Earnings Management and Stock Exposure to Exchange Rate Risk

Transcription

1 Earnngs Management and Stock Exposure to Exchange Rate Rsk Feng-Y Chang a, Chn-Wen Hsn b, and Shn-Rong Shah-Hou c JEL classfcaton: F31, G30 Keywords: Exchange rate exposure, Earnngs Management, Theory of optmal hedgng a Graduate School of Management, Yuan Ze Unversty b Department of Fnance, Yuan Ze Unversty, Taoyuan 32003, Tawan, Tel: ext. 2662, Fax: , e-mal: fncwshn@saturn.yzu.edu.tw (correspondng author) c. Department of Fnance, Yuan Ze Unversty, Tawan.

2 Abstract Ths study proposes an earnngs management hypothess to explan the lack of prevalng evdence of sgnfcant relatonshp between exchange rate rsk and stock returns of U.S. corporatons. In partcular, we examne whether the earnngs management actvtes undertaken by managers mtgate frm exchange rate exposure. Consstent wth our hypothess, we fnd that frms engaged wth greater earnngs management actvtes, especally those of ncome smoothng, tend to exhbt sgnfcantly lower exchange rate exposure. Ths assocaton tends to be partcularly strong when the currency movements are adverse to the frm earnngs. Our results also have mplcatons for the nformaton contents about earnngs qualty shed from the correlaton between accrual changes and cash flow changes. 1

3 1. Introducton Foregn currency exchange rate rsk affects a frm s operatng and fnancng actvtes and therefore the assocated cash flows. Most pror studes, however, faled to fnd a sgnfcant relaton between foregn exchange rate fluctuatons and stock returns of U.S. corporatons [e.g., Joron (1990), Amhud (1994) and Bodnar and Gentry (1993), Bartov and Bodnar (1994), Grffn and Stulz (2001) and Dodge et al. (2002)]. 1 Varous explanatons have been suggested to resolve ths paradox, ncludng the use of hedgng nstruments [Rawls and Smthson (1989), Dolde (1993), Geczy et al. (1997), He and Ng (1998), and Allayanns and Ofek (2001)], the delayed reacton of the stock prce to the exchange rate fluctuatons [Bartov and Bodnar (1994), Chow et al. (1997a,b), Grffn and Stulz (2001), and Bodnar and Wong (2003)] and possble sample selecton problems [Bartov and Bodnar (1994)]. Nonetheless, only lmted success has been reached. Ths paper attempts to offer another explanaton to ths puzzle, thanks to the recent advances n the earnngs management lterature. In partcular, we argue that earnngs management by corporate managers contrbutes to the nsgnfcant assocaton between exchange rate changes and stock returns. Exchange rate changes affect a frm s fnancal poston. Gans or losses due to favorable or adverse exchange rate movements appear n a frm s fnancal report and lead to earnngs performance devatng from corporate target. Although the mpact of unexpected exchange rate changes can be mtgated through the usage of foregn currency dervatves, however, certan frms also argue n ther accountng report that they cannot and do not want to fully elmnate ther exchange rate rsk. For ths reason, a frm has ncentves to reduce the mpact of unexpected exchange rate fluctuatons through earnngs management strateges when ts currency rsk exposure s not hedged or not well-managed. If a frm does employ earnngs management to wndow dress reported operatng results, nvestors wll not be able to tmely and 1 One excepton s the evdence presented by Wllamson (2001), whch fnds sgnfcant compettve effects of exchange rate shocks between Japanese and the U.S. automotve ndustres. In terms of nternatonal evdence, Bodnar and Gentry (1993) and He and Ng (1998) fnd more sgnfcant exchange rate exposure for Canadan and Japanese frms. Dodge et al. (2002) fnd that exchange rate movements are economcally sgnfcant n terms of frm value for nternatonal markets. Grffn and Stulz (2001) fnd that the common ndustry effect across countres s several tmes larger than the competton effect of exchange rate shocks on U.S. ndustres. The exchange rate effects for other countres are larger, but are stll generally small. 2

4 accurately evaluate ts exchange rate exposure because the actual mpact of exchange rate changes on frm value wll not even be released n the future. Ths then leads to the absence of a sgnfcant relaton between exchange rate fluctuatons and stock returns. In vew of ths possblty, we hypothesze that those frms employng greater levels of earnngs management wll reduce more of ther exchange rate exposure that s assessed by the stock market. Further analyses wll be developed later n ths paper to present such possblty. Dumas (1978), Hodder (1982), and Adler and Dumas (1984) defne a frm s exchange rate exposure as the senstvty of a frm s market value to unexpected exchange rate varatons. Followng Adler and Dumas (1984), we measure a frm s exchange-rate exposure as the regresson coeffcent of the frm s stock returns on the foregn exchange rate(s) changes whle controllng for the Fama-French three factors (1993). The lagged as well as the contemporaneous exchange rate change s ncluded n the regresson model to ncorporate the possble delay n the market s reacton to exchange rate fluctuatons. Usng a sample of non-fnancal COMPUSTAT frms, we fnd that only about 22 percent of our sample frms have sgnfcant contemporaneous exposure coeffcents and about 18 percent of the sample frms have sgnfcant lagged exposure coeffcents. Ths prelmnary result s consstent wth exstng studes and also conforms to our earnngs management hypothess, whch mples that the mpact of exchange rate changes wll not be fully reflected from ether current or future stock prces f contnued earnngs management s at work. The combnaton of contemporaneous and lagged exposure coeffcents s then used to measure a frm s combned exchange rate exposure as suggested by Bartov and Bodnar (1994). Followng the study of Leuz, Nanda, and Wysock (2003), we apply four earnngs management measures to capture the extent to whch a frm manpulates reported earnngs. 2 Two of these measures, denoted as EM1 and EM2 n the paper, are developed to capture the degree of earnngs management ntended to smooth a frm s reported earnngs. A thrd measure, EM3, consders the average magntude of 2 Leuz, Nanda, and Wysock (2003) develop four country-level earnngs management measures whch do not adapt to our frm-level analyses. 3

