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

359

Table 6 Results of estimation of foreign exchange exposure Table 6 presents the results of the estimation of the nonlinear model specified by Eq. (1). Panel A shows the test result of the linear model specified by Eq. (4). The exponential model corresponds with the use of a transition function specified by (2). The logistic model corresponds with the use of a function as specified by (3). Panel B shows the results of the estimations. The nonlinear model + + was only estimated if the linearity test rejected the linear model. N5% and N10% indicate the number of firms with positive and − − statistically significant exposure, respectively, at 5% and 10% levels of significance. N5% and N10% indicate a negative and statistically significant exposure. N indicates the total number of firms used in the analysis. Panel A – results of the linearity tests for exchange rate exposure Number of companies in which H0: linear exchange rate is rejected

H1: Exponential model H1: Logistic model

72 35

Panel B – results of estimation of foreign exchange exposure Coefficient

Mean

Median

Maximum

Minimum

0.0001 0.554

Standard deviation 0.0092 0.296

Constant Market return (βim) Linear exchange rate exposure (βiS) Nonlinear exchange rate exposure (βiS')

−.0012 0.560

0.050 1.454

− 0.052 − 0.557

− 0.222

− 0.182

0.512

1.714

− 2.836

−.0278

− 0.0238

0.056

0.164

− 0.202

+ N5% 9 10

− N5% 29 45

+ N10% 13 15

− N10% 44 59

N 196

Panel C – synthesis of the companies' exchange rate exposure Number of companies Number of companies Number of companies Number of companies

with statistically significant linear and nonlinear exposure (5%/10%) with statistically significant linear or nonlinear exposure (5%/10%) which have only statistically significant nonlinear exposure (5%/10%) with statistically significant linear exposure (5%/10%)

12/20 81/111 43/54 26/37

significance level. Similar results were found by Aysun and Guldi (2011b) using a nonparametric approach. The authors found an increase in the number of companies' exposed to the exchange rate fluctuations when nonlinear models are used to estimate companies' exposure compared to linear models. In other words, in line with Muller and Verschoor (2006a, 2006b), the results confirm that the use of nonlinear models increases the number of firms with statistically significant foreign exchange exposure. Besides the statistical significance, the use of nonlinear models has an economically relevant result. On average, companies have a nonlinear negative foreign exchange exposure, i.e., movements of greater magnitude of the exchange rate exacerbate problems related to domestic currency devaluation. Thus, strong exchange rate movements usually linked to financial crises have a larger impact on the foreign exchange exposure leading to a greater fall in return for the firms than with small exchange rate movements. Panel C, in Table 6, summarizes the estimation of the linear model. Again using the 5% level of significance, 12 companies (6.12% of the sample) have statistically significant linear and nonlinear exposures. Confirming the importance of the exchange rate for Brazilian firms, about 41% of the sample (81 companies) has some kind of exchange rate exposure. Finally, the importance of nonlinear exposure can be seen as 43 companies have only this type of exposure compared to 26 that only have a linear foreign exchange exposure. 5. The main determinants of companies' exchange rate exposure International activities (exports, import inputs), industry competitive structure, operational hedging, and corporate financial policies, especially foreign currency borrowing and use of derivatives, are among the factors identified by the literature as exerting impact on companies' exchange rate exposure. Jorion (1990) and Dominguez and Tesar (2006) analyze the role of foreign activities on companies' exchange rate exposure. Allayannis and Ihrig (2001) and Williamson (2001) study the importance of competitive structure and Allayannis and Ofek (2001), Allayannis et al. (2001), Nguyen et al. (2007), Muller and Verschoor (2006b), Bartram et al. (2010) and Aysun and Guldi (2011b) analyze the impact of hedging activities in the determination of companies' foreign exchange exposure. Focusing on emerging markets, Aysun and Guldi (2011a) analyze the relationship between the use of derivatives and companies'

