Can the use of foreign currency derivatives explain variations in foreign exchange exposure?

Can the use of foreign currency derivatives explain variations in foreign exchange exposure?

J. of Multi. Fin. Manag. 13 (2003) 193 /215 www.elsevier.com/locate/econbase Can the use of foreign currency derivatives explain variations in forei...

183KB Sizes 1 Downloads 75 Views

J. of Multi. Fin. Manag. 13 (2003) 193 /215 www.elsevier.com/locate/econbase

Can the use of foreign currency derivatives explain variations in foreign exchange exposure? Evidence from Australian companies Hoa Nguyen a, Robert Faff b,* a

School of Economics and Finance, RMIT University, GPO Box 2476V, Melbourne, Victoria 3001, Australia b Department of Accounting and Finance, Faculty of Business and Economics, P.O. Box 11E, Monash University, Clayton, Victoria 3800, Australia Received 27 May 2001; accepted 24 April 2002

Abstract We investigate the role of foreign currency derivatives (FCD) in alleviating foreign exchange rate exposure of Australian firms. While there is some evidence that the use of FCD reduces the level of ex-post short-term exposure, such an effect is absent with regard to the degree of foreign operations. Our results support the view that FCDs are used to hedge existing exchange rate exposures and that Australian firms, generally, are extensively exposed to currency fluctuations in the long run. While monthly exposure appears to be a function of a firm’s size and financial hedging, exchange rate exposure of shorter horizons (1 and 3 months) appears to be negatively related to a firm’s price earnings ratio (proxying growth opportunities) */thereby supporting the ‘underinvestment’ hypothesis. Further, the exposure of longer horizons (12 and 24 months) is positively related to a firm’s liquidity, supporting the view that liquidity is a substitute for hedging. # 2003 Elsevier Science B.V. All rights reserved. JEL classification: G30; G32 Keywords: Foreign currency derivative use; Foreign currency exposure

* Corresponding author. Tel.: /61-3-9905-2477; fax: /61-3-9905-2339. E-mail address: [email protected] (R. Faff). 1042-444X/03/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S1042-444X(02)00051-8

194

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

1. Introduction The strong emergence of derivative financial instruments in the last decade as the most cost-effective way to manage risks has triggered considerable interest among financial markets’ participants. The contemporary finance discipline is also becoming more and more focused on hedging activities and risk management practices of corporations. Nevertheless, in order to make informed hedging decisions, it is imperative that financial managers are aware of the nature and extent of risks to which the firm is exposed. As far as Australian firms are concerned, foreign exchange rate exposure has been a topic of particular interest since the flotation of the Australian dollar (AUD) in late 1983. The extent to which Australian firms are exposed to foreign exchange rate risks has since been an important empirical issue in relation to the development of a more comprehensive set of financial reporting standards which is deemed desirable in improving the transparency in derivative reporting. Existing empirical work has shown that the use of derivatives as a hedging mechanism can be a value enhancing exercise. Derivatives can generally be used to reduce expected taxes (Nance et al., 1993), to alleviate the cost of expected financial distress (Bessembinder, 1991) and to minimize the underinvestment cost resulting from liquidity constraints (Geczy et al., 1997). While it is generally assumed that derivatives are used to hedge an existing exposure, it is not obvious whether this is the case. Derivative disasters such as the recent Metallgesellshaft and Baring Bank losses1 show that derivatives can also be used to speculate on market movements. A priori, hedging, by definition, will reduce the level of risk to which a firm is exposed. On the other hand, speculation will increase risks. To identify the motives behind the use of foreign currency derivatives (FCDs), (Allayannis and Ofek, 2001) examine the effect of the use of FCDs on the estimated exchange rate exposure of all the S&P 500 non-financial firms in 1993. On the basis of the assumption that the level of foreign exchange exposure that a firm bears is determined simultaneously by the firm’s foreign operations and financial hedging, they develop a model that incorporates proxies for these two factors as the determinants of foreign exchange exposure. Their results show that a firm’s absolute exchange rate exposure is positively related to its ratio of foreign sales to total sales, and negatively related to its ratio of FCDs to total assets. They, therefore, conclude that the firms in their sample use FCDs with a view to hedging an existing exposure rather than to speculating in the derivative market. In this paper, we aim to test the generalizability of these findings by using data obtained for a sample of Australian firms. In particular, we test the hypothesis that the use of FCDs reduces a firm’s foreign exchange exposure if FCDs are used to hedge. Alternatively, the use of FCDs for speculation purposes can increase exposure. To focus on the relationship between exposure and the use of FCDs, it 1 For details, see Beder, 1996, ‘Lessons from derivatives losses’, Derivatives Risk and Responsibility, Irwin Publisher.

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

195

is assumed that long-term natural hedges associated with operating and financing decisions are predetermined. We extend the model to control for the effect of hedging by including in the analysis the variables that are proxies for a firm’s incentives to hedge. We also differentiate between industrial and resource stocks to see whether exchange rate exposure is an industry specific feature. Although we uncover some evidence that the use of FCDs reduces the magnitude of monthly exchange rate exposure, in contrast to Allayannis and Ofek (2001), we fail to find a consistent relationship between exposure and the ratio of foreign sales to total sales. The industry-sector test further reveals that the level of foreign involvement does not impact on industrial and resource sector firms differently. However, we observe that industrial versus resource sector firms do tend to be engaged in different hedging practices, as the effect of financial hedging on firms in the two industry sectors is quite different. Specifically, financial hedging plays a more influential role in altering exchange rate exposure in the resource sector than in the industrial sector, a result consistent with the view that resources firms tend to hedge their production more extensively. Consistent with the notion that long horizon returns may be associated with economic exposure we also include in our study a long horizon exposure analysis. Similar to other research in the area, we find that Australian firms tend to be more exposed to exchange rate fluctuations as the return horizon increases. Evidently, firms are much less effective in hedging long-term exchange rate exposure. We, however, go further by identifying specific firm characteristics that can potentially explain variations in the long horizon foreign exchange exposure. Our findings show that for relatively shorter return horizons (1 and 3 months), a firm is more exposed to foreign currency risk, the lower is its price earnings ratio (PER). As such, the ‘underinvestment’ hypothesis is supported. For relatively longer return horizons (12 and 24 months), however, a firm tends to be more exposed to exchange rate risk the more liquid it is, thereby supporting the role of liquidity as a substitute for hedging. The remainder of the paper proceeds as follows. Section 2 offers an overview of the Australian economy. Section 3 presents our methodology, Section 4 discusses the results and Section 5 concludes.

