The Exchange Rate and the Performance of Japanese Firms: A Preliminary Analysis Using Firm-level Panel Data

Transcription

1 Prelmnary The Echange Rate and the Performance of Japanese Frms: A Prelmnary Analyss Usng Frm-level Panel Data Takash Hanagak and Masahro Hor (Economc and Socal Research Insttute, Cabnet Offce, JAPAN) July 31, 2015 Abstract Usng a large frm-level panel dataset coverng about 360,000 Japanese frms from the md-1990s to FY2013, ths paper presents the results of a fact fndng study eamnng the effects of echange rate changes on the performance of Japanese frms, especally n recent years. Whle our results are very prelmnary, we frst confrm that yen deprecaton has a postve effect on the performance of Japanese frms (through the sales channel) n general, and, as epected, t s eport-orented large and medum-szed frms that beneft the most. Though our fndngs on ther own cannot substantate the vew that the yen deprecaton caused the strong performance of frms, the frm performance n 2013 s grossly n lne wth or an etenson of the estmated relatonshp between the yen rate and the performance of Japanese frms n the past 20 years. Moreover, we fnd that, for some reason, the postve effect of yen deprecaton on the eportng frms was smaller n 2013 than n prevous years, but our prelmnary analyss dd not produce evdence supportng the hypothess that eportng frms overseas transfer of producton s reducng ther gans from yen deprecaton. JEL classfcaton: F31, F62, L25 Keywords: Sales growth, ROA, Echange rate, Globalzaton Ths paper s prepared for our presentaton at the ESRI Internatonal Conference ttled To ensure Japan s economc growth n Tokyo on July 31st, Specal thanks go to the Mnstry of Economy, Trade and Industry for provdng us wth the mcrodata from the Basc Survey of Japanese Busness Structure and Actvtes. Please address correspondence to: Masahro Hor. 1

2 1. Introducton In order to pull the Japanese economy out of prolonged deflaton and stagnaton, the Prme Mnster, Shnzo Abe, ntroduced a set of polces n December 2012 summarly referred to as Abenomcs. Abenomcs conssts of three arrows: unconventonal monetary polcy (the frst), epansonary fscal polcy (the second), and economc growth strateges to encourage prvate nvestment (the thrd arrow). The frst two of the three arrows were mplemented quckly, whle the thrd arrow (structural reforms) s n the process of mplementaton and epected to take tme before effects appear. The stock and foregn echange markets reacted very favorably to the new polcy (see Fgure 1(a) and 1(b)). 1 The real economy (Japan s macro fundamentals) s also showng some sgns of mprovement. Whle markets tend to react favorably when they epect a recovery n terms of macro fundamentals, the tmng of events snce the ntroducton of Abenomcs appears to suggest that the mprovements n Japan s macro fundamentals lkely are the result of the market response rather than the other way around. The purpose of ths paper s to nvestgate the effects of echange rate changes (one of the two market reactons mentoned above) on the performance of Japanese frms. In a zero nterest rate envronment, n whch tradtonal ways to drectly nfluence the course of the economy through monetary polcy are not necessarly avalable, currency deprecaton turns out to be the most tangble channel through whch to stmulate prvate busness actvty. 2 Despte the hstorcal fact that developments n Japan s economy are closely lnked wth trends n the echange rate of the yen (see Fgure 1(c)), not everyone welcomes a weaker yen, wth some argung that t makes t dffcult for smaller frms manly servng the domestc market to pay for more epensve foregn goods. Some argue that even eportng frms are not benefttng much, 1 Fukuda (2015) nvestgates why the two markets reacted to Abenomcs so favorably, and reports that the dramatc market responses under Abenomcs happened only n tme zones when foregn nvestors were actve. 2 Shoj (2015) argues that the currency deprecaton snce December 2012 could turn out to be useful for endng Japan s long battle wth fallng prces. 2

3 snce many of them have shfted substantal parts of ther producton overseas. Although only tme wll tell, many economsts epect that the volume of eports by Japanese frms wll not ncrease very much, snce they have not cut ther eport prces n dollar terms (rase them n yen terms) despte the fallng value of yen (see Fgure 2). 3 Aganst ths background, ths paper presents the results of a mcrodata based fact fndng study eamnng the effects of echange rate changes on the performance of frms n Japan. Whle there s a substantal lterature on the mplcatons of echange rate swngs for the real economy n other countres, 4 and despte ts mportant polcy mplcatons, there s lttle frm-level evdence on the effect echange rate changes on the performance of Japanese frms, ecept for a study on Japanese frms nvestment by Hote (2012). 5 Basng on a large frm-level panel dataset of 359,000 non-fnancal frms (2.2 mllon total observatons) n Japan from FY1994 to FY2013, ths paper nvestgate the relatonshp between the real effectve echange rate and Japanese frms performance. There are at least a few reasons to thnk that a frm-level study s approprate for studyng the relatonshp between echange rate and frms performance. Echange rate changes could affect frm performance through a number of dfferent channels, such as the prce of eports relatve to foregn compettors, the cost of mported nputs relatve to other factors of producton, or the cost of nternal/eternal borrowng, etc. Snce these channels are epected to work dfferently on frms wth dfferent characterstcs, a frm-level data based analyss, enrched wth detaled nformaton about frms characterstcs, permts a better understandng of the transmsson channels. A frm-level panel based analyss also allow us to control for unobservable ndvdual effects, whch are lkely to be 3 Basng on the auto-regressve dstrbuted lag model as well as tme-varyng parameter estmaton of echange rate pass-through n Japanese eports, Shmzu and Sato (2015) argues that the slow recovery of Japanese trade balance n response to the yen deprecaton can be eplaned by Japanese frms prcng behavor as well as the actve overseas operaton caused by the unprecedented level of yen apprecaton before Abenomcs. 4 See, for eample, Baggs et al. (2009) for a study on Canadan frms, Nucc and Pozzolo (2001, 2010) for studes on Italan frms, Fung and Lu (2009) for a study on Tawanese frms, and Dhasmana (2013) for a study on Indan frms. 5 If we wden our scope to studes that used aggregated or sectoral data, there appears to be more studes on Japan avalable. See Matsubayash (2011) as an eample. 3

