“Swimm ing Ducks Forecast the Com ing of Spring”—The predictability of aggregate insider trading on futuremarket returns in the Chinesemarket

Chafen Zhu,LiW ang,Tengfei Yang

aDepartment of Accounting and FinancialM anagement,Schoolof M anagement,Zhejiang University,China

bSchool of Accountancy,University of Akron,USA

cCollege of Humanities and Social Sciences,National University of Defense Technology,China

“Swimm ing Ducks Forecast the Com ing of Spring”—The predictability of aggregate insider trading on futuremarket returns in the Chinesemarket

Chafen Zhua,*,LiW angb,Tengfei Yangc

aDepartment of Accounting and FinancialM anagement,Schoolof M anagement,Zhejiang University,China

bSchool of Accountancy,University of Akron,USA

cCollege of Humanities and Social Sciences,National University of Defense Technology,China

A R T IC L E IN F O

Article history:

Received 20 Ju ly 2012

Accep ted 14 April 2014

Availab le online 16 Sep tember 2014

Aggregate insider trading

Large shareho lder trading

Information hierarchy

Corporate governance

Emergingmarket

This study systematically exam ines the ability o f aggregate insider trading to predict futuremarket returns in the Chinese A-sharemarket.A fter contro lling for the contrarian investment strategy,aggregate executive(large shareho lder) trading conducted over the past sixmonths can p redict66%(72.7%)ofmarket returns twelvemonths in advance.Aggregate insider trading p redicts future market returns very accurately and is stronger for insiderswho have a greater information advantage (e.g.,executives and controlling shareholders). Corporate governance also af ects the p redictability o f insider trading.The predictability of executive trading isweakest in centralstate-owned com panies, probab ly because the“quasi-o f cial”status of the executives in those companies ef ectively curbs their incentives to benef t from insider trading. The p redictive power o f large shareho lder trading in private-owned com panies is higher than that in state-owned com panies,probab ly due to their stronger prof tmotivation and higher invo lvement in business operations.This study comp lements the literatureby exam ining an emergingmarketand investigating how the institutional context and corporate governance afect insider trading. ?2014 Production and hosting by Elsevier B.V.on behalf o f China Journalo f

Accounting Research.Founded by Sun Yat-sen University and City University o f Hong Kong.

1.Introduction

Insiders include a company’s corporate o f cers,directors,supervisors(all referred to as“executives”hereafter)and large shareholders(thosew ith ownership o f 5%ormore o f the company’s stock).Insider trading based on a superior information advantage can inf uence the ef ciency and fairness o f the f nancialmarket, and thus presents a thorny regu latory challenge.

When the Chinese A-sharemarketwas initially established,it had a dual share structure comp rising tradable and non-tradable shares.Sharesowned by large shareho ldersand executiveswere basically non-tradable as they were prohibited from being traded on the secondary stock market.As such,insider trading was very rare at that tim e.Since the end of 2005,w ith them arket refo rm o f non-tradab le shares,m arket segm entation between tradable and non-tradable shareshasbeen gradually rem oved.To m itigate the supp ly pressu re on the secondary m arket,the reform p reven ted non-tradab le shares from being so ld imm ediately,and instituted a lock-up period of one to three years.Some locked-up shares have been“lifted”since the beginning o f 2007.By the end o f 2011,almost all of the non-tradable shares became tradab le and insidersbegan to trade their stocksmore frequently,a new occurrence thathasattracted a lotof attention.The A-sharemarket isstill emerging and its ef ciency requires im provement.The quality of its corporate in formation disclosure is generally not high,its information intermediaries(analysts)have yet to mature and the information asymmetry between insiders and outsiders is still relatively large.These lim itations have undoubted ly provided insiders with more trading opportunities.Furthermore,a regulatory system has not yet been developed,making insider trading a potentially serious and comp licated challenge.

Based onm edia reports,insiders from diferen t com paniesoften trade in the sam e direction during the sam e period.W hen the m arket index is high or rap id ly rising,insiders often consistently decrease their ho ld ings. When themarket index is in a slump,insiders often uniform ly increase their holdings.The“mainstream”aggregate insider trading matches“cleverly”w ith marketmovements,suggesting that it is ideal in terms o f market tim ing.

Using A-sharemarket insider transaction data from January 2007 to August 2011,we empirically exam ine the predictability o f aggregate insider trading onmarket returns.We fnd that after controlling for contrarian trading,the past six months of aggregate executive(large shareho lder)trading can p redict 66%(72.7%)o f market returns twelvemonths in advance.W e also exam ine the ef ect o f information hierarchy on predictive power and fnd that the p redicative power o f aggregate trading is signif cantly higher for insidersw ith more business operation invo lvem en t and a higher position in the inform ation hierarchy(e.g.,executives and contro lling shareho lders)than fo r insiders w ith a lower position(e.g.,supervisors and im portan t shareho lders). This evidence strongly supports that an inform ation advantage,beyond being a sim p le contrarian trading strategy,is the root cause o f the strong p redictive power of aggregate insider trading.Stemm ing from their high involvement in businessoperations,insiders can aggregately form a stronger ability to predictmacroeconom ic trends and detect deviations in systematic valuation in the stock market(referred to as the“macro in formation advantage”)and use this advantagewhen trading.

As a typical form o f agency con f ict,insider trading behavior is afected systematically by corporate governance(Gunny et al.,2008).Corporate governance afects the in formation content of insider transactions in two ways.First,the ownership structure af ects the distribution of contro l rights and decision-making power between shareho ldersand executives,thusaf ecting the information distribution and advantagesof specif c insiders.Second,insider trad ing behavior ism onitored diferen tly under diferen t corporate governance structures,and corporate governance afects them o tives o f insiders and thus the possibility that insiders w ill use their info rm ation advan tage in trading.

In this study,we exam ine the ef ect of corporate governance on the predictive power o f aggregate insider trading.Depending on the nature o f the largest shareho lder’sownership,companies can be divided into three categories:private,localstate-owned and centralstate-owned.Localstate-owned companiesare contro lled by the local State-owned Assets Supervision and Adm inistration Comm ission(SASAC).Central state-owned companies are contro lled by either the central SASAC or central government departments.These three types of companieshavesystematically dif erent agency problemsand governancestructures.As investor p rotection isweak in the A-share stock market,the largest shareholder in most privately owned companies has high ly concentrated ownership and even contro lling ownership for self-p rotection purposes.Because these largeshareholdersare actively invo lved in businessoperationsand have strong incen tivesand the ability to m onito r m anagers,there ison ly a smallagency con f ictbetween shareholdersandmanagers.However,the largeprivate shareholder hasa strong incentive to expropriate the interestso f small shareholders.He or she can gain lucrative private benef tso f control via insider trading in the secondarymarket,related party transactionsand the pyram id shareho lding structure.In state-owned companies,theultimateshareho lderscan be traced up to local or central SASAC,which always delegates bureaucrats as board chairmen,resulting in serious governance p rob lems such as investor phantom s and manager control.M anagement occupies a large number of seats on theseboards,creating a self-monitoring problem.Despite their high concentration of ownership,largestate shareholders have an in formation disadvantage due to their low participation in business operations.There are also signif cant d if erencesbetween the cen traland local state-owned com panies.Centralstate-owned companies are usually super-large com panies tightly controlled by the central SASAC in strategically im portan t industries,such as China M obile,the four largest state-owned banks,China Petro leum and China Railway G roup.These large central state-owned companies tend to have a corresponding adm inistrative hierarchy and servemore po litical purposes,such asachieving fnancial and price stability.For example,China Petroleum and China Petrochem icalarem inisterial-level companiesand their executivesoften have“quasi-o f cial”status along with opportunities for promotion to the provincial orm inisterial government levels.Therefore, these executives are under recessive but powerfu l government adm inistrative control.