2. The Australian economy

*

/

an overview

Australia is a member of the organization for economic cooperation and development (OECD) and as such is classified as one of the leading industrialized nations in the world. According to the Australian bureau of statistics (ABS), in 2000, Australia’s GDP adjusted for purchasing power parity was USD 445.8 billion, equivalent to USD 23,200 per capita. Australia is rich in natural resources and therefore a major exporter of agricultural products, minerals (predominantly coal), metals and fossil fuels. Commodities account for 57% of the total value of Australia’s exports. As a result, like many developing countries, the economy is vulnerable to price fluctuation in the world commodities markets and to inflation in its main supplier markets. Imports, on the other hand, are mainly manufactured

196

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

products including machinery and transport equipment, computers, telecommunication equipment and parts. While primary production plays a central role in the country’s exports, in terms of the domestic economy it has become far less significant. Agriculture and mining account for only 3 and 4%, respectively, of the GDP while services represent a massive 71% (1999 estimate). In terms of volume, Japan and the U.S. are the two major trading partners. Jointly, they represent 28% Australia’s exports and account for 36% of its imports. The AUD was floated in 1983 and its value has been determined by the supply and demand of the currency since. From the beginning of 1995, the dollar has been fluctuating around the USD 0.75 mark, with a high of 0.8203 occurring on the 2nd of December 1996. However, from mid 1997, the dollar followed a falling trend. By the end of 1997, the dollar was trading in the vicinity of USD 0.65, a rate that was sustainable for 2 years. In terms of the TWI, the AUD fell from a value of 56.2 in early 1995 to a low of 48.2 at the end of 2000. The depreciation of the AUD, despite the efforts of the Reserve Bank of Australia (RBA), is the result of the financial markets seeing more profitable investment opportunities brought about in part by the interest rate differential. Although the economy is generally sound, the value of the currency has plunged as speculators see the AUD as an attractive currency to short. The severe loss in value of the AUD while benefiting exporters, hurts importers badly and opens up the potential risk of Australian firms being taken over by overseas corporations.

3. Data and methodology 3.1. Data Our sample is selected from the Connect4 database consisting of over 500 Australian companies that are listed on the Australian stock exchange (ASX). On the basis of the assumption that firms with foreign sales are likely to have inherent foreign exchange exposure, we include in our sample firms that have overseas sales as reported in the segment reporting of their financial reports for the fiscal year 1999. Firms whose foreign sales exceed 10% of total sales are required by the accounting standard to report audited footnote segment information. Although this selection criterion addresses the self-selection bias problem arising from the inclusion of firms with limited international linkages, it does not guarantee that these firms respond in the same direction to a change in the exchange rate (see Bartov and Bodnar, 1994 for a discussion of the sample selection procedure). Financial firms are excluded from the sample because the nature of the business causes them to use derivatives for purposes other than hedging. This approach produces 144 firms in our final sample with mean foreign sales of AUD 750 million which, on average, accounts for approximately 40% of total sales. To examine the impact that the use of FCDs might have on exposure, we first estimate exposure for the period from January 1997 to December 1999. Consistent with Allayannis and Ofek (2001), we believe that a 3-year return period surrounding

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

197

the year in which derivative data are disclosed is appropriate to measure the contemporaneous impact of FCDs on a firm’s exchange rate exposure. Further, data on year-end notional value of FCD contracts are obtained from the notes to the financial statements as disclosed in Connect4 for the year 1999. FCDs include forwards, futures and options contracts.2 Out of the 144 firms in our sample, 77 firms (53.47%) report the use of FCDs. Data regarding individual firms’ return, return on a market index as measured by the ASX All Ordinaries Index and return on a trade-weighted index are obtained from Datastream International. 3.2. Empirical framework 3.2.1. Stage one: estimation of exchange rate exposure Consistent with Jorion (1990), Loudon (1993) and Allayannis and Ofek (2001), foreign exchange rate exposure (b2i )is estimated for the period from January 1997 to December 1999 using the following model:3 Rit  b0i b1i Rmt b2i Rxt o it ;

(1)

where Rit is the return on stock i in period t; Rmt is the return on the ASX All Ordinaries Index in period t; and Rxt is the rate of change in the trade weighted index value of the AUD in period t , measured in foreign currency per one unit of AUD. The use of a trade-weighted index incorporates the assumption that the pattern of a firm’s international linkages is similar to that of the national trade with foreign countries. Therefore, changes in the trade-weighted index are assumed to affect firms uniformly. Of course this is not necessarily the case. Nevertheless, trade-weighted indices are used widely in prior research as a measure of foreign exchange rate. Furthermore, given the trade composition of Australia, the use of a trade-weighted index appears to be more appropriate than using a single exchange rate, for example AUD/JPY or AUD/USD to address the fact that ASEAN countries are becoming

2 Currency swaps are not included in the aggregate measure of FCDs for the reason that it is not a popular instrument and is mostly used in conjunction with a foreign debt. 3 It is worth noting that if Purchasing Power Parity (PPP) holds, this model may fail to explain the mechanism underlying the exposure. Since exchange rate changes are tied to changes in interest rates (as predicted by PPP), firm value is more likely to be related to changes in interest rate rather exchange rate. The reason for this is that the interest rate affects the discount rate and thus the value of the firm (being the discounted value of future cash flows). For example, if domestic interest rates increase and the currency appreciates, a highly levered firm might see its stock losing value. The firm appears to have negative exposure while in fact, it has positive exposure.

198

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

increasingly important to Australia trade, for which a single exchange rate does not exist.4 Based on the approach adopted in Di Iorio and Faff (2001), theoretical predictions regarding the direction of exchange rate exposure of the sample Australian firms have been developed. A firm is defined as having negative (positive) exchange rate exposure if its stock return decreases as the AUD appreciates (depreciates). It is assumed that a firm’s exchange rate exposure is determined solely by its import/export portfolio. As a result, we expect a firm will have negative (positive) exposure if its exports exceed (are exceeded by) imports whereas positive exposure is to result if imports exceed exports.5 It is predicted that out of the 144 sample firms (24 ASX industries), 65 firms (nine industries) are predicted to have negative exposures, 61 firms (10 industries) to have positive exposures while 18 firms (five industries) remain indeterminate. 3.2.2. Stage two: cross-sectional regression Once the exchange rate exposure is estimated, the basic relationship between exposure versus foreign operations and the use of FCDs is tested using the cross-sectional regression framework of Allayannis and Ofek (2001). However, to address the issue of non-normality in the disturbances, Eq. (2) is estimated using the generalized method of moments (GMM) approach.6 GMM is a robust estimator in that it does not require information of the exact distribution of the error term.     FS FCD b2i a0 a1 a2 li ; (2) TS i TA i where: b2i is the exposure coefficients estimated in Eq. (1); FS/TS is the ratio of foreign sales to total sales; FCD/TA is the amount of FCDs used by a firm scaled by the firm’s total assets for the financial year 1999.

4 It is, however, important to note that while the trade composition is important it does not necessarily reflect the source of exchange rate risk to which an importer/exporter is exposed. For example, if trade to ASEAN countries are resource related and proceeds are in USD (as commodity prices are, more often than not, quoted in USD), then it is the AUD/USD exchange rate which is of concern. To address this potential complication, we include a robustness check in Section 4.4 (a) regarding exposure to changes in the AUD/USD and AUD/JPY exchange rates. We would like to thank an anonymous referee for focusing our attention on this issue. 5 The imports/exports coefficients are estimated using the Australian Bureau of Statistics (ABS) data (see Di Iorio and Faff, 2001 for details). 6 Initially, we estimated all regressions using OLS and tested the normality of the residuals. In all cases, there was a strong rejection of normality. We thank an anonymous referee for alerting us to this potential problem.