360

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

exchange rate exposure. Rossi (2011) study the relationship between corporate financial policies and companies' exchange rate exposure. With the exception of Aysun and Guldi (2011b) none of these papers analyzed the main causes that might drive nonlinearities in companies' exchange rate exposure. The literature presents different theoretical arguments for the presence of nonlinearities in companies' exchange rate exposure. Various reasons related to the behavior of the firm and investors can generate a nonlinear relationship between the exchange rate and the value of the firm. 5.1. Risk management policy – foreign exchange hedge According to Muller and Verschoor (2006a, 2006b), one of the main factors to cause nonlinearity in the exchange rate exposure of firms is the policy of risk management either through the use of financial derivatives or hedging. The use of financial derivatives can generate nonlinear payoffs caused by the exchange rate fluctuations, thereby influencing in a nonlinear way the firm's cash flow and, consequently, its value. The use of options, for example, allows the firm to make asymmetric gains, thereby influencing the firm's currency exposure in accordance with the magnitude of the exchange rate fluctuation. Several authors (Allayannis and Ofek (2001), Bartram et al. (2010), Rossi (2011), Aysun and Guldi (2011b) have demonstrated the existence of a relationship between exchange rate exposure and the risk management policy of firms. 5.2. Incorrect pricing of assets Investors may find it difficult to interpret the impact of exchange rate fluctuations on the firm's value. As discussed by Muller and Verschoor (2006a, 2006b), several reasons may be behind the mistakes of investors. The difficulty of interpreting the persistence of shocks can lead investors to incorrect conclusions. They are unable to correctly identify whether a shock is permanent or temporary, and therefore find it difficult to determine the real impact of the shock on the firm. The lack of transparency in reports about the policy of protecting companies could also produce incorrect estimates of the firm's exposure, thus creating errors of assessment. Moreover, uncertainties about the future strategy to be adopted by the company may also lead to errors of assessment by investors. Overall, the authors argue that these inaccuracies result in investors ignoring the effect of lower magnitude exchange rate fluctuations and, in contrast, reacting more strongly to movements of greater magnitude. This is particularly true with negative events, thus justifying the appearance of a nonlinear relationship between the exchange rate fluctuations and the firm's value. 5.3. Pricing policy and market structure Firms may react asymmetrically to exchange rate fluctuations with respect to the policy of pricing their products. One possibility raised by international financial literature is that companies have a pricing-to-market policy, that is, they fix their prices in the currency of the country to which they are selling their product and not the domestic currency. Thus, there would not be an automatic transfer (pass-through) of exchange rate movement to the price charged by the exporting firm, but there would be a variation in the company's profit margin in line with the exchange rate movement. Since Krugman (1986), the behavior of companies' passthrough of exchange rate variation to the price versus to the variation in the company's profit margin is studied, other examples are Knetter (1992) and Marston (1990). Zorzi et al. (2007) analyze the pass-through in emerging markets. In this case, the firm's passthrough of the exchange rate movements to the prices charged would depend on the magnitude of the exchange rate fluctuation: small exchange rate fluctuations would be absorbed by the firm and they would only change their prices with higher fluctuations, thereby creating nonlinear cash flow. Works such as Pollard and Coughlin (2003) confirm that the transfer of exchange rate movement to prices is nonlinear, depending on the magnitude of fluctuation.