In this study,we exam ine how the nature o f ownership afects the predictive power o f insider trading.W e f nd that the predictive power o f executive trading in central state-owned companies is the weakest.The“quasi-of cial”statuso f theexecutives in this type of companyweakens itsmotivation to p rof t through trading,forcing it to give up opportunities to trade based on an inform ation advantage.The pred ictive power o f large shareho lder trading in private-owned com panies is signif cantly higher than that in state-owned com panies,as the shareholders in p rivate-owned com panies have a stronger p rof tmotivation and higher invo lvem ent in business operations.

Thisstudymakes the fo llowing contributions to theinsider trading literature.First,mosto f the literaturehas focused on theinformation contento f insider trading from am icro perspective.Ithasobserved that insiderscan use their f rm-levelinformation advantage to time tradesand that their trading activitiescan p redict their companies’future stock movem ents.W e exam ine the in formation content of aggregate insider trading from a m acro perspectiveand f nd thataggregate insider trading can predict futuremarketmovementsand that inform ation hierarchy and corporategovernanceaf ect thep redictability o f aggregate insider trading.Second,most insider trad ing studies focus on m aturem arkets,especially the U.S.m arket,and p rovide lim ited insights into em ergingm arketsw ith d if erent institutionalenvironm en ts.This study com p lem en tsp revious stud iesby focusing on an em ergingm arket and provides interesting f nd ings that are diferen t from those observed in m atu re m arkets.Thegeneral f nding inmaturemarkets is that insiderbuying hashigher information content than insider selling and that large shareho lder trading has lower information content than executive trading.However, we obtain dif erent fndings for the A-sharem arket due to its dif erent institutional context and governance structure.Third,this study expands the research related to how corporate governance af ects insider trading. Insider trading activitiesare not only in fuenced by statutory laws and law enforcement at the national level, butalso afected by corporategovernance.Thediverse typesof corporategovernancestructureson theChinese A-sharemarket p rovidean excellent setting for studying theef ectsof corporategovernanceon insider trading. Fourth,our fndingsprovide valuab le insights into form ing investment strategies.Asaggregate insider trading has signif cant power in p redicting fu ture stock retu rns,external investors can use insider trading activities to fo recast futu re stock m ovem ents.As the fam ous Chinese poem says,“A lthough w inter w ind blow s brisk ly, ducks forecast the com ing o f sp ring and begin to sw im in the lake.”M uch like sw imm ing duckshelp to forecast the change in season,aggregate insider trading helps to p redict themovement of themarket.

2.Literature review

Insiders are special traderswho possess high ly accurate information at a very low access cost.Company executives aremore fam iliar w ith their own companies than any analyst on W all Street.They know when a new productw ill be launched,an inventory w ill begin to stack,p rof tmarginsw ill expand or p roduct costs w ill improve.As Seyhun(1998)observes,if you want to fnd smart investors,these are smart investors.

Insiders have diferent k inds o f in form ation advan tages.First,they know in advance which m ajo r even ts will afect stock prices.This type of advantage disappears after the events are announced,usually w ithin a short time.Furthermore,direct use o f event-related inside information iso ften p rohibited inm any countries. Second,insiders can better assess the com pany’searnings prospectsand grow th potential than outsiders.This type of information advantage is usually long term and does not rely on specif c events.Third,insiders can better assess the intrinsic value of a company than outsidersand can thus identify and exploit opportunities when the stock marketovervaluesor undervalues the company.Fourth,insidershave a better sense of industry and macro-economy trends and thus tend to p redict futuremacroeconom ic directionsmore accurately. The f rst three typesof advantagescan increase insiders’ability to predict f rm-levelp ricemovementm orep recisely.They seem to always be able to cash in when the p rice is high and pu rchasewhen the price is low.The fourth advan tage increases aggregate insiders’ability to tim e them arket.

W hether and how insiders use their info rm ation advantage to trade is the co re issue o f em pirical studies. These questionsmay be answered by analyzing the relationship between stock pricemovements and insider trades,and exam ining the tim ing and pro f tability o f the trades.Studies w ith a m icro perspective exam ine the relationship between insider trades and the corresponding company’s stock pricemovements to verify whether insiders use f rm-level p rivate in formation to trade.In contrast,studies w ith a macro perspective exam ine the relationship between aggregate insider trading and futuremarketmovements to verify whether insiders have an overall ability to tim e themarket.

2.1.M icro perspective:can insider trades predict future stock returns?

Them icro perspectivewidely involves theevent studymethod to investigatea company’ssharep ricemovements shortly before and after an insider trade ismade to determ ine the insider’s ability to obtain abnormal returns.Insiders accurately tim e trades over the sho rt term(Friedrich et al.,2002).Furtherm o re,m any studies fnd that insider trad ing can p redict a com pany’s fu ture long-term p ricem ovem ents.Stock p rices tend to go up (down)in the long run after insider purchases(sales)are m ade.This p redictability suggests that insiders exploit important private in formation in advance that afects future stock p rices(Givo ly and Palmon,1985; Seyhun,1998;Pettit and Venkatesh,1995;Lakonishok and Lee,2001;Jeng et al.,2003;Ravina and Sapienza,2010).

Research has verifed in dif erentways that an information advantage accounts for the accurate tim ing and abnormal returns associated w ith insider trades.The pro f tability o f insider trades is closely related to the in formation hierarchy that insiders belong to;the higher the in formation hierarchy,the higher the p rof tability.In the U.S.,large shareho lders’trading prof tability is found to be signif cantly lower than that of executives.Because large shareho lders o ften do not own large p roportions o f the shares and are rarely directly involved in business decisions,they have less o f an in fo rm ation advantage than executives(Lakonishok and Lee,2001).Even execu tives have an in form ation hierarchy,w ith CEOs at the top,o ther m anagers in them idd le and board directorsat the bottom.The trading pro f tability o f diferent groups is positively correlated w ith such an in formation hierarchy(Seyhun,1998).However,Jeng etal.(2003)fnd that although CEOs havemore information,their trading activitiesaremore likely to bemonitored by shareholdersand regu lators, and therefore their trading pro f tability isnot necessarily signif cantly higher than that o f othermanagers.In addition,studies consistently f nd that the p rof tability of executives’sales isweaker than that of purchases. Executives’purchases aremain ly based on private inform ation.In contrast,w ith the popularity of equity incentives,executivesoften sellstocks tomeet liquidity or diversif cation needs,and thisnon-information-driven selling dilutes the overall in formation content for sales.

2.2.M acro perspective:can aggregate insider trading predict futuremarket trends?

There is some evidence that aggregate insider trading coincides w ith market price movements.Seyhun (1990)f nds through a case study that sho rtly after the stock m arket crash in October 1987,insiders aggregately pu rchased shares and accu rately pred icted them arket rebound.Seyhun(1992)also f nds that during 1975–1989,the net insiders’pu rchase index of the past 12m onths pred icted 16%(61%)of the next 6(12) monthso fmarket returns.It seems thatwhen futuremarket returns are positive(negative),aggregate insidertrading activities exhibit op tim istic purchases(pessim istic sales).Lakonishok and Lee(2001)design diferen t net purchasemeasuresand contro l for theef ectsof a simple contrarian strategy.They also f nd thataggregate insider trading can accurately predict futuremarket returns.The one-year-ahead market return gap between themonthsw ith the highest and lowest net purchases index isabout 11%,and the aggregate trading o f executivesexhibitsbetter p redictability than that of largeshareholders.Overall,empiricalstudiesshow that aggregate insider trading is an important leading indicator o f future stock marketmovements and can accurately p redict futuremarket returns.