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

199

The equation incorporates both a firm’s foreign operations (as proxied through foreign sales) and financial hedging (as measured by the use of FCDs)*/two factors that are believed to simultaneously determine the level of exchange rate exposure. If firms use FCDs to hedge exchange rate exposure and assuming that FCDs are effective in hedging this exposure, it is predicted that the more FCDs a firm uses, the less exposed it is to exchange rate risk. In particular, the use of FCDs should decrease exposure for firms with positive exposure and increase (decrease in absolute value) exposure for firms with negative exposure. The ratio of foreign sales to total sales, on the other hand, is predicted to increase exposure. Jorion (1990) finds some evidence that exposure is positively related to this ratio. Following the approach adopted by Chow and Chen (1998), we extend Eq. (2) and control for the effect of hedging by including in the regression variables that proxy for firms’ incentives to hedge. The cross-sectional regression is specified as follows: b2i a0 a1 LEVi a2 SIZEi a3 FSTSi a4 FCDi a5 LIQi a6 PEi ui ; (3) where b2i is the exchange rate exposure estimated using Eq. (1); LEV is the ratio of total debt to firm size; SIZE is calculated as the sum of market value of equity and total debt; FSTS is the ratio of foreign sales to total sales; FCD is the ratio of the total contract value of FCDs scaled by total assets; LIQ is calculated as cash and cash equivalents scaled by firm size and PE is the firm’s PER. As noted above, hedging, especially an operational hedge demonstrates economies of scale. Larger firms are more likely to possess the necessary financial and human resources to undertake a hedging program. Under the assumption that the incentive of hedging is positively related to the effectiveness of hedging, the short horizon exchange rate exposure would be smaller for small firms because they are more likely to hedge transaction exposure than larger firms.7 Larger firms, in contrast, have a comparative advantage in administering an operational hedge, and thus are more likely to hedge economic exposure. As a result, while having larger exposure in the short run, larger firms are less exposed in the long run. Firms that employ a high level of debt in their capital structure are more prone to financial distress and therefore have more incentive to hedge to reduce the variance of their earnings. Similarly, a firm with growth options hedges to reduce the cost of underinvesment by ensuring that the firm will have sufficient funds to undertake all positive NPV projects that are available. We use the PER as a proxy for growth opportunities and predict that the higher leverage and the more growth options a

7 Smaller firms are more likely to be involved in hedging because financial distress is relatively more costly to them. The expected cost of financial distress increases less than proportionally as firm size increases. Small firms also tend to face a greater degree of information asymmetry. See Ang et al. (1982) for a detailed discussion of why small firms face higher costs of financial distress.

200

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

firm has, the less exposure it should have. Liquidity is normally viewed as a substitute for hedging. A high level of liquidity lessens the pressure on the firm to use derivatives to smooth earnings. Therefore, a higher level of liquidity should be associated with a higher level of exposure in absolute terms.

4. Regression results 4.1. Exchange rate exposure of Australian firms and the use of foreign currency derivatives Table 2 reports the descriptive statistics for foreign exchange rate exposure estimated using Eq. (1) for the full sample and the two industry sector sub-samples. The results in Panel A show that out of the full sample of 144 firms, only 21 firms (14.58%) have significant weekly exposure, which is slightly higher than has been found in previous research. Jorion (1990) found that 5.2% of his sample had significant exposure while Loudon (1993), using a sample of 141 firms listed on the ASX, reports that only 6.4% of his sample are exposed to exchange risk. Further, we observe that 25.93% of the resource stocks turn out to have significant exposure. Nevertheless, this significance disappears when monthly returns were used instead of weekly returns. This result suggests that while resources firms are sensitive to weekly changes in exchange rates, monthly exposures are well hedged against by means of either natural or financial hedges. Industrial stocks, on the other hand, yield a consistent 12.82% of cases with significant exposure in both weekly and monthly analysis. The weekly results show that there are slightly more firms with positive exposure (54.17% as opposed to 45.83%) suggesting that Australian firms generally benefit when the AUD appreciates (contrary to our predictions). Nevertheless, the monthly regression produces supporting results with 56.25% of the firms having negative exposure. Our non-parametric sign test reveals that except for monthly measures of all firms and industrial stocks that show a tendency of firms being negatively exposed, positive and negative exposure are quite evenly distributed. Panel B of Table 1 reports the results for a test of difference in exposure between FCD users and non-users. As expected, non-users as a group show a consistently higher mean exposure in both raw exposure and absolute exposure settings although the differences are not statistically significant in the case of weekly exposures. FCD users and non-users, however, show a statistically significant difference in monthly exposures. When only the magnitude of the exposure (absolute value of the exchange rate exposure), not the direction of exposure, is considered, statistical significance gets stronger as indicated by the P -values. Obviously, in a univariate sense, nonFCD users do not get the benefits of financial hedging and thus, are more exposed to changes in exchange rates. Generally speaking, exchange exposure is created via day-to-day foreign operations of the firms and is reduced through financial hedging to achieve the target risk level. However, using the ratio of foreign sales to total sales (FS/TS) as a measure of

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

201

Table 1 Descriptive statistics for foreign exchange exposure coefficients Panel A: Descriptive statistics of the estimated foreign exchange exposure All firms

Number of observations Mean Median Standard deviation Minimum Maximum No. of positive cases No. of significant cases No. of negative cases No. of significant cases % of significant cases Test statistica

Resources

Industrials

Weekly

Monthly

Weekly

Monthly Weekly Monthly

144 0.078 0.037 0.585 /3.352 2.649 78 12 66 9 14.58% 1.4142

144 0.043 /0.122 1.088 /2.081 4.961 63 5 81 10 10.34% /2.1213**

27 0.004 /0.123 1.088 /3.352 1.198 15 5 12 2 25.93% 0.8165

27 0.155 0.052 1.101 /1.957 2.914 14 0 13 0 0% 0.2722

117 0.095 0.038 0.507 /1.347 2.649 63 8 54 7 12.82% 1.1767

117 0.008 /0.167 1.088 /2.081 4.961 49 5 68 10 12.82% /2.4841**

Panel B: Test for the difference in exposure between FCD users and non-users

Mean Mean Mean Mean

weekly raw exposure weekly absolute exposure monthly raw exposure monthly absolute exposure

FCD users Non users (n/77) (n/67)

t -statistic P -value

0.0178 0.3157 /0.1110 0.5237

1.0837 0.6710 1.7484 3.4509

0.1166 0.3744 0.2046 0.9706

0.2803 0.4141 0.0825 0.0007

Exchange rate exposures b2i are estimated from the following equation for the period from January 1997 to December 1999 using OLS: Rit b0i b1i Rmt b2i Rxt o it ; where Rit is the return on stock i in period t ; Rmt is the return on the ASX All Ordinaries Index in period t ; Rxt is the rate of change in the trade weighted index value of the AUD in period t ; oit is the error term. a The test statistic is a z statistic determined by a two tailed test of equality between the number of positive and negative exposure observations and is calculated as: p1  p2 ; z pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pq((n1  n2 )=n1 n2 ) where p1 is the proportion of positive cases; p2 is the proportion of the negative cases; n1 /n2 is the sample size; p/(x1/x2)/(n1/n2), where x1 is the number of positive cases and x2 is the number of negative cases; q/1/p . * Test statistic is significant at 10% level. ** Test statistic is significant at 5% level. *** Test statistic is significant at 1% level.