4 correlated wth frm s response to echange rate fluctuatons. Whle the fndngs of ths paper are very prelmnary, constraned by the short observaton perod after the start of Abenomcs, we could confrm that yen deprecaton generally has a postve effect on the performance of Japanese frms through sales channel, and that t s large eport-orented frms that beneft the most. Favorable performance of Japanese frms n 2013 was grossly n lne wth or an etenson of the estmated reduced form relatonshp between the yen rate and the performance of frms n Japan n the past 20 years. Moreover, we fnd that, for some reason, the postve effect of yen deprecaton on the eportng frms was smaller n 2013, but our prelmnary analyss dd not produce evdence supportng the hypothess that eportng frms overseas transfer of producton s reducng ther gans from yen deprecaton n recent years. The remander of the paper s organzed as follows. Secton 2 descrbes the frm-level dataset used n ths study and ts two data sources,.e. the Basc Survey of Japanese Busness Structure and Actvtes (BSBSA) and the Orbs database. Net, Secton 3 brefly eplans our emprcal methodology to eamne the effects of echange rate changes on the performance of frms n Japan. Secton 4 then presents the results, whle Secton 5 summarzes the fndngs and lsts up ssues left for the future research. 2. Datasets and summary statstcs To eamne the effects of echange rate changes on the performance ndcators (sales growth & ROA) of Japanese frms from the md-1990s to FY2013, we used the followng two frm-level panel datasets: Basc Survey of Japanese Busness Structure and Actvtes (BSBSA) The Mnstry of Economy, Trade and Industry (METI) conducts ths survey once a year to acqure a 4

5 quanttatve understandng of the actual condtons and actvtes of Japanese enterprses, and to obtan basc data for the mplementaton of ndustral polces. The survey each year covers about 30,000 enterprses wth 50 or more employees and pad-n captal of over 30 mllon yen. We were able to obtan the BSBSA mcrodata from 1994 to Orbs database Orbs s a product of Bureau van Djk Electronc Publshng, a world leadng frm of prvate company nformaton, and contans nformaton on over 160 mllon companes worldwde. We were able to obtan the nformaton on Japanese frms n Orbs (roughly about 200,000 frms every year) from 2004 to March Whle Orbs has the advantage that t provdes a larger sample and more recent coverage, t goes back only to the md-2000s and does not cover detaled nformaton on frms global actvtes such as ther nternatonal trade, overseas operatons, etc. To take advantage of the BSBSA s long coverage (from the md-1990s) and the rch nformaton on frms global actvtes, we bascally use the BSBSA mcrodata to analyze the performance of large and medum-szed frms. We use the Orbs database smply to etend the sales and ROA data to FY2013 and to wden the coverage of our analyss to nclude small frms (wth fewer than 50 employees), assumng that smaller frms nternatonal trade and overseas operatons were neglgble. The number of observatons eamned s roughly 2.2 mllon, coverng 359,000 frms n total, whch consst of 46,000 (large and medum-szed) frms n the BSBSA and 314,000 (small) frms n Orbs. The basc statstcs of the sample are reported n Table 1. We classfed our sample frms nto three sze categores based on the number of regular employees: 300 or more (large frms), 50 to 299 (medum-szed frms), and fewer than 50 (small frms); and four ndustral categores: 5

6 manufacturng, wholesale & retal, constructon, and others. Comparng frms average performance ndces durng FY1995-FY2013 perod by frm sze, we see that larger frms performed slghtly better than medum-szed frms (see the BSBSA result n Table1-2-(1)). Whle we cannot strctly compare the performance of small frms wth that of large and medum-szed frms snce ther sample perods are dfferent, the performance of the small frms does not look as good as that of the larger frms. Table 1-2-(2) compare the performance ndces between the average of FY2011-FY2012 and that of FY2013, n order to see how frms performance were dfferent n FY2013 from prevous years. We can see the average sales growth n FY2013 are hgher than that of FY2012-FY2011 regardless of frm sze or ndustry. As for ROA, we can confrm better performance n FY2013 for our small frm observatons, though we do not see a notceable mprovement n the proft rate for our large and medum-szed frms. To further eamne the stuaton n 2013 (.e., after the start of Abenomcs) wth precedng years, Fgure 3 presents kernel denstes of the dstrbutons of performance ndcators by year (4th quarter). The estmated dstrbutons ndcate that n terms of frms sales growth (from the prevous year) and ROA, 2013 was among the best years n the past decade. On the other hand, the level of frms sales has only just recovered from the slump n the wake of the global fnancal crss. Comparng the performance of large and medum-szed frms vs. small frms (Fgure 3.2), whle the performance of ndces for small frms vary wdely as epected, we can see the favorable performance n 2013 not only for large and medum-szed frms but also for small frms. Comparson between eportng frms and mportng frms (Fgure 3.3) suggests that the dfferences n the 2013 performance between them are not as outstandng as they were antcpated. Fnally, Fgure 3.4, whch compares the performance between frms nsde and outsde large ctes, shows that frms are performng relatvely well n 2013 regardless of frm's locaton. In order to more formally nvestgate the causes of the observed favorable busness 6