5 dscretonary accruals over a specfc perod. The fourth measure, EM4, assesses a frm s tendency to avod reportng small losses. The descrptve analyss ndcates that all four earnngs management measures are strongly related to frm exchange rate exposure, nonetheless n dfferent drectons. Frms that are more aggressve n smoothng ther ncomes, as measured by EM1 and EM2, or that are more averse to report small losses, as assessed by EM4, tend to show lower exchange rate exposure, the fndng of whch s consstent wth our hypothess. To the contrary of our expectaton, those frms wth hgher level of dscretonary accruals (EM3) exhbt hgher exchange rate exposure. Further analyss s requred to resolve ths nconsstency. After controllng for other corporate factors of exchange rate exposures, the panel regressons fnd that a more aggressve earnngs management as revealed by the level of ncome smoothng (EM1 and EM2) or by the tendency of avodng small loss reportng (EM4) does show to be assocated wth lower exchange rate exposures. Ths evdence s consstent wth our hypothess that earnngs management mtgates frm exchange rate rsk exposures. However, the other earnngs management measure that proxes for the magntude of dscretonary accruals (EM3) stll demonstrates a postve relatonshp between earnngs management and stock return exchange rate exposure, a result nconsstent wth our hypothess. These results are not materally changed when other sample s consdered, e.g. large frms, small frms and frms wth zero foregn sales reportng. We re-examne the relaton between frm exchange rate exposure and the earnngs management measures at a ed pont n tme,.e., for sngle sample wndow one at a tme. The two ncome smoothng measures become the only measures exhbtng a prevalngly sgnfcant result, a result consstent wth our hypothess. Fnally, we examne whether the effect of earnngs management on frm exchange rate exposure remans the same durng favorable and unfavorable perods of currency movements. Two hypotheses are proposed to explan a frm s earnngs management behavors and ther mpact on the observed exchange rate exposure. Our evdence ndcates that the effects of earnngs management actvtes, especally those 4

6 amng to smooth reported ncome, become more pronounced durng adverse operatng condtons. Ths s consstent wth the opportunstc earnngs management hypothess, whch suggests that frms tend to engage n more earnngs management to stablze ther reported earnngs durng unfavorable perods, however, they may choose not to mask the benefcal operatng results of favorable perods. Ths paper contrbutes to the lterature by offerng a plausble explanaton for the falure of fndng strong evdence of prced exchange rate rsk for U.S. stocks. In addton, our emprcal results provde mplcatons for the nformaton contents of the correlaton between accruals and cash flows. To a certan extent, the result from our second ncome smoothng measure s consstent wth the argument that a greater correlaton between accruals and cash flows mples earnngs management. The remander of ths paper s organzed as follows. Secton 2 descrbes our methodology to measure frm exchange rate exposure and earnngs management. Secton 3 examnes the effect of earnngs management on frm exchange rate exposure and Secton 4 concludes. 2. Measurng exchange-rate exposure and earnngs management Inconsstent wth common ntuton and expectaton, exstng emprcal studes do not fnd sgnfcant evdence that the currency exchange rate rsk s prced for U.S. corporatons. Ths paper proposes to explan ths paradox wth the earnngs management actvtes employed by frms. Snce the behavors of manpulatng earnngs would mask a frm s real performance subject to unexpected exchange rate movements, nvestors may not evaluate accurately and tmely the mpact of exchange rate changes on a frm s market value. Ths then leads to the observed weak lnkage between exchange rate changes and contemporaneous stock returns. Ths secton descrbes the measure of frm exchange rate exposure and the proxy varables to assess the extent to whch managers manpulate earnngs usng ther fnancal reportng decsons. 5

7 2.1 Estmatng a frm s exchange-rate exposure Dumas (1978), Hodder (1982), and Adler and Dumas (1984) defne a frm s exchange rate exposure as the senstvty of a frm s market value to unexpected exchange rate varatons. 3 Adler and Dumas (1984) thereupon suggest that a frm s exchange-rate exposure can be best measured as the regresson coeffcent of the frm s market value on the contemporaneous foregn exchange rate(s). Joron (1990) further ncludes the market return n the regresson equaton to measure the exchange rate exposure. 4 Ths study modfes the prevous market-model based regresson and nstead employs the three-factor model suggested by Fama and French (1993). Moreover, as Amhud (1994) and Bartov and Bodnar (1994) document that lagged changes n foregn currency exchange rate demonstrate a sgnfcant effect on contemporaneous stock returns, we also nclude the lagged exchange rate change n the regresson equaton to ncorporate the possble delay n the market s reacton to exchange rate fluctuatons. The regresson model used n ths study to estmate a frm s exchange-rate exposure s as follows. R t 0 lagged - R ft = + β FX t + β FX t 1 β + β ( R - R ) + β SMB + β HML + ε m mt ft SMB t HML t t (1) In the regresson, R t s the stock return of frm n perod t, R ft s the rsk-free rate n perod t, FX t s the percentage change of the trade-weghted exchange rate ndex, measured as foregn currency per one unt of U.S. dollar n perod t, FX t 1 s the lagged percentage change n the exchange rate, R mt s the return on the market portfolo, SMB t s the dfference n the returns between value-weghed portfolos of small and bg stocks, and HML t s the dfference n the returns between value-weghted portfolos of hgh 3 Adler and Dumas (1984) defne exchange rate exposure wthout suggestng there exsts certan causaton between stock prces and exchange rates, nstead, both varables are endogenously determned. However, t can be safely assumed that exchange rates are exogenous for an ndvdual frm [see also He and Ng (1998) and Allayanns and Ofek (2001)]. 4 Many other related studes appled smlar methodology. See, e.g. Joron (1990, 1991), Amhud (1994), Bodnar and Gentry (1993), Allayanns (1996), He and Ng (1998), Allayanns and Ofek (2001) and Dodge, Grffn and Wllamson (2002). 6