Why doesaggregate insider trading efectively p redict futuremarket trends?Seyhun(1998)f nds that insidersasa group can correctly anticipate the direction of the future real economy and react to changes in future econom ic trends nearly a year ahead of the stock m arket.Jiang and Zam an(2007)study the causes o f aggregate insider trad ing’s p redictability o f futu rem arket retu rns and fnd that although such trad ing is strongly correlated w ith cash fow in form ation unexpected by them arket,expected cash fow in form ation doesnot lead to largewavesof insider trading.These studies suggest thataggregate insider trading isnot sim ply contrarian trading,but is rather based on insiders’efective assessment of futuremacroeconom ic trends.

M ostof the research on China focuseson them icro perspective.Studiesconsistently fnd that insider trades can predict both short-and long-term stock movements,and that insidershave the ability to time trades and earn high abnormal returns.In term so f executive trading,researchers fnd that executive salescan predict the fall of a company’s stock pricew ithin a short timew indow(Zeng,2008;Zhang and Zeng,2011).Zhu et al. (2011a,b)investigate the long-term p rof tability of executive trading and fnd that executive sales and purchasesexhibit a strong p redictability of stock pricemovement sixmonths in advance.In termso f large shareholder trading,studies also f nd that large shareho lder sales are accu rately tim ed w ithin a shortw indow(Zhu et al.,2011a,b;Caiand W ei,2009;Lin and Qu,2010;W u and W u,2010;Shen et al.,2011).However,the timing ability of shareho lder purchases is not obvious and purchase activities are not strongly correlated w ith future performance indicators(Liet al.,2011).

TheseChinesestudieshave adop ted them icro perspectiveand often focuson executivesor large shareho ldersasa single group,or on“buy or sell”asa one-way transaction.This is the f rst study to exam ine the inform ation contento f aggregate insider trading in the Chinese A-share stockmarket from am acro perspective.It comp rehensively compares the information content of trades in dif erent directionsand tradesmade by dif erent types o f insiders,and of ers in-depth analysis o f how inform ation hierarchies and corporate governance afect the in formation content of insider trading.

3.Research hypotheses

A t f rst glance,thepub lic and insidersboth observe the same changes inmacroeconom ic po licies.A lthough insiders havemore private company in formation,it seems that they should on ly be ab le to p redict the stock m ovem ent of their own companies.Insidershave no additionaladvantage overall for p redicting stock market m ovem ent.

In fact,insidersat the front lineo f operationscan often perceivemacroeconom ic trend changesearlier.For examp le,they can observe changes in in fation earlier through com pany material and product pricemovem ents.In contrast,the public has to wait until theend of themonth for such inf ation statistics to be released. M ore importantly,even if insiders and the pub lic were to observe the samemacroeconom ic policy changes, based on their experience w ith business operations,insiders have a deeper understand ing o f the efect o f changes on the com pany,the industry and fu ture m acroeconom ic perform ance.Thus,insiders have better analytical capabilities than ou tside investo rs in interp reting m acroeconom ic info rm ation.For exam p le, although insiders and the public observe the sam e changes in interest rates,insiders can better interp ret the efecto f the changeson the com pany and industry’s futureperformance.W ithout know ing the company’sspecif c f nancing structure,outsiderswou ld havedif culty calculating the efect o f the changeson the company’s p rof t(Seyhun,1992).As insidersare involved at the forefront of businessoperations,they are ab le to detect m acro and industry changesearlier andm oreaccurately interp ret theef ecto f these changeson the futureperformance of the company,industry and macro-economy,and can better predict futuremarketmovement and fo llow their p redictionswhen trading.Seyhun(1998)f nds that insider trading can refect future changes in the real economy one year earlier than the stock market.

In addition to better pred ictability o f m acroeconom ics and industry changes,insiders are m o re able to detect system ic bias in their stock m arket valuations.Because investors can be emotional and irrational at times,themarket sometimessystematically overvaluesor undervaluesstocks,providing uniform opportunities for insider trades.In an overheated bullmarket,when the prices ofmost stocks are high,the insiders of different companies appear to decrease their holdings uniform ly and thus poo l into selling waves.Themarket eventually returns to rationalvaluations in the future,resulting in negative futuremarket returns.Insidersalso conversely take advantage of systematicmarket undervaluations to uniform ly increase holdings.Aggregate insider trading reacts to both systematicmarket overvaluations and undervaluations.

In summary,insidershave an evident in formation advantagewhen predictingmacroeconom icmovements and detecting system atic deviations in m arket valuation.Thism acro inform ation advan tage is no t a resu lt o f insiders ob taining m acro private in form ation(e.g.,changes in in terest rates)sooner than ou tsiders.Rather,it stem s from insiders’high involvem ent in business operations,which gives them a superior ability to analyze and interp retmacro in formation,p redictmacroeconom ic trendsand identify systematic deviations in market valuation.M ost insiders use theirmacro inform ation advantagewhen trading and aggregate insider trading can be used to p redict futuremarketmovements.

The p redictability o f aggregate insider trading is not a simp le summary o f insiders’ability to time trades based on private f rm-level information.Assum ing that insiders use on ly f rm-level inform ation to trade,different companieswou ld report good and bad news during the same period.As such,insiders from diferent companies would make purchases and sales at the same tim e,and therefore the aggregated trading shou ld be zero.However,if themajority of the insiders uniform ly choose to buy or sell during the same period,it is likely that the inform ation they are using is based on their pred iction of fu tu rem acro-levelm ovem en ts.

Hypothesis1.Aggregate insider trading can predict futuremarketmovementsand this predictability ref ects insiders’macro in formation advantage.

Insidersare divided into two m ajor categories:large shareholders and execu tives.Large shareholders can be subd ivided into the largest shareho lder and im portant shareho lders(m ore than 5%stock ownership bu t excluding the largest shareho lder).Executives can be subdivided into board directors,managersand supervisors.There are signif cant dif erences in operational involvement and the level of in formation advantage among insiders.The largest shareho lders have an important inf uence on strategic decisions and can directly appoint directorsandmanagers.Therefore,their operational invo lvementand level of in formation advantage are stronger than those o f important shareho lders.Among executives,directors(especially non-executive and independent directors)are main ly responsible for monitoring managers and making strategic decisions. Because theymeeton ly a few timesa year,their involvement in businessoperationsand thus their information advantage is less than that o fmanagers.Chinese companiesnormally add another position to the executives group:that is superviso rs,whose p rim ary responsibility is to perform a com p liance check afterm ajo r decisions arem ade.M ost supervisors are staf delegates or outside retired persons wo rking part tim e.Thus,their operational involvement is signif cantly less than that ofmanagers and directors(Zhu et al.,2011a,b).

Insiders’macro information advantage stems from their superior ability to interpretmacro information and predictmacroeconom ic trends as a resu lt of their high involvement in business operations.Therefore, themore insidersare involved in businessoperationsand thehigher their position in the in formation hierarchy (e.g.,largest shareho ldersandmanagers),themoreobvious theirmacro-in formation advantage and thestronger the p redictive power o f their trading activities for futuremarket returns.

