UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Unseasoned equity issues and institutional trading : an empirical analysis Vandemaele, Sigrid N. 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1994-0645.pdf [ 1.09MB ]
Metadata
JSON: 831-1.0087461.json
JSON-LD: 831-1.0087461-ld.json
RDF/XML (Pretty): 831-1.0087461-rdf.xml
RDF/JSON: 831-1.0087461-rdf.json
Turtle: 831-1.0087461-turtle.txt
N-Triples: 831-1.0087461-rdf-ntriples.txt
Original Record: 831-1.0087461-source.json
Full Text
831-1.0087461-fulltext.txt
Citation
831-1.0087461.ris

Full Text

UNSEASONED EQUITY ISSUES AND INSTITUTIONAL TRADINGAN EMPIRICAL ANALYSISbySIGRID N. VANDEMAELEBachelor of Science in Commerce, Katholieke Universiteit Leuven, 1990A THESIS SUBMITI’ED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESFaculty of Commerce and Business AdministrationWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAOctober 1994© Sigrid N. Vandemaele, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.,r, ji4/4apRr*IRnI of aiic/ /3iWii iZ%t7The University of British ColumbiaVancouver, CanadaDate//9f9DE-6 (2/88)AbstractIn the first part of this study, the focus is on the long-term performance of two IPO samples withissue years 1990 and 1991, respectively. The performance of particular IPO portfolios has beenstudied in a series of articles. There seems to be a consensus that investors in IPOs earn lowreturns. For the sample of 70 firms with issue year 1990, I find evidence of underperformanceconsistent with Ritter’s (1991) findings: the three-year buy-and-hold return on a portfolio formedby investing $1 in each IPO at the end of the first listing date is 23.2%, compared to anequivalent return for the index of 45.8% and to an equivalent return for the matched firmportfolio of 48.6%. In addition to the standard buy-and-hold analysis, I show that thisunderperformance disappears if the IPO portfolio is rebalanced to equal weights on eachanniversary of the issue date. I conclude that the difference in performance of a buy-and-holdportfolio and of a portfolio with yearly rebalancing could be due to overreaction by investors inIPOs : after a certain period, poor performers tend to be undervalued (possibly due tooverreaction) and/or good performers tend to be overvalued (possibly due to overreaction).Rebalancing to equal weight is equivalent to putting more weight in past losers and less weightin past winners. If there is a tendency to mean reversion, the rebalancing effect on the portfolioreturn will be positive. In contrast, however, for the sample of 224 IPOs of 1991, I do not findevidence of underperformance. Neither is there evidence of a positive rebalancing effect on theIPO portfolio return.In the second part of the study, I focus on the trading behavior of institutions in the aggregatein IPO markets. In particular, I examine how the quarterly changes in institutional holdings11relate to quarterly past, current and future returns. I find evidence of a strong positive feedbacktrading strategy. The positive association between current adjusted performance and changesin holdings could be evidence of a rational update in expectations by institutions on the lessinformed side of the market. The pattern of buying current winners and selling current loserscould be strengthened by a concern about reputation or over the ‘fiduciary duty’-aspect of amoney managers’ job. The positive association between past adjusted performance and changesin institutional holdings could be driven by a belief that past trends are likely to continue.However, there is evidence of at least part of it being the result of an intertemporal associationbetween money managers’ trades : managers, observing the holdings of their colleagues at theend of the quarter, don’t want to miss the boat and buy a positive stake the next quarter. Theevidence of a positive association between current (past) positive adjusted performance andchanges in institutional holdings is stronger than has been documented by Lakonishok et al(1992) for the average common stock institutions trade in. The observed trading patterns couldbe destabilizing in nature and could result in an undervaluation of past underperformers and anovervaluation of past overperformers. To attribute the finding of a positive rebalancing effecton the return of the IPO portfolio to institutional trading behavior in IPOs, we should have anidea of the importance of institutions’ trading in IPO markets. The percentage of IPO sharesoutstanding that institutions hold at the end of the first and second post-issue year may give anindication that their presence in IPO markets could be important. The results, however, do notallow us to make any conclusive statement about a possible causal relationship betweeninstitutional IPO trading and the evidence of overreaction in IPO prices.111Table of Contents PageAbstract iiTable of Contents ivList of Tables viChapter I : Literature Review and Statement of Objective 11.1 An IPO Puzzle 11.2 The Rise of Institutions 31.3 Objective and Main Results of This Study 9Chapter II: Data Description 13Chapter III Documentation of Long Run IPO Performance 17111.1 Evidence 17111.2 Discussion 22Chapter IV : Institutional Trading Patterns in IPOs 24IV. 1 Analysis of Quarterly Aggregate Institutional Holding Changes and Returns 24IV. 1.1 Descriptive Statistics 25IV.1.2 Changes in Institutional Holdings and Contemporaneous Returns 27IV. 1.3 Changes in Institutional Holdings and Lagged Returns 31IV. 1.4 Combining Contemporaneous, Lagged and Future Returnsin One Regression 35IV.1.5 Discussion of The Results 38ivIV.2 Institutional Trading One Year after The IPO 40IV.3 Exploring Portfolio Returns 40IV.4 Concentration versus Multiple Institutions Hypothesis 44Chapter V : Conclusion 50Bibliography 53VList of Tables PageTable I : Sample Offers Categorized by Industry 15Table II : Distribution of Sample Offers across Size DedilesDefined by The Amex-NYSE Stock Universe 16Table ifia: One-, Two- and Three-Year Buy-and-Hold Returns (%) 19Table Tub : One-, Two- and Three-Year Buy-and-Hold Returns (%) 19Table IVa : Annual Percentage Returns per Post-Issue Yearon Yearly Rebalanced Portfolios 21Table IVb : Annual Percentage Returns per Post-Issue Yearon Yearly Rebalanced Portfolios 21Table V : Spearman Correlation Coefficients between The Changein Aggregate Institutional Holdings during The Second Quarterand Concurrent, Lagged and Future Adjusted Returns 25Table VI: Spearman Correlation Coefficients between The Changein Aggregate Institutional Holdings during The Fourth Quarterand Concurrent, Lagged and Future Adjusted Returns 26Table VII : Spearman Correlation Coefficients between The Changein Aggregate Institutional Holdings during The Sixth Quarterand Concurrent, Lagged and Future Adjusted Returns 27Table VIII : Regression of The Rank of The Changein Aggregate Institutional Holdings over The Second Quarteron Contemporaneous Performance Dummies 32Table IX : Regression of The Rank of The Changein Aggregate Institutional Holdings over The Fourth Quarteron Contemporaneous and Lagged Performance Dummies 33Table X : Regression of The Rank of The Changein Aggregate Institutional Holdings over The Sixth Quarteron Contemporaneous and Lagged Performance Dummies 34viTable XI: Regression of The Change in Aggregate Institutional Holdingsfrom The End of Quarter i to The End of Quarter j onContemporaneous, Lagged and Lead Returns Combined 37Table XII : Average Change in Percentage Ownership Held by Institutionsover A Quarter Condtional on Current and Past Performance 38Table XIII: Regression of The Rank of the Change in Institutional Holdingsover The First Quarter of The Second Post-Issue Year on The Amex-NYSEPerformance Dummies over The Past Year 41Table XIV : Average Level of Percentage Ownership Held by Institutions atThe End of The First And Second Post-Issue Years by Past Performance Quartiles 42Table XV : Regression of The Rank of The Change in The Number of InstitutionsHaving a Positive Stake in The Firm over The Second Quarter onContemporaneous Performance Dummies 47Table XVI : Regression of The Rank of The Change in The Number of InstitutionsHaving a Positive Stake in The Firm over The Fourth Quarter on Contemporaneousand Lagged Performance Dummies 48Table XVII : Regression of The Rank of The Change in The Number of InstitutionsHaving a Positive Stake in The Firm over The Sixth Quarter on Contemporaneousand Lagged Performance Dummies 49viiChapter I : Literature Review and Statement of Objective1.1 An IPO PuzzleAt a given moment in time, wealth constraints and risk aversion may make managers of aclosely held firm decide to go public. New claims against the firm’s assets are marketed to thepublic, usually via an underwriting done by investment bankers in the primary market. Newshares are allocated to investors on the basis of interest shown, and in the case ofoversubscription, the investment bankers involved in the issue process apply rationing rules todistribute the shares’. When these new shares are listed on an exchange, any investor canacquire (or get rid of) the desired amount through the regular market process. As potentialportfolio holdings, these newly issued shares are different, in a certain sense, from othercommon equity assets for two reasons. First, their valuation is perceived as being more difficultthan