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

361

5.4. Hysteresis in international trade Sharp exchange rate movements may cause companies to invest in new markets, especially exporters after major domestic currency depreciation. If the investment is made in a place with investment irreversibility (high entry cost associated with a high cost in reduction of invested capital) the problem of hysteresis arises (Baldwin and Krugman, 1989). Even after the currency's appreciation, firms are required to maintain their investment in the same place even though operating losses may take place. This creates a situation where the exchange rate movements have a negative impact on the firm's value. As discussed by Baldwin and Krugman (1989), the phenomenon of hysteresis creates nonlinearity in foreign exchange exposure of firms. This would only happen after greater magnitude exchange rate movements, since small movements would have no impact on the decision of entry and exit of firms. 5.5. Asymmetries due to government interference Another reason for the existence of asymmetry is the possibility of government interference in the foreign exchange market, which mainly occurs through respective Central Banks. Government interference is an indirect form of aid to companies, thereby limiting the risk of exchange rate fluctuations and, therefore, limiting the company's exchange rate exposure. The government would only act on the exchange rate movements in situations where the exchange rate exceeded a certain critical point, limiting the currency's volatility. An example would be the effect of depreciations on a company's foreign currency debt. By acting on the foreign exchange market, the government would limit the risk of non-payment of debts. This would then reduce the negative impact of depreciation on the company. 6. Results for the main determinants of companies' exchange rate exposure Table 7 shows the results for the estimation of the main determinants of companies' foreign exchange exposure. Panel A shows a logit estimation where the dependent variables are dummy variables that assume the value of one if the company's exposure is statistically significant at a 10% level of significance. Results are similar if a 5% level of significance is used. The table analyzes the determinants for companies that have any type of exposure – linear or non-linear – coefficient statistically significant (Exposure), companies that present the linear exposure statistically significant (Linear) and companies that present a nonlinear foreign exchange exposure (Nonlinear). Results in Table 7 indicate that firms that export their products and that hold foreign currency denominated debt are more likely to be exposed to fluctuations of the exchange rate. These results confirm that international trade and companies' financial policies are important determinants of their exchange rate exposure. Results in Table 7 also show that although the other variables included in the analysis present the expected sign, none of them are statistically significant in explaining the likelihood of a firm be exposed to fluctuations of the exchange rate. Results in Table 7 evidence that once we consider the firms that present the linear coefficient statistically significant only foreign currency denominated debt is statistically significant. This confirms the results of Rossi (2011) that indicate the debt expressed in foreign currency as the main variable determining the linear coefficient of companies` exchange rate exposure. Muller and Verschoor (2006b) argue that the use of derivatives by firms would be one of the main factors to cause nonlinearity in the exchange rate exposure. Results present in Table 7 confirm this hypothesis. The variable for the use of currency derivatives is statistically significant indicating that users of derivatives are more likely to have a nonlinear coefficient statistically significant. Another advantage of using Brazilian data is that is possible to divide the users of derivatives into different types. The fourth column in Table 7 separates the firms between users of currency swaps and other derivatives (forward, futures and options). The results indicate that the use of options, futures and forwards exerts a positive impact on the likelihood of a firm to present a nonlinear foreign exchange exposure and the use of currency swaps is statistically insignificant. These results are consistent with Muller and Verschoor (2006b) that argue that nonlinear payoffs caused by the exchange rate fluctuations, thereby influencing in a nonlinear way the firm's cash flow and, consequently, its value. Therefore

362

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

Table 7 Results for the estimation of main determinants of companies' exchange rate exposure Table 7 shows the results of the estimation of the main determinants of companies' exchange rate exposure. Panel A shows the Logit estimation where Exposure, Linear and Nonlinear respectively stands for a dummy variable that assumes the value of 1 if the firm has a statistically significant exchange rate exposure (linear or nonlinear, linear and nonlinear) and 0 otherwise. Panel B is a weighted least square estimation where the dependent variable is the estimated exchange rate exposure of each company and the weight is the standard deviation of the coefficients. Linear and Nonlinear stand respectively for the linear and nonlinear coefficients. Size is the logarithm of companies' total assets. Exporter is a dummy variable that assumes the value of 1 if the firm is classified as an exporter and 0 otherwise. Importer is a dummy for firms classified as an importer. Foreign Currency debt is a dummy for companies that hold foreign currency denominated debt. Derivatives is a dummy for users of currency derivatives. Swap is a dummy for firms that use currency swaps. Other derivatives is a dummy for firms that use forwards, options and futures. Foreign operations is a dummy for firms that have foreign subsidiaries. N indicates the total number of firms used in the analysis. * and ** indicate, respectively, 5 and 10% level of significance. Panel A