Hypothesis2(information hierarchy hypothesis).The aggregate trading o f insiders w ith higher levels o f operational involvement and in formation hierarchy positions(e.g.,largest shareholders and managers)has more predictive power for future m arket returns than that of insiders w ith lower in formation hierarchy positions(e.g.,important shareholders and supervisors).

As a type of agency conf ict and wealth m isappropriation,insider trading is afected by corporate governance.The essence o f corporate governance is to lim it insiders’self-in terest behavio r and m o tivate them to m axim ize the wealth o f their com panies through a variety o f form al and in form al con trol m echanism s (G unny et al.,2008).Good corporate governance can help decrease illegal trades and the inapp ropriate use of an in formation advantage.A t the company level,corporategovernance can reinforce the advance approvalo f insider trades,develop a com p rehensive process fo r in fo rm ation m anagem ent and d isclosure,set a strict lockout period before news announcements and impose harsh punishments for ru le vio lations.Ravina and Sapienza(2010)con f rm that corporate governance can signif cantly afect the p rof tability of insider trades. A lthough insiders of companieswith good corporate governance are not able to obtain signif cant abnormal returns,thoseof companiesw ith poor corporategovernance can obtain abnorm al returnsup to a levelof 21%. Fidrmuc et al.(2006)fnd that the in formation content o f executive trading for companieswith external large shareholder oversight is less than that for companiesw ith dispersed ownership.

In theChinese A-sharemarket,executive trading isdom inated by sales.Executives’excessivesalescan have a signif cant negative ef ect on stock p ricesand market con fdence and can be detrimental to the interestso f large shareho lders.In addition,due to the histo rical low popu larity of equity incen tives,o riginal execu tive holdings are norm ally low and large decreases in executive holdings decrease the congruence o f in terest between execu tives and shareholders,weakening large shareho lders’con trol over execu tives and increasing agency conf icts.Therefore,large shareholdersmust control the decreases in executive ho ldings w ithin an accep table range.

The largest shareholders play a key oversight role in executive trading.In private-owned companies,the largest shareho lders are more concerned about their own interests and stock p rices,and are more w illing and havea stronger ability to explicitly or implicitly contro lexecutives’excessive salesand illegal trading.This m ay constrain executives’freedom to exploit their information advantage when trading.Rather,due to investor phantoms and insider control,shareho lders’oversight of executive trading is relatively weak in state-owned companies and thus self-interested executive trading is expected to be more serious than that in p rivate-owned com panies.

State-owned com panies can be further separated in to cen traland local state-owned com panies.Despite the weak oversight of their large shareho lders,central state-owned companies have a strong alternative adm inistrative contro l over executives.Executiveso f central state-owned com panieshave the statuso f quasi-governm ento f cials.The appointment,p romotion and demotion o f these executivesoccur under a strict and opaque adm inistrative contro l system and are not determ ined by the pro fessionalmarket o f corporate executives. Character and integrity are important aspects o f how executives are evaluated.Prof table executive trading m ay damage the reputation o f political executives and their“clean”image,negatively af ecting their evaluation by governmental organizations and thus their po litical futures.W ith this hidden but powerful adm inistrative contro l,executives are likely to restrain their prof tab le trading activities.Chen et al.(2011)p rovide em pirical evidence that suppo rts this argum en t.They exam ine the executive stock options held by execu tives in large red-chip state-owned com panies listed on the Hong K ong Stock Exchange(m ain ly cen tral stateowned com panies).They fnd that lucrative stock op tions are rarely exercised and state-owned execu tivesgive up the opportunity to pro f t legally from stock op tions.These executives face a number of“hidden rules.”Because they are not only executives but also quasi-government of cials,theym ust comply w ith a strict personnel system designed for“cadres.”As a result,the quasi-government of cial ro le o f these executives is expected to weaken theirmotive to trade for pro f t,forcing them to give up opportunities to use their inform ation advantages(including their macro in formation advantage)when trading.This in turn decreases the p redictive power o f aggregate executive trading in central state-owned companies.

Severalexpectationsarise.In localstate-owned companies,especially thosebelow thep rovincial level,executivesaremoreakin to pro fessionalmanagersw ith fewer“hidden”adm inistrative rules to follow.In addition, because large shareholder oversight of them anagem ent in this type o f com pany is relatively weak,execu tive trading in local state-owned com panies is higher than execu tive trading in private-and central state-owned com panies.Execu tives in local state-owned com panies arem ore likely to exp loit their info rm ation advan tage when trading,including their macro in formation advantage.Consequently,the aggregate trading of these executives has the highest predictive power.

Hypothesis3(effect of corporate governance on executive trading predictability).The executive trading of local state-owned com panies has the highest p redictive power due to ineffective large shareho lder and adm inistrative oversight of these executives.In contrast,the executive trading o f central state-owned companies has the lowest p redictive power,as the“quasi-o ff cial”status o f these executives lim its their m otivation to p rof t from their information advantage.

Large shareholder trading is signif can tly af ected by the ownership nature o f the shareho lders.There are signif cant dif erencesamong the large shareholders of state-owned versus p rivate-owned companies in terms of interest orientation and operational invo lvement.These dif erences can signif cantly afect themotivation and ability o f large shareholderswho are trading for p rof t.Large shareholders of p rivate-owned companies havea strongermotivation to benef t from trading in the secondarymarket.In contrast,large shareho lderso f state-owned companies cannot truly retain any p rof tgained from trading.Because the prof tsaredelivered to the SASAC,the large shareho lders’motivation to trade for pro f t isweaker than that of private-owned shareholders.In addition,because large shareho lders of central state-owned companies have po litical objectives, they cannot focusonly on p rof tability goalswhen trading.For example,during the 2008 fnancial crisis,large central state-owned listed com panies actively responded to the call for m ain taining f nancialm arket stability by increasing their ho ld ings of listed com panies.This type of“po litically d riven”pu rchase isnot p rof t d riven. The cen tral governm en t also requires state-owned shareholders to preserve their con trolling holdings in strategically im portant industries and gradually phase out their ho ldings in non-strategic,competing industries. Therefore,large shareho lders of state-owned companies cannot be comp letely free in their trading decisions. From the perspective o f operational involvement,large shareho ldersof private-owned companieso ften serve as executive directors or have fam ilym embers serve asexecutive directorsand managers so that they can be actively invo lved in major strategic decisions and daily operations.In contrast,large shareholders o f stateowned companies often delegate their control and decision-making authority to managers and indirectly acquire their in formation advantage from directorswho are appointed to the companies.From both perspectiveso f p rof tmotivation and operational invo lvement,large shareholders o f p rivate-owned companieshave m ore incentives to tim e their trad ing and m axim ize retu rns based on their ability to p redict futu rem acroeconom ic trends and identify system atic m arket value deviations.

Hypothesis4(effectof corporate governanceon trading predictability of large shareholders).Due to thegreater operational invo lvement and stronger p rof tmotivation,tradesmade by large shareholders of p rivate-owned com panies have a higher pred ictive power than thosem ade by large shareho lders o f state-owned com panies.