8 book-to-market stocks and low book-to-market stocks. 5 To estmate stock return exchange-rate exposure, the selecton of an approprate measurement of foregn currency exchange rate, e.g. a trade-weghted exchange rate ndex or a blateral currency exchange rate, has been a debatable queston. It s common for a frm exposed to more than one currency [e.g., see Adler and Dumas (1984)]. However, the smultaneous ncluson of multple blateral exchange rates could encounter a mult-collnearty problem. Ths paper thus chooses to use the Bank of England nomnal trade-weghted exchange rate ndex to estmate the exchange rate exposures. The nomnal, nstead of real, exchange rate s appled here. Ths follows the consderatons that the gans and losses from foregn currency transactons recognzed n the accountng report are determned prmarly based on nomnal exchange rate changes and that the strateges employed by frms to mtgate the effect of adverse exchange rate fluctuatons are more lkely consdered on nomnal bass. Adler and Dumas (1984) suggest that the exchange-rate exposure may vary over tme. 6 Ths study estmates each frm s exchange-rate exposure over a three-year sample wndow wth monthly return data and re-estmates the exposures as the 3-year wndow rolls forward. Fve-year sample wndows are also tested and the results are not materally changed. In equaton (1), β measures the contemporaneous exchange rate exposure of frm for the specfc three-year sample wndow whle β measures the lagged response of exchange rate changes on the lagged stock return. 7 Bartov and Bodnar (1994) demonstrate that lagged dollar changes are a sgnfcant varable n explanng current abnormal returns of ther sample frms and suggest to use the combnaton of β and β lagged to measure the possble effect of gven dollar changes. In lght of ths, we use the sum of β and β to measure the exchange rate exposure of frm. Note that an ncrease (a decrease) n the lagged 5 We would lke to thank Professor Fama and French for sharng the three-factor data on French s webste. 6 See also Jorn (1990) and Allayanns (1997) 7 The coeffcents n equaton (1) are estmated based on ordnary least-squares regresson, and the standard errors are adjusted for heteroskedastcty and autocorrelaton. 7

9 BOE trade-weghted ndex ndcates a U.S. dollar apprecaton (deprecaton). Frms wth postve combned exposures are regarded as net mporters snce the dollar apprecaton s expected to show favorable (postve) effect on the frm stock returns. Conversely, frms wth negatve combned exposures are regarded as net exporters snce the dollar apprecaton s expected to have adverse (negatve) effect on the frm stock returns. 2.2 Measures of the Engagement of Earnngs Management Ths paper attempts to offer another explanaton, based on frms engagement of earnngs management, for the lack of strong evdence of sgnfcant relatonshp between dollar fluctuatons and stock returns. Exchange rate shocks affect the sales and costs of products and thus a frm s long-term operatng cash flows. Exchange rate movements also affect the domestc value of a cash flow when t s denomnated n foregn currency. A frm may adopt busness hedgng or fnancal hedgng schemes to reduce the effect of exchange rate shocks on operatng performance. The lterature on optmal hedgng theory has offered frm characterstcs that are assocated wth those frms wth strong motvaton to hedge. However, frms often cannot or wll not fully hedge away ther exchange rate rsk due to ether mplementaton constrants or possble negatve effect on frm performance. CEOs of these frms thus have strong ncentves to conceal the mpact of unexpected exchange rate fluctuatons through earnngs management. If managers do undertake earnngs management actvtes to wndow-dress ther reported operatng performance, nvestors cannot accurately and tmely evaluate a frm s exchange rate exposure. Ths then leads to the documented nsgnfcant contemporaneous relatons between exchange rate changes and stock returns. We thus hypothesze that those frms engaged wth greater magntude of earnngs management expect to experence wth lower exchange rate exposure. One needs frst to defne approprate measures of earnngs management to test ths earnngs management hypothess. In the lterature of earnngs management, the modfed Jones (1991) model s 8

10 wdely appled to measure the magntude of earnngs management when earnngs management s examned n relaton to partcular corporate events 8. Snce foregn currency actvtes are long-term ongong events for a frm, the modfed-jones model s apparently napproprate and dffcult to mplement. Alternatvely, Leuz, Nanda, and Wysock (2003) proposed four frm-level measures of earnngs management, whch capture dfferent scopes of frm earnngs management, ncludng that managers can take actons to manpulate earnngs and that managers can avod reportng frms actual actvtes. In ths study we modfy ther measures to assess the magntude of earnngs management at frm level. The followng descrbes the four earnngs management measures used n our emprcal analyss. The frst and second measures, EM1 and EM2, evaluate a frm s degree of earnngs management based on the smoothng of ts reported earnngs. A thrd measure, EM3, assesses a frm s level of dscretonary accruals. Leuz et al. use the rato of absolute value of accruals deflated by cash flows from operatons to estmate the relatve level of manpulated accruals. Ths rato however does fully reflect the actual magntude of dscretonary accruals. Ths paper thus calculates the dscretonary accruals generated from the modfy Jones model to replace the EM3 beng used n the study of Leuz et al.. EM1 Degree of ncome smoothng: the relatve volatlty of operatng earnngs and operatng cash flow When exchange rate fluctuatons result n losses n frm operatng ncome, managers may be motvated to ncrease reported earnngs through accrual adjustments. It follows that frm exchange rate exposure may be reduced through earnngs smoothng. To modfy the EM1 from Leuz, Nanda, and Wysock (2003), whch used the measure to evaluate the country-wde earnngs management, we calculate for each frm a rato of the standard devaton of quarterly operatng earnngs over the standard devaton of quarterly operatng cash flow. These two standard devatons are estmated from a three-year wndow 8 The modfed Jones model has been wdely appled for testng those events ncludng IPO, SEO, LBO, Reverse LBO among others. Those studes usually assume that pror to an event a frm has a normal magntude of accruals. Based on the modfed Jones model, the degree of earnngs management s proxed by a frm s dscretonary accruals, whch s defned as the dfference between a frm s actual accruals and ts expected accruals. In the applcaton, the expected accruals of a frm are estmated by the accruals of a set of matched frms. 9