the valuation of seasoned securities: estimating earnings is no simple task for an establishedcompany, let alone for a firm with no track record in public markets. Second, there is a seriesof studies in the (academic) literature that document that investors in initial public offerings(IPOs) earn low returns. Ritter (1991), for instance, documents that 1526 IPOs marketed in theU.S. from 1975 to 1984 exhibit significant negative abnormal performance of 26 percent in thethree years following the offering date. The measure used by Ritter is a 3-year buy-and-holdreturn : if one had bought a portfolio of IPOs at the closing price on the first day of trading andheld that portfolio for the subsequent 3 years, the portfolio would have been worth 26% less‘I’m describing the U.S. system here. See Smith (1986) or Beatty and Ritter (1986) for amore detailed description of the underwriting process.1than if it had been invested in a portfolio of comparable firms matched by size and industry.Ritter also documents the fact that younger companies and companies that go public in highvolume years perform worse than average. At least 3 academic studies, plus a series of articlesin Forbes magazine, dealing with the long-run performance of IPOs marketed in the U.S. werepublished before Ritter’s study. Stoll and Curley (1970), focusing on 205 small offers, find that,“in the short run, the stocks in the sample show remarkable price appreciation... .In the long run,investors in small firms do not fare so well....” (pp. 314-315.) Ibbotson (1975), using oneoffering per month for the 10-year period 1960-69, computes excess returns on IPOs with anoffer price of at least $3.00 per share. He concludes that the “results generally confirm thatthere are no departures from market efficiency in the aftermarket.” (p. 265.) However, he doesfind evidence that there is “generally positive performance the first year, negative performancethe next 3 years, and generally positive performance the last [fifth] year,” (p.252), although thestandard errors of his estimates are high enough to make it difficult to reject market efficiency.He reports that IPOs underperform relative to a matched sample of firms by an average ofapproximately I % per month in the second through fourth years of public trading, with positiveexcess returns in the first and fifth years. Buser and Chan (1987) evaluate two-year performanceof over 1078 NASDAQ/National Market System eligible initial public offerings in 1981-1985.Their sample has a positive average initial return of 6.2% and a mean 2-year market-adjustedreturn of 11.2% (exclusive of the initial return) where they use the NASDAQ Composite Indexfor their market adjustment. An article in the December 2, 1985 Forbes magazine reports theresults of a study that finds, for the period from January 1975 through June 1985, that IPOsunderperform the market in the long run. Stern and Bornstein (1985) find as reported on p. 1522that “from its date of going public to last month, the average new issue was down 22% relativeto the broad Standard & Poor’s 500 Stock Index.” Forbes analyzes 1922 issues with an offeringprice of $1.00 or more. Unlike academic research, which typically uses event time, Forbes usescalendar time, so the excess returns are computed over a period of anywhere from 10 years toa few months.In summary, Stoll and Curley (1970), Ibbotson (1975), Stern and Bornstein (1985) and Ritter(1991) present evidence which suggests that at some point after going public the abnormalreturns on IPOs may be negative. Only the Buser and Chan (1987) study does not find evidenceof negative aftermarket performance.There are three potential explanations for the negative abnormal long-run performance of IPOs:i) risk mismeasurement; ii) ex-post bad luck; and iii) irrational overoptimism for the IPOs at thetime of the initial public offering. Ritter rules out risk mismeasurement : although the exactmagnitude of the underperformance of issuing firms depends on the chosen risk benchmark, theIPOs in Ritter’s sample underperform all of a number of reasonable benchmarks. Given thewidespread occurence of the phenomenon across industries and because returns are lowestfollowing heavy issue periods, which would suggest that issuers take advantage of ‘windows ofopportunity’, Ritter favors the explanation of persistent misvaluation at the time of going public.Consistent with this interpretation, Mikkelson and Shah (1993) report that for 284 firms thatwent public in the years 1980 through 1983, the average operating cash flow to assets ratio fallsdramatically between the year before going public and subsequent years. Thus, investors appearto value the firms going public as if their earnings per share will continue to grow rapidly, whenin fact the rapid earnings growth prior to going public often ends at the time of the IPO.3Investors seem to systematically misestimate the autocorrelation of earnings growth.Similar results on IPO long-run underperformance have been obtained for other countries.Keloharju (1993), for example, shows that Finnish IPOs underperform the Helsinki StockExchange value-weighted index by 26 % over three years and Levis (1993) reports similarpatterns of long-run underperformance for UK IPOs, although the average underperformance ofnew UK issues is not as excessive as that recorded for their U.S. counterparts, and ranges from6% to 12% for a three year holding period, depending on the benchmark used.1.2 The Rise of InstitutionsDuring recent years, institutional investors have become an ever growing portion of the totalinvestment population. In 1955, institutions owned 23% of US equities; by 1990 the institutionalshare had grown to 53% and the share of individual investors had fallen to only 47%. In 1990,institutional trading volume accounted for 70% of total trading volume on the New York StockExchange (Lakonishok et al [19921). Institutional investors are large investors who exercisediscretion over the investments of others, and include investment advisors, investmentcompanies, banks, insurance companies, private and public pension funds, colleges anduniversities, and foundations, among others2. Because of institutions’ overwhelming presencein asset markets, it has become an important task for capital market researchers to understandinstitutional and other features that affect money managers’ trading behavior and their potentialimpact on asset prices.2 For purposes of SEC filing requirements, institutional investors are defined as entities,other than natural persons, with investment discretion of at least $100 million in equitysecurities.4There are several reasons why the trading behavior of institutions may differ from that ofindividual investors. First, institutions act as agents for others, and hence their incentives maydiffer from those of individual investors. If fund managers believe that a “better-looking”portfolio will increase investment in the fund and, consequently, increase their compensation ifit is based on total fund assets, they may engage in “window dressing” by selling securities thatrelatively underperformed in the past quarter and buying securities that did relatively well3. Thebelief that investors prefer investing in a portfolio of stocks that have not performed poorlypresumes that managers assume investors believe that returns are correlated. Anotherinstitutional detail that could influence institutions’ investment behavior is that, typically,institutions are evaluated against a common benchmark, and hence, against each other. To avoidfalling behind a peer group by following a unique investment strategy, they have an incentiveto hold the same stocks as other money managers. Scharfstein and Stein (1990) develop aformal model and show that, in certain circumstances, it can be rational from the perspectiveof money managers who are concerned about their reputation in the labour market, to simplymimic the investment decision of other managers, thereby ignoring substantial privateinformation. Institutions may also all react to the same information set, such as changes individends or analysts’ recommendations and pile into and out of the same stocks at the same time“Window dressing” is a response to the requirement that institutions are required to reportholdings only at the end of the quarter, and refers to the trading practice, as described byLakonishok et al [1992], in which institutions buy and sell securities at the end of the quarterto change the market’s perception of the trading strategy of the institution, either on the basisof the performance of the portfolio or the riskiness of the shares held.5(or “herd”, as it is often called) as a result4.Second, institutions differ from individuals in that the former generally serve in a fiduciarycapacity, and are held to a prudent-person standard in their investment decisions5. The PrudentMan Rule, the traditional trust law standard of reasonable care, requires that a fiduciary“conduct himself faithfully and exercise a sound discretion” based on observations of “how menof prudence, discretion and intelligence manage their own affairs in regard to the permanentdisposition of their funds” (Droms [19921). Therefore, if performance affects the perceivedriskiness of a stock (e.g. a large loss is associated with increased bankruptcy risk), institutionsmay feel an obligation to sell the stock of firms that have performed poorly.A third reason why institutional investors may differ from individual investors is the size of theirholdings. Size could have implications for both monitoring activities by institutions and theirlevel of informedness. As large shareholders, institutions may have more powerful incentivesto monitor managers and enhance shareholder wealth. Shleifer and Vishny (1986) and Admati,Pfeiderer and Zechner (1991) provide formal models in which a single large shareholdermonitors management because the return on its shares is sufficient to cover monitoring costs andavoid the free rider problem faced by atomistic shareholders. Institutional monitoring andactivism may take some time to affect performance. However, in an efficient market, if thereThis is also true for individuals; they may also react to the same exogeneous signals. Butbecause the signals reaching institutions are typically more correlated than the signals that reachindividuals, institutions might herd more.