202

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

foreign operations and the use of FCDs as a proxy for financial hedging, we only find weak supportive evidence. Specifically, in Panel A of Table 2 we regress the raw value of the foreign currency betas obtained in Eq. (1) against FS/TS and FCD in a cross-sectional equation (Eq. (2)) and find that although the coefficient on FS/TS is positive, signaling that the presence of foreign sales increases the exposure, it is not statistically significant. Similarly, the coefficient on the FCD variable is positive and insignificant. For positive exposures, a positive relationship reveals that the use of FCDs actually increases the level of exchange rate risk to which a firm is exposed. Conversely, a positive relationship in the context of negative exposures means that the use of FCDs reduces such risk. To address this ‘sign confusion effect’, we use the absolute value of the estimated foreign currency betas instead of the raw value betas. As shown in Panel B of Table 1, while FCDs have no influence on the weekly exposure, they have a statistically significant lessening effect on the monthly exchange rate exposure. For a percentage increase in the ratio of FCD/TA used, exposure decreases by an average of 0.045 in magnitude. What remains puzzling, however, is the role of foreign sales in the weekly regression. It appears that higher foreign sales are associated with a lower level of weekly exposure. Taken together, these results provide some weak support for the evidence found by Allayannis and Ofek (2001). Using the same estimation framework, they found overwhelmingly supportive evidence based on a sample of U.S. companies. In particular, they report that exposure increases as the ratio of foreign sales to total sales increases and as the ratio of FCD/TA decreases. These results conform to theoretical predictions. To further address the ‘sign confusion’ issue, we regress the exchange rate beta separately against a set of positively and negatively exposed firms. The results of these regressions are presented in Panels C and D, respectively. According to the results, neither the ratio of FS/TS nor the use of FCDs has any power in explaining the variation in weekly exchange rate exposure. On the other hand, while the coefficients on FS/TS of the monthly regressions remain insignificant, the statistically significant negative coefficient of FCD/TA in Panel C and the positive coefficient in Panel D indicate that the use of FCDs, indeed, reduces the magnitude of monthly exposure, an effect representative of hedging behavior. Finally, Panel E of Table 2 reports regression results for a sample that consists of only FCD users. This specification is to account for the fact that regression results might be affected in some way by the inclusion of firms that do not use FCDs in the sample. However, this modification introduces a further complication. In particular, the lessening effect of the use of financial hedging is no longer observed. In contrast, the use of FCDs appears to be related to an increase in exchange rate exposure. There are two possible explanations. First, the result observed is a manifestation of the ‘sign confusion effect’ addressed earlier. Second, FCD users tend to have high exchange rate exposure to start with and while financial hedging via derivatives reduces exposure, FCD users still have more exposure than non-users who naturally

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215 Table 2 Tests for the relationship between foreign currency exposure and the use of currency derivatives a0

a1

a2

Panel A: Regression of raw value of FC betas on FS/TS and FCD/TA-whole sample results Weekly 0.0480 0.0093 0.0446 SE 0.1000 0.1804 0.0333 t -statistic 0.4797 0.0516 1.3397 P -value 0.6322 0.9589 0.1825 Monthly SE t -statistic P -value

/0.0356 0.1627 /0.2189 0.8270

0.1859 0.3995 0.4648 0.6428

0.0088 0.0149 0.5910 0.5554

Panel B: Regression of absolute value of FC betas on FS/TS and FCD/TA-whole sample results 0.0306 Weekly 0.4702 /0.3653 SE 0.0772 0.1390 0.0249 t -statistic 6.0875 /2.6285 1.2320 P -value 0.0000 0.0095 0.2200 Monthly SE t -statistic P -value

0.6190 0.1268 4.8801 0.0000

0.3366 0.3006 1.1198 0.2647

Panel C: Regression of positive FC betas on FS/TS and FCD/TA Weekly 0.3519 0.1204 SE 0.0841 0.2104 t -statistic 4.1844 0.5724 P -value 0.0001 0.5688 Monthly SE t -statistic P -value

0.6699 0.2854 2.3475 0.0222

0.5953 0.6615 0.8999 0.3718

/0.0448 0.0192 /2.3385 0.0208

0.0029 0.0092 0.3193 0.7504 /0.0482 0.0204 /2.3659 0.0212

Panel D: Regression of negative FC betas on FS/TS and FDC/TA 0.1461 Weekly /0.3790 SE 0.1402 0.2304 t -statistic /2.7034 0.6341 P -value 0.0088 0.5283

0.3148 0.2287 1.3748 0.1741

Monthly SE t -statistic P -value

0.1760 0.0546 3.2252 0.0018

/0.6004 0.0881 /6.8149 0.0000

/0.1354 0.1937 /0.6989 0.4867

Panel E: Regression of raw value of FC betas on FS/TS and FCD/TA-FCD users only results Weekly 0.0219 0.1683 0.0251 SE 0.0577 0.1580 0.0082 t -statistic 0.3804 1.0650 3.0523 P -value 0.7047 0.2903 0.0032 Monthly SE t -statistic

/0.1498 0.1218 /1.2298

0.1242 0.2970 0.4184

0.0245 0.0113 2.1696

203

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

204

Table 2 (Continued ) a1

a0 P -value

0.2227

a2 0.6769

0.0332

The relationship is estimated using the following model: b2i a0 a1

  FS TS

i

a2

  FCD TA

li ;

i

where b2i is the foreign currency (FC) exposure coefficient (beta) estimated in Eq. (1); FS/TS is the ratio of foreign sales to total sales; FCD/TA is the amount of FCDs used by a firm scaled by the firm’s total assets for the financial year 1999; li is the error term. have no or low exposure. However, this line of argument is in conflict with our univariate analysis.8 We also adopt a ‘response surface’ model in which the t -statistics for b2i estimated from Eq. (1) are used as the dependent variable in Eq. (2). This procedure is to address the concern that exchange rate exposures are estimated with a differing degree of precision due to the differing estimation variance. Table 3 reports the response surface regression results under different sample settings. For simplicity of interpretation we focus on the regression that uses the absolute value of t -stats. As shown in Panel B of Table 3, FS/TS and FCD/TA do not demonstrate a significant ‘response surface’ relationship.9 In general, our cross-sectional analyses show that there is some evidence that the use of FCDs is associated with a reduction in exchange rate exposure, thus supporting the hypothesis that the use of FCDs is for hedging purposes. Nevertheless, the evidence is generally weak to the extent that the relationship lacks consistency. The results also vary depending on the sampling interval upon on which exposures are estimated, for example, weekly versus monthly exposures.