7 condtons, we run mcrodata-based reduced-form regressons n the followng sectons. 3. Emprcal strategy To emprcally eamne the mpact of echange rate changes on frms performance, we run the followng reduced-form regressons for frm performance varable z, where z s ether the log of frms sales or ROA: Dz,t = α + + β 1,k k = 0,12 2,k k = 0,12 1,k k = 0,12 β D ln EER D ln EER t k t k t k *(( X M 2,k k = 0,12 ) / S )* TSDum t k 3,k k = 0,12 + γ 1 D ln DRt 12 + γ 2D ln DU t 12 + γ 3 D ln DWt + γ 4D ln DWt * XDum + γ 5D ln DEt + γ 6 D ln DEt * XDum + ρ D ln POIL + ρ D ln POIL * MDum + ε + β,t D ln EER t k *(( X M ) / S )*(1 TSDum ) (1), where EER t s the real effectve echange rate at perod t, epressed n terms of foregn currency unts per yen, so that an ncrease n the rate amounts to a yen apprecaton (see Append at the tal of ths paper for the defntons of varables and data sources). (X -M )/S s the sze of frm s trade surplus relatve to ts sales. We nclude TSDum, a dummy varable that takes 1 when (X -M )>0 holds for frm, to allow for asymmetrc responses between frms wth trade surplus and frms wth trade defct. As control varables, we nclude the annual growth rates of domestc prvate demand (DR t ), domestc publc demand (DU t ), the US and European economes (DW t ), and Asan economes (DE t ), as well as rate of change of the ol prce (POIL t ). XDum and MDum are dummy varables that take 1 f frm eports or mports, respectvely. If yen deprecaton has a postve effect (after controllng for other factors) on the performance of Japanese frms (regardless of whether they are eporters or mporters, or both, or none), the coeffcent β s epected to be negatve. If frms wth a trade surplus (defct) are 1, k affected more postvely (negatvely) from yen deprecaton, β ( β ) s also epected to be 3, k 2, k negatve. If mpacts on the trade surplus frms and those on the trade defct frms are symmetrc, 7

8 β = holds true. As for the demand-related control varables, we generally epect that the 2, k β 3,k coeffcent γ should be postve. Whle changes n the ol prce are epected to have a negatve j effect (ρ<0) on frms ROA, the effects on sales growth are uncertan, snce a hgher ol prce may cause some ncreases n sales prces. Further to see whether the effects of echange rate changes on frms performance dffer across frms wth dfferent characterstcs, we would compare (1) large vs. medum-szed vs. small frms, and (2) manufacturng vs. wholesale and retal vs....(n secton 4.2). We also try the followng epanded regressons to check whether the effects of yen deprecaton n FY2013, under Abenomcs, were dfferent from those observed n the past (secton 4.3): Dz,t = α + + 1,k k= 0,12 2,k k= 0, β,t D ln EER D ln EER t t t t k t k + β D ln EER * D2013 *(( X M ) / S )* TSDum + t + β D ln EER *(( X M ) / S )* TSDum * D2013 3,k k= 0,12 + β D ln EER *(( X M ) / S )* ( 1 TSDum )* D ε β t β t D ln EER t k *(( X M ) / S )* ( 1 TSDum ) (2). D2013 t s a dummy varable that takes 1 f perod t belongs to fscal year β 4, β 5 and β 6 wll be sgnfcantly dfferent from zero f the patterns n 2013 were dfferent from those n the prevous years. In secton 4.4, we wll further eamne the relatonshp between the effects of echange rate changes and the global actvtes of Japanese frms. The globalzaton of Japanese frms advances rapdly, as reflected n the hke of eport/mport shares n total frm sales, n ncreasng frms overseas operatons, as well as n the upward trend of Japanese frms nvestment and loans to ther overseas afflated frms (see Fgure 4). Therefore t s reasonable to assume that the effect of echange rate changes on the performance of frms changes under ther nfluences. We wll specfcally eamne (1) how the profts (n foregn currency terms) of frms overseas busness are 8

9 reflected n the performance of Japanese frms; and (2) how the acceleratng shft to overseas producton s changng the pattern of (eportng) Japanese frms gans from yen deprecaton. 4. Emprcal fndngs 4.1 General evaluaton The results of our baselne regresson are reported n Table 2 (all estmated coeffcents reported n tables of ths paper are from medan regressons). We frst run the regresson usng the sample of large and medum-szed frms from the BSBSA, about 450,000 observatons (45,000 frms). We also run regressons usng an etended dataset wth smaller frms from the Orbs database, about 1.7 mllon observatons (350,000 frms). Despte the dfference n sample sze between the BSBSA dataset and the etended dataset, the obtaned parameters generally look very smlar for the two datasets. The parameters obtaned follow the epected patterns n general. As for the effects on Sales growth, ) the sales of Japanese frms (regardless of whether they are eporters or mporters, or both, or none) ncrease sgnfcantly when the yen deprecates ( β 0 ); a 30% deprecaton of 1, k < the yen, roughly equvalent to a yen deprecaton after the start of Abenomcs, results n a 3.9 (= ) % ncrease n frms sales; ) the postve effects of a yen deprecaton are sgnfcantly larger not only for frms wth a trade surplus ( β 0 ) but also for frms wth a trade 2, k < defct ( β 0 ); ) the demand-related control varables enter the sales regresson as epected 3, k > ( γ > 0 ); and v) an ol prce ncrease leads to an ncrease n Japanese frms sales ( ρ > 0 ), j probably due to some sales prce ncreases. As for the effects on ROA, ) the ROA of Japanese frms ncreases n the year that the yen deprecates ( β 0 ); a 30% deprecaton of the yen results n a roughly 0.3% ncrease n 1,0 < 9

10 frms ROA; ) whle the (postve) effect n the year of yen deprecaton s sgnfcantly larger for frms wth a trade surplus ( β 0 ), that for frms wth a trade defct s sgnfcantly smaller or 2, k < even turns to negatve ( β 3, k < 0 ); nverson of sgn on coeffcent β from postve n the sales 3, k regressons to negatve n the ROA regressons suggests that the sales ncrease observed for mportng frms s not enough to cover the cost ncrease due to a yen deprecaton; ) the demand varables enter the ROA regressons postvely ( γ > 0 ) ; and v) an ol prce ncrease leads to a lower ROA of Japanese frms ( ρ < 0 ), as epected. j 4.2 Comparson by sze and ndustry Table 3 reports the results of the regressons by frm sze. Whle we see some mnor dfferences between large and medum-szed frms, the estmated coeffcents look generally the same as those reported n Table 2 (our baselne result). However, the coeffcents for small frms are qute dfferent from those for the larger frms. Among other thngs, 1) the postve effects of a yen deprecaton look grossly neglgble for small frms, and 2) the sales of small frms are negatvely affected by an ncrease n the ol prce. Table 4 reports the results of the regressons by ndustry (and frm sze). The postve sales effects of a weaker yen are most pronounced for large/medum-szed (eportng) manufacturng frms. However, postve sales effects can also be observed for large/medum-szed frms n other ndustres and small frms n non-manufacturng ndustres. Whle the ROA of eportng large/medum-szed frms are affected postvely by yen deprecaton, those of mportng frms are sometmes affected negatvely, and the ROA gans for small frms look largely neglgble. 4.3 Effects n