- Fiduciary responsibilities fall into two generic categories-the duty of undivided loyalty,covering rules and regulations designed to minimize conflicts of interest, and the standard ofreasonable care. The fiduciary responsibilities of trust managers are drawn in the main fromstate law on personal trusts, from case law, from Scott’s Law of Trusts and from the AmericanLaw Institute’s Restatement of Trusts. (see Droms [1992] for further details)6are monitoring benefits associated with increased institutional ownership, one would see a pricejump reflecting these future benefits at the moment of institutional acquisition of the stock.Institutions may also have better access to information, either because of their access tomanagement as large shareholders or because their size provides justification for largerinvestment in research. Recently, Brennan and Cao (1994) developed a noisy rationalexpectations model of the type of Hellwig (1980) where the assumption of informationasymmetry is sufficient for the less informed agents to optimally choose a trend chasing strategy,while the rational, better informed agents act as contrarians. Positive feedback trading, or trendchasing, is the strategy of buying winners and selling losers. The opposite of trend chasing isbeing contrarian, a strategy of buying those assets whose price goes up and selling those assetswho experience a price decline. In the paper of Brennan and Cao, the measure of an agent’sinformedness is the precision of his private signal about the final asset payout. As is commonin the Hellwig type rational expectations model, both the private signal and the price enter theagents’ demand equations. Less informed investors, however, rely more on the signal revealedby the price, and therefore buy when the price goes up and sell when it goes down. Betterinformed agents do the opposite: they act as contrarians. Based on the Brennan and Cao model,if institutions have access to superior information, one would see them buy assets whose pricesmove down and sell assets whose prices move up. One would see the opposite trading patternif, in a particular market, institutions are on the less informed side.6Little empirical evidence exists on the trading behavior of institutions. Kraus and Stoll (1972),6 Brennan and Cao (1994) are the first to demonstrate that feedback trading strategies canbe a result of fundamental analysis, as opposed to technical analysis. In the rest of the paper,I will talk about fundamental feedback trading as opposed to technical feedback trading.7one of the first studies on this topic, address the question of ‘parallel trading’ (which is the sameas herding) by institutions using data from a SEC study of institutional investors on monthlychanges in holdings. They find little evidence of herding and weak evidence of acontemporaneous relationship between price changes and excess demand by institutions. In amore recent study, Lakonishok, Shleifer and Vishny [1991] examine the relation betweencontemporaneous and lagged stock returns and changes in holdings to test for herding andwindow dressing in a sample of 769 equity pension funds (representing about 18% of all activelymanaged pension funds) which voluntarily report quarterly holdings to SET mc, a largeconsulting firm in financial services for institutional investors. Their results suggest that pensionfund managers herd relatively little in their trades in large stocks. There is some weak evidence,however, of more herding in smaller stocks. Lakonishok, Shleifer and Vishny find weakevidence of an association between returns and institutional trading, limited to the smallest twoquintiles of NYSE/AMEX stocks. They interpret this last result as window dressing : moneymanagers dump losers among small stocks to dress up their portfolios. The strategy of dumpingsmall stock losers makes sense if sponsors are less sensitive to holdings of poorly-performingblue chips than to holdings of poorly-performing unknown stocks.Previous tests of institutions’ degree of informedness have focused on the performance ofportfolios managed by professional portfolio managers. The vast majority of studies has focusedon the performance of mutual funds. Earlier studies, such as the classic papers by Treynor(1965), Sharpe (1966), and Jensen (1968) find that mutual funds do not systematicallyoutperform benchmark portfolios. In a more recent paper, Lehman and Modest (1987)document that the measures of abnormal mutual fund performance are sensitive to the benchmark8chosen to measure normal performance. Other papers, for example Hendricks, Patel andZeckhauser (1993) suggest that at least some mutual funds outperform benchmark portfolios, andthat there is some short-run persistence in relative performance of mutual funds. To date, theliterature is inconclusive as to whether the average mutual fund manager provides superiorinvestment talent.Empirical studies on institutional monitoring activities have focused primarily on institutionalvoting patterns in proposals that reduce shareholder wealth. Brickley, Lease and Smith (1988)document that institutional investors vote more actively on antitakeover amendments and findthat opposition by institutions is greater when the proposal is detrimental to shareholders’interests. Agrawal and Mandelker (1990) find a positive relation between the shareholder wealtheffects of antitakeover charter amendments and institutional ownership. So, the evidence isconsistent with there being some wealth increasing monitoring activities associated withinstitutional ownership.1.3 Objective and Main Results of This StudyThe objective of this study is two-fold. In the first part of the study, I take a closer look at thereturns of a sample of US firms that went public in 1990 and 1991. I hope to shed someadditional light on the previously described IPO puzzle. If IPOs are systematically overpricedrelative to their long-term value, this would suggest that the long-term investors in IPOs do notevaluate all available information when they purchase the IPOs. Such a conclusion wouldundermine the widely held belief that the market is efficient, as well as bring into question themotivations of the long-term investors in IPOs. For the sample of 70 firms with issue year91990, I find evidence of underperformance consistent with Ritter’s (1991) findings: the three-year buy-and-hold return on a portfolio formed by investing $1 in each IPO at the end of thefirst listing date is 23.2%, compared to an equivalent return on the index of 45.8% and to anequivalent return for the matched firm portfolio of 48.6%. In addition to the standard buy-and-hold analysis, I show that the underperformance disappears if the portfolio is rebalanced to equalweights on each anniversary of the issue date. I conclude that the difference in performance ofa buy-and-hold portfolio and of a portfolio with yearly rebalancing could be due to overreactionby investors in IPOs: after a certain period, poor performers tend to be undervalued (possiblydue to overreaction) and/or good performers tend to be overvalued (possibly due tooverreaction). Rebalancing to equal weights is equivalent to putting more weight in past losersand less weight in past winners. If there is a tendency to mean reversion, the rebalancing effecton the portfolio return will be positive.In contrast, for the sample of 224 IPOs of 1991, I do not find evidence of underperformance.Neither is there evidence of a positive rebalancing effect on the IPO portfolio return.In the second part of this study, I take a closer look at the trading behavior of institutionalinvestors during the 2 years following the IPO event. The average of the aggregate institutionalholdings in the sample IPOs increases from 17.7% at the end of the first post-issue quarter to27.7% at the end of the first post-issue year and to 32.8% at the end of the second post-issueyear7. It would be interesting to gain insight into this process of building up institutionalpositions and, in particular, to find out how the changes in institutional holdings relate to the‘ If it is true that institutional investors own approximately 53% of U.S. equities, we wouldexpect this percentage to increase even further.10performance of the IPOs. Moreover, I believe IPOs form a particularly interesting sample