8 We also perform an ‘interaction effect’ test between FS/TS and the use of FCDs. In particular, we use the following model:

b2i a0 a1

  FS

        FCD FCD FCD FCD a21 D1 a22 D2 a23 D3 a24 D4 vi ; TS i TA i TA i TA i TA i

where the definitions of all variables are the same as in Eq. (2), with the addition of: D1 /1 if FS/TS is in the first quartile and zero otherwise; D2 /1 if FS/TS is in the second quartile and zero otherwise; D3 /1 if FS/TS is in the third quartile and zero otherwise; D4 /1 if FS/TS is in the last quartile and zero otherwise. However, we find little evidence that the use of FCDs has different effects on the level of exchange rate exposure of a firm for various levels of foreign sales. More specifically, our results (not reported) show that a21, a22, a23 and a24 are neither significant nor statistically different from each other. 9 The results in Panel A that uses the raw value of t -stats, once again demonstrates a positive relationship between the use of FCDs and exchange rate exposure that we believe to be related to the ‘sign confusion effect’.

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

205

Table 3 Foreign currency exposure and the use of FCDs-response surface regression results a0

a1

a2

Panel A: Regression of raw value of t -statistics on FS/TS and FCD/TA-whole sample results Coefficient 0.0771 0.1362 0.0488 SE 0.1516 0.2956 0.0242 t -statistic 0.5089 0.4609 2.0180 P -value 0.6116 0.6456 0.0455 Panel B: Regression of absolute value of t -statistics on FS/TS and FCD/TA-whole sample results Coefficient 0.9344 /0.1263 /0.0177 SE 0.0881 0.1668 0.0145 t -statistic 10.6016 /0.7574 /1.2277 P -value 0.000 0.4501 0.2216 Panel C: Regression of positive value of t -statistics on FS/TS and FCD/TA Coefficient 1.0720 /0.3280 SE 0.1444 0.2466 t -statistic 7.4250 /1.3301 P -value 0.0000 0.1875

/0.0006 0.0081 /0.0758 0.9398

The estimation model is: tstati a0 a1

  FS TS

i

a2



 FCD TA

li ;

(2?)

i

where t-stati is the t-statistic of b2i estimated in Eq. (1) using weekly data; FS/TS is the ratio of foreign sales to total sales; FCD/TA is the amount of FCDs used by a firm scaled by the firm’s total assets for the financial year 1999; li is the error term. 4.2. Industry effect on exchange rate exposure There is some evidence in the existing literature pointing to the fact that exchange rate exposure might be an industry specific phenomenon. Bodnar and Gentry (1993), for example, argue that exchange rate fluctuations affect some industries differently than others because some are more export or import dependent than others. In particular, they document a significant positive exposure for industries such as apparel, transport equipment, motor freight transportation, general merchandise stores and miscellaneous retail while industries such as metal mining, heavy construction, petroleum refining, wholesale trade and durable goods and business services are reported to have significant negative exchange rate exposure. Similarly, Shin and Soenen (1999) study a set of U.S. manufacturing firms and find that only one industry (electrical equipment) has significant positive exposure and primary metal records a significant negative exposure. Given Australia’s international pattern of trade, it is expected that the resources sector will have ‘revenue exposure’ (negative exposure) while the industrial sector will have a ‘cost exposure’ (positive exposure). Australia is one of the world largest exporters of gold and mineral products such as coal, oil and natural gas. As such, resources firms have foreign currency denominated revenue that is contingent on

206

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

world commodity prices and therefore will be adversely affected as the AUD appreciates. Industrial firms, on the other hand, have a large component of their costs denominated in foreign currency, most notably the EURO and the USD, being the imports of semi-manufactured and manufactured industrial products. Therefore, industrial firms have a net positive exchange rate exposure and benefit when the AUD appreciates. Loudon (1993), for example, finds that resource stocks tend to have negative exposure while industrial stocks portray a generally positive exposure. In addition to the preceding argument, in the Australian setting several authors have noted the wide differential in performance between resource and industrial sector stocks, (see for example, Ball and Brown, 1980; Ball, 1986; Dolan, 1997; Ord, 1998). A final reason for considering a possible industry effect is that different sectors may also engage in different hedging practices. For example, it is widely believed that resource/mining companies are more likely to hedge extensively. As shown in Table 1, there is some univariate evidence that resource stocks and industrial stocks respond differently to changes in the value of the trade-weighted index. Accordingly, we extend Eq. (2) by adding two dummy variables proxying for the industry effect as follows:         FS FS FCD FCD b2i c1 c2 Dind c3 Dres c4 Dind c5 Dres TS i TS i TA i TA i ui ;

(4)

where Dres (Dind) is a dummy variable that is equal to unity if firm i belongs to the resource (industrial) sector and zero otherwise. This equation is designed to examine the potential differential impact that foreign operations and financial hedging have on exchange rate exposure for each industry sector. Additionally, in this analysis the absolute values of exposure were used. Results with regard to the weekly and monthly exposures are presented in Panels A and C of Table 4, respectively. Generally, the relative extent of foreign operations does not appear to have any impact on the level of exposure as indicated by the statistically insignificant coefficient estimates. The insignificant impact of FS/TS on exchange rate exposure is also uniform between industrial and resources sectors. As shown in Panels B and D, the coefficient tests fail to reject the null hypothesis that exposure is more influenced by the degree of international operation in one industry sector than the other at the 10% significance level. Nevertheless, in accordance with our expectations, industrial and resource firms appear to engage in different hedging practices and as a result derive a different level of hedging effectiveness. Since resource sector firms hedge more extensively, these firms achieve a greater degree of monthly exposure reduction (0.0296 compared to 0.1977) and by the same token, more severely adversely affected by an increase in weekly exposure. However, it is our belief that the monthly results are more robust and reliable to the extent that monthly returns are less affected by short-term return fluctuations brought about by factors other than changes in exchange rates. The coefficient equality tests reported in Panels B and D further strengthen our conclusions that the use of FCDs has a greater impact on resource sector firms than on industrial sector firms.

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

207

Table 4 Tests for the relationship between foreign currency exposure and the use of currency derivatives-industry sector specification c1

c2

c3

c4

c5

Panel A: Dependent variable is the absolute weekly exposure coefficient Coefficient 0.2390 0.3610 /0.0337 SE 0.0738 0.2919 0.0997 t -statistic 3.2388 1.2319 /0.3380 P -value 0.0018 0.2203 0.7363

/0.4941 0.2070 /2.3869 0.0197

0.6537 0.1105 5.9154 0.0000

Null hypothesis

Chi-statistic (Prob)

F -statistic (Prob)

Panel B: Tests for equality of coefficients c2 /c3 2.6419 (0.1085) 2.6419 (0.1041) 20.6692 (0.0000) 20.6692 (0.0000) c4 /c5 c2

c4

c5

Panel C: Dependent variable is absolute monthly exposure coefficient Coefficient 0.7044 0.0756 0.4621 SE 0.1318 0.3572 0.3945 t -statistic 5.3441 0.2118 1.1714 p -value 0.0000 0.8326 0.2434

/0.0296 0.0163 /1.8127 0.0720

/0.1977 0.0799 /2.4752 0.0145

Null hypothesis

Chi-statistics (Prob)

c1

c3

F -statistics (Prob)

Panel D: Tests for equality of coefficients 0.8099 (0.3691) 0.8099 (0.3681) c3 /c4 c5 /c6 3.8561 (0.0516) 3.8561 (0.0496) Model: b2i c1 c2 Dind

  FS TS

i

c3 Dres



 FS TS

i

c4 Dind



 FCD TA

i

c5 Dres

  FCD TA

ui ; i

where b2i is the weekly exchange rate exposure; FS/TS is the ratio of foreign sales to total sales; FCD/TA is the amount of FCDs used by a firm scaled by the firm’s total assets for the financial year 1999; Dres (Dind) is an industry dummy variable that is equal to unity if firm i belongs to the resource (industry) sector and zero otherwise; ui is the error term.