11 Table 5 reports the regresson results eamnng whether the mpact of yen deprecaton n 2013 dffered from that n the past. We fnd (n Table 5(a)) a negatve and sgnfcant coeffcent ( β < 4 0 ) on the cross-term, whch potentally suggests that the postve mpact of a yen deprecaton was larger n However, t s also possble that Japanese frms performed well n 2013 for other reasons that are not controlled n our reduced-form regressons. We tred varous alternatve regressons to eclude the effects of other factors such as the surge n demand before the consumpton ta rate ncrease n Aprl 2014 and ncreases n publc spendng under Abenomcs (see Table 5(b)), but the estmated coeffcent ( β 4 ) s not greatly affected. Another fndng s that the postve effects of yen deprecaton on the performance (sales growth, ROA) of frms wth a trade surplus n 2013 were sgnfcantly smaller ( β > 5 0 ) than n prevous years. The smaller gan from yen deprecaton for eportng frms s consstent wth our macro-based fndng that the eport volume overall has not ncreased much n 2013 despte yen deprecaton. In sum, whle Japanese frms appear to have been favorably affected by the deprecaton of the yen n 2013, the channel through whch they dd so seems to have shfted away from the classcal boost to eport. 4.4 Echange rate effects and globalzaton Fnally, Table 6 reports the results eamnng the relatonshp between the effects of echange rate changes and the globalzaton of Japanese frms. We frst etended our regresson specfcaton to nclude cross-terms to see whether the postve effect of yen deprecaton s larger for frms wth overseas busness establshments. Yen deprecaton s epected to have a postve effect on the sales and profts of Japanese frms wth overseas busness, snce the yen value of ther overseas sales and profts ncreases wth a 11

12 deprecaton of the yen. The results reported n Table 6(a) support ths conjecture; at the same tme, the effect of yen deprecaton remans postve even after controllng for ths factors (and even for frms wthout eports). We also eamned whether the smaller gans for eportng frms n 2013 (reported n Table 5) result from Japanese frms globalzaton (.e., from the transfer of producton overseas). If the smaller gans for eportng frms n 2013 are due to changes n the behavor of frms that transferred producton overseas, the ncluson of a cross-term wth a dummy for frms that transferred producton should reduce the sgnfcant postve coeffcent ( β 5 > 0 ) n Table 5. Although the hypothess that the transfer of producton s reducng the gans from yen deprecaton has ganed wde currency and sounds plausble, our result (the coeffcent on the cross-term wth the overseas transfer enhancement dummy n Table 6(b)) appears not to support t. As frm-level eport data for 2013 s not yet avalable, we unfortunately need to wat another year to settle ths ssue. 5. Concludng remarks The purpose of ths paper was to eamne the effects of echange rate changes on the performance of Japanese frms, especally n recent years. Whle our results are very prelmnary, the fndngs can be summarzed as follows: (1) The sales growth and ROA of Japanese frms n FY2013 were among the best n the past decade, rrespectve of frm sze or locaton of the frm; (2) Yen deprecaton appears to have a postve effect on the performances of Japanese frms through the sales channel n general, and t s large eportng frms that beneft the most as epected; (3) In contrast, yen deprecaton affects ROA of mportng frms negatvely through the cost channel; (4) The frm performance n 2013 s grossly n lne wth or an etenson of the estmated relatonshp between the yen rate and the performance of Japanese frms n the past 20 years, whle our 12

13 fndngs on ther own cannot substantate the vew that the yen deprecaton caused the strong performance. (5) The postve effect of yen deprecaton on the eportng frms appears to be smaller, for some reason, n 2013, though our analyss do not produce evdence supportng the hypothess that frms overseas transfer of producton s reducng ther gans from yen deprecaton. The fndngs reported n ths paper are very prelmnary, so further analyses are necessary to arrve at frmer conclusons. We frst would lke to note that our analyss lacks a strong theoretcal foundaton, whch are necessary to provde a frmer bass for our analyss. One promsng drecton would be to eamne the relatonshp between frms' partcpaton n global supply chans and echange rate effects. (For ths purpose t would be more frutful to focus on a sample of frms wth certan characterstcs, whle the current study cast ts net wder, coverng all frms.) Secondly, our observaton perod s too short to evaluate the mpact of Abenomcs. As tme goes by, etendng our observaton perod to cover one more year wll help to obtan frmer and rcher conclusons. Fnally, t would also be frutful to conduct smlar studes on frms n other countres to draw more general conclusons that apply not only to Japan, but more broadly. References Baggs, Jen, Eugene Beauleu, and Loretta Fung (2009) Frm survval, performance, and the echange rate, Canadan Journal of Economcs, Vol.42, No.2, , May Dhasmana, Anubha (2013) Real Effectve Echange Rate and Manufacturng Sector Performance: Evdence from Indan frms, Indan Insttute of Management, Bangalore, Workng Paper No.412, May Fukuda, Shn-ch (2015) Abenomcs: Why Was It So Successful n Changng Market 13

14 Epectatons? CIRJE Dscusson Papers CIRJE-F-969, March Fung, Loretta and Jn-Tan Lu (2009) The mpact of real echange rate movements on frm performance: A case study of Tawanese manufacturng frms, Japan and the World Economy, Vol.21, Hote, Masak (2012) The Effects of the Echange Rate on Japanese Frms Investment: An Analyss wth Frm-Level Data, Publc Polcy Revew, Vol.8, No.5, Polcy Research Insttute, Mnstry of Fnance, Japan, , November Matsubayash, Yoch (2011) Echange Rate, Epected Proft, and Captal Stock Adjustment: Japanese Eperence, Japanese Economc Revew, Vol.62, No.2, Japanese Economc Assocaton, Nucc, Francesco and Alberto F. Pozzolo (2001) Investment and the echange rate: An analyss wth frm-level panel data, European Economc Revew, Vol.45, Nucc, Francesco and Alberto F. Pozzolo (2010) The echange rate, employment and hours: What frm-level data say, Journal of Internatonal Economcs, Vol.82, Shmzu, Junko and Kyotaka Sato (2015) Abenomcs, Yen Deprecaton, Trade Defct, and Eport Compettveness, RIETI Dscusson Paper Seres 15-E-020, February Shoj, Etsuro (2015) Tme varyng pass-through: Wll the yen deprecaton help Japan ht the nflaton target? JSPS Grants-n-Ad for Scentfc Research (S) Understandng Persstent Deflaton n Japan Workng Paper Seres No.068, March