totest the implications of some of the above mentioned factors for institutional trading behavior.The particular nature of IPOs, and in particular the uncertainty regarding their correct value,could strongly affect institutions’ trading behavior because of institutional (e.g. those that leadto window dressing and parallel trading) and/or regulatory (e.g. the Prudent Man Rule) details.Also, because not much information is available on IPOs, data on institutional IPO trading isparticularly interesting to examine whether institutions have access to superior information onfuture IPO payoffs.My data is holding and trading data for institutions in the aggregate, so portfolios and portfoliochanges of individual institutions are not known. Theoretical models on monitoring (Shleiferand Vishny [1986]; Admati, Pfeiderer and Zechner [1993]) show that a relatively large stake ina particular stock is necessary in order to make it a profitable activity for the monitoringshareholder. Thus, a potential price effect would occur if a particular institution, possiblyalready holding some stock of the firm, acquires additional shares so as to reach the thresholdlevel of percentage ownership that makes monitoring profitable. Once the threshold level isreached, a price effect could also accompany a further increase in holdings because the furtherincrease in holdings implies an increase in monitoring benefits. Because my data is aggregatedata, it is not possible to reliably test for a monitoring effect on prices and I will not invoke themonitoring argument to explain any result in the empirical analysis.The results of the analysis of the association between changes in holdings and performance areconsistent with institutions following a strong positive feedback trading strategy. The positiveassociation between current adjusted performance and changes in holdings could be evidence of11a rational update in expectations by institutions on the less informed side of the market. Thepattern of buying current winners and selling current losers could be strenghtened by a concernabout reputation or over the ‘fiduciary duty’-aspect of a money managers’ job. The positiveassociation between past adjusted performance and changes in institutional holdings could bedriven by a belief that past trends are likely to continue. However, there is evidence of at leastpart of it being the result of an intertemporal association between money managers’ trades:managers, observing the holdings of their colleagues at the end of the quarter, don’t want tomiss the boat and buy a positive stake the next quarter. The evidence of a positive associationbetween current (past) positive adjusted performance and changes in institutional holdings isstronger than that documented by Lakonishok et al (1992) for the average common stockinstitutions trade in8. The observed trading pattern could be destabilizing in nature and couldresult in an undervaluation of past underperformers and/or an overvaluation of pastoverperformers. To attribute the finding of a positive rebalancing effect on the return of the IPOportfolio, documented in the first part of this study, to institutional trading behavior in IPOs, weshould have an idea of the importance of institutions’ trading in IPO market. The percentageof IPO shares outstanding that institutions hold at the end of the first and second post-issue yearmay give an indication that their presence in IPO markets could be important. But the resultsdo not allow us to make any conclusive statement about a possible causal relationship betweeninstitutional IPO trading and the evidence of overreaction in IPO prices.B The ‘average’ stock is likely to be a large stock (in terms of market capitalization):Lakonishok et al. (1992) document that, for pension funds, only 1.3% of the 26292 quarterlychanges in stock holdings (each quarter-stock is an observation) are in the smallest size quintilestocks, where size quintiles are determined from the universe of NYSE and AMEX stocks. Incontrast, 70% of the changes are in the two largest quintiles.12Chapter II: Data DescriptionThe sample consists of 70 firms that went public in 1990 and 224 firms that went public in1991. These firms meet the following criteria(1) the offer price is $1.00 per share or more;(2) the offering involves common stock only (unit offers are excluded);(3) the offering is not a reverse LBO;(4) the company is listed on the CRSP daily Amex-NYSE or NASDAQ tapes within 3 monthsof the offer date;(5) quarterly institutional shareholdings are available for the first two years after the IPO.Institutional ownership data have to be available within 4 months after the listing date.Reverse LBOs are excluded from the sample because they may have different characteristicsfrom the average IPO. Reverse LBOs tend to be larger than the average IPO and, because thefirm has been public in the past, there may be less uncertainty as to their future prospects’°.The imposition of criterion (5) may lead to an upward bias in the returns of the IPOs in mysample. Of the original sample of 276 IPOs in 1991, 16 dropped out because no institutionaldata were available in any one of the quarters studied, and 7 were excluded because institutionaldata ceased to exist after a few quarters. For the initial sample of 92 IPOs in 1990, theThe invasion of Kuwait and the related uncertainty could be a possible explanation for thelow number of firms that went public in 1990.10 See Degeorge and Zeckhauser (1993) for more details on reverse LBOs.13corresponding figures were 4 and 1. As there are relatively few omissions, I feel that the mainresults of this paper are not altered significantly by this potential bias.Quarterly data on institutional ownership are obtained from the SEC disks which contain the datafrom the quarterly SEC filings of all public companies required to file”. These are publiccompanies that (i) have at least 500 shareholders and (ii) have at least 5 million dollars worthof total assets.Quarterly holdings are expressed as the percentage of firm shares outstanding held by institutionsat the end of each quarter. Because data on individual institutions’ holdings are not available,all institutions are treated as one investor. Thus, the aggregate net institutional holding changesover a given quarter are known, but we do not know the level of trading of certain subgroupsof institutions that was countered by opposite demand shifts of other subgroups. The data alsodo not include (aggregate) intraquarter round-trip transactions. Because data on the number ofinstitutions are also available, it is possible, however, to get an idea of the average holding ofthe institutions and the number of institutions following a particular stock.All returns data are extracted from the CRSP, NASDAQ or Amex-NYSE tapes.Table I segments firms with issue years 1990 and 1991 by industry classifications based uponthree-digit Standard Industrial Classification Codes (SIC). Where two or more SIC codesrepresent industries that are very similar, I have grouped them into a single industry’2. The15 industries for which there were at least 3 IPOs in one of the subsamples (as defined by issueyear) are listed, with the remaining firms grouped together.“ The data are available on CD ROM from Disclosure Incorporated.12 The procedure follows the classification used by Ritter (1991).14Table I : Sample Offers Categorized by IndustryNumber of Number ofIndustry SIC codes offers offersin 1990 in 1991Oil and Gas Extraction 131, 138 8-Chemicals and Allied 283, 287 3 32Industrial, Commercial Machinery, 356, 357 10 10Computer EquipmentElectric Equipment 363, 365, 7 15366, 367Meas Instruments, Photo Goods, 382, 384, 4 17Watches 386Miscellaneous Manufacturing 394, 398, 7Industries 399Motor Freight Transport, Warehouse 420-422 1 4Communications 481, 483 1 5Wholesalers 501-5 19 8 9Retailers 526-599 3 22Financial Institutions 614-679 3 17Hotels, Other Lodging Places 701, 703- 3Computer Programming, Data 737 5 28ProcessingHealth Services 801-809 4 16Engineering, Account, Research, 871-874 2 9Management, Related ServicesAll Other Firms 11 30All Firms 70 22415Table II: Distribution of Sample Offers across Size Deciles Defined by The Amex-NYSE StockUniverseSize Deciles Number of Number of(000s) offers in 1990 offers in 19911-10556 7 910556-27772 24 727772-50887 31 1650887-91225 53 1491225-159080 49 14159080-276754 38 4276754-534919 19 5534919-1001049 2 11001049-2514810 1 02514810-74529168 0 0Table II gives the distribution of the sample firms across size deciles. The cut-off points for sizedeciles are determined from the universe of NYSE and AMEX stocks on December 31, 1990.The categorization of the IPOs into these size deciles is done on the basis of their marketvalueat the end of the first post-issue quarter.As one would expect, the IPOs are not evenly distributed across size deciles, and are, onaverage, relatively small firms : 85.7 % of the IPOs with issue year 1990 and 73.2 % of theIPOs with issue year 1991 fall into the five lower deciles.16Chapter III : Documentation of Long Run IPO Performance111.1 EvidenceRitter (1991) documents that the underperformance of IPOs varies with the issue year: highvolume years (in terms of the number of firms going public) are followed by more severeunderperformance. As I deal with two years that differ greatly