11 durng whch a frm s exchange rate exposure s estmated. To control for the effect of frm sze, both operatng earnngs and cash flows from operatons are scaled by lagged total assets. A lower EM1 rato ndcates a greater accountng dscreton exercsed by managers and thus a hgher degree of earnngs management. EM1 s calculated as follows: IB σ t TA, t 1 EM1 = OANCF σ t TA, t 1, (2) where IB s the quarterly earnngs before extraordnary tems and dscontnued operatons, OANCF s the quarterly operatng cash flow, and TA s quarterly total assets. EM2 Degree of ncome smoothng: the correlaton between changes n accruals and changes n operatng cash flows Managers may use accruals to dmnsh economc shocks to the frm s cash flows from operatons. Snce cash flow shocks may be adjusted by accountng accruals, there s a negatve correlaton between changes n accruals and cash flows from operatons. Our second measure (EM2) s the contemporaneous correlaton between changes n accruals and changes n cash flows from operatons. A larger EM2 value ndcates a hgher correlaton n absolute value, whch suggests that managers have exercsed greater accountng dscreton to smooth earnngs. The expected relaton between EM2 and exchange rate exposure s not as clear as EM1 snce there has been no consensus on the nformaton contents about earnngs qualty conveyed by the correlaton between accruals and cash flows. Dechow and Dchev (2002) suggest that the negatve correlaton between accruals and cash flows s assocated wth earnngs qualty. However, there s no concluson as to whether EM2 s postvely or negatvely related to earnngs qualty [see, e.g. Dechow and Dchev (2002), Leuz, Nanda, and Wysock (2003) and Wysock (2005)]. If EM2 s postvely assocated wth earnngs qualty, we would expect ts correspondng estmated exchange rate exposure should be hgh and thus result n a negatve 10

12 relaton between exchange rate exposure and EM2. In contrast, f the absolute value of EM2 can be nterpreted as an ndcator of earnngs management, t would be expected that a frm wth more the absolute value of EM2 wll take lower exchange rate exposure. That s, t results n a postve relaton between the absolute value of EM2 and exchange rate exposure. In lght of ths, our emprcal evdence thus has mplcatons for the nformaton contents about earnngs qualty shed from the correlaton between accruals changes and cash flow changes. Quarterly data of total accruals and operatng cash flows are used to calculate EM2 of ndvdual frm for the 3-year wndow durng whch a frm s exchange rate exposure s estmated. Both changes n accruals and changes n cash flows are scaled by lagged total assets. EM3 - Average magntude of dscretonary accruals In addton to earnngs smoothng, managers may use dscretonal accruals to manpulate the frm performance. Manageral dscreton n the accruals system provdes managers wth opportuntes to manpulate earnngs. Through accrual adjustments, managers may ncrease current reported earnngs whle have the dscrepances recovered from future reported earnngs. For the purpose of analyss, total accruals (TAC) can be decomposed nto nondscretonary accruals (NDAC) that are correlated wth frm performance, and dscretonary accruals that may cause earnngs to be systematcally managed. Our thrd measure of earnngs management (EM3) s to measure the degree of earnngs management based on the magntude of dscretonary accruals (DAC). Here, dscretonary accruals are defned as the dfference between total accruals and nondscretonary accruals 9. We use the modfed cross-sectonal Jones (1991) model to estmate those nondscretonary accruals that are dctated by frm condtons and ndependent of manageral manpulaton. 10 An ordnary least squares regresson of current accruals for a gven year s regressed on the change n sales and the level of property, plant and equpment (PP&E) for that year usng all matchng 9 Nondscretonary accruals present the normal level of total accruals that s assocated wth sales and nvestments of ed assets. 10 Ths cross-sectonal model has been wdely appled to estmate expected accruals, ncludng Teoh et al. (1998a, 1998b), Dechow, Sloan, and Sweeney (1995), and Rchardson (2000). 11

13 frms n the same two-dgt SIC code (?). Ths ntra-ndustry cross-sectonal regresson s estmated for each year over the sample perod from 1992 to All varables, ncludng the ntercept term, n the cross-sectonal regresson are deflated by the total assets of pror year to reduce the heteroskedastctes. The regresson s as follows: TAC TA jt j, t 1 1 REVjt PP & E jt = α 0( ) + α1( ) + α2( ) + ε TA TA TA j, t 1 j, t 1 j, t 1 jt (3) where TA j,t-1 s the total assets n year t 1, REV j,t s the change n sales n year t for frm j, TAC j,t s total accruals of frm j and s the dfference between net ncome and cash flows from operaton, and PP&E j,t s the gross property, plant, and equpment of frm j n year t. The nondscretonary total accruals scaled by assets (NDAC) and the dscretonary total accruals scaled by assets (DAC) are computed as: NDAC t 1 REV & ˆ ˆ t ARt PP E = 0( ) + 1( ) + ˆ t α α α2( ) (4) TA TA TA, t 1, t 1, t 1 DAC t TAC ( NDAC (4) t = ) TAt t where AR t s the change n account recevables for sample frm, ˆ α0, ˆ α1, and ˆ α 2 of equaton (4) denote estmates of α 0, α1, and α 2 from equaton (3). The DAC s our thrd choce of earnngs management measure - EM3. EM3 s expected to be postvely related to a frm s estmated exchange rate exposure f managers frequently manpulate ther frm s reported earnngs through adjusted accruals. EM4 Degree of small loss averson Hayn (1995) demonstrates that the volume of frms reportng small profts largely domnate those 12

