Transcript
When heirs become major shareholders: Evidence onpyramiding nanced by related-party sales Sunwoo Hwang a , Woochan Kim b, ⁎ a Kenan-Flagler Business School, University of North Carolina at Chapel Hill, NC 27599, USA b Korea University Business School, Seoul 02841, Republic of Korea a r t i c l e i n f o a b s t r a c t Article history: Received 18 January 2016Received in revised form 20 August 2016Accepted 25 August 2016Available online 26 August 2016 This study investigates how related-party sales are used as a means to nancially support the rms in which heirs become major shareholders and allow them to strengthen control overother rms in the group through pyramiding. From a universe of Korean chaebol rms during2000 – 2011, we identify a subset of rms where heirs become major shareholders (treatmentgroup) and compare them against their propensity-score-matched rms (control group) beforeand after the ownership change. A series of difference-in-differences tests with rm xed ef-fects reveal that treatment group rms experience greater related-party sales, bene t fromthem in terms of earnings, and gain importance in controlling other rms in the group. How-ever, we do not nd these results when non-heirs (e.g., controlling shareholders and other rel-atives) become major shareholders.© 2016 Published by Elsevier B.V. Keywords: Family rmBusiness groupChaebolSuccessionRelated-party transactionControl-enhancement 1. Introduction Family business owners who wish to see their business transferred to the next generation need to complete two tasks. One isto appoint their chosen heir to a top management position (hereafter “ managerial succession ” ) and the other is to transfer equitystakes to the heir (hereafter “ ownership succession ” ). This second task is not trivial. Assuming that the family rm repeatedly re-lies on external equity nancing, the equity stakes the heir inherits may not be large enough to ensure control over the rm. Theexistence of estate tax may further aggravate the problem. It would be interesting to investigate how these challenges affect thebehavior of family rms and what alternative mechanisms may be devised to preserve control within the family when ownershipsuccession is prohibitively costly. The variations in such mechanisms across countries with different legal settings also deserve se-rious investigation.The existing literature, however, is uneven. Research, to date, has mainly focused on the appointment of family members totop management positions and its impact on rm performance (Smith and Amoako-Adu, 1999; Pérez-González, 2006;Bennedsen et al., 2007; Bertrand et al., 2008; Cucculelli and Micucci, 2008; Mehrotra et al., 2013). Exceptions to this includeEllul et al. (2010) and Tsoutsoura (2015), who investigate how inheritance law or succession tax in uences the investment deci-sions of family rms. Another exception is Villalonga and Amit (2009), who document the use of dual-class shares, trusts, foun- dations, and limited partnerships by US family rms. They suggest that founders may have introduced these mechanisms toperpetuate family control over subsequent generations. Journal of Corporate Finance 41 (2016) 23 – 42 ⁎ Corresponding author. E-mail addresses:
[email protected] (S. Hwang),
[email protected] (W. Kim). http://dx.doi.org/10.1016/j.jcorp n.2016.08.0130929-1199/© 2016 Published by Elsevier B.V. Contents lists available at ScienceDirect Journal of Corporate Finance journal homepage: www.elsevier.com/locate/jcorpfin In this study, we investigate how related-party transactions are used as a means to nancially support the rms in which heirsbecome major shareholders, and allow them to strengthen control over other rms in the group through pyramiding. In otherwords, we explore the possibility of another mechanism, besides dual-class shares, that may be employed by business groupsto preserve family control when ownership succession is prohibitively costly. It is true that pyramiding is less ef cient thandual class shares as a control-enhancement mechanism. According to Villalonga and Amit (2009), the wedge between controland cash ow rights in US family rms mainly comes from dual-class shares. They also show that indirect ownership throughtrusts, foundations, limited partnerships, and other corporations rarely creates a wedge. In certain jurisdictions, however, corpo-rate pyramiding may be the only available alternative.Consider a country where inheritance tax rates are high but dual-class shares are unavailable. In this setting, corporatepyramiding nanced by intra-group, related-party transactions between member rms may