4.Sample selection and descriptive statistics

4.1.Sample selection

From a macro perspective,this study exam ines the ability of aggregate insider trading to p redict future market returnsbased on trading data taken from the Chinese A-sharemarket during January 2007 to August 2011.W emeasuremarket returns based on the Shanghai Stock Exchange A-share Index.We obtained executive trading data from the Exchangewebsite.1D ata are availab le on the Exchange’s website under“L isted Com pany Creditability Records”and in the co lum n titled“D irectors, Supervisors and M anagers Changes in Holdingso f Company Shares.”A fter removing executive transactionsbelow 20,000 yuan and adding up mu ltiple transactionsmade by the sam e executive on the same day,9384 transactions remained in the f nalsam pleof executive transactions.Largeshareho lder trading datawere taken from theW ind database. W e began by m anually coding the ownership natu re o f large shareholder transactions.W e also d if erentiated con tro lling shareho lders from o ther im po rtan t shareholders.Controlling shareholders include the largest shareholder and u ltim ate owner,and o ther im portan t shareho lders include shareholders w ith m o re than a 5%stock ownership(non-contro lling shareho lders)and shareholderswho are closely related to the controlling shareholder,even if they have less than 5%ownership.For the large shareho lder transaction data,we deleted transactions through a block trading platform2On April20,2008,theCommission issued its“Guidanceon Transferring Lifted Restrictive Sharesof Listed Com pany,”which requires that certain lifted restrictive shares be traded through the stock exchange block trad ing system when the trading shares in a month are projected to bem ore than 1%o f the total shares.The block trading p latform system is d iferent from the secondarym arket’s centralized auction trading system.M embership is required to participate in the b lock trading platform system.The transaction time is lim ited to weekdays from 15:00 to 15:30.The participantsaremostly institutionsand large shareholders.A transaction application is required and the trading price is not included in themarket index.Because only a few en tities are involved,the block trading p latform system is not active and lacks liqu idity.and retained transactions through the secondarymarket.W ethen deleted sm all transactions below 50,000 yuan and transactions w ith long trad ing periods(over 90 days from the transaction’s start to end).There are 5553 shareho lder transactions in the f nal sample.Due to the large diference in transaction scale,we exam ine the executive and shareho lder trading samples separately.

4.2.Empiricalmodel

W em easu re aggregate insider trading using the net pu rchase ratio(NPR).First,based on them on th o f the transaction,we combine all of the executive(large shareholder)buying and selling transactions conducted w ithin thesamemonth to obtain the totalm onth ly purchases(BUYk)and sales(SELLk)amounts for theexecutives(large shareholder).Using the follow ing formu la,we then calcu late the NPR:

NPR6 aggregates the insider trading activities performed over the past six months.It is a comprehensive m acro-level indicator,calculated as the net of the buying and selling amount divided by the total amount o f the two types o f activities.NPR can be calculated for three,six or twelvem onths.W emain ly p rovide the results for six months(NPR6).NPR6 is calculated for executives and large shareho lders separately. According to our hypothesis,NPR6iw ill be calculated for a specif c subgroup o f companies or subclass o f insiders(i),such as NPR6-p rivate,which is calculated using on ly the trading data from p rivate-owned companies.

We use a model sim ilar to that used by Lakonishok and Lee(2001).Themodel controls for the reversal feature of m arket retu rns over a long period to m ore accurately m easure whether aggregate insider trading has increm en tal p redictive power fo r fu tu re m arket retu rns.Ou r m odel also con trols fo r the inertia featu re o fm arket returns w ithin a sho rt interval.

Themodel is equated as fo llows:

Our insider trading data covers January 2007 to August 2011.A lthough therewould normally be 56NPR6tobservations,because data arem issing from the FR6 and FR12 calculations,the actual number of observations in the f nal sample is less than 56.

4.3.Descriptive statistics

Fig.1 shows the relationship between the aggregate insider trading NPR6 and the Shanghai A-share Index trend during thesampleperiod,and visually demonstrates thataggregate insider trading has theability to time themarket.NPR6 is calculated for both the executives and large shareholders for each month.As seen in Fig.1,the two lines representing NPR 6-Executive and NPR 6-Shareholder are alm ost parallel to each o ther, exhibiting sim ilar trends.In addition,aggregate insider trading activities are inversely related to the m arket index.W hen the m arket index is low(high)in a m onth,NPR 6 increases(decreases).For exam p le,during Septem ber to December 2008,when the index d rops to a historic low point o f 2000,NPR6 for executivesand large shareholders spikes.During the period of M ay to July 2010,when the index is at a localized low, NPR6 spikes again.In hindsight,we can see that insiders can successfu lly time them arket when trading in most cases.

Figure 1.Aggregate insider trading NPR and the Shanghai A-sharemarket index.

Fig.2 presents the relationship between aggregate insider trading NPR6 and futuremarket returns FR6. Net insider purchasing activitiesmove in the sam e direction as future market returns consistently during the samp le period.When NPR6 increases,there is a corresponding increase in FR6.On the contrary,when NPR 6 is close to-1,FR 6 ism ostly negative.Therefore,the NPR 6 ind icator can ef ectively pred ict fu turem arket returns.

As shown in Tab le 1,there are 6925 executive sales transactionsworth a total o f 25.525 billion yuan and 2459 executive purchase transactions worth a total o f 1.153 billion yuan.The ratio o f purchases to sales is 1:22.14.On average,the amount for each selling transaction isalsomuch higher than that for each purchasing transaction.Therefore,executive sales are both much higher in terms o f frequency and am ount than purchases.

Asshown in Tab le 2,there are 5048 large shareholder sales transactionsworth a total of 259.4 billion yuan and only 505 purchase transactions worth a total of 16.27 billion yuan.The ratio of purchases to sales is 1:16.3.Furthermore,the am ount for each sales transaction ishigher than that for each purchase transaction. The prevalence o f sales and the inactivity of pu rchases in the A-share m arket is due to the rem oval o f restrictions on non-tradab le shares.

Figure 2.Aggregate insider trading NPR6 and futuremarket returns FR6.

Table 1 Executive transaction analysis.

Table 2 Large shareholder transaction analysis.

Table 3 M arket returns and NPR analysis.

In Table3,themean andmedian o f the futuremarket returnsare both negativeand the standard deviation is large due to the sudden sw itch from a bu llmarket to a bear market during 2007–2011 and the long bear m arket period since 2008.The means and medians of the aggregate executive and shareholder trades are all negative,indicating net sales for both typeso f insiders.The NPRsaremore negative for shareho lders than for executives,suggesting that net sales are relatively stronger for shareho lders.

A ccord ing to ou r hypo thesis,NPR 6 m ust be calculated for the subgroup of com panies or subclass o f insiders.As shown in Tab le 4,the m ean and m ed ian o f NPR6-M anager are less negative than that o f NPR 6-D irecto r and NPR 6-Superviso r,ind icating that execu tive net sales are relatively weaker form anagers. Table 4 also shows that themean andmedian o f NPR6-Private aremore negative than those of NPR6-Local and NPR6-Central,indicating thatexecutivenet salesarestronger in private-owned companies.Them ean and m edian of NPR6 are sim ilar for the three dif erent company sizes.

As reported in Tab le5,themean andmedian of NPR6-Importantaremorenegative,indicating that thenet selling tendency is stronger for important shareho lders than for controlling shareho lders.Tab le5 also reports that themean andmedian of NPR6-Privatearemore negative than those of NPR6-Localand NPR6-Central, indicating that thenet saleso f largeshareho ldersarestronger in private-owned companies than in state-owned companies.Themean and median of NPR6 are sim ilar for the three dif erent company sizes.

Table 4 NPR for executive trad ing by executive type,ownership structure and company size.