4.3. Determinants of exchange rate exposure In Table 5, we introduce a set of controlling variables that are incentives for hedging, namely: firm size, leverage, liquidity, and growth opportunities (as measured by the PER) together with FS/TS and FCD/TA. Chow and Chen (1998), for example, find that Japanese firms with high leverage, low liquidity and high cash dividends have high exchange rate exposure. Conforming to the notion that hedging exhibits economies of scale, for return horizons greater than 1 month, they also find that larger firms tend to have smaller exposure. Using the absolute

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

208

Table 5 Determinants of foreign exchange exposure

Constant Leverage Size FSTS FCD Liquidity PE R -squared

Coefficient

SE

t -statistic

P -value

0.6086 0.1414 /1.64E/08 0.3267 /0.0426 0.3345 /0.0004 0.0421

0.1954 0.4479 7.91E/09 0.2740 0.0179 0.2374 0.0006

3.1139 0.3157 /2.0685 1.1924 /2.3779 1.4093 /0.7336

0.0022 0.7527 0.0405 0.2352 0.0188 0.1610 0.4645

The cross-sectional regression is specified as follows: b2i a0 a1 LEVi a2 SIZEi a3 FSTSi a4 FCDi a5 LIQi a6 PEi ui ; where b2i is the absolute value of the estimated monthly foreign currency exposure from Eq. (1); LEV is the leverage ratio calculated as total debt over firm size; SIZE is calculated as the sum of the market value of equity and total debt; FSTS is the ratio of foreign sales over total sales; FCD is the ratio of FCDs over total assets; LIQ is the liquidity ratio calculated as the ratio of cash and cash equivalents over firm size; PE is the price earnings ratio; ui is the error term.

value of monthly exposure, we find that both firm size and the use of financial hedging, as proxied by FCDs, are associated with a reduction of exchange rate exposure. This result relating to FCDs adheres to theoretical expectations and suggests that firms use FCDs with a view to hedging short-term exchange rate exposure, a result consistent with the findings of the Wharton School survey of derivative usage (1998). The insignificant role of foreign operations in explaining the variation in exchange rate exposure further provides an indication that while foreign costs and revenues represent a source of potential risks as the exchange rate fluctuates, such risks can be hedged away using various instruments including FCDs. The negative relationship between firm size and exposure is intuitive. In light of hedging theory, using derivatives is the game of the ‘big firms’ since only they have sufficient human and financial resources to seriously participate in the derivative market. Consequently, they are more effective in hedging risks, leading to the results we observe. Although this result contradicts our prediction that smaller firms tend to have smaller exposure in the short run because they hedge more extensively against costly financial distress situations, it suggests that larger firms are not only more efficient in performing long-term economic hedges but also short-term financial hedges. All in all, we have achieved a moderate success in identifying and explaining exchange rate exposure. Although there is a lack of evidence that the degree of overseas operations increases a firm’s stock return sensitivity to fluctuations in exchange rate changes, the available empirical evidence suggests that the use of FCDs is associated with a reduction in exposure. It also appears that the monthly exposure analysis better corresponds to theoretical predictions than the weekly

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

209

exposure analysis. As mentioned earlier, weekly stock returns are more likely to be affected by short-term fluctuations that are not attributable to fluctuations in exchange rates. 4.4. Robustness checks 4.4.1. Exposure to the USD and JPY The use of a trade weighted index to proxy for the exchange rate can be criticized on the grounds that changes in the index are assumed to affect individual firms uniformly and that a firm’s international linkages are similar to that of the national trade. As a result, the use of a trade-weighted index may cause aggregation biases which undermine the effort to estimate firm specific exchange rate coefficients for individual currencies. To address this issue, we re-estimate the monthly exchange rate exposure using two cross currency exchange rates that are most relevant to Australian firms, namely the Japanese Yen and the US Dollar. Panel A of Table 6 reports the descriptive statistics of the JPY and USD exchange rate exposure. In general, 10.42 and 11.11% of the sample have significant exposure to the USD and the JPY, respectively. Although the test statistics of the sign test suggest that USD and JPY exchange rate exposure is evenly distributed in terms of positivity and negativity, the number of significant observations indicates otherwise. It appears that Australian firms are more likely to be positively exposed to the USD and negatively exposed to the JPY. This reflects the trading pattern of Australian firms whereby they import more from the USA (22% of total imports) and export more to Japan (19% of total exports). These results generally confirm the earlier findings that only a limited number of Australian firms have significant exchange rate risk exposure. The remaining firms may have effectively managed their exposures via financial hedging or are not inherently impacted by exchange rate fluctuations in the first place. Consistent with the above analyses, in Panels B and C, we report the results of the cross-sectional regressions using the absolute value of monthly USD and JPY exposure as the dependent variables. Interestingly, while the use of FCDs reduces exposure to changes in the AUD/USD exchange rate, such effect is absent in the case of the JPY. Given that the USD and the JPY are the two most important currencies to Australia trade and most widely hedged against, the exposure reduction effect that we observe in the Section 4.3 must have come from the reduction in the USD, not the JPY. 4.4.2. Long horizon exposure analysis It is argued by many that short-term exchange rate exposure might be difficult to identify because using short horizon stock returns tends to measure operational exposure which can easily be alleviated by firms through various hedging techniques (see Amihud, 1993; Bodnar and Gentry, 1993; Bartov and Bodnar, 1994; Chow et al., 1997). Economic exposure, on the other hand, is long-term in nature and more difficult to hedge and therefore firms tend to be more exposed to currency fluctuations in the long run. Based on the assumption that using long horizon

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

210

Table 6 Exposure to fluctuations of the USD and JPY exchange rates and the use of FCDs USD Panel A: Descriptive statistics for exchange rate exposure Mean 0.1051 Median 0.0105 Standard deviation 0.5868 Maximum 3.1465 Minimum /1.2428 No. of positive cases 74 No. of significant cases 10 No. of negative cases 70 No. of significant values 5 % of significant cases 10.42% Test statistica 0.4714 Constant

JPY

0.0154 0.0130 0.3022 1.4768 /1.0935 76 5 68 11 11.11% 0.9428 FS/TS

FCD/TA

Panel B: Regression of absolute value of FC betas (USD) on FS/TS and FCD/TA /0.0363 Coefficient 0.1243 /0.0249 SE 0.0830 0.1287 0.0129 t -statistic 1.4981 /0.1062 /2.8174 P -value 0.1363 0.9156 0.0055 Panel C: Regression of absolute value of FC betas (JPY) on FS/TS and FCD/TA Coefficient 0.2370 0.0478 SE 0.0455 0.1208 t -statistic 5.2103 0.3955 P -value 0.0000 0.6931