14 frms reportng small losses. Degeoreg et al (1999) and Burgstahler and Dchev (1997) suggest that the overwhelmngly majorty of small-profts reportng can be attrbuted to earnngs management. Burgstahler and Dchev (1997) and Leuz, Nanda, and Wysock (2003) use the rato of frms reportng small profts to frms wth small reported losses to measure the degree of earnngs management on the country level. However, Dechow, Rchardson and Tuna (2003) make a cauton of ths as they cannot fnd evdence that frms utlze dscretonary accruals to prevent small-loss reportng (check). In order to adapt ths measure to our frm level assessment of earnngs management, ths study uses the rato of the numbers of quarters wth small loss reportng to the total number of quarters wth non-mssng profts and losses data as a measure of a frm s tendency to avod reportng small losses. To calculate EM4, we requre the sample frm to have at least ten quarters wth non-mssng profts or losses data wthn the sample wndow correspondng to the estmaton of exchange rate exposure. Ths measure s desgned based on the followng two assumptons: frst, frms are unwllng to recognze losses n ther accountng report. Second, a small loss s much easer to be altered than a large loss. Based on the two assumptons, those frms that do not avod recognzng small losses n ther accountng report are expected to engage less n earnngs management. A hgher score of EM4 thus mples a lower level of earnngs management and expects to be assocated wth a hgher exchange rate exposure. Small losses are defned to be n the range [ 0.01, 0.00) and small profts are defned to be n the range [0.00, 0.01]. In order to compute a relable rato, we requre at least ten quarterly observatons of losses and profts for a frm. 2.3 Sample selecton and descrptve statstcs Our sample covers those non-fnancal frms n COMPUSTAT. To be ncluded n the sample, frms must have average total assets greater than 500 mllon dollars over the sample perod and have postve foregn sales. The total sample perod extends from 1992 to It s dvded nto nne overlapped sub-perods each coverng three years. Only those frms wth non-mssng data of prce seres, total assets, 13

15 total sales, and market value over the total sample perod are ncluded n the fnal sample. Prevous studes have documented sgnfcant evdence of earnngs management at the tme of IPOs [see Teoh, Welch and Wong (1998a) and Teoh,Wang, and Rao (1998)]. Our sample excludes those frms that went publc after January 1, 1991 and only those wth non-mssng data on prce seres are ncluded n the fnal sample. Ths crteron mtgates the potental bas due to the earnngs management actvtes before IPOs. Each sample frm s exchange rate exposure s estmated by equaton (1) usng monthly return data, whch are collected from the CRSP database. We use the CRSP value-weghted ndex return of the NYSE, AMEX, and NASDAQ composte ndexes to proxy for the market return. The US dollar trade-weghted exchange rate by Bank of England s retreved from the Datastream database. Followng earler studes, our sample ncludes those frms wth non-zero foregn sales, a prelmnary ndcator for a frm s foregn nvolvement. Usng a sample wth a profle smlar to prevous studes allows us to compare the emprcal results and test our hypothess. Descrptve statstcs Table 1 reports the descrptve statstcs of the exchange-rate exposures estmated from equaton (1) and the frm factors consdered for determnng frm exchange rate exposures. The absolute value of the sum of β and β represents a frm s exchange rate exposure. Panel A shows that the mean and lagged medan absolute value of combned exposure, β + β 14 lagged, for all frms s equal to and As shown n Panel B and Panel C, the mean combned exposure for frms wth postve combned exposure (net mporters) and for frms wth negatve exposure (net exporters) s and 1.119, respectvely. Note that the foregn sales rato, export sales rato and frm sze are roughly the same between postve and negatve exposure frms. Table 1 also reports the summary statstcs for our four earnngs management measures. The two ncome smoothng measures (EM1 and EM2) of postve exposure frms are slghtly greater than those of negatve exposure frms. Snce hgher scores of EM1 and EM2 mply lower degree of

16 earnngs management, ths result ndcates that net mporters on average engage less n earnngs smoothng than net exporters. The other two earnngs management measures (EM3 and EM4) do not materally dffer between postve exposure frms and negatve exposure frms. [ Insert Table 1 Here ] Earnngs management and exchange rate exposure: a prelmnary analyss Pror researches dd not fnd a prevalng evdence of a sgnfcant relatonshp between exchange rate fluctuatons and contemporaneous stock returns even when the samples were lmted to frms wth certan level of foregn nvolvement. Ths study tests the hypothess that the observed weak lnkage between exchange rate changes and stock returns s at least partly attrbutable to the earnngs management. [ Insert Table 2 Here ] Table 2 presents the prelmnary analyss, wthout controllng for other factors, of the relatonshp between earnngs management and exchange rate exposure. We partton the sample frms nto quartles (Q1, Q2, Q3 and Q4) based on the scores of each earnngs management measure. Note that for the two ncome smoothng measures (EM1 and EM2), hgher scores mply less earnngs management. On the contrary, for the magntude of dscretonary accruals measure (EM3), hgher scores mply a more aggressve earnngs management. We put those frms wth the hghest level of earnngs management n Q4 and those wth the lowest level of earnngs management n Q1. For the small-loss averson measure (EM4), the sample frms are parttoned nto two groups, namely Zero and Non-Zero. The Zero group contans those frms never reportng small quarterly losses durng the 3-year wndow and the Non-zero group conssts of those frms reportng small losses n more than one quarter durng the same perod. Frms n the Zero group are regarded as conductng an aggressve earnngs management and those n the Non-Zero group are regarded as practcng a conservatve earnngs management. Based on our hypothess that earnngs management reduces frm exchange rate exposure, we expect those frms n Q1 and Non-zero 15