15 Fgure 1. Developments n macroeconomc ndcators for Japan (2010=100) 1(a ) Echange rates Dollar to yen rate Real effectve rate 0 Jan-95 Aug-95 Mar-96 Oct-96 May-97 Dec-97 Jul-98 Feb-99 Sep-99 Apr-00 Nov-00 Jun-01 Jan-02 Aug-02 Mar-03 Oct-03 May-04 Dec-04 Jul-05 Feb-06 Sep-06 Apr-07 Nov-07 Jun-08 Jan-09 Aug-09 Mar-10 Oct-10 May-11 Dec-11 Jul-12 Feb-13 Sep-13 Apr-14 Nov-14 1(b ) Japanese stock prce ndces TOPIX JASDAQ Jan-95 Aug-95 Mar-96 Oct-96 May-97 Dec-97 Jul-98 Feb-99 Sep-99 Apr-00 Nov-00 Jun-01 Jan-02 Aug-02 Mar-03 Oct-03 May-04 Dec-04 Jul-05 Feb-06 Sep-06 Apr-07 Nov-07 Jun-08 Jan-09 Aug-09 Mar-10 Oct-10 May-11 Dec-11 Jul-12 Feb-13 Sep-13 Apr-14 Nov-14 1(c) Real effectve echange rate, shpments, and operatng proft Real effectve rate Shpments nde Operatng proft Jan-95 Aug-95 Mar-96 Oct-96 May-97 Dec-97 Jul-98 Feb-99 Sep-99 Apr-00 Nov-00 Jun-01 Jan-02 Aug-02 Mar-03 Oct-03 May-04 Dec-04 Jul-05 Feb-06 Sep-06 Apr-07 Nov-07 Jun-08 Jan-09 Aug-09 Mar-10 Oct-10 May-11 Dec-11 Jul-12 Feb-13 Sep-13 Apr-14 Nov-14 15

16 Fgure 2. Annual percent changes n the dollar/yen echange rate and trade ndcators Eport volume Unt eport prce Dollar/yen echange rate

17 Fgure 3. Kernel densty of performance ndcators by year (4th quarter) 3.1 All observatons Sales growth Sales (level) ROA All frms(lms) All frms(lms) All frms(lms) Large & medum-szed frms vs small frms Large and medum-szed frms: BSBSA Sales growth Sales (level) ROA BSBSA frms(lm): Large & Mddle Small frms: Orbs Orbs frms(s): Small BSBSA frms(lm): Large & Mddle Orbs frms(s): Small BSBSA frms(lm): Large & Mddle Orbs frms(s): Small

18 Fgure 3. (contnued) 3.3 Eportng frms vs mportng frms Eportng frms n BSBSA Sales growth Sales (level) ROA Eportng BSBSA frms(ex) Eportng BSBSA frms(ex) Eportng BSBSA frms(ex) Importng frms n BSBSA Importng BSBSA frms(im) Importng BSBSA frms(im) Importng BSBSA frms(im) Frms nsde and outsde large ctes Frms nsde large ctes Sales growth Sales (level) ROA Frms nsde large ctyzone(c) Frms nsde large ctyzone(c) Frms nsde large ctyzone(c) Frms outsde large ctes Frms outsde large ctyzone(nc) Frms outsde large ctyzone(nc) Frms outsde large ctyzone(nc) kdensty roa_egyo_master

19 Fgure 4. Globalzaton of Japanese frms seen n the BSBSA (% ) Fgure 4(a). Share of frms wth overseas establshments n the BSBSA sample frms Fgure 4(b). Developments of nvestment and loans to overseas afflated frms by the BSBSA sample frms (trllon yen) Total amount of nvestment and loans to overseas afflated frms Share of frms wth nvestment and loans to overseas afflated companes (BSBSA) (% )