serve as an alternative mechanismof preserving family control over subsequent generations. This involves setting up a small, privately held company with theheir as a major shareholder and instructing other rms in the group to purchase goods and services from that rm. Increasedsales and earnings will enlarge the rm's asset size, eventually allowing it to acquire controlling equity stakes in other member rms. In the end, this pyramiding allows the heir to take control over the entire business group.We test this possibility through a study of family-controlled business group rms in Korea, known as “ chaebol ” rms. We useKorean data for several reasons. First, Korea has a legal setting where ownership succession is costly (i.e., with inheritance and gifttax rates of 50%), dual-class shares are prohibited by law, and indirect ownership through trusts and foundations are heavily reg-ulated, making corporate pyramiding the only remaining option. Second, family-controlled business groups dominate the Koreaneconomy, and the two key elements of the control preservation mechanism that we examine in this study – pyramidalshareholdings and related-party transactions – are prevalent in Korea. Third, Korea is one of the few countries where data oninter-corporate shareholdings and inter-corporate related-party transactions are available, even for non-listed rms. Such infor-mation is indispensable for addressing our research question.From a universe of Korean chaebol rms during 2000 – 2011, we identify a subset of rms in which heirs have become majorshareholders (treatment group) and compare them against their propensity-score-matched rms (control group) before and afterthe ownership change. A series of difference-in-differences tests with rm xed effects reveal several results consistent with ourpredictions.This study makes a number of contributions to the literature. Most importantly, to the best of our knowledge, this is the rststudy to explore the possibility that corporate pyramiding nanced by related-party sales is used as a means of preserving familycontrol when ownership succession is costly. As mentioned, research on family succession has focused mainly on managerial suc-cession rather than on ownership succession. Moreover, research on ownership succession has focused mainly on dual classshares as an alternative mechanism when ownership succession is costly. By focusing on corporate pyramiding, we hope toshed light on a different mechanism that has not received much attention.Second, our study contributes to family rm performance studies, which have gained in popularity since Anderson and Reeb(2003). 1 We contribute to this area by highlighting the importance of related-party transactions in performance assessments, es-pecially when family rms are a part of a business group. More remotely, our study is also related to studies on managerial own-ership and rm performance (Morck et al., 1988). Again, in a business group setting, the relationship between ownership andperformance cannot be assessed without considering related-party transactions.Third, we enrich the literature on business group tunneling (Bae et al., 2002; Bertrand et al., 2002; Baek et al., 2006; Cheung etal., 2006; Black et al., 2015). 2 We report empirical evidence that related-party transactions could bene t rms in which heirs be-come major shareholders, presumably at the expense of minority shareholders in counterparty rms. Our evidence on tunneling,however, remains indirect, as in other tunneling studies. 2. Pyramiding backed by related-party sales A business owner wishing to hand over controlling equity stakes to the next generation generally faces two challenges: therisk of dilution and the risk of taxation.If a family rm repeatedly relies on external equity nancing, the equity stake that later generations inherit may not be largeenough to warrant control over the rm (Helwege et al., 2007). In certain jurisdictions, this challenge is overcome by the use of dual-class shares or voting agreements (Villalonga and Amit, 2009). Business owners holding shares with multiple voting rights orpossessing contracts that revoke the voting rights of other shareholders may be free from the risk of dilution.The risk of taxation is another major challenge. Although some jurisdictions have abolished inheritance tax, many others stillimpose it. 3 In the US, the tax rate is as high as 35%. Moreover, even in jurisdictions that have abolished inheritance tax, capitalgains tax may still apply upon succession. Similarly, while some jurisdictions have abolished the gift tax, several jurisdictionshave retained it. In certain jurisdictions, this challenge of taxation is resolved with the use of trusts or private foundations thatreceive shares as donations (Thomsen, 1999; Villalonga and Amit, 2009). As charitable entities, they are exempt from taxationbut are still governed by family members who serve as trustees or are on their board of directors. 