Panel B

Exposure

Linear

Nonlinear

Nonlinear

Linear

Nonlinear

0.148 (0.152) 0.147 (0.623) − 0.294 (0.47) 1.27 (0.62)* 0.446 (0.470) –

0.108 (0.134) − 0.389 (0.595) 0.459 (0.441) − 0.359 (0.535) 0.769 (0.446)** –

0.077 (0.137) − 0.683 (0.626) 0.421 (0.445) 0.025 (0.556) –

Swap

0.193 (0.140) 1.10 (0.586)** 0.128 (0.456) 0.384 (0.108)* 0.490 (0.427) –

0.098 (0.237) 2.52 (0.51)* - 1.35 (0.30)* − 1.51 (0.59)* 2.24 (0.54)* –

0.023 (0.010)* 0.072 (0.024)* 0.031 (0.029) − 0.075 (0.047) − 0.067 (0.038)** –

Other derivatives

–

–

–

–

–

Foreign Operations

0.735 (0.677) Yes 0.127 – 196

− 0.063 (0.11) Yes 0.110 – 196

0.270 (0.569) Yes 0.097 – 196

1.74 (0.53)* Yes – 0.905 196

− 0.068 (0.031) Yes – 0.613 111

Size Exporter Importer Foreign currency debt Derivatives

Sector dummies Pseudo-R2 R2 N

0.078 (0.451) 0.969 (0.454)* 0.202 (0.579) Yes 0.105 – 196

derivatives like options are more likely to generate a nonlinear relationship between movements of the exchange rate and firm's value. Panel B in Table 7 shows the role of the different factors that have impact on the value of companies' exchange rate exposure. In panel B the dependent variable is the value of the coefficient of company's exchange rate exposure. The column Linear (Nonlinear) analyzes shows the results for the estimation when the linear (nonlinear) coefficient of company's exchange rate exposure is our dependent variable. The results in Table 7 indicate that different factors have a significant impact on the magnitude of company's foreign exchange exposure. Table 7 gives evidence that exports affect positively companies' exchange rate exposure. In opposition, imports have a negative and statistically significant impact on companies' exchange rate exposure, confirming that importers do not benefit from depreciations of the home currency. The results do confirm the hypothesis that foreign currency denominated debt exerts a negative impact on companies' exposure. This result indicates the importance of the negative balance sheet effects on companies' value. The results confirm the idea that foreign currency denominated debt exposes Brazilian companies to a significant source of risk, and that the negative effect of the interaction between foreign debt and exchange rate fluctuations is not negligible Results in Table 7 also show that the use of currency derivatives does alleviate companies' exposure to exchange rate fluctuations corroborating the results by Aysun and Guldi (2011a). The results indicate that the use of currency derivatives have a positive effect on the companies' exchange rate exposure. Finally, results in Table 7 shows that operational hedges have positive effects on the size of companies' exchange rate exposure. A dummy that assumes the value 1 if the firm has foreign subsidiaries is positive and statistically significant.

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

363

Table 7 indicates that three factors have an impact on the magnitude of companies' nonlinear exposure coefficient: Size, export orientation and the use of derivatives. Table 7 indicates that there is a positive relationship between size proxied by the logarithm of total assets and the nonlinear exchange rate exposure. It might be the case that larger companies are able to manage fluctuations of the exchange rate better, especially larger fluctuations, leading to a positive relationship between nonlinear exposure and the size of the company. Indeed, if foreign currency denominated debt is the main risk for Brazilian firms a better management of companies' exposure or a better access to financial markets would alleviate the negative impact of large devaluations of the home currency. Exporters are also more likely to benefit from large swings in the exchange rate exerting a positive impact on companies' nonlinear exchange rate exposure. Results in Table 7 confirm that foreign revenues also alleviate the possible negative impact of large devaluations of the home currency on the liability side of companies' balance sheet. Differently than Aysun and Guldi (2011b) that found that the use of derivatives have an impact only on companies' linear exposure, results in Table 7 show that the use of derivatives exert an impact on companies' linear and nonlinear exposures. The results indicate that, although being effective to reduce the negative impact of devaluations of the exchange rate on companies' value, the use of currency derivatives has a negative impact on companies' exposure for larger variations of the exchange rate. Different factors might be behind these results. One possibility is that if a firm uses derivatives mainly options, forward and futures as a way to hedge its foreign exposure especially the exposure that comes from its foreign currency denominated debt payments it may be more susceptible to moments of illiquidity of foreign exchange markets in order to renew its hedging positions. If moments of crisis are associated with more illiquid foreign exchange markets and larger swings of the exchange rate, this firm will have its cash flow strongly affected in these moments, generating a nonlinear relationship between exchange rate movements and its value. Another possibility is that companies use foreign exchange markets to other reasons besides hedging. Speculation might be irrelevant for movements of the small size of the exchange rate, but it might generate significant payoffs during larger movements of the exchange rate. This would be consistent with the results present by Geczy et al. (2007) that have observed that risk management is the main reason why companies use derivatives. However, once the fixed costs of hedging are paid, firms would extend their positions based on the assumption that they can generate profits (but not increase risk) using derivatives. 7. Conclusion This paper analyzes the exchange rate exposure of Brazilian non-financial firms during the period of flexible exchange rates from 1999 to 2009 and, unlike other studies, examines the importance of nonlinearity in determining the exposure of Brazilian companies to exchange rate fluctuations. The results indicate that the use of a nonlinear smooth transitions (STAR) model increases the number of firms with statistically significant exposure to exchange rate movements and is thus more suitable for the analysis of Brazilian firms. The results show that although the proportion of firms with linear exposure is significant, when nonlinearity is added to the model, the number rises substantially. Finally, the results show that depending on the level of significance used, about 50% of the sample has some type of exchange rate exposure. The results show that domestic currency devaluation is problematic for Brazilian firms. This is indicated by the negative value of the mean exchange rate exposure of firms as well as by the distribution of foreign exchange exposure where there a greater number of firms experience a reduction in their returns after devaluation of the Real. Furthermore, the results suggest that movements of greater magnitude in the exchange rate exacerbate this problem because of the significance of nonlinearity. The paper also analyzes the main determinants of companies' exchange rate exposure. The results indicate that an exporter or foreign currency denominated debtor are more likely to have a linear foreign exchange exposure and that the use of currency derivatives increases the likelihood of a firm to present a nonlinear exposure. The paper indicate that the magnitude of companies' exchange rate exposure is positively affected by their foreign revenue and negatively affected by their import expenses and foreign currency denominated debt and that derivatives are able to offset this negative impact.