in volume, I calculateaftermarket performance by year of issuance.To evaluate the long-run performance, two benchmarks are used : (i) the CRSP value-weightedAmex-NYSE index and (ii) the returns on a matching firm, where matching is based on industryand size. To choose a matching firm, on December 31 of the pre-issue year, all common stockslisted on the CRSP, Amex-NYSE and Nasdaq tapes which have been listed for at least 3 yearsare ranked by their market capitalization. The firm in the same three-digit industry with themarket value of equity closest to that of the IPO on the first listing date is chosen as thematching firm for the issuing firm. If a matching firm is delisted before the ending date of thereturn interval considered, a second (and, if necessary, third, fourth, etc) is spliced in after thedelisting date of the first matching firr&3. The replacement firm is the firm in the same three-digit industry with market value on the original ranking date second closest to that of the IPOon the first listing date.I document the relative performance of the IPO firm for 1 year, 2 years and, if possible, 3 years13 When a firm is delisted, it does not necessarily mean that it went bankrupt. It is possiblethat it is taken over or that it simply doesn’t satisfy the listing requirements of a certain exchangeanymore. For the entire sample, 38 second matched firms, 4 third matched firms and 2 fourthmatched firms were spliced in.17after the offer date’4. The measures of performance used are the 1-, 2- and 3-year buy-and-hold returns for both the set of IPOs and the appropriate benchmarks. Returns include dividendsand capital gains. The buy-and-hold returns are calculated from the first listed post-issue closingprice’5 to the appropriate anniversary date of the offering. The construction of the sample issuch that the issuing firm survives at least 2 years after the issue. If a firm of the sample withissue year 1990 is delisted prior to its third anniversary date’6, its total return is truncated onthat date. The same procedure is applied to the matching firm and the index.In Tables lila and Tub, the holding-period (=T) return on an equally weighted portfolio of thefirms going public in calendar year t is listed, together with the returns on an equivalentportfolio of the matched firms, and the returns on the Amex-NYSE Index. The buy-and-holdreturn is measured asRTable lila shows that the 3 year buy-and-hold return on an equally weighted portfolio of firmsthat went public in 1990 was 22.6% less than the equivalent return on the Amex-NYSE index,and 25.4% less than the equivalent return on a portfolio of matching firms. These figures aresimilar to the 26% given by Ritter for a sample of IPOs from 1975 to 1984. Although the‘‘ For the firms that went public in 1991, only 2 years of returns data are available.‘ The initial return from the offering price to the first listing price is excluded becauseinvestors can’t be sure to get allocated the number of shares they want through the initialdistribution of the shares, whereas the market price represents a price that is available for animplementable portfolio strategy.6 This is the case of only one firm in the sample.18Table lila: One-, Two- and Three-Year Buy-and-Hold Returns (%). The first column gives thebuy-and-hold returns on a portfolio of IPOs issued in 1990. The second and third columns givethe equivalent returns on the index and on a porfolio of matching firms, respectively. The Pvalues for the difference in the returns between the IPO portfolio and the Amex-NYSE Indexand between IPO portfolio and the matched firm portfolio are given in the last column.t=1990, N70IPOs NY/Am Match IPO-NY/Am IPO-Match P-ValuesT=l year 9.3 14.1 8.9 -4.8 0.4 0.59/0.94T=2 years 10.2 28.0 17.9 -17.8 -7.7 0. 17/0.67T=3 years 23.2 45.8 48.6 -22.6 -25.4 0.33/0.53Average YearlyComp. Return 7.2 13.4 14.1Table Tub: One-, Two- and Three-Year Buy-and-Hold Returns (%). The first column gives thebuy-and-hold returns on a portfolio of IPOs issued in 1991. The second and third columns givethe equivalent returns on the index and on a porfolio of matching firms, respectively. The Pvalues for the difference in the returns between the IPO portfolio and the Amex-NYSE Indexand between IPO portfolio and the matched firm portfolio are given in the last column.t=199l, N=224IPOs NY/Am Match IPO-NY/Arn IPO-Match P-ValuesT=l year 5.3 12.3 13.5 -7.0 -8.2 0.07/0. 14T=2 years 26.1 27.0 30.7 -0.9 -4.6 0.92/0.74Average YearlyComp. Return 12.2 12.7 14.319economic magnitude seems important, the 22.6% and 25.4% are not statistically significant.Table Tub shows that a portfolio of stocks that went public in 1991 has a performance verysimilar to that of the Amex-NYSE index and to that of a portfolio of matched firms.The 3-and 2-year buy-and-hold returns obscure an interesting pattern in average annualizedreturns, obtained when, on each anniversary of the issue date, the portfolios are rebalanced toequal weights and the average buy-and-hold return for the next year is calculated. Tables IVaand IVb report these annualized returns for the first, second and third post-issue year, togetherwith their geometric mean.Table IVa shows that an equally weighted portfolio of IPOs from 1990, held from the beginningof the second post-issue year to the beginning of the third post-issue year, underperforms bothbenchmarks. A newly invested, equally weighted portfolio of the same IPOs, held over thesubsequent year, outperforms both benchmarks. The under- and overperformance lack statisticalsignificance due to the high standard deviation of the IPO returns. The economic magnitude,however, seems important: ranging from 9.2% to 10.7% for the underperformance, dependingon the benchmark used, and from 7.4% to 14.1 % for the overperformance, depending on thebenchmark used. The average yearly compounded return for IPOs rises from 7.2% in Table lilato 13.0% in Table IVa because of the positive effect of rebalancing the portfolio at the end ofthe second post-issue year. Note that the average compounded yearly return on the indexportfolio does not change from Table lila to Table IVa, while the average compounded returnon the portfolio of matched firms only changes by 0.2%.Table IVb shows similar patterns for the IPOs issued in 1991. During the first post-issue year,an equally weighted portfolio of these IPOs underperforms the index benchmark and the matched20Table IVa Annual Percentage Returns per Post-Issue year on Yearly Rebalanced Portfolios.The first, second an third columns give the returns on the portfolio of IPOs from 1990, on theindex and on the portfolio of matched firms, respectively. The P-values for the difference inthe yearly average returns between IPOs and the Amex-NYSE Index and between IPOs and thematched firm portfolio are given in the last column.t=l990, N70IPOs NY/Am Match IPO-NY/Am IPO-Match P-values1st Post-Issue Yr 9.3 14.1 8.9 -4.8 0.4 0.59/0.942nd Post-Issue Yr 3.8 13.0 14.5 -9.2 -10.7 0.26/0.303rdPost-IssueYr 27.2 13.1 19.8 14.1 7.4 0.30/0.54Geom. Mean 13.0 13.4 14.3Table IVb : Annual Percentage Returns per Post-Issue year on Yearly Rebalanced Portfolios.The first, second an third columns give the returns on the portfolio of IPOs from 1990, on theindex and on the portfolio of matched firms, respectively. The P-values for the difference inthe yearly average returns between IPOs and the Amex-NYSE Index and between IPOs and thematched firm portfolio are given in the last column.t=199l, N=224IPOs NY/Am Match IPO-NY/Am IPO-Match P-values1st Post-Issue Yr 5.3 12.3 13.5 -7.0 -8.2 0.07/0. 142nd Post-Issue Yr 21.0 13.0 12.4 8.0 8.6 0.35/0.27Geom. Mean 12.9 12.6 12.9firm benchmark by 7.0% and 8.2% respectively, while during the second post-issue year a21newly invested, equally weighted portfolio of these IPOs outperforms both benchmarks by 8.0%and 8.6% respectively. Only the 7% underperformance is statistically significant at anacceptable level of 7%. The average compounded yearly return only rises from 12.2% to12.9% through the rebalancing procedure.111.2 DiscussionThe previous analysis indicates an underperformance of on average 23% for the 1990 sampleand no underperformance for the 1991 sample. What is remarkable for the 1990 sample,though, is the sensitivity of the result to the trading rule adopted. An equally weighted portfolioof the sample of firms that went public in 1990, bought at the closing price on the first listingdate and held for 3 years achieves an average yearly compounded return of 7.2%. If, however,the portfolio is rebalanced at each anniversary of the issue date, the average yearly compoundedreturn rises to 13%, a return similar in magnitude to the 13.4% and 14.3% on the index and thematched firm portfolio, respectively.’7 This suggests that at the end of the second post-issueyear, IPOs that performed badly in the past were on average undervalued and/or IPOs thatperformed well over the past period were on average overvalued. Rebalancing to equal weightsat a given moment is equivalent to putting more weight (relative to a buy-and-hold strategy) inpast losers and less weight (relative to a buy-and-hold strategy) in past winners. If such astrategy earns superior returns relative to a buy-and-hold, it suggests that past losers were onaverage undervalued and/or that past winners were on average overvalued. The