17 group to have greater exchange rate exposure whle those frms n Q4 and Zero group to have lower exchange rate exposure. Panel A of Table 2 reports the dstrbuton of exchange rate exposure across earnngs management quartles for the entre sample frms. All earnngs management measures, wth the excepton of EM3, exhbt a relatonshp wth exchange rate exposure consstent wth our hypothess. Takng the example of our frst ncome smoothng measure, EM1, the mean exchange rate exposure s equal to 1.380, 1.050, and for those frms n the group of Q1 (most conservatve), Q2, Q3, and Q4 (most aggressve), respectvely. Smlar patterns exst when β and β lagged are consdered separately. The evdence from ncome smoothng measures s consstent wth our hypothess that a more aggressve earnngs management wll decrease a frm s observed exchange rate exposure. The results of the small-losses averson measure (EM4) also support our earnngs management hypothess. The mean exchange rate exposure, β + β lagged, s equal to and for Non-zero (conservatve) group and Zero (aggressve) group. Unlke other measures, the statstcs of EM3, the most commonly used ndcator of earnngs management, exhbts dfferent pattern and are contrary to our hypothess. The mean exchange rate lagged exposure, β + β, for Q1 (most conservatve), Q2, Q3, and Q4 (most aggressve) s equal to 0.828, 0.984, and Ths ndcates that frms wth hgher magntude of dscretonary accruals tend to have hgher observed exchange rate exposure. Same tendency s found when β and β lagged consdered separately. The paradoxcal results thus call for the testng the effect of earnngs management on frm exchange rate exposure when other related varables are accounted for. are 3. Exchange-rate exposure and earnngs management: a regresson model The prelmnary analyss n prevous secton reveals that there exsts lnkage between earnngs management and estmated exchange rate exposure, although varous earnngs management measures proxy 16

18 for dfferent earnngs management behavors exhbt nconsstent patterns. In ths secton, we examne the mpact of earnngs management on a frm s estmated exchange rate exposure whle controllng for other varables related to frm exchange rate exposure. We hypothesze that earnngs management reduces a frm s estmated exchange rate exposure f earnngs management masks a frm s real performance. 3.1 Exstng Hypotheses/Theores on Frm Factors Determnng Exchange Rate Exposure Due to the falure n uncoverng a sgnfcant relaton between exchange rate changes and stock returns, many studes had focused on the determnants of exchange-rate exposure [e.g. Joron (1991), Amhud (1994), Bartov and Bodnar (1994), He and Ng (1998), Allayanns and Ofek (2001), Marston (2001) and Dodge et al. (2002)]. Buldng on exstng studes, two groups of factors are suggested and found as mportant determnants affectng a frm s exchange-rate exposure: the level of foregn nvolvement and hedgng actvtes. The level of foregn nvolvement It s a consensus that the level of a frm s foregn nvolvement or nternatonal actvtes s an mportant determnant of frm exchange rate exposure. Many studes have documented that frms wth greater foregn nvolvement are exposed to hgher exchange rate rsk [e.g. Joron (1991), He and Ng (1998), Allayanns and Ofek (2001) and Dodge et al. (2002)]. Consderng the data avalablty, ths study uses frm sze (SIZE), the export rato (EXPORT), and the foregn sales rato (FSALE) as control varables representng a frm s foregn nvolvement. Theory of optmal hedgng Dumas (1978) s one of the earlest researchers suggestng that the actual mpact of exchange rate rsk on frm value s not only determned by a frm s nternatonal actvtes but also by ts engagement of 17

19 hedgng, both fnancal and operatng hedgng actvtes. Indeed, a company may utlze fnancal nstruments as well as operatng dversfcaton to mtgate the cash flow fluctuatons due to exchange rate changes. In an mperfectly compettve market, stock prces should reflect a frm s expected hedgng actvtes. A negatve relaton expects to be observed between the exchange rate exposure and the hedgng actvtes. Exstng studes have substantated the hypothess and fnd that U.S. corporatons use fnancal nstruments to mtgate ther exposure to exchange rate fluctuatons. [see Rawls and Smthson (1989), Dolde (1993), Allayanns and Ofek (2001), and Allayanns and Weston (2001)]. Pror lterature assocated wth theores of optmal hedgng has suggested several hypotheses to explan why a frm engages n hedgng. Frst, hedgng can reduce the transacton costs of fnancal dstress and then mtgate the probablty of bankruptcy. Frms wth greater probablty of or hgher transacton cost from fnancal dstress are thus more motvated to undertake hedgng actvtes. [see Smth and Warner (1979), Smth and Stulz (1985), Nance et al. (1993), and Geczy et al. (1997)]. Ths study ncludes fnancal leverage, ROA and dvdend yeld to control a frm s hedgng ncentves for mtgatng ts costs or probablty of fnancal dstress. Frms wth hgher fnancal leverage, lower ROA 11 and hgher dvdend yeld are classfed as beng suffered hgher fnancal dstress cost or probablty and are expected to undertake more hedgng actvtes to reduce exchange rate exposure. Second, Froot et al (1993) extends Myers (1977) agency cost hypothess and suggests that the undernvestment problem can be mtgated by hedgng as hedgng actvtes reduce the cash flow volatlty and thus make external fnancng less costly. Ths study employs the followng varables to control for a frm s ncentves to reduce ts undernvestment problems: the market-to-book rato (MB), R&D expendture as a rato of sales (R&D) 12 and dvdend yeld (dv). Frms wth hgher market-to-book rato, hgher R&D expendture rato and lower dvdend yeld are expected to engage n more hedgng actvtes and thus are assocated wth lower exchange rate exposure. 11 See Nance et al. (1993) and He and Ng (1998). 12 The R&D expendture s mostly mssng for our sample frms. We also use captal expense rato to proxy for frm growth opportunty, and the results generally reman unchanged. 18