364

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

The paper shows that size, export revenue and the use of derivatives play a role on the determination of the magnitude of nonlinearities of companies' exchange rate exposure. For large movements of the exchange rate size and export revenues alleviate the negative impact on companies' value of such movements, but the use of currency derivatives seems to exacerbate the problem. These results are important for economic policy makers. They can rationalize the possibility of intervention in the foreign exchange market since movements of greater magnitude in the rate lead to an increase in exchange exposure of firms. Therefore, this opens the possibility that one of the objectives of policy is the smoothing of movements in the exchange rate in order to reduce the effect on firms. As a future research, it is important to know if there is a reverse causality where the desire to smooth the movement of the exchange rate leads firms to neglect their policy of managing risk, as discussed by Chang and Velasco (2006). Other contribution is that for future research the result of the use of derivatives should be more scrutinized. The paper points out that the use of derivatives is efficient for reducing companies' negative exchange rate exposure but this impact is reduced for large devaluations of the home currency. Several reasons might be behind these results and they have to be analyzed for further analyses. Appendix A

Fig. 1. Trajectory of exchange rate R$/US$Fig. 1 shows the trajectory of the exchange rate of R$/US$ from February 1999 to December 2009. Source: Central Bank of Brazil.

References Allayannis, G., Ihrig, J., 2001. Exposure and markups. Review of Financial Studies 14, 805–835. Allayannis, G., Ihrig, J., Weston, J., 2001. Exchange rate hedging: financial versus operational strategies. The American Economic Review 91, 391–395. Allayannis, G., Ofek, E., 2001. Exchange rate exposure, hedging and the use of foreign currency derivatives. Journal of International Money and Finance 20, 273–296. Amihud, Y., 1994. Exchange rates and the valuation of equity shares. In: Amihud, Y., Levich, R.M. (Eds.), Exchange Rates and Corporate Performance. Irwin, New York, pp. 49–59. Aysun, U., Guldi, M., 2011a. Derivatives market activity in emerging markets and exchange rate exposure. Emerging Markets Finance and Trade 47, 46–67. Aysun, U., Guldi, M., 2011b. Exchange rate exposure: a nonparametric approach. Emerging Markets Review 12, 321–337. Bartov, E., Bodnar, G., 1994. Firm valuation, earnings expectation, and the exchange rate exposure effect. Journal of Finance 44, 1755–1786. Bartram, S., 2004. Linear and nonlinear foreign exchange rate exposure of German nonfinancial corporations. Journal of International Money and Finance 23, 673–699. Bartram, S., Bodnar, G., 2012. Crossing the lines: The conditional relation between exchange rate exposure and stock returns in emerging and developed markets. Journal of International Money and Finance 31, 766–792.