finding is17 A rough test indicates that the difference of 5.8% between the average yearly compoundedreturn on the buy-and-hold portfolio and the equivalent return on the rebalanced portfolio is notstatistically significant.22consistent with the evidence of overreaction presented by De Bondt and Thaler (1985) for asample of New York Stock Exchange common stocks. De Bondt and Thaler (1985) documentthat prices systematically overshoot, and that their reversal is predictable from past return dataalone. Investors seem to extrapolate recent earnings trends into the future, ignoring the manyrandom walk elements in earnings patterns as well as ignoring the tendency of most divergencesfrom average earnings performance to correct themselves over time by a familiar process knownas reversion to the mean. It seems to me that, the greater the uncertainty involved in estimatingthe present value of the future payment stream - and, as pointed out, this uncertainty is relativelybig in the case of IPOs - the worse the problem of overreaction could be. To say it in Hamlet’swords, “The undiscover’d country... makes us rather bear those ills we have than fly to othersthat we know not of.”Another interesting result is the large year-to-year variability in the returns on the portfolio ofIPOs, and the fact that the relative over- or underperformance is characterized by absolutecalendar years during the calendar year l99ll99218, both the IPOs with issue year 1990 andthose with issue year 1991 underperform their benchmarks on average, while during the calendaryear 1992199319, they outperform their benchmarks on average. This could point towards an‘IPO-factor’ in the returns data. This possibility is not explored in this study, and left for futureresearch. For now, the observation is important because it suggests yet another source of riskassociated with an investment in IPOs.‘ The calendar year 1991-1992 is by approximation the second post-issue year for the IPOsissued in 1990, and the first post-issue year for IPOs issued in 1991.19 The calendar year 1992-1993 is by approximation the third post-issue year for the IPOsissued in 1990 and the second post-issue year for the IPOs issued in 1991.23Chapter IV : Institutional Trading Patterns in IPOsThe empirical analysis of the association between institutional holdings and returns during thefirst two post-issue years is limited to the sample firms that went public in 1991. This isbecause quarterly returns and holdings are defined relative to the issue year, while theperformance tends to be characterized by absolute calendar years. I chose 1991 because morefirms went public in that year compared to the issue year 1990.Sections IV. 1 and IV.2 take a closer look at the trading behavior of institutions in the aggregateduring the 2 years following the IPO event. In sections IV. 1, IV.2 and IV.3, institutions aretreated as one big investor. Only section IV.4 will focus on the number of institutions involvedin IPO trading.IV. 1 Analysis of Quarterly Aggregate Institutional Holding Changes and ReturnsIn this section, I first focus on the relation between changes in institutional holdings of IPOs ina given quarter and IPO returns over the same quarter. Subsequently, I examine the relationbetween changes in institutional holdings and performance in previous and subsequent periodsto provide more detailed descriptive evidence on the relation between firm performance andchanges in institutional holdings, and to provide insight into the potential reasons for therelation.24IV. 1. 1 Descriptive StatisticsTable V gives Spearman correlation coefficients for the second post-IPO quarter change ininstitutional holdings (PIN21), quarterly contemporaneous adjusted returns (QADJ forAmex/NYSE-adjusted returns and QADJMA for matched firm adjusted returns), and lagged(postcript -LA) and lead (postscript -LE) adjusted quarterly returns. Adjusted returns aredefined as raw returns minus the return on the Amex-NYSE Index over the same time-periodor raw returns minus the return on the matched firm over the same time-period. Tables VI andVII give analogeous data for the fourth and sixth quarter, respectively. These tables areTable V : Spearman Correlation Coefficients between The Change in Aggregate InstitutionalHoldings during The Second Quarter (PIN21) and Concurrent, Lagged and Future AdjustedReturns. The P-values are in italics below the corresponding correlation coefficients.PIN21 QADJ QADJ QADJ QADJ QADJ QADJLA LE MA MALA MALEPIN21 1.0000 0.2193 0.1556 0.1667 0.1447 0.1342 0.10090.0 0. 0009 0. 0190 0. 0123 0. 0307 0. 0449 0.1340QADJ 1.0000 0.0458 0.0222 0.6628 0.1095 -0.03320.0 0.4936 0. 7405 0. 0001 0.1028 0. 6228QADJ 1.0000 0.0043 -0.0182 0.6399 -0.1489LA 0.0 0.9490 0. 7871 0. 0001 0.0265QADJ 1.0000 0.1255 0.0250 0.5361LE 0.0 0. 0619 0.7114 0. 0001QADJ 1.0000 0.0169 0.0290MA 0.0 0.8023 0.6676QADJ 1.0000 -0.0756MALA 0.0 0.2619QADJ 1.0000MALE 0.025Table VI: Spearman Correlation Coefficients between The Change in Aggregate InstitutionalHoldings during The Fourth Quarter (PIN43) and Concurrent, Lagged and Future AdjustedReturns. The P-values are in italics below the corresponding correlation coefficients.P1N43 QADJ QADJ QADJ QADJ QADJ QADJLA LE MA MALA MALEP1N43 1.0000 0.3545 .3399 -0.0828 0.1825 0.3315 -0.01880.0 0. 0001 0. 0001 0.2171 0. 0052 0.0001 0. 7819QADJ 1.0000 -0.0459 -0.0520 0.6687 0.1417 -0.04910.0 0.4947 0.4397 0. 0001 0. 0353 0.4711QADJ 1.0000 -0.1156 0.0196 0.5361 -0.0537LA 0.0 0.0851 0.7735 0.0001 0.4301QADJ 1 .0000 -0.0169 -0.0621 0.6494LE 0.0 0. 8046 0.3592 0. 0001QADJ 1.0000 0.0470 -0.0310MA 0.0 0. 4886 0. 6495QADJ 1.0000 -0.0414MALA 0.0 0.5431QADJ 1.0000MALE 0.0representative for all eight quarters considered. Of particular interest are the significant positivecorrelations of the PINji’s (PJNji represents the change in aggregate institutional holdings fromthe end of quarter i to the end of quarter j) with concurrent and lagged returns and the generallack of a relationship between the PINji’s and lead returns : over all eight quarters studied, thecorrelations of changes in holdings with concurrent and lagged returns are significantly positive;the correlations with future returns are everywhere insignificant, except for the second quarter,where it is significantly positive.26Table VII : Spearman Correlation Coefficients between The Change in Aggregate InstitutionalHoldings during The Sixth Quarter (P1N65) and Concurrent, Lagged and Future and AdjustedReturns. The P-values are in italics below the corresponding correlation coefficients.P1N65 QADJ QADJ QADJ QADJ QADJ QADJLA LE MA MALA MALEP1N65 1.0000 0.3639 0.1971 0.0455 0.2770 0.2085 0.04200.0 0. 0001 0. 0031 0.4987 0. 0001 0.0019 0.5391QADJ 1.0000 -0.0261 -0.0572 0.6404 0.0330 0.01170.0 0.6987 0.3955 0. 0001 0. 6281 0.8643QADJ 1.0000 0.0658 0.0091 0.6731 0.0406LA 0.0 0.3278 0. 8935 0. 0001 0.5531QADJ 1.0000 0.0160 0.0462 0.6698LE 0.0 0.8189 0.4977 0.0001QADJ 1.0000 0. 1042 0.0314MA 0.0 0.1250 0.6465QADJ 1.0000 0.0794MALA 0.0 0.2455QADJ 1.0000MALE 0.0IV. 1.2 Changes in Institutional Holdings and Contemporaneous ReturnsThe analysis of the relation between changes in institutional holdings from the end of quarter ito the end of quarter j (PINji) and returns over the same quarter presents an inference problem.If trading takes place at the beginning of the quarter, a positive association could be indicativeof superior access to information. If the changes in holdings occur after the price movemene°,20 As has been pointed out, the Brennan and Cao (1994) model is a rational expectationsmodel : it posits a price function and demonstrates that it is consistent with market clearing. Asprice setting and market clearing are assumed to happen at the same time, ‘after’ is not to betaken literally here.27a positive association could be the result of a rational update in expectations by institutions thatare on the less informed side of the market, as shown in the Brennan and Cao (1994) model.A positive association could also be the result of institutions’ concern over their fiduciary dutyto invest prudently or of institutions engaging in window dressing. Another possibleinterpretation of a positive association is price pressure by institutional buying or selling.Because the time of trade is not known, it is not possible, at this point, to distinguish betweenthese hypotheses. A negative association between quarterly holdings and concurrent returnscould be interpreted as a contrarian trading strategy, and could be indicative of institutions beingon the better informed side of the market. Tables V to VII show that the PINji’s are positivelycorrelated with concurrent returns, with correlation coefficients ranging from 0.14 to 0.36. Allcorrelations with concurrent returns over all eight quarters are statistically significant.The relation could be nonlinear, however. Attention could be focused on extreme under- oroverperformers, resulting in heavy trading in, for example, the top 25% outperformers or thebottom 25% underperformers. To examine this possibility, I measure the level of institutionaltrade across quartiles of firm performance measures. For each quarter, I estimate two (using2 different performance measures) regressions of the rank of the change in institutional holdingsfrom the beginning of quarter i to the beginning of quarter j (RKPINji) on dummy variablesindicating performance quartiles 2 through 4 during the same quarter (quartile 1 is the poorestperforming quartile; quartile 4 the best performing quartile). Rank regressions are estimatedbecause the objective is to explain relative changes in institutional holdings by relativeperformance measures. The coefficients on the dummy variables compare the rank of thechange in institutional holdings for the performance quartile to that of firms with performance28measures in the first quartile. The performance measures used are the Amex-NYSE Indexadjusted returns and the matched firm adjusted returns. The average percentage of ownershipper institution at the beginning of the quarter (PINAV the aggregate institutional ownershippercentage divided by the number of institutions that hold stocks of that firm) is included as anadditional explanatory variable. It is included because the size of the ‘typical’ institutionalholding in the firm may be an important determinant of the institutions’ price sensitivity : if, onaverage, institutions own a relatively large fraction of a firm’s stock, we would expect them tobuy relatively less when the firm performs well than if they held less stock. In contrast, wewould expect them to sell relatively more when the firm performs poorly than if they held lessshares. Because many of the firms going public are young growth firms in need of funds, andbecause growth is an ongoing process, we would expect some of them to return to the stockmarket to obtain additional financing during the post-issue period. It is possible that institutionsget priority to acquire new shares during the secondary offering. Often, such secondaryfinancing is done in high valuation periods or following recent high valuation periods. BecauseI want to capture the changes in institutional holdings that occur through the regular marketprocess and relate them to market returns, the percentage change in the number of sharesoutstanding over the quarter, adjusted for stock splits and stock dividends, is included as anadditional control variable (SHRCH).More formally, for each quarter, the following regression models are estimatedRKPINji= a + b PINAV + c SHRCH + d2 DUMAJ2 + d3 DUMAJ3 + d4 DUMAJ4 +eRKPINji= a + b PINAV + c SHRCH + d2 DUMAJMA2 + d3 DUMAJMA3 (1)+ d4 DUMAJMA4 + e29where DUMAJ2 (DUMAJMA2) through DUMAJ4 (DUMAJMA4) are dummy variablesindicating the second through fourth quartile of the index adjusted (matched firm adjusted)performance measure. The first two columns of Tables VIII to X give the results of these rankregression for quarters 2, 4 and 6, respectively. As expected, the coefficients on PINAV aresignificantly negative in all regressions. The coefficient on the percentage change in sharesoutstanding is insignificantly positive for the second quarter, significantly positive for the fourthquarter, and insignificantly negative for the sixth quarter. A positive sign on the coefficientsuggests that institutions in the aggregate acquire a larger fraction of new shares than isnecessary to maintain their initial (i.e. before the secondary offering) ownership percentage inthe firm. A negative sign suggests the opposite. Because the dummies related to theperformance adjusted by the Amex-NYSE Index typically contain less noise than thecorresponding dummies related to the matched firm adjusted performance (the latter also containa portion of matched firm-specific risk), the regressions using the former dummies typically havegreater explanatory power. The pattern of the coefficients on the dummy variables is veryinteresting and similar across quarters and performance dummies: the coefficients areeverywhere monotonically increasing in the quartiles of performance measures. For example,according to the first column of Table X, the rank of the change in institutional holdings relativeto performance quartile 1 is 18.50 higher for the second quartile , 48.78 higher for the thirdquartile, and 59.90 higher for the fourth quartile21. For all regressions, an F-statistic is211 estimated similar regressions as the ones specified under (1), but with dummies for thefirst (lowest), second and third performance quartiles. The coefficients on these dummies wouldthen compare the rank of the change in institutional holdings for firms in the first, second andthird performance quartiles to that of firms in the highest performance quartile. As expected,the dummy coefficients are all significantly negative and more negative for the lower30computed to test for equality of the dummy coefficients. Only in the second quarter is thehypothesis of equality not rejected.IV.l.3 Changes in Institutional Holdings and Lagged ReturnsThe Spearman correlation coefficients in Tables V to VII show that the correlation betweenchanges in institutional holdings and previous quarter returns is always significantly positive.By focusing on lagged variables, it is possible to examine a setting in which superior access toinformation or price pressure are not relevant. The suggested positive relation could be due totechnical trend chasing22, driven by a belief that past trends are likely to continue. Becausethere is no evidence of a (positive) autocorrelation in (adjusted) returns over subsequent quarters(see Tables V to VII), a more plausible explanation is an intertemporal relation between moneymanagers’ trades. Managers, infering information from or reassured by the positions of theircolleagues as presented in the quarterly statements, mimick these colleagues’ trades during thesubsequent quarter. Given that managers who trade on current price movements follow apositive feedback trading strategy, mimicking them in the subsequent quarter induces a positiverelation between changes in institutional holdings and past quarter’s adjusted returns. Toexamine the relation between changes in holdings and past adjusted returns more carefully, aperformance quartiles. For example, using NYSE-Amex adjusted performance, the dummycoefficients for the fourth quarter are (to be compared with the first column of Table X) : -60.45for the first quartile, -38.24 for the second quartile and -8.87 for the third quartile.22 As explained in footnote 15, the fundamental trend chasing strategy as explained byBrennan and Cao (1994) results in a contemporaneous positive association between changes inholdings and returns. Fundamental trend chasing can be ruled out as an explanation for thepositive association between changes in holdings and past performance.31Table VIII : Regression of RKPIN21 (= the rank of the change in aggregate institutionalholdings over the second quarter) on Contemporaneous Performance Dummies.PINAV is the ratio of the aggregate institutional ownership percentage to the number ofinstitutions holding the stock at the beginning of the quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.DUMAJ is a dummy related to performance adjusted by the Amex-NYSE index; DUMAJMAis a dummy related to performance adjusted by a matched firm (P-values are in parentheses).Regression CoefficientsIndepend Var Concurrent PerformanceIntercept 108.19 115.49(0.0001) (0.0001)PINAV -9.24 -10.85(0.0310) (0.0129)SHRCH 23.27 26.95(0.4873) (0.4214)DUMAJ2 9.03(0.45 83)DUMAJ3 23.23(0.0594)DUMAJ4 32.30(0.0096)DUMAJMA2 0.52(0.9663)DUMAJMA3 17.53(0.1554)DUMAJMA4 23.64(0.0556)R2 0.07 0.06Pvalue(*) 0.1613 0.1688(*) for the difference in dummy coefficients32Table IX : Regression of RKPIN43 (= the rank of the change in aggregate institutional holdingsover the fourth quarter) on Contemporaneous and Lagged Performance Dummies.PINAV is the ratio of the aggregate institutional ownership percentage to the number ofinstitutions holding the stock at the beginning of the quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.DUMAJ is a dummy related to performance adjusted by the Amex-NYSE index; DUMAJMAis a dummy related to performance adjusted by a matched firm (P-values are in parentheses).Regression CoefficientsIndepend Var Concurrent Performance Lagged PerformanceIntercept 110.87 135.56 111.50 115.91(0.0001) (0.0001) (0.0001) (0.0001)PINAV -19.80 -21 .90 -20.80 -21 .48(0.0001) (0.000 1) (0.000 1) (0.0001)SJ{RCH 56.11 67.52 57.14 63.49(0.0967) (0.0562) (0.0899) (0.0633)DUMAJ2 23.59 20.91(0.0362) (0.0629)DUMAJ3 24.42 29. 17(0.0301) (0.0103)DUMAJ4 58.87 56.89(0.0001) (0.0001)DUMAJMA2 -4.14 13.60(0.7203) (0.2272)DUMAJMA3 -1.36 30.05(0.9066) (0.0077)DUMAJMA4 21.18 50.49(0.0680) (0.0001)R2 0.22 0.12 0.21 0.20P-value (*) 0.0022 0.0680 0.0038 0.0060(*) for the difference in dummy coefficients33Table X : Regression of RKPIN65 (= rank of the change in aggregate institutional holdings overthe sixth quarter) on Contemporaneous and Lagged Performance Dummies.PINAV is the ratio of the aggregate institutional ownership percentage to the number ofinstitutions holding the stock at the beginning of the quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.DUMAJ is a dummy related to performance adjusted by the Amex-NYSE index; DUMAJMAis a dummy related to performance adjusted by a matched firm (P-values are in parentheses).Regression CoefficientsIndepend Var Concurrent Performance Lagged PerformanceIntercept 101.072 115.40 117.43 117.23(0.0001) (0.0001) (0.0001) (0.0001)PINAV -13.93 -14.56 -13.30 -13.27(0.0027) (0.0025) (0.0064) (0.0066)SFIRCH -7.38 -6.62 -19.93 -12.64(0.8006) (0.8258) (0.5263) (0.6859)DUMAJ2 18.50 6.702(0.1116) (0.5809)DUMAJ3 48.78 27.18(0.0001) (0.0273)DUMAJ4 59.90 26.40(0.0001) (0.0305)DUMAJMA2 9.98 8.40(0.3942) (0.4874)DUMAJMA3 16.99 24.03(0. 