1 For studies on the relationship between family ownership and rm performance, see Maury (2006), Miller (2007), and Andres (2008). 2 Theliteraturealso documents the bright side ofintragrouptransactions — helping to overcome marketfrictions; see Gopalan et al., 2007; Buchuketal. (2014) and Gopalan et al. (2014). 3 Some jurisdictions use the term “ estate tax ” instead of “ inheritance tax. ” 24 S. Hwang, W. Kim / Journal of Corporate Finance 41 (2016) 23 – 42 What would happen if dual-class shares were legally prohibited, the counterparties to voting agreement dif cult to nd, andtrusts or private foundations heavily regulated? Founding families would then have a strong incentive to seek alternative ways of handing over controlling equity stake to their descendants. For families controlling business groups, one solution is to make itsholding company or its de facto holding company (in the case of groups with no obvious holding companies due to circularshareholdings) issue new shares privately to the heir at a heavily discounted price. This would enable the heir to acquire a con-trolling equity stake in the holding company at minimal nancial cost. 4 This is not possible for publicly traded companies, how-ever, where the preemptive rights of existing shareholders are typically well protected. For privately held companies, taximplications will prevent the use of such a scheme.An alternative solution is pyramiding nanced by related-party sales – setting up a privately held rm where the heir is amajor shareholder and instructing other rms to purchase goods and services from that rm. Increased sales and earnings willenlarge the rm's asset size, eventually allowing it to acquire controlling equity stakes in other member rms. 5 In this manner,pyramiding allows the heir to take control over the entire business group.Anecdotal evidence of pyramiding nanced by related-party transactions abounds among Korean chaebols. An exemplary caseis Hanwha S&C, an integrated IT service rm of Hanwha Group (see Fig. 1). Originally, it was wholly owned by Kim Seung-youn(33.3% share), the group Chairman of Hanwha Group and Hanwha Corp (67.7% share). However, by 2007, the shares of HanwhaS&C were sold to the chairman's three sons, with each owning 50%, 25%, and 25% of all shares respectively. 6 Afterwards, HanwhaS&C ′ s sales to member rms soared from 117 billion Korean won (approximately 117 million US dollars) in 2007 to 319 billionKorean won in 2010. Its earnings before interest and taxes (EBIT) also jumped from 11 billion Korean won in 2007 to 24 billionKorean won in 2010. This improved nancial strength enabled Hanwha S&C to acquire shares in other member rms. As of 2012,it holds shares of Hancomm (70%), Hanwha Corporation (2.2%), Hanwha General Insurance (0.37%), Human Power (100%),Hanwha Solar Energy (20%), Hanwha Total Energy (100%), and Yeosu Cogeneration System (100%). Prior to 2007, Hancommwas the only rm in which Hanwha S&C held shares. 7 3. Research design In this study, our aim is to quantify the effect of ownership change on a rm's related-party sales, earnings, and control overother rms. An obvious challenge is the endogenous nature of our treatment variable (i.e., ownership change). This may be ad-dressed by difference-in-differences ( DID ) or an instrumental variable ( IV ) approach using an exogenous shock to ownershipchange. However, such a shock is absent in our sample of Korean chaebol rms from 2000 to 2011.Hence, we employ the next best approach of using propensity-score-matched control group rms. First, we identify rms thatexperienced a major increase in their respective heir's ownership. We label this set of rms the “ treatment group ” rms. Second,for each treatment group rm, we identify a set of rms that are in the same industry, do not experience any major change infamily ownership during the entire sample period, and are not af liated with the same chaebol group. 8 Given the dominanceof manufacturing rms in Korea, we use 4-digit Korean Standard Industrial Classi cation (SIC) codes for manufacturing and 2-digit SIC codes for others. From this set of rms, we identify up to ve whose propensity scores are closest to that of the treated rm (i.e., ve nearest neighbors or 5-NN) within a caliper (maximum propensity score distance) of 0.0008. 