20 Table 1-1. Sample statstcs (Levels) Total number of observatons BSBSA + Orbs ( Total number of frms ) Sample of large and medum-szed frms from the Basc Survey of Japanese Busness Structure and Actvtes (BSBSA) FY1994-FY2013 Sample of small frms from Orbs (Bureau Van Djk Database) FY2005-FY2013 Obs. Mean Medan S.D. Obs. Mean Medan S.D. Captal (mllon yen) Total Sample 2,249,471 ( 359,641 ) 525,083 1, ,836 1,724, By frm sze: Large (300 or more employees) 110,186 ( 8,324 ) 110,186 6, ,259 Medum (50 to 299 employees) 414,897 ( 37,556 ) 414, ,830 Small (fewer than 50 employees) 1,724,388 ( 313,761 ) n.a. n.a. n.a. n.a. 1,724, By ndustry: Manufacturng 390,102 ( 42,720 ) 255,194 1, , , Wholesale and retal 418,331 ( 54,256 ) 182, , , Constructon 1,129,372 ( 204,844 ) 8,553 1, ,371 1,120, Other 311,666 ( 57,821 ) 78,674 1, , , Total assets (mllon yen) Total Sample 2,233,526 ( 359,641 ) 509,139 21,416 3, ,717 1,724, ,992 By frm sze: Large (300 or more employees) 107,414 ( 8,324 ) 107,414 83,524 19, ,577 Medum (50 to 299 employees) 401,725 ( 37,556 ) 401,725 4,810 2,796 9,524 Small (fewer than 50 employees) 1,724,387 ( 313,761 ) n.a. n.a. n.a. n.a. 1,724, ,992 By ndustry: Manufacturng 381,789 ( 42,720 ) 246,881 22,634 3, , , ,146 Wholesale and retal 413,037 ( 54,256 ) 177,368 16,319 4, , , ,468 Constructon 1,129,102 ( 204,844 ) 8,283 13,302 3,983 64,445 1,120, ,090 Other 309,598 ( 57,821 ) 76,607 30,173 2, , , ,385 Number of regular employees Total Sample 2,202,265 ( 359,535 ) 508, ,334 1,693, By frm sze: Large (300 or more employees) 107,270 ( 8,324 ) 107,270 1, ,756 Medum (50 to 299 employees) 401,031 ( 37,556 ) 401, Small (fewer than 50 employees) 1,693,964 ( 313,655 ) n.a. n.a. n.a. n.a. 1,693, By ndustry: Manufacturng 380,626 ( 42,718 ) 246, , , Wholesale and retal 411,392 ( 54,250 ) 177, , Constructon 1,118,086 ( 204,794 ) 8, ,109, Other 292,161 ( 57,773 ) 76, , , Amount of sales (mllon yen) Total Sample 2,239,123 ( 359,641 ) 514,745 23,940 4, ,502 1,724, ,750 By frm sze: Large (300 or more employees) 108,303 ( 8,324 ) 108,303 88,043 24, ,316 Medum (50 to 299 employees) 406,442 ( 37,556 ) 406,442 6,858 3,677 14,945 Small (fewer than 50 employees) 1,724,378 ( 313,761 ) n.a. n.a. n.a. n.a. 1,724, ,750 By ndustry: Manufacturng 384,783 ( 42,720 ) 249,877 21,462 3, , , ,225 Wholesale and retal 414,733 ( 54,256 ) 179,064 30,453 7, , ,669 1, ,814 Constructon 1,129,100 ( 204,844 ) 8,285 15,589 4,908 62,362 1,120, Other 310,507 ( 57,821 ) 77,519 17,773 3, , , ,460 Operatng proft (mllon yen) Total Sample 2,232,004 ( 359,571 ) 507, ,987 1,724, By frm sze: Large (300 or more employees) 107,163 ( 8,310 ) 107,163 2, ,122 Medum (50 to 299 employees) 400,453 ( 37,500 ) 400, Small (fewer than 50 employees) 1,724,388 ( 313,761 ) n.a. n.a. n.a. n.a. 1,724, By ndustry: Manufacturng 381,338 ( 42,714 ) 246, , , Wholesale and retal 412,976 ( 54,254 ) 177, , , Constructon 1,129,102 ( 204,844 ) 8, ,872 1,120, Other 308,588 ( 57,759 ) 75,596 1, , ,

21 Table 1-2-(1). Sample Statstcs (Rato)- from FY1995-FY2013 Sales growth from prevous FY (%) Total number of observatons (BSBSA + Orbs ) Sample of large and medum-szed frms from the Basc Survey of Japanese Busness Structure and Actvtes (BSBSA ) Sample of small frms from Orbs (Bureau Van Djk Database) FY1995-FY2013 FY2005-FY2013 Obs. Mean Medan S.D. Obs. Mean Medan S.D. Total Sample 1,728, , % 0.2% 19.5% 1,283, % 0.0% 38.1% By frm sze: Large (300 or more employees) 95,764 95, % 1.0% 18.7% Medum (50 to 299 employees) 349, , % 0.0% 19.8% Small (fewer than 50 employees) 1,283,605 n.a. n.a. n.a. n.a. 1,283, % 0.0% 38.1% By ndustry: Manufacturng 321, , % 0.1% 20.4% 101, % -0.1% 30.2% Wholesale and retal 335, , % 0.1% 17.8% 181, % 0.2% 27.9% Constructon 844,907 7, % -0.3% 20.8% 837, % -0.3% 41.0% Other 227,038 64, % 0.3% 20.2% 162, % 0.6% 37.1% ROA (Operatng proft/total assets) Total Sample 1,722, , % 2.7% 6.7% 1,284, % 1.5% 12.4% By frm sze: Large (300 or more employees) 94,605 94, % 3.3% 6.4% Medum (50 to 299 employees) 343, , % 2.5% 6.7% Small (fewer than 50 employees) 1,284,798 n.a. n.a. n.a. n.a. 1,284, % 1.5% 12.4% By ndustry: Manufacturng 317, , % 2.8% 6.6% 101, % 1.8% 9.3% Wholesale and retal 334, , % 2.2% 5.7% 181, % 1.7% 8.2% Constructon 845,488 7, % 2.5% 4.9% 838, % 1.3% 13.5% Other 225,697 62, % 4.2% 8.5% 163, % 2.1% 12.2% 21

22 Table 1-2-(2). Sample Statstcs (Rato)- from FY2011-FY2013 Sales growth from prevous FY (%) Sample of large and medum-szed frms from the Basc Survey of Japanese Busness Structure and Actvtes (BSBSA) Sample of small frms from Orbs (Bureau Van Djk Database) FY2011, FY2012 FY2013 FY2011, FY2012 FY2013 Obs. Mean Medan S.D. Obs. Mean Medan S.D. Obs. Mean Medan S.D. Obs. Mean Medan S.D. Total Sample 51, % 1.0% 17.7% 19, % 2.6% 17.3% 398, % 1.8% 38.3% 173, % 2.5% 35.8% By frm sze: Large (300 or more employees) 11, % 0.9% 16.8% 4, % 3.2% 16.6% Medum (50 to 299 employees) 40, % 1.0% 17.9% 15, % 2.3% 17.4% Small (fewer than 50 employees) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 398, % 1.8% 38.3% 173, % 2.5% 35.8% By ndustry: Manufacturng 23, % 0.2% 19.3% 9, % 1.8% 18.4% 29, % 2.2% 29.2% 13, % 0.8% 28.0% Wholesale and retal 16, % 1.4% 15.3% 6, % 3.3% 16.2% 50, % 1.6% 27.2% 23, % 2.0% 25.8% Constructon % 2.8% 18.0% % 6.6% 18.1% 258, % 2.2% 41.7% 109, % 3.9% 39.4% Other 11, % 1.4% 17.5% 3, % 2.7% 15.7% 60, % 1.1% 35.4% 27, % 1.0% 31.1% ROA (Operatng proft/total assets) Total Sample 51, % 3.1% 7.0% 13, % 3.2% 5.8% 398, % 1.5% 13.2% 173, % 2.2% 12.8% By frm sze: Large (300 or more employees) 11, % 3.6% 6.7% 3, % 3.8% 5.4% Medum (50 to 299 employees) 40, % 3.0% 7.0% 10, % 3.0% 5.9% Small (fewer than 50 employees) n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 398, % 1.5% 13.2% 173, % 2.2% 12.8% By ndustry: Manufacturng 23, % 2.8% 6.5% 6, % 2.8% 5.4% 29, % 1.8% 9.5% 13, % 2.1% 9.3% Wholesale and retal 16, % 3.0% 6.0% 4, % 2.9% 4.9% 50, % 1.7% 8.8% 23, % 1.8% 8.4% Constructon % 3.1% 5.5% % 4.1% 6.0% 258, % 1.3% 14.4% 109, % 2.2% 14.1% Other 10, % 4.4% 8.7% 2, % 5.0% 7.5% 60, % 2.1% 12.3% 27, % 2.5% 12.1% 22