Table 5 NPR for shareholder trading by shareholder type,ownership structure and com pany size.

5.Empirical results

5.1.The information content of aggregate insider trading

Tables 6 and 7 exam ine the predictive power o f aggregate executive trading for futurem arket returns. Table 6 does not con trol for historical m arket returns BR 6 and BR 12,and Table 7 does.In Tab le 6, NPR 6 exp lains25%o f FR 6 and 58.4%of FR 12.As such,aggregate execu tive trad ing exhibits fairly good p redictive power for fu turem arket retu rns.In Table 7,after contro lling for BR 6 and BR 12,NPR 3 and NPR 6 con tinue to p rovide signif cant increm en tal exp lanatory power of futurem arket returns,indicating that executivesuse theirmacro in formation advantage to time themarket.W hen comparing the resultso f NPR3 and NPR6,we can see that the t statistics and Ad j-R2values are higher for NPR6 than NPR3.Therefore,aggregation over the longer term increases the p redictability o f insider trading.

Furtherm ore,according to Tab le7,as the forecast period isextended from three to twelvemonths,thep redictability of NPR increases,which is refected in the increasing NPR t statisticsand Ad j-R2values.It isworth mentioning that the NPR6modelexplains66.9%of FR12.The coef cientsof BR6 are signif cantly positive in all of themodels,consistentw ith the short-term market inertia exp lanation.A ll of the BR12 coef cientsare signif cantly negative,consistentw ith the long-term m arket reversal exp lanation.

Tables8 and 9 exhibit the power of aggregate shareho lder trading to p redict futu rem arket returns.Regardlesso fwhether BR 6 and BR 12 are con trolled fo r,the NPR coef cientsare consistently signif can tly positive in allo f them odels,ind icating that aggregate shareho lder trad ing has a signif can t ability to pred ict futurem ar-ket returns.Furtherm ore,thep redictability of NPR increasesas the forecast period isextended(from three to twelvemonths).The NPR6 model explains almost 73%of the variation in FR12,suggesting that aggregate large shareho lder trading has a strong ability to predict futuremarket returns.

Table 6 Aggregate execu tive trading and futuremarket returns w ithout controlling for BR 6 and BR 12.

Table 7 Aggregate execu tive trading and futuremarket returns controlling for BR 6 and BR12.

Table 8 Aggregate shareholder trading and futurem arket returns without contro lling fo r BR6 and BR 12.

The fndings shown in Tables 8 and 9 dif er from p rior empirical evidence identif ed for the U.S.market. Literature focusing on the U.S.m arket has unanimously found that large shareholder transactions have low in formation content,mainly because ownership ismore dispersed and the holdingso f large shareho lders are generally low in the U.S.market.Large shareholdersare usually not invo lved in businessoperations,but rely on compensation contracts,the corporate m anagement job m arket,independent directors and other m eans to indirectly con trolm anagem en t.As such,they are m ostly passive shareho lders and have a m uch weaker in form ation advan tage than execu tives.However,due to the lack of investo r pro tection in the Chinese A-sharem arket,concen trated ownership is an alternative fo r shareholders’self-p rotection.Fu rtherm ore,due to the im perfections that co rporate m anagers and independent directo rs face in the job m arket and othergovernancemechanisms,largeshareholdersare usuallymoreactively involved in strategic decisionsand operations.Thisallows them to ensure that they have a f rm grasp of the control rights so that other shareho lders or managers do not harm their interests.These large shareholders have both the ability and willingness to obtain access to in formation.In particular,the largest shareholder o ften p laysa central role in corporate governance,as he o r she can appoin t board m em bers and m anagers.Therefore,his or her trad ing should have relatively high in form ation conten t.

Table 9 Aggregate shareholder trading and futurem arket returns controlling for BR 6 and BR 12.

In Fig.3,we use the estim ated coef cients of the NPR 6 m odel(includ ing BR 6 and BR 12)to calculate the predicted FR6 for eachmonth and then compare the predicted FR6 curvew ith the corresponding actual FR6 curve.Asshown in Fig.3,the predicted FR6 valuesbased on both the executiveand large shareholdermodels match closely w ith theactual FR6 value.The resu lts suggest that thepredictability of aggregate insider trading is very high.

5.2.Empirical tests of the direction of trade

To further determ inewhether buying or selling hasmore predictive power,we divide the transactions into“buy”and“sell”and replace the NPR with LNBUY 6 and LNSELL6,as shown in Table 10.Because theo riginal yuan am oun t o f the pu rchase and sales transactions fo r each m on th variesw idely,we use the natu ral logarithm of the yuan amount.

Figure 3.Comparison of actualand projected futuremarket returns.Note:True-FR6:theactual FR6(six-month-ahead market returns). M gpredict-FR6:predicted FR6(six-month-ahead market returns)using the executive NPR6model.Shpredict-FR6:predicted FR6(sixmon th-ahead m arket returns)using the large shareho lder NPR 6 model.

Table 10 Predictive power by trading direction for the execu tive sam p le.

Table10 compares theexecutives’purchasesand sales.The coef cientsof LNSELL6 aresignif cant in both m odels2 and 4,w ith orw ithout BR6 and BR12.In contrast,the coef cientso f LNBUY 6 arenot signif cant in m odels1 and 3.Next,we use the Vuong test(1989)to exam ine the signif cance of the diferences in R2values. The results suggest that aggregate executive selling activities havemore predictive power than buying activities.These f ndingsare diferent from p rior studies thatuseU.S.market data.Prior studiesof the U.S.market have consistently found that executive buying hasmore information content than selling.In the U.S.,equity incentives are o ften the largest com ponents o f executive com pensation.Execu tives often decrease their ho ldings for liquidity or diversif cation reasons,which d ilutes the in form ation conten t of selling.In con trast,their buying is usually d riven by private in form ation and has high in form ation con tent.However,executives in the Chinese A-sharem arket do not have equivalent equity incentives like their counterparts in the United States. Executive ho ldingsare also often very low,3The average stock ownership of board directors,executivesand supervisors in com panies listed on the Shenzhen Stock Exchangewas only 0.06%in 2008.so selling ismore likely to bebased on an information advantage. In contrast,executive buying is sparse and o ften occurs for diverse reasons.For example,executivesmay increase their holdings for political reasons,4Fo r exam p le,during the 2008 fnancial crisis,large cen tral state-owned listed com panies actively responded to the call to m aintain f nancialmarket stability by increasing their insider holdingso f listed companies.to increase the stock p rice shortly before a seasoned equity of ering5Com paniesmaywant to take advantageofmarketparticipantswho follow executivepurchases to increase theirho ldingsand therefore the stock price sho rtly before a seasoned equity ofering.or because doing so ismandated by equity incentive p lans.6Executive equity incentive programs sometimesm andate that annual bonuses be used to purchase a com pany’s stock.These typeso f purchasesundoubted ly dilute the in form ation content of executive buying(Zhu et al.,2011a,b).

Table 11 com pares the large shareho lders’pu rchasesand sales.A lthough the coef cien tso f the large shareholder LNBUY 6 aresignif cantand the Ad j-R2valuesofmodels1 and 3 arenot low,the higher coef cientso f LNSELL6 and the higher Adj-R2values in the LNSELL6models suggest that large shareholder selling has a much stronger p redictive power than buying.Large shareholders and especially those in central state-owned companiesmay respond to po litical calls to increase their holdingsduring an econom ic crisisor in strategically important industries.Largeshareholdersmay also increase their ho ldings during a control rights com petition. These non-in formation-driven buying activities dilute the information content of aggregate shareho lder buying.