/0.0023 0.0073 /0.3131 0.7547

The relationship is estimated using the following model: b2i a0 a1

  FS TS

i

a2

  FCD TA

li ;

i

where b2i is the foreign currency (FC) exposure coefficient (beta) estimated in Eq. (1); FS/TS is the ratio of foreign sales to total sales; FCD/TA is the amount of FCDs used by a firm scaled by the firm’s total assets for the financial year 1999; li is the error term. a Refer to Table 1 for a definition of the z test statistic. Exposure to changes in theUSD and JPY cross rate is estimated using Eq. (1) whereby changes in the trade weighted index are replaced by changes in the AUD/USD and AUD/JPY cross rates, respectively.

returns can yield a measure of economic exchange rate exposure, in this section we focus on a firm’s long-term exposure and the role that an individual firm’s characteristics play in explaining the variation in such exposure. For this purpose, we estimate exchange rate exposure at the 1-, 3-, 6-, 12- and 24-months (overlapping) intervals for an extended period from January 1995 to December 1999. This procedure, as noted by Chow and Chen (1998), causes the error term to be autocorrelated with order T/1. To address this issue, we once again use Hansen’s

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

211

(1982) GMM technique to estimate longer term exchange rate exposure. This method adjusts the variance /covariance matrix of the estimated coefficients for heteroskedasticity and autocorrelation. Panel A of Table 7 presents a summary of the exchange rate exposures over the specified time intervals. Generally speaking, our results support previous findings that the number of firms having significant exposure increases with the time horizon. In particular, at the 1 month horizon 15.28% of all firms have significant exchange rate exposure while 58.33% are exposed for the 24 months horizon. It also appears that Australian firms are overwhelmingly positively exposed to exchange rate changes and the number of firms that are positively exposed increases with the time horizon. These results indicate that in the long run returns on Australian firms’ stock generally decline when the AUD depreciates. Given that Australia has traditionally been a net importer, we believe that our results truly reflect the pattern of impacts that changes in the exchange rate have on Australian firms. Panel B of Table 7 indicates that there is a strong and positive correlation between exchange rate exposures of different horizons. The correlation tends to be stronger as the time gap narrows. Our univariate test for the difference in mean exchange rate exposure between FCD users and non-users, as presented in Panel C, reveals that the mean exposure of FCD users is not statistically different from that of non-users across different return horizons. In other words, regardless of the use of FCDs ex-post, firms appear to be exposed to a similar level of exchange rate risk. As noted above, this might be the ‘equilibrium’ level of risk that firms are willing to take after having successfully transferred the unwanted risks to another party. The results of a cross-sectional regression are presented in Table 8. The independent variables are calculated as the 2-year average of 1996 and 1999. Due to an inability to obtain a continuous measure for the use of FCDs in the year 1996, we use a dummy variable that equals unity if a firm indicates the use of FCDs in both 1996 and 1999. The results show that for relatively shorter return horizons (1 and 3 months) and for the 24 months return horizon, exchange rate exposure is negatively associated with a firm’s PE ratio. Since our descriptive statistics indicate that the firms in the sample generally have positive exposure, the discussion will be focused on positive exposure. Hedging theories suggest that generally a firm is more likely to hedge if it has more growth options (as proxied by the PE). This is so because a firm with more growth options faces higher potential underinvestment costs if growth is hindered by a lack of financial resources. Therefore, under the assumption that hedging instruments utilized by the firm include FCDs and that the use of FCDs is effective in alleviating foreign currency exposure, firms with a high PE will have lower exchange rate exposure. As a result, this result seems plausible in light of hedging theories. For relatively longer return horizons (12 and 24 months), exchange rate exposure appears to be positively related to liquidity (consistent with our predictions). In light of hedging theories, liquidity is a substitute for hedging. As a result, a higher level of liquidity tends to be associated with a lower degree of hedging and thus higher exposure. This result is, however, inconsistent with Chow and Chen (1998) who report that a higher level of liquidity is associated with lower exchange rate

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

212

exposures. Moreover, it can be argued that a firm’s hedging incentive to manage cash flows is relatively short-term in nature and therefore the hedging employed cannot explain long-term exposure. In the absence of a bridging theory that links liquidity Table 7 Long horizon exchange rate exposure Horizon (months)

1

3

Panel A: Exchange rate exposure of firms across return horizons Number of observations 144 144 Mean 0.1940 0.3365 Median 0.1893 0.3166 Standard deviation 0.6922 1.0661 Minimum /2.4850 /5.895 Maximum 2.4492 3.8993 No. of positive cases 87 97 No. of significant cases (at a /10%) 13 31 No. of negative cases 57 47 No. of significant cases (at a /10%) 9 11 % of significant cases 15.28% 29.17% Test statistica 3.5355*** 5.8926***

6

12

24

144 0.4702 0.4680 1.2894 /6.0422 4.5851 99 39 45 10 34.03% 6.3640***

144 0.9924 0.7689 1.9695 /3.3174 8.0152 103 58 41 14 50.00% 7.3068***

144 1.3729 0.9769 2.7314 /6.1013 9.9889 106 66 38 18 58.33% 8.0139***

0.2929 0.1456 0.3753 1.00

0.3878 0.4107 0.5599 0.6804 1.00

Panel B: Correlation coefficients between exposure betas of different horizons 1 1.00 0.6754 0.6084 3 1.00 0.9049 6 1.00 12 24 FCD users (n/77)

Non-users (n/67)

t -statistic

Panel C: Test for the difference in exposure between FCD users and non-users Mean exposure (1 month) 0.1334 0.2636 1.1268 Mean exposure (3 months) 0.3904 0.2897 0.5641 Mean exposure (6 months) 0.4920 0.4513 0.1885 Mean exposure (12 months) 0.9095 1.0646 0.4699 Mean exposure (24 months) 1.1970 1.5259 0.7194

P -value

0.2617 0.5738 0.8507 0.6391 0.4730

Exchange rate exposures are reestimated for longer time horizon returns, from January 1995 to December 1999 using a variation of Eq. (1). Ri;t;tT b0i b1i Rm;t;tT b2i Rx;t;tT o i;t;tT ; where Ri , t , tT is the continuous compound return on stock i in period t , t/T ; Rm ,t , tT is the return on the ASX All Ordinaries Index in period t , t/T ; Rx , t , tT is the rate of change in the trade weighted index value of the AUD in period t , t/T ; oi , t , tT is the error term. * Test statistic is significant at 10% level. ** Test statistic is significant at 5% level. a Refer to Table 1 for a definition of the z test statistic. *** Test statistic is significant at 1% level.