23 Table 2. Effects of echange rate and control varables on the performance of Japanese frms Dep: Sales growth ( ln S,t) Dep: ROA,t BSBSA BSBSA (FY1995-FY2013) BSBSA BSBSA (FY1995-FY2013) plus plus (FY1995-FY2013) small frms n Orbs (FY1995-FY2013) small frms n Orbs (FY2005-FY2013) (FY2005-FY2013) Annual changes n real effectve echange rate (G_REER t ) *** *** *** *** G_REER t *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** *** Annual growth of prvate demand t *** *** *** *** Annual growth of publc demand t *** *** *** *** US and European GDP growth t (G_WEST t) *** *** *** *** G_WEST t Eportng frm dummy *** *** *** *** Asan GDP growth t (G_ASIA t ) *** *** *** *** G_ASIA t Eportng frm dummy *** *** *** Annual changes n ol prce (G_OILP t ) *** *** G_OILP t *** *** *** *** Importng frm dummy G_OILP t *** *** *** *** Importng frm dummy G_OILP t *** *** *** *** Number of observatons 445,343 1,728, ,151 1,306,343 Pseudo R Notes: Coeffcents are from medan regressons. Regressons also nclude 1997 VAT dummy and a constant. *** (** / *) ndcates the coeffcent s sgnfcantly dfferent from zero at the 1 (5 / 10) % level. 23

24 Table 3. Effects of echange rate and control varables on the performance of Japanese frms by frm sze Dep: Sales growth ( ln S,t) Dep: ROA,t BSBSA Orbs BSBSA Orbs (FY1995-FY2013) (FY2005-FY2013) (FY1995-FY2013) (FY2005-FY2013) Large frms Medum-szed frms Small frms Large frms Medum-szed frms Small frms Annual changes n real effectve echange rate (G_REER t ) *** *** G_REER t *** *** ** *** *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t ** *** *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t ** *** Annual growth of prvate demand t *** *** *** *** *** *** Annual growth of publc demand t *** *** ** *** *** US and European GDP growth t (G_WEST t) *** *** *** *** *** *** G_WEST t Eportng frm dummy *** *** *** *** *** *** Asan GDP growth t (G_ASIA t ) *** *** *** *** *** *** G_ASIA t Eportng frm dummy *** *** *** *** *** Annual changes n ol prce (G_OILP t ) *** *** *** *** G_OILP t *** *** *** *** *** *** Importng frm dummy G_OILP t *** *** *** *** *** *** Importng frm dummy G_OILP t *** *** *** *** *** Number of observatons 95, ,579 1,283,605 83, , ,192 Pseudo R Notes: Coeffcents are from medan regressons. Regressons also nclude a 1997 VAT dummy and a constant. *** (** / *) ndcates the coeffcent s sgnfcantly dfferent from zero at the 1 (5 / 10) % level. ( / ) ndcates the coeffcent s sgnfcantly dfferent from that for large frms at the 1 (5 / 10) % level. 24

25 Table 4. Effects of echange rate on the performance of Japanese frms by ndustry BSBSA (FY1995-FY2013): Large frms & medum-szed frms Orbs (FY2005-FY2013) Small frms Wholesale Wholesale Manufacturng Constructon Others Manufacturng Constructon Others & retal & retal Dep: Sales growth ( ln S,t) Annual changes n real effectve echange rate (G_REER t ) *** *** *** *** *** *** * G_REER t *** *** ** *** *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t * *** Annual growth of prvate demand t *** *** *** *** *** ** Annual growth of publc demand t *** *** *** *** *** *** ** *** Number of observatons 219, ,633 7,041 64, , , , ,551 Pseudo R Dep: ROA,t Annual changes n real effectve echange rate (G_REER t ) *** *** ** G_REER t *** *** * *** *** *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** Trade surplus frm dummy Trade surplus rato G_REER t * *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t ** *** Annual growth of prvate demand t *** *** Annual growth of publc demand t *** *** *** *** *** ** ** Number of observatons 188, ,849 5,934 51,607 76, , , ,605 Pseudo R Notes: Coeffcents are from medan regressons. Regressons also nclude the same control varables as n the regressons reported n Tables 2. *** (** / *) ndcates the coeffcent s sgnfcantly dfferent from zero at the 1 (5 / 10) % level. ( / ) ndcates the coeffcent s sgnfcantly dfferent from that for manufacturng frms at the 1 (5 / 10) % level. 25