9 We label this set of 4 Thesamepurposecouldbeservedwithaconvertiblebondorabondwithwarrant,withaheavilydiscountedconversionratioorexerciseprice.Agoodexampleof this isthecaseof Samsung Everland,which issued convertiblebondsto the sons and daughters oftheSamsungGroupChairmanin1996. Samsung Everland was con-sidered a de facto holding company of Samsung Group for many years. This transaction, however, triggered a lawsuit that lasted for several years, until 2008 ( The Fi-nancial Times , “ Samsung Prosecutor Demands Jail for Lee, ” July 10, 2008). 5 Another possibility is merging the rm in which the heir is a major shareholder with another that already has signi cant equity stakes in other rms. A good ex-ample is the merger between Cheil Industry (formerly Samsung Everland) and Samsung C&T in July 2015. Before the merger, the heirs were holding 42.15% of CheilIndustry shares but had no holdings in Samsung Electronics. Samsung C&T, on the other hand, was holding 4% of Samsung Electronics shares. Elliott Associates, a UShedgefund,opposedthemerger,arguingthatthemergerratiowasoverlybene cialtoCheilIndustryshareholders( TheWallStreetJournal , “ Samsung-ElliottFightWillGo More Rounds, ” July 1, 2016). 6 In May 2010, the shareholders of Hanwha Corp. led a derivative suit against the directors of Hanwha Corp. for selling Hanwha S&C shares below the discountedcash ow(DCF)value( TheKoreaTimes ,HanwhaSuedOverWealthTransfer,May19,2010).Inthiscivilcharge,theshareholdersaskedforacompensationof45billionKoreanwon(approximately,45millionUSdollars).InOctober2013,SeoulCentralDistrictCourtorderedthedirectorstoreturnonly8.9billionKoreanwontothecom-pany,whichwaswellbelowthedamagesoriginallyestimated.OnNovember6,2013,SeoulHighCourtreversedthedistrictcourt'srulinganddismissedallthechargesagainstChairmanKim.Atthetimeofwriting,thecaseisbeforetheSupremeCourt.Inaseparatecriminalcase(embezzlement),ChairmanKimwassentencedtoathree-yearprisontermwitha ve-yearsuspension( nalizedinFebruary2014).However,hewasacquittedofthechargeofsellingHanwhaS&CsharesbelowtheDCFvalue.These results indicate how dif cult it is to prevent tunneling with ex post legal remedies in Korea. 7 From Fig. 1, one can see that Hanwha S&C reduced its holdings in three rms that itacquired earlier.For Hancomm, the stake fellfrom 100% to 69.87% in2007 asHanwhaS&Csold31.13%ofHancommsharestoChairmanKim'sspouse.In2015,100%ofHancommsharesweresoldtoOricom,anadvertisingcompanycontrolledbyacompletelyseparatebusinessgroup(Doosan).TheproceedingsfromthesalesofHancommsharesin2007and2015musthaveallowedHanwhaS&Ctostrengthenitscontrol over other member rms.For Dangjin Technopolis, the stakefell from 80% to 0%in2009asHanwhaS&C soldall ofits sharesinDangjinTechnopolis to DreamPharma (Hanwha Group's pharmaceutical rm wholly owned by Hanwha Galleria, in which the controlling family has no direct equity stake). However, DangjinTechnopolis was liquidated in2011.WesuspectthatHanwha S&C's preemptive divestmentofDangjin Technopolis shares was meantto minimize theloss thatcouldhave accrued to the heirs. For Hanwha ITC, the stake fell from 100% to 0% in 2009 as Hanwha ITC merged with Hanwha S&C. 8 A major family ownership change refers to a change in the net ownership of heirs, controlling shareholders, or other remote relatives by N 10 percentage pointscumulatively over the entire sample period. 9 In our unreported analyses (available upon request), we con rm that our results are robust to different matching algorithms: nearest neighbor matching (1-NNwithout a caliper) and radius matching (using all rms within a caliper). When using 1-NN without a caliper, we have slightly lower t-values compared against thebaseline algorithm of 5-NN with a caliper of 0.0008. This suggests that using multiple matching rms helps reduce coef cient standard errors. We also nd that ourkey results survive even if we mandate publicly traded (privately held) treated rms to be matched only with publicly traded (privately held) rms.25 S. Hwang, W. Kim / Journal of Corporate Finance 41 (2016) 23 – 42 rms the “ control group ” rms and expect that such matching signi cantly lowers the risk of self-selection bias. Third, byconducting difference-in-differences test, we compare these two groups of rms before and after the treatment.To estimate propensity scores, we run a probit regression where we regress a binary treatment variable (1 for the treated rmin the year of treatment and 0 otherwise) on the lagged values of rm pro tability (EBITDA/sales) and rm size (natural loga-rithm of assets size). When using propensity scores for matching, we match more than one nearest neighbor ( “ oversampling ” )because it reduces variance by using more information to construct the counterfactual. 