20 Fnally, the hedgng costs also affect a frm s decson to use hedgng nstruments. Due to the large start-up costs, a frm s hedgng ncentves are subject to economes of scale. Sze, already ncluded as a proxy for a frm s foregn nvolvement, s also consdered as a control varable for a frm s relatve hedgng costs. Under ths consderaton, frms wth larger sze are assocated wth lower hedgng costs and are more motvated to engage n hedgng. Busness hedgng Ths study also recognzes that the hypotheses of hedgng ncentves and actvtes derved from optmal hedgng theores also apply to operatng hedgng. A frm may typcally mtgate ts exchange rate exposure by geographc or busness dversfcaton. COMPUSTAT provdes two knds of dversfcaton data, the number of geographc segments and busness segments n whch a frm s engaged. We focus on busness segments snce the number of geographc segments tends to be sgnfcantly correlated wth foregn sales rato and frms wth more geographc segments tend to have hgher exchange rate exposure. We would expect that frms engagng n greater number of busness segment can dversfy ther currency rsk and thus have lower exchange rate exposure. When a frm s engaged n multple busness segments, the dummy varable, Dseg, s set to 1 and 0 otherwse. 13 Ths varable proxes for the dversfcaton of a frm s cash flow sources, and thus the degree of busness hedgng. It s expected that Dseg wll be negatvely correlated wth frm currency rsk exposure The Panel Regresson A two-stage framework smlar to He and Hg (1998) and Allayanns and Ofek (2001) s employed to further document the mpact of earnngs management on a frm s exchange rate exposure whle controllng other factors related to exchange rate exposure. Dfferng from He and Hg (1998) and Allayanns and Ofek 13 Lang and Stulz (1994), Berger and Ofek (1995), and Allayanns and Weston (2001) have appled a smlar measure to serve as a proxy for a frm s ndustral dversfcaton, and examne ts relatonshp wth frm value. Allayanns and Weston (2001) fnd that ndustral dversfcaton ncreases frm value, whch s consstent wth our hypothess. 19

21 (2001), we apply the approach repeatedly over a longer sample span, nstead of a sngle ed perod. That s, for each year t when the control varables proxy for a frm s foregn nvolvement and hedgng actvtes are recorded, we estmate each frm s exchange rate exposure and the four earnngs management measures from the three years wndow (year t-1 to year t+1) surroundng year t. In the frst stage, Eq. (2) s used to estmate ndvdual frm s exchange rate exposure for each three-year wndow. Earnngs management measures are also calculated for the correspondng wndow over whch exchange rate exposure s estmated. Snce our sample perod covers 11 years, we construct a total of 9 partal overlappng three-year wndows, , , , , , , , and respectvely. In the second stage, we regress a frm s exchange rate exposure estmated from frst stage on the earnngs management measures and the control varables wth a panel regresson model as follows: ˆ β + γ J ˆ lagged, j + β, j = j = 1 6, j MB, j + γ 7, j RD, j D γ j + γ 0, j 8, j + γ 1, j SIZE EXPORT, j, j + γ + γ 9 j 2 j Dseg Fsale, j, j + γ + γ 3, j 10 j DAT EM, j, j + γ + ε, j 4, j ROA, j + γ 5, j DIV, j, (5) where ˆ β s the measure of combned exchange rate exposure as estmated from equaton (1) lagged j + ˆ, β, j for frm durng wndow j. Ths method takes advantage of tme-varyng estmates of exchange rate exposure and earnngs management measures generated from our movng wndow calculatons, and yelds greater testng power due to the expanded observatons. A Feasble GLS s mplemented to account for possble autocorrelatons n the resduals Emprcal results Table 3 presents the regresson results of equaton (3), whch tests the relaton between earnngs management and estmated exchange rate exposure whle controllng for varables related to a frm s 20

22 foregn nvolvement and hedgng ncentves. In the frst column of Table 3, the ndependent varables are ordered as factors assocated wth a frm s foregn nvolvement, hedgng ncentves, and earnngs management. We hypothesze that the market or nvestors cannot accurately evaluate a frm s exchange rate exposure f the frm engages n earnngs management to mask ts true performance subject to exchange rate changes. The estmated exchange rate exposure s thus expected to be negatvely related to earnngs management level. Our analyses focus on the sgn and sgnfcance of the four earnngs management measures. The second column n Table 3 lsts the expected sgn of the four earnngs management measures). Panel A of Table 3 lsts the results for all frms. Consstent wth our expectaton, the two ncome smoothng measures, EM1 and EM2, are sgnfcantly and postvely related to estmated combned exchange rate exposure no matter other earnngs management measure(s) s/are ncluded or not. Snce lower scores of EM1 and EM2 mply hgher degree of earnngs management and are hypotheszed to be assocated wth lower exchange rate exposure. The postve relaton between ncome smoothng measures and exchange rate exposure thus ndcate that the earnngs management actvtes for ncome smoothng purposes do reduce exchange rate exposure beng estmated by stock market. The results of small-loss avodance measure, EM4, also sgnfcantly support our hypothess. We fnd that EM4 s also postvely related to exchange rate exposure n all regresson combnatons. On the other hand, the sgnfcance of the magntude of dscretonary accruals measure, EM3, s much lower than other earnngs management measures. Ths s consstent wth our expectaton. However, the sgn of EM3 are postve n all regresson combnatons, ths s stll a confusng results ndcatng that the hgher magntude of dscretonary accrual s assocated wth the hgher estmated exchange rate exposure. [ Insert Table 3 Here ] Panel B of Table 3 shows the results for postve exposure frms (net mporters). Consstent wth the hypothess, frms wth hgher EM1, EM2 and EM4 are postvely related estmated exchange rate exposure. On the contrary, the results of EM3 stll do not support our hypothess. Panel C of Table 3 shows the results 21

23 for negatve exposure frms (net exporters). The results are smlar to all frms and postve exposure frms. Other varables also exhbt sgnfcantly explanatory abltes, e.g., Sze, dvdend yeld, R&D and fnancal leverage. We also fnd some evdence n Table 3 whle comparng postve and negatve exposure frms. Among the foregn nvolvement factors, both export rato and foregn sales rato are postvely and sgnfcantly related to exchange rate exposure for net exporters (frms wth negatve exposure). Ths s consstent wth our expectaton that hgher foregn nvolvement wll ncrease frm exchange rate exposure. On the contrary, the two foregn nvolvement varables are nsgnfcantly related to exchange rate exposure for net mporters (frms wth postve exposure). Snce the foregn revenues n excess of foregn costs wll offset the benefts (losses) form the U.S. dollar apprecaton (deprecaton) and thus mtgate a net mporter s exchange rate exposure, foregn sales s expected to be negatvely or at least nsgnfcantly related to a net mporter s exchange rate exposure. The evdence thus confrms our methodology to partton sample frms nto net mporters or net exporters. Ths s mportant snce some conjectures below depend on the correct partton of net mporters and net exporters. 4. Summary It has long been puzzled that studes do not offer a prevalng evdence of sgnfcant relatonshp between exchange rate rsk and stock returns of U.S. corporatons. Ths paper proposes that earnngs management may be an mportant determnant of frm exchange rate exposure n that the market partcpants cannot correctly evaluate the mpact of exchange rate changes on frm value when a frm s engaged n earnngs manpulaton. We thus hypothesze that frms employng greater levels of earnngs management wll reduce more of ther exchange rate exposure assessed by the stock market. Ths study estmates a frm s exchange rate exposure usng the Fama-French three-factor model. The sum of current and lagged exposure coeffcents s used to measure a frm s exchange rate exposure. We 22