Horizon

1 month

3 months

6 months

12 months

24 months

Constant Leverage Size FSTS FCD Liquidity PE

0.3287 (2.3218) ** /0.2069 (/0.5875) /1.25E/08 (/1.1838) 0.2537 (1.1573) /0.1196 (0.3082) 0.1274 (0.2905) /0.0029 (/3.0587)***

0.2598 (1.1717) 0.4055 (0.4648) /1.87E/08 (/1.1323) 0.3490 (1.0135) /0.1444 (/0.7864) 0.5127 (0.7439) /0.0026 (/1.6893)*

0.4256 (1.5673) 0.1383 (0.2042) /1.75E/08 (/0.8647) 0.1943 (0.4607) /0.0661 (/0.2938) 0.9759 (1.1563) /0.0022 (/1.1875)

1.0434 (2.5528)** /0.3555 (/0.3475) /4.30E/08 (/1.4084) /0.2123 (/0.3335) 0.1754 (0.5163) 2.5876 (2.0310)** /0.0031 (/1.1136)

0.9205 (1.6587)* 1.1425 (0.8251) /5.68E/08 (/1.3739) 0.3016 (0.3499) 0.1824 (0.3966) 3.8594 (2.2375)** /0.0079 (/2.1025)**

The cross-sectional regression model is specified as follows: b2i a0 a1 LEVi a2 SIZEi a3 FSTSi a4 FCDi a5 LIQi a6 PEi ui ; where b2i is the foreign exchange exposure estimated using Eq. (1); LEV is the 2-year average (1996 and 1999) leverage ratio calculated as total debt over firm size; SIZE is calculated as the 2-year average (1996 and 1999) sum of the market value of equity and total debt; FSTS is the 2-year average (1996 and 1999) ratio of foreign sales over total sales; FCD is a dummy variable equaling to unity if a firm uses FCDs in both 1996 and 1999 and zero otherwise; LIQ is the 2-year average (1996 and 1999) liquidity ratio calculated as the ratio of cash and cash equivalents over firm size; PE is the 2-year average (1996 and 1999) price earnings ratio; ui is the error term. * Test statistic is significant at 10% level. ** Test statistic is significant at 5% level. *** Test statistic is significant at 1% level.

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

Table 8 Cross-sectional regression analysis of long horizon foreign currency exposure

213

214

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

with long-term exchange rate exposure, it is not so clear why liquidity can explain variations in long-term economic exposure. On the other hand, if a firm manages its cash flows using different hedging techniques (including derivatives) with a longterm view then our argument with regard to liquidity remains valid. Despite the belief that long-term exchange rate exposure is more difficult to hedge and that larger firms tend to be more likely to have the resources required to action such hedges, we do not find any evidence regarding the impact of firm size on the exchange rate exposure. Additionally, our failure to document a relationship between the use of FCDs and long horizon exposure lends support to the idea that using long horizon stock returns captures economic exposure and that such exposure can only be hedged via major operational restructuring, not a financial hedge using derivatives. Finally, leverage and the existence of foreign sales do not appear to have any power in predicting exchange rate exposures across different return horizons.

5. Conclusion In this paper we seek to document both the short- and long-term exchange rate exposure of Australian companies and more importantly, the effect of the use of FCDs on such exposures. The major contribution of the paper lies in the fact that it is the first paper that thoroughly examines both short- and long-term exchange rate exposure in a cross section of Australian firms (rather than a portfolio or an index) and more importantly documents the role of financing hedging in influencing the level of firms exchange rate exposure. Although the estimation model is not new in itself, there is a certain degree of innovation in the selection of the independent variables as well as the estimation techniques. Consistent with previous studies, we find that ex-post, Australian firms have minimal short-term exposure to exchange rate fluctuations. Moreover, we achieve a degree of success in providing empirical evidence in support of the hypothesis that the use of FCDs reduces short-term exchange rate exposure (especially the monthly exposure). As the return horizon lengthens FCDs, however, appear to be less effective in hedging exchange rate risks. Our results lend support to the argument that using short return horizons (weekly and monthly) only captures the extent to which firms are exposed to transaction exposure and since transaction exposure can be effectively hedged via financial hedging, only a small number of firms are exposed to fluctuations in the exchange rate in the short run. Our long horizon analyses reveal that Australian firms are generally exposed to foreign exchange risks in the long run. This finding is consistent with Chow et al.’s (1997) conjecture that economic exposure is difficult to hedge */while Australian firms appear to be very effective in hedging short-term exchange rate exposure, they have limited success in hedging long-term exposure. Finally, while leverage, foreign sales and the use of FCDs do not have any effect on exchange rate exposures of different return horizons, exchange rate exposures of shorter horizons appear to be negatively related to a firm’s PER, while exposures of longer horizons are positively

H. Nguyen, R. Faff / J. of Multi. Fin. Manag. 13 (2003) 193 /215

215

related to a firm’s liquidity. The negative role of the PER (as a proxy for growth opportunities) supports the ‘underinvestment’ hypothesis while the positive role of liquidity supports the view that liquidity and hedging are substitutes.

Acknowledgements The authors gratefully acknowledge the helpful comments of an anonymous referee.

References 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., 1993. Evidence on exchange rates and valuation of equity shares. Recent Advances in Corporate Performance, Business One, Irwin, Homewood, IL. Ang, J.S., Chua, J.H., McConnell, J.J., 1982. The administrative costs of corporate bankruptcy: a note. Journal of Finance 37, 219 /226. Ball, R., 1986. Risk and return from equity investments in the Australian mining industry: 1974 /8. JASSA 3, 10 /12. Ball, R., Brown, P., 1980. Risk and return from equity investments in the Australian mining industry: January 1958 /February 1979. Australian Journal of Management 5, 45 /66. Bartov, E., Bodnar, G.M., 1994. Firm valuation, earnings expectations, and the exchange rate exposure effect. Journal of Finance 49, 1755 /1786. Bessembinder, H., 1991. Forward contracts and firm value: investment incentive and contracting effects. Journal of Financial and Quantitative Analysis 26, 519 /532. Bodnar, G.M., Gentry, W.M., 1993. Exchange-rate exposure and industry characteristics: evidence from Canada, Japan and the U.S. Journal of International Money and Finance 12, 29 /45. Chow, E., Chen, H., 1998. The determinants of foreign exchange rate exposure: evidence on Japanese firms. Pacific-Basin Finance Journal 6, 153 /174. Chow, E.H., Lee, W., Solt, M., 1997. The economic exposure of U.S. multinational firms. Journal of Financial Research 20, 191 /210. Di Iorio, A., Faff, R., 2001. A test of the stability of exchange rate risk: evidence from the Australian equities market. Global Finance Journal 12, 179 /203. Dolan, P., 1997. Tilting the balance: how to put the ‘value’ in value management. JASSA, 24 /29 (Summer). Geczy, C., Minton, B.A., Schrand, C., 1997. Why firms use currency derivatives. Journal of Finance 52, 1323 /1354. Hansen, L., 1982. Large sample properties of general method of moment estimators. Econometrica 50, 1029 /1054. Jorion, P., 1990. The exchange rate exposure of U.S. multinationals. Journal of Business 63, 331 /345. Loudon, G., 1993. The foreign exchange operating exposure of Australian stocks. Accounting and Finance 33, 19 /32. Nance, D., Smith, C.W., Smithson, C.W., 1993. On the determinants of corporate hedging. Journal of Finance 48, 267 /284. Ord, T., 1998. Diggers, dreamers and lady luck. JASSA, 2 /7 (Autumn). Shin, H., Soenen, L., 1999. Exposure to currency risk by U.S. multinational corporations. Journal of Multinational Financial Management 9, 195 /207.