26 Table 5. Effects of echange rate on the performance of Japanese frms n (BSBSA sample only) Dep: Sales growth ( ln S,t ) Dep: ROA,t All ndustres Manufacturng Wholesale Constructon Others All ndustres Manufacturng Wholesale Constructon Others & retal & retal 5(a ) Results based on the sample untl the end of FY2013. Annual changes n real effectve echange rate (G_REER t ) *** *** *** ** *** *** *** ** G_REER t *** *** *** * *** *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** *** *** ** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** ** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** *** *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** ** *** 2013FY Dummy t G_REER t *** *** *** *** *** *** *** *** *** *** 2013FY Dummy t Trade surplus frm dummy Trade surplus rato G_REER t *** *** ** ** 2013FY Dummy t ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t * Number of observatons 445, , ,633 7,041 64, , , ,849 5,934 51,607 Pseudo R (b ) Results based on the sample ecludng the 4th quarter of FY2013. Annual changes n real effectve echange rate (G_REER t ) *** *** *** ** *** *** *** * ** G_REER t *** *** *** * ** *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** *** *** ** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** ** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** *** *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** ** *** *** *** FY Dummy t G_REER t *** *** *** *** 2013FY Dummy t Trade surplus frm dummy Trade surplus rato G_REER t *** *** * ** ** FY Dummy t ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** ** Number of observatons 433, , ,609 6,816 62, , , ,124 5,781 50,199 Pseudo R Notes: Coeffcents are from medan regressons. Regressons also nclude the same control varables as n the regressons reported n Tables 2. *** (** / *) ndcates the coeffcent s sgnfcantly dfferent from zero at the 1 (5 / 10) % level. ( / ) ndcates the coeffcent s sgnfcantly dfferent from that for manufacturng frm at the 1 (5 / 10) % level. 26

27 Table 6. Frms' global actvtes and the effects of echange rate changes (BSBSA sample only) Dep: Sales growth ( ln S,t ) Dep: ROA,t All ndustres Manufacturng Wholesale All ndustres Manufacturng Wholesale & retal & retal 6(a ) Is the postve effects of yen deprecaton larger for frms wth overseas busness establshments? Annual changes n real effectve echange rate (G_REER t ) *** *** *** *** *** ** G_REER t *** *** *** *** *** *** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** *** *** ** Trade surplus frm dummy Trade surplus rato G_REER t *** *** *** * ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** ** *** *** *** *** ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t *** *** ** *** Overseas busness establshments dummy t G_REER t *** ** *** Overseas busness establshments dummy t G_REER t ** *** *** *** Overseas busness establshments dummy t *** *** ** Number of observatons 414, , , , , ,614 Pseudo R (b ) Does the acceralaton of overseas transfer lead to the smaller gans through the eport channel n 2013? 2013FY Dummy t G_REER t *** *** *** *** *** *** 2013FY Dummy t Trade surplus frm dummy Trade surplus rato G_REER t *** *** ** 2013FY Dummy t ( 1 -Trade surplus frm dummy ) Trade surplus rato G_REER t Overseas transfer enhancement dummy 2013FY Dummy t G_REER t Overseas transfer enhancement dummy 2013FY Dummy t Trade surplus frm dummy Trade surplus rato G_REER t Number of observatons 359, , , , , ,686 Pseudo R Notes: Coeffcents are from medan regressons. Regressons n Table 6(a ) nclude the same controllng varables as those reported n Table 2. Regressons n Table 6(b ) also nclude the same eplanatory varables as those ncluded n Table 6(a ). *** (** / *) ndcates the coeffcent s sgnfcantly dfferent from zero at the 1 (5 / 10) % level. ( / ) ndcates the coeffcent s sgnfcantly dfferent from that of manufacturng frm at the 1 (5 / 10) % level. 27

28 Append. Defntons of varables and ther data sources. Varable Abbrevaton Deffnton Source Captal Amount of frm 's captal. BSBSA, Orbs Total assets Amount of frm 's total assets ncludng fed assets and current assets. BSBSA, Orbs Number of regular employees Frm 's employees ecept part-tme workers. BSBSA, Orbs Sales S,t Amount of frm 's total sales BSBSA, Orbs Operatng proft Operatng proft =Sales-Operatng cost BSBSA, Orbs ROA Operatng proft / Total assets (end of the preceedng year) Author Eport value of frm n perod t X,t Amount of frm 's sales (eports) to abroad BSBSA Eportng frm dummy XDum Dummy varable that takes 1 f frm eports BSBSA, Author Import value of frm n perod t M,t Amount of frm 's purchace (mports) from abroad BSBSA Importng frm dummy MDum Dummy varable that takes 1 f frm mports BSBSA, Author Trade surplus rato (X -M )/S The (average) share of frm 's net eports n ts total sales BSBSA, Author Trade surplus frm dummy TSDum Dummy varable that takes 1 f frm s eport s larger than ts mport on BSBSA, Author average. FY2013 dummy D2013t Dummy varable that takes 1 f perod t belongs to fscal year BSBSA, Author Overseas busness establshments dummy Dummy varable that takes 1 f frm have at least one overseas busness establshment. BSBSA, Author Overseas transfer enhansment dummy Dummy varable that takes 1 f frm ncreaseas nvestment and loans for overseas afflated companes by more than 50%, from average to average BSBSA, Author Real effectve echange rate EER t Real effectve echange rate; G_REERt = ln EERt - ln EERt-12 BOJ, Author Real domestc prvate demand DR t Real domestc prvate demand ncludng prvate consumpton, prvate resdental and non-resdental nvestment, and cange n prvate nventores. SNA, Author Real domestc publc demand DU t Real domestc prvate demand ncludng government consumpton, publc nvestment and change n publc nventores. SNA, Author Real GDP of the advanced western countres DWt Share-weghted real GDP of U.S., Canada, UK, Swtzerland, Denmark, Sweden, Norway, Iceland, Austra, Belgum, Fnland, France, Germany, Ireland, Italy, Luembourg, Netherlands, Portugal, Span; G_WESTt = ln DWt - ln DWt-12. IFS, Author Real GDP of Asan economes DE t Real GDP Asan countres ecludng Japan; G_ASIAt = ln DE t - ln DE t-12. Ol prce POIL t Inde of average crude prce of Duba Fateh, U.K. Brent and West Teas Intermedate (spot, US $ / barrel); G_OILPt = ln POIL t - ln POIL t-12 IFS, Author IFS, Author (Note) 1. Macroeconomc varables are 12-month backward movng averages. If the monthly data s not avalable, we calculate monthly-spltted seres from quarterly data by usng frequency converson method (quadratc-match average) n Evews Frms whch do not meet followng crtera are removed from sample n order to eclude outlers. - Obs. >= 3 - Sales data s avalable. - Total employees > 0 - Total assets > 0 - Captal >= 3 mllon yen - ROA <= 3σ 28