24 calculate four earnngs management measures at frm level by followng the study of Leuz, Nanda, and Wysock (2003). Consstent wth our hypothess, frms engaged wth more aggressve earnngs management actvtes, especally those amng at ncome smoothng, tend to have sgnfcantly lower exchange rate exposure. A further analyss fnds that the effect of earnngs management on frm exchange rate exposure becomes strengthened when the currency movements are adverse to frm earnngs. Ths evdence supports the opportunstc earnngs management hypothess that managers only smooth ther reported earnngs durng adverse perods whle enjoy the profts durng favorable perods. Our results also have mplcatons for the nformaton contents about earnngs qualty shed from the correlaton between accruals changes and cash flow changes. Our second ncome smoothng proxy, measured as the correlaton between accruals changes and cash flow changes, s found sgnfcantly postvely assocated wth exchange rate exposure. Ths ndcates that to a certan extent, greater correlaton between changes of accruals and changes of cash flow mples earnngs management. 23

25 References Adler, M., and B. Dumas (1984), Exposure to Currency Rsk: Defnton and Measurement, Fnancal Management, Vol. 13, pp Allayanns, G. (1997), The tme varaton of exchange rate exposure: An ndustry analyss, workng paper, Unversty of Vrgna. Allayanns, G. and E. Ofek (2001), Exchange-Rate Exposure, Hedgng, and the Use of Foregn Currency Dervatves, Journal of Internatonal Money and Fnance, Vol. 20, pp Allayanns, G. and J. Weston (2001), The Use of Foregn Currency Dervatves and Frm Market Value, Revew of Fnancal Studes, Vol. 14, pp Amhud, Y. (1994), Exchange Rates and the Valuaton of Equty Shares, n Yakov Amhud and Rchard M. Levch, eds: Exchange Rates and Corporate Performance, (New York, Irwn). Bartov, E. and G.M. Bodnar (1994), Frm Valuaton, Earnngs Expectatons, and the Exchange-rate Exposure Effect, Journal of Fnance, Vol. 49, pp Berger, P. and E. Ofek (1999), Causes and Effects of Corporate Refocusng Programs, Revew of Fnancal Studes, Vol. 12, pp Bodnar, G.M. and M. Wong (2003), Estmatng Exchange Rate Exposures: Issues n Model Structure, Fnancal Management, Vol. 32, pp Bodnar, G.M. and W.M. Gentry (1993), Exchange Rate Exposure and Industry Characterstcs: Evdence from Canada, Japan, and the USA, Journal of Internatonal Money and Fnance, Vol. 12, pp Burgstahler, D. C. and I. D. Dchev (1997), Earnngs Management to Avod Earnngs Decreases and Losses, Journal of Accountng and Economcs, Vol.24, pp Chow, E.H., W.Y. Lee, and M.E. Solt (1997a), The Exchange-rate Rsk Exposure of Asset Returns, Journal of Busness, Vol. 70, pp Chow, E.H., W.Y. Lee, and M.E. Solt (1997b), The Economc Exposure of U.S. Multnatonal Frms, Journal of Fnancal Research, Vol. 20, pp

26 Dechow, P. (1994), Accountng earnngs and cash flows as measures of frm performance: The role of accountng accruals, Journal of Accountng and Economcs, Vol.18, pp Dechow, P. M., R. G. Sloan and A. P. Sweeney (1995), Detectng Earnngs Management, The Accountng Revew, Vol.70, pp Dechow, P. and Dchev, I. (2002), The qualty of accruals and earnngs: The role of accrual estmaton errors, The Accountng Revew, Vol.77 (Supplement), pp Dechow, P. M., S. A. RICHARDSON and I. TUNA (2003), Why Are Earnngs Knky? An Examnaton of the Earnngs Management Explanaton, Revew of Accountng Studes, Vol.8, pp Degeorge, F., J. Patel and R. Zeckhauser (1999), Earnngs Management to Exceed Thresholds, The Journal of Busness, Vol.72, pp D Ioro, A. and R. Faff (2000), An Analyss of Asymmetry n Foregn Currency Exposure of the Australan Equtes Market, Journal of Multnatonal Fnancal Management, Vol. 10, pp Dolde, W. (1993), The Trajectory of Corporate Fnancal Rsk Management, Journal of Appled Corporate Fnance, Vol. 6, pp Dodge, C.A., J. Grffn, and R. Wllamson (2002), Does Exchange Rate Exposure Matter?, SSRN Workng Paper. Dumas, B. (1978), The Theory of the Tradng Frm Revsted, Journal of Fnance, Vol. 33, pp Fama, E.F. and K.R. French (1989), Busness Condtons and Expected Returns on Stocks and Bonds, Journal of Fnancal Economcs, Vol. 25, pp Froot, K., D. Scharfsten, and J. Sten (1993), Rsk Management: Coordnatng Corporate Investment and Fnancng Polces, Journal of Fnance, Vol. 48, pp Geczy, C., B.A. Mnton, and C. Schrand (1997), Why Frms Use Currency Dervatves?, Journal of Fnance, Vol.52, pp Grffn, J. M. and R.M. Stulz (2001), Internatonal Competton and Exchange Rate Shocks: A Cross-Country Industry Analyss of Stock Returns, Revew of Fnancal Studes, Vol. 14, pp. 25