5.3.The ef ect of an information hierarchy on the predictive power of insider trading

Executives can be divided into three categories:managers,directorsand supervisors.Each of these categories has a dif erent position on the information hierarchy.As shown in Table 12,based on the resu lts o f the Vuong tests and the signif cance o f the coef cients,regard less o f whether BR6 and BR12 are controlled for, managershave thestrongest trading p redictability,followed by directorsand then supervisors.These fndings coincidew ith the order o f the execu tive info rm ation hierarchy d iscussed earlier.Invo lvem en t in businessoperations and the in form ation hierarchy signif can tly af ect the pred ictive power o f insider trading.This fnding strongly suppo rts the argum ent that insider trading isbeyond sim p le contrarian trad ing and is strongerm ain ly due to themacro information advantage insiders have based on their operational involvement.

We now divide large shareholders into two categories:controlling and important shareholders.These two categories have diferent positions on the in formation hierarchy.Small shareholdersw ith less than 5%ownership and no relation to the largest shareho lders as a benchmark are also included,as they normally do not have an information advantage(see Tab le 13).

Regardlesso fwhether BR6 and BR12 are controlled for,thep redictability of controlling shareholder trading is signif cantly stronger than that o f important shareho lder trading.When BR6 and BR12 are not contro lled for,the Ad j.R2value o f the important shareholdermodel is on ly 6.2%,and that of the contro lling shareholder m odel is 35.5%.Because NPR 6-Sm allsh is not signif can t,sm all shareholders’trading activities have no info rm ation con tent.The descending p redictive power of trad ing belonging to con trolling,im portan tand sm all shareholders coincides w ith the order o f the shareho lders’info rm ation hierarchy.This f nd ing fu rther validates theargument that the predictability of insider trading isa resu lto f the insiders’macro in formation advantage.

Table 11 Predictive power by trading direction for the large shareho lder samp le.

Table 12 Predictive power of trading by executive type.

Table 13 Predictive power of trading by shareholder type.

5.4.Efect of corporate governance on the predictive power of insider trading

To investigate the ef ect of corporate governance on executive trading,we divide companies into three categories:p rivate-,local state-and central state-owned.As shown in Table 14,based on the signif cance of the coef cientsand the Vuong testsof the Adj.R2values,the p redictive power o f executive trading is theweakest for centralstate-owned companies.Furthermore,it isslightly higher for localstate-owned companies than for private-owned companies,but the dif erence isnot signif cant.Tab le 15 exhibits the resu ltsof a sensitivity test where theNPR ratio iscalculated using the number o f transactions instead of theamountof yuan.The regression resu lts are consistentw ith those shown in Tab le 14.

The preceding f ndingsare partly consistentw ith Hypothesis3.Asexpected in Hypothesis3,given thehidden but strong adm inistrative contro l,the executives in cen tral state-owned com panies are restrained from trading fo r p rof t.The“quasi-o f cial”status of these execu tivesweakens their p rof tm o tivation,forcing them to give up oppo rtunities to p rof t from their m acro info rm ation advantage.This exp lains the relatively low predictive power o f aggregate trading by this type o f executive.

The p redictive power o f executive trading between local state-and p rivate-owned companies isnot significantly dif erent.Because trading based onmacro information advantagewou ld notharm the interestsof large shareholders as seriously as trading based on company-level p rivate in formation,large shareho lders do not monitor it,even in private-owned companies.Therefore the dif erence in shareholder oversight between local state-owned companiesand private-owned companieshasno signif cant efect on executive trading based on a macro in formation advantage and the p redictability of executive trading.However,as reported by Fidrmucetal.(2006),the diference in largeshareholder oversightmay afect executive trading based on company-level insider inform ation.W hen there is strong large-shareho lder oversight,the pro f tability of executive trading is lower than that for executives in companiesw ith dispersed ownership.

Table 14 Efect of corporate governance on the predictive power of executive trading.

Table 15 E fect of corpo rate governance on the pred ictive power of executive trading(sensitivity test).

Table16 exam ines the efectof corporategovernanceon largeshareholder trading.The predictivepower o f large shareho lder trading in p rivate-owned companies is strongestwhen historical returns are not contro lled for.As shown in model1,when theNPR6-Private isused to predict FR6,themodel can exp lain 54.4%o f the portfo lio returns.In contrast,the NPR6-Localm odel(2)and the NPR6-Centralmodel(3)explain on ly 18.3% and 34.4%of the variation in the portfo lio returns,respectively.W hen historical returnsare con trolled fo r,the NPR 6-Privatem odel(4)stillhashigher exp lanato ry power than the NPR 6-Localm odel(5).H owever,there is no signif can t diference in the Ad j.R2values betw een the NPR 6-Private m odel(4)and the NPR 6-Cen tral m odel(6),partially due to the increased exp lanation power of BR6 in the NPR6-Centralmodel(6).These f ndingsare consistentw ith Hypothesis4.Because largeshareho lders in private-owned companieshave stronger p rof tmotivationsand active roles in business operations,the in formation content of their trading is signif cantly higher than that of large shareholders in state-owned companies.

5.5.Efect of company size on the predictive power of insider trading

Several studies exam ine the efect of company size on the in formation content of insider trading(Seyhun, 1992;Lakonishok and Lee,2001).To be consistent w ith those studies,we exam ine the size efect and the results are shown in Tables 17 and 18.

Thep redictive power of large company executive trading is signif cantly lower than that of sm all-andmedium-sized com panies.This fnd ing is consistentw ith the prior literatu re.Because large com panies tend to of er m ore public d isclosures and can attractm ore atten tion from the pub lic and p rofessional investo rs,the chance o f their stock beingm isp riced is sm aller.As a result,execu tives in large com panies have less private in form ation and fewerm ispricing oppo rtunities.In add ition,the literatu re consistently fnds that the executive tradingo f large companies,whichmay besubject tomore rigorousm onitoring,has less information content than that o f small companies.

Table 16 Efect of corpo rate governance on the pred ictive power of large shareholder trading.

Table 17 Efect of com pany size on the predictive power of executive trading.

Table 18 E fect of company size on the predictive power of large shareholder trading.

W ithout controlling for historical returns,the predictive power of large shareho lder trading in large companies is signif cantly smaller than that ofmedium-sized companies.When BR6 and BR12 are controlled for, company size hasno ef ect on the predictive power o f shareho lder trading.Thus,the ef ect of company size is m ore noticeab le in executive trading than in shareho lder trading.

We conducta variety of sensitivity tests.First,weuse thenumber of shares traded and thenumber of transactionsasalternativem easures to the am oun t o f yuan in calcu lating the NPR index,and the conclusions stay the sam e.Second,we use FR 9 and FR 12 as alternative dependen t variab les,and the em pirical resu ltsare consistentw ith the use of FR6.Third,given the signif cant diferences in turnover,company size and listing history between themain board,small-cap and grow th enterp risemarkets,we remove the small-cap and grow th enterprisem arkets,leaving only themain board market,and f nd that allo f the regression resultsare robust. W euseanothermethod to consider the efectsof dif erentmarketson the results.Weuse thesmall-capmarket index to calcu late the future and p riormarket returns,and f nd that insider trading in companies listed on the sm all-cap and grow th enterp risemarkets also has strong p redictability for future small-cap market returns. Fourth,Zhang and Zeng(2011)fnd that the relativesof executives can serveasan alternate and imp licit insider trading method to avoid strictmonitoring and that relative trading has a sim ilar tim ing ability.We conduct a sensitivity test by using NPR 6-Relatives in the m odels and fnd that relative trad ing cannot p redict fu turem arket retu rns.A lthough relatives do no t seem to have am acroeconom ic in form ation advan tage,their in fo rm ation advantagem ay be lim ited to the com pany level.