J.L. Rossi Júnior / Emerging Markets Review 13 (2012) 352–365

365

Bartram, S., Brown, G., Minton, B., 2010. Resolving the exchange rate puzzle: the many facets of the exchange rate exposure. Journal of Financial Economics 95, 148–173. Baldwin, R., Krugman, P., 1989. Persistent trade effects of large exchange rate shocks. Quarterly Journal of Economics 54, 635–654. Bodnar, G., Gentry, W., 1993. Exchange rate exposure and industry characteristics: evidence from Canada, Japan and the USA. Journal of International Money and Finance 12, 29–45. Chang, R., Velasco, A., 2006. Currency mismatches and monetary policy: a tale of two equilibria. Journal of International Economics 69, 150–175. Dominguez, K., Tesar, L., 2006. Exchange rate exposure. Journal of International Economics 68, 188–218. Geczy, C., Minton, B., Schrand, C., 2007. Taking a view: corporate speculation, governance and compensation. Journal of Finance 62, 2405–2443. Jorion, P., 1990. The exchange rate exposure of US multinationals. Journal of Business 63, 331–345. Koutmos, G., Knif, J., 2004. Exchange rate exposure: evidence from Finnish stock returns. Working Paper. Fairfield University. Koutmos, G., Martin, A., 2003a. Asymmetric exchange rate exposure: theory and evidence. Journal of International Money and Finance 22, 365–383. Koutmos, G., Martin, A., 2003b. First and second moment exchange rate exposure: evidence from U.S. stock returns. Financial Review 38, 455–471. Luukkonen, R., Saikkonen, P., Teräsvirta, T., 1988. Testing linearity against smooth transition autoregressive models. Biometrical 75, 491–499. Merlotto, J., Pimenta, T., Rosifini, V., 2008. A exposição ao risco de câmbio e o valor das empresas: uma análise no mercado de ações brasileiro no período de 1999 a 2003. Revista de Administração Eletrônica 1, 1–12. Muller, K., Reuer, J., 1998. Asymmetric corporate exposures to foreign exchange rate changes. Strategic Management Journal 19, 1183–1191. Muller, A., Verschoor, W., 2006a. Asymmetric foreign exchange risk exposure: evidence from U.S. multinational firms. Journal of Empirical Finance 13, 495–518. Muller, A., Verschoor, W., 2006b. European foreign exchange risk exposure. European Financial Management 12, 195–220. Nguyen, H., Faff, R., Marshall, A., 2007. Exchange rate exposure, foreign currency derivatives and the introduction of the Euro: French evidence. International Review of Economics and Finance 16, 563–577. Pollard, P., Coughlin, C., 2003. Size matters: asymmetric exchange rate pass-through at the industry level. Working Paper. The Federal Reserve Bank of St. Louis. Priestley, R., Odegaard, B., 2004. Exchange rate regimes and the price of exchange rate risk. Economics Letters 82, 181–188. Rossi, J., 2009. Corporate financial policies and the exchange rate regime: evidence from Brazil. Emerging Markets Review 10, 279–295. Rossi, J., 2011. Corporate foreign vulnerability, financial policies and the exchange rate regime: evidence from Brazil. Emerging Markets Finance & Trade 47, 67–89. Tai, C., 2005. Asymmetric currency exposure of U.S. bank stock returns. Journal of Multinational Financial Management 15, 455–472. Teräsvirta, T., 1994. Specification, estimation, and evaluation of smooth transition autoregressive models. Journal of the American Statistical Association 89, 208–218. Teräsvirta, T., 1998. Modeling economic relationships with smooth transition regressions. In: Ullah, A., Giles, e D.E.A. (Eds.), Handbook of Applied Economic Statistics. Marcel Dekker, Inc, New York, pp. 507–552. Van Dijk, D., Teräsvirta, T., Franses, P., 2002. Smooth transition autoregressive models – a survey of recent developments. Econometric Reviews 21, 1–47. Williamson, R., 2001. Exchange rate exposure and competition: evidence from automotive industry. Journal of Financial Economics 59, 441–475. Zorzi, M., Hahn, E., Sanchez, M., 2007. Exchange rate pass-through in emerging markets. ECB Working Paper Series 739.