10 At the same time, we impose a caliperto avoid bad matches and hence raise our matching quality. We set the caliper level suf ciently narrow to avoid bad matches butnot to a degree that considerably increases the number of treated rms with no match. These two criteria led us to choose a cal-iper of 0.0008 11 (also, a caliper of 0.008 is roughly half of the propensity score's standard deviation calculated over the entire sam-ple). We allow an untreated rm to be a match for two different treated rms (matching with replacement).It is worth noting that matching by industry and pro tability helps us to rule out the alternative hypothesis that heirs delib-erately increase ownership in rms with better industry prospects and such a choice subsequently results in the rm's higherearnings or stronger control over other rms. Such matching also reduces the risk of breaching the parallel trends assumption,which is the key identifying assumption for the consistency of DID estimators.Eq. (1) speci es the DID regression to verify whether related-party sales (as a fraction of total revenue) in treatment group rms increase after the treatment, relative to that in control group rms: RPS it ¼ α þ β 0 TG i þ β 1 TP it þ β 2 TG i TP it t ≥ k i ð Þþ X Φ þ μ i þ υ t þ ε it ð 1 Þ RPS it is the related-party sales (as a fraction of total revenue) of rm i with other member rms in year t . We explain the de-tails of its measurement in Section 4.3. TG i is a treatment group dummy variable that takes a value of 1 if rm i is treated (i.e.,experiences a major increase in heir's ownership during 2000 – 2011) and 0 otherwise. We explain precisely what we mean by “ a major increase in heir's ownership ” in Section 4.3. TP it is a treatment period dummy variable for rm i that takes a value of 1 during the treatment period ( t ≥ k i ) and 0 otherwise. Note that k i is the year at which the treatment is being switched on in rm i . Firm i and its match share a common treatment period dummy. X is a column vector of control variables. μ i and υ t represent the xed effects of rm and year respectively. The coef cient of interest is β 2 , which captures the increase in related-party sales (as a fraction of total revenue) of treatment group rms after thetreatment relative to that in control group rms. Consistent with our prediction, we expect this coef cient to be positive and sta-tistically signi cant. Since the same rms appear multiple times in this panel regression, we use coef cient standard errors Hanwha S&C Chairman & Hanwha Corp. (~ 2005) 33.3% + 66.7% = 100% Chairman’s Three Sons (2005-6) 66.7% + 16.67% + 16.7% = 100%(2007 ~) 50% + 25% + 25% = 100%Derivative suit filed against the directors of Hanwha Corp. for selling Hanwha S&C shares below the DCF value Related-party SalesEBIT 117(2007)319billionKRW(2010)11(2007)24billionKRW (2010) 2005 Hancomm(100%)2007 DangjinTechnopolis(80%) 69.87% since 2007 2007 HanwhaCorporation (2.2%)2007 Hanwha Total Energy (70%)2008 First Fire & Marine Insurance(0.99%)2008 Hanwha ITC (100%) Hanwha General Insurance(0.37%)2012Hanwha Energy (100%)0% since 20090% since 2009 2010 YeosuCogeneration System (100%)2011 Human Power (100%)2011 Hanwha Solar Energy (20%) Fig. 1. Hanwha S&C's use of related-party transactions for control-enhancement. This gure depicts how a major change in Hanwha S&C's ownership structureprompted an increase in its related-party sales, earnings, and control over other member rms in the group. Hanwha S&C's sales to member rms soared from117 billion Korean won (approximately 117 million US dollars) in 2007 to 319 billion Korean won in 2010. Its earnings (EBIT) also jumped from 11 billion Koreanwon in 2007 to 24 billion Korean won in 2010. This improved nancial strength enabled Hanwha S&C to acquire shares in other member rms. In 2012, it heldshares of Hancomm (70%), Hanwha Corporation (2.2%), Hanwha General Insurance (0.37%), Human Power (100%), Hanwha Solar Energy (20%), Hanwha Total En-ergy (100%), and Yeosu Cogeneration System (100%). Prior to 2007, Hancomm was the only rm in which Hanwha S&C held shares. 10 In our unreported analyses (available upon request), we con rm that our key results are robust to the number of nearest neighbors within a caliper of 0.0008(i.e., using 2-NN, 3-NN, 4-NN, and 5-NN). 11 Inour unreported analyses (available upon request),wealsocon rmthat our key ndingsare robust to different choices ofcaliper levels, ranging from 0.0001 to0.0010.26 S. Hwang, W. Kim / Journal of Corporate Finance 41 (2016) 23 – 42