6.Conclusions

U sing Chinese A-sharem arket data taken from January 2007 to August 2011,we exam ine the predictive power of aggregate insider trading on futuremarket returns from a macro perspective.W e fnd that aftercon tro lling fo r the con trarian investm ent strategy,aggregate execu tive(large shareholder)trad ing from the prior six months p redicts 66%(72.7%)of market returns twelvemonths in advance.The predictive power is caused by insiders’in formation advantage in p redicting macroeconom ic trends and detecting deviations in stock market valuation.Compared w ith the results found in the U.S.market,aggregate insider trading in the A-share market has a stronger predictability,which may high light the severity of insider trading in the A-sharesmarket or indicate that them arketmay bemore prone to systematic valuation bias to create more opportunities for insider trading.Furthermore,we fnd that the information hierarchy and corporate governance have a signif cant efect on the in formation content o f aggregate insider trading.The trading o f insiderswho aremore invo lved in business operations and at higher positions in the information hierarchy have signif can tly m o re p redictive power.The corporate ownership structure also has a signif cant efect on insider trading.The p redictive power o f executive trad ing is weakest for central state-owned com panies. A lthough there is a lack of ef ective large-shareho lder oversight,the im p licit adm inistrative con trol in cen tral state-owned companies becomes an alternativemechanism for curbing executive trading.Large shareho lders in private-owned companieshave stronger incentives to benef t from insider trading and havem ore power to contro l businessoperations.As such,the predictive power of their trading is signif cantly higher than that o f large shareho lders in state-owned companies.

The relationship between insider trading and in formation advantage and how external regulation and corporate governance af ect insider trading have been important issues for regu lators,academ ia and investors. This study com plements p rior studies by exam ining an emergingmarket and investigating how institutional contextand corporate governance af ect insider trading.It provides interesting f ndings that are diferent from those observed in m atu rem arkets.For exam p le,it fnds that the trading activities o f large shareho lders in the Chinese A-share m arket havem uch higher inform ation con tent than those in m aturem arkets.The fndings difer mainly due to China’s institutional background and corporate governance.Due to theweak investor protection in the Chinese A-sharemarket,large shareho lders tend to be actively invo lved with businessoperationsand thusbecome important insiderswho cannot be ignored.This study also further exam ines the relationship between corporate governance and insider trading.We f nd that private-and state-owned companies face diferent agency p rob lem sand governance structures,resulting in signif cant dif erences in insider trading behavior.The in form ation content o f the trading of executives in central state-owned companies is lowest because they are subject to imp licit but strong adm inistrative control.This suggests that executive trading is in fuenced by not only large shareho lders’oversight,but also another implicit inform almonitoring system such as the adm inistrative contro l o f an o rganization.

Acknowledgments

This paper is supported by the follow ing p rojects including“Trading Bans Po licy on Insider Trading:Policy Ef ectiveness and Econom ics Consequences”supported by NSFC(National Natural Science Foundation of China)(No.71302059),“Research on Controlling Shareholder’s Trading Behavior and Regu lation Imp lication”supported by Research Foundation for Young Teachers by M inistry o f Education o f China(No. 11YJC790313),and“Research on Executive Trading Behavior and Regulation Imp lication”supported by Zhejiang Provincial Natural Science Foundation o f China(No.LY 12G 02022).

Cai,N.,Wei,M.,2009.Earningsmanagement associated w ith large shareholder selling.Audit Res.2,40–49.

Chen,Z.,Guan,Y.,Ke,B.,2011.Fakestock op tions?An empiricalanalysiso fmanagerialstock option com pensation in state-controlled Chinese f rm s listed in Hong Kong.＜http://www.ssrn.com＞.

Fidrmuc,J.P.,Goergen,M.,Renneboog,L.,2006.Insider trading,new s releases,and ownership concentration.J.Finance 6,2931–2973.

Friedrich,S.,G regory,A.,Tonks,I.,2002.Short-run returns around the trades of corporate insiders on the London stock exchange. Europ.Finan.M anage.8,7–30.

Givoly,D.,Palmon,D.,1985.Insider trading and the exploitation o f inside information:some em pirical evidence.J.Bus.58(1),69–87.

Gunny,K.,K e,B.,Zhang,T.C.,2008.Corporate governance and opportunistic insider trade.In:China International Conference in Finance.

Jiang,X.Q.,Zaman,M.A.,2007.Aggregate insider trading and the predictability ofm arket returns:contrarian strategy or m anagerial tim ing?＜http://ssrn.com/abstract=970987＞.

Jeng,L.A.,M etrick,A.,Zeckhauser,R.,2003.Estimating the returns to insider trading:a performance-evaluation perspective.Rev.Econ. Stat.85,453–471.

Lakonishok,J.,Lee,I.,2001.A re insider trades info rm ative?Rev.Financ.Stud.14,79–111.

Li,J.,W ang,D.,Zhang,T.,2011.An analysis of the efects and m otivations o f large shareholder pu rchasing.Chin.Soc.Sci.4,95–110.

Lin,Z.,Qu,W.,2010.The pricing and timing of large shareholder selling–an empiricalstudy based on block trades in the Shanghaiand Shenzhen stock exchange.Secur.M arket Herald 10,71–77.

Pettit,R.R.,Venkatesh,P.C.,1995.Insider trad ing and long-run return perfo rm ance.F inanc.M anage.24(2),88–103.

Ravina,E.,Sapienza,P.,2010.W hat do independent directors know?Evidence from their trading.Rev.Financ.Study 23,962–1003.

Seyhun,H.N.,1990.Overreaction or fundamentals:some lessons from insiders’response to themarket crash o f1987.J.Finance45,1363–1388.

Seyhun,H.N.,1992.W hy does aggregate insider trading predict future stock returns?Quart.J.Econ.107,1303–1331.

Seyhun,H.N.,1998.Investm ent Intelligence from Insider Trading.M IT Press,pp.107–134.

Shen,H.,Li,J.,Tu,Y.,2011.An em piricalanalysiso f the factorswhich infuence large shareholder selling.Chin.Ind.Econ.1,109–119.

Wu,Y.,W u,S.,2010.Large shareholder selling and tunneling.Chin.Ind.Econ.5,122–130.

Zeng,Q.,2008.Can insiders choose the tim ing of trade correctly in China?Finance Res.10,117–135.

Zhang,J.,Zeng,Y.,2011.A study on insider relative trading.Finance Res.3,121–133.

Zhu,C.,Yao,Z.,Jevons,L.,2011a.Can executive trading p redict future stock returns?M anage.W orld 9,141–152.

Zhu,C.,Jevons,L.,Chen,C.,2011b.A study on market reaction and the tim ing o f large shareholder selling in the A-sharemarket.J. Zhejiang Univ.41(3),159–169.

*Corresponding author.

E-mail address:zhuchafen@163.com(C.Zhu).

http://dx.doi.org/10.1016/j.cjar.2014.08.001

1755-3091/?2014 Production and hosting by Elsevier B.V.on behalf of China Journalof Accounting Research.Founded by Sun Yat-sen U niversity and City University of H ong K ong.