1514) (0.0482)DUMAJMA4 48.01 28.63(0.0001) (0.0174)R2 0.17 0.11 0.07 0.07Pvalue(*) 0.0017 0.0044 0.1808 0.2388(*) for the difference in dummy coefficients34regression similar to equation (1) is estimated, but now, performance quartiles are defined withrespect to last quarter’s performance. Results for quarters 4 and 6 are given in the last twocolumns of Tables IX and X23. The coefficients on PINAV are, again, all significantlynegative. For the fourth quarter, the dummy coefficients show the same monotonicallyincreasing pattern as the dummy coefficients in the regressions on concurrent performancequartiles, suggesting that part of the trading strategy consists of buying past quarter’s top 25%performers and/or selling past quarter’s bottom 25% performers. The hypothesis of equality ofthe dummy coefficients is strongly rejected. For the sixth quarter, the third and fourth dummycoefficients are similar in magnitude and both significantly positive. This suggests a tradingstrategy less focused on the extreme performance quartiles. Still, the pattern is consistent withbuying last quarter’s top half performers and/or selling last quarter’s bottom half performers.In this case, however, the hypothesis of equality of the dummy coefficients cannot be rejected.IV. 1.4 Combining Contemporaneous, Lagged and Future Returns in One RegressionThe following models examine the extent to which trade is related to past (postscript -LA forlagged) versus current versus future (postscript -LE for lead) performance:PINji= a + b SHRCH + c PINAV + d QADJ + f QADJLA + g QADJLE + e (2)PINji= a + b SHRCH + c PINAV + d QADJMA + f QADJMALA + g QADJMALE + e23 do not regress second quarter holdings changes on first quarter performance quartilesbecause the first ‘quarter’ spans the period between the first listing date and the first date onwhich institutional holdings data are available, and this varies from a few weeks to 4 months.35Table Xl provides evidence that both past and current adjusted returns contribute in explainingthe crosssectional variation in changes in institutional IPO holdings.A negative coefficient on the lead return could indicate that institutional buying drives prices upand/or institutional selling depresses them; this could explain a reversal in the subsequent period.On the other hand, if institutions trade because they have private information about futurereturns, institutional buying in a quarter should be associated with positive adjusted returns insubsequent quarters.For all quarters, only one correlation coefficient between changes in institutional ownership andlead adjusted returns is significantly positive (see Table V : correlation between PIN21 andQADJLE). All other coefficients are not significantly different from zero. None of thecoefficients on the lead adjusted returns in the regression models specified in (2) is significantlydifferent from zero, suggesting that neither price pressure nor superior information is likely tobe the driving force of the observed patterns.For the fourth quarter, the explanatory power of the models specified under (2) is large : forexample, the index-adjusted returns, together with the control variables, explain 29% of thechanges in institutional portfolio holdings of IPOs.36Table XI: Regression of PINji’s (= change in aggregate institutional holdings from the end ofquarter i to the end of quarter j) on Contemporaneous, Lagged (postscript -LA) and LeadReturns (postscript -LE) Combined.PINAV is the ratio of the aggregate institutional ownership percentage to the number ofinstitutions holding the stock at the beginning of the quarter.SHRCFI is the percentage change in the number of shares outstanding during the quarter.QADJ are the Amex-NYSE Index-adjusted returns; QADJMA are the matched firm adjustedreturns (P-values are in parentheses).Independ Var Dependent VariablesP1N43 P1N65Intercept 4.28 4.63 2.14 1.81(0.0001) (0.0001) (0.0037) (0.0185)PINAV -1.92 -2.21 -0.65 -0.44(0.0001) (0.0001) (0.1756) (0.3697)SHRCH 4.24 6.81 0.61 1.46(0.2232) (0.0639) (0.8245) (0.6024)QADJ 5.04 4.62(0.0001) (0.0001)QADJLA 8.05 2.96(0.0001) (0.0035)QADJLE 1.77 0.13(0.1861) (0.8848)QADJMA 1.88 3.69(0.0668) (0.0005)QADJMALA 4.88 2.04(0.0001) (0.0381)QADJMALE 1.08 0.063(0.3459) (0.9371)R2 0.29 0.22 0.12 0.0837IV.1.5 Discussion of The ResultsBased on the previous regressions, it is not possible to determine whether institutions buy past(current) winners heavily or sell past (current) losers heavily or do both. Some factors,however, affect institutions’ buying vs selling behavior unevenly. For example, a concern overtheir fiduciary duty could affect institutions’ selling decisions but it has no implications for theirbuying decisions. In Table XII, averages of changes in institutional holdings during a quarterare computed by current and past quarter’s performance quartiles. The table illustrates thatreactions to current and past positive price signals are more important than reactions to negativeprice signals. This is what one would expect given that the average of the aggregate institutionalownership percentage over all 224 IPOs increases from 17.7% to 27.7% over the first post-issueyear and from 27.7% to 32.8% over the second post-issue year.Table XII : Average Change in Percentage Ownership Held by Institutions over A QuarterConditional on Current and Past Quarter Performance. An *sign means significance at the10%-level.Current Quarter Second Fourth Sixth Past Quarter Fourth SixthPerformance Quarter Quarter Quarter Performance Quarter QuarterQuartiles Quartiles1 (lowest) 0.99 -0.96 2.09* 1 -1.11 -0.382 1.70 1.57 0.52 2 0.88 0.433 3.40* 1.41 2.69* 3 2.05* 1.99*4 (highest) 4.96* 5.31* 3.65* 4 535* 2.76*38On the basis of the regression results and the results presented in Table XII, institutions’ tradingstrategy could be described as follows : Institutions, having no access to superior information,face a geat deal of uncertainty associated with the valuation of IPOs and weigh therefore currentnews heavily when making their trading decisions. This makes them heavy buyers of currentoverperformers and in some quarters (for example, the sixth post-issue quarter) heavy sellersof current underperformers. Some of the money managers are concerned over their fiduciaryduty or simply want to present an ‘acceptable’ statement of holdings at the end of the quarter,which results in additional selling of current losers. Possibly, some managers believe that pasttrends are likely to continue, and therefore engage in technical trend chasing. Given the lackof (positive) autocorrelation in successive (adjusted) quarterly returns, it is improbable that theheavy buying of past quarter’s winners can be totally explained by a technical trend chasingstrategy. Alternatively, the strong relation between positive returns and institutional buying inthe subsequent quarter could reflect an intertemporal relation between money managers’ trades.Current positive price signals make some money managers buy the stock. Other managers, whodon’t want to ‘miss the boat’, infer information from their colleagues’ holdings at the end of thequarter and buy a positive stake in the subsequent quarter. Additional evidence, consistent with.this explanation, is given in section IV.4.Although the significant positive association between currrent changes in institutional holdingsand current performance could be totally explained by technical trend chasing or windowdressing, I do believe that part of it is due to a rational update in expectations. The fact that thecoefficients on the adjusted lead returns in the regression combining current, lagged and leadadjusted returns is positive (although not significantly different from zero) suggests that39institutions’ trading decisions result in normal returns. This is what one would expect ifinstitutions do not have superior access to information but rationally incorporate the availableinformation into their trading decisions.IV.2 Institutional Trading One Year After The IPOThe previous analysis was limited to quarterly holdings and returns data. It is conceivable that,with the hindsight of one year IPO performance, institutions take a different strategy than theone suggested by the quarterly analysis. To investigate this conjecture, I estimate a regressionmodel with the rank of the change in holdings during the first quarter of the second post-issueyear as the dependent variable and Amex-NYSE adjusted performance dummies related to thereturn over the first post-issue year as independent variables. The same control variables asbefore are included in the regression: the average institutional holdings at the end of the firstpost-issue year and the percentage change in the number of shares outstanding during the firstquarter of the second post-issue year, adjusted for stock splits and stock dividends. Table XIIIreports the evidence that buying past year’s top 25 % winners and selling past year’s bottom 25% losers is, again, an important part of the institutions’ investment strategy in IPOs.IV.3 Exploring Institutional PortfoliosIt has been mentioned previously that the average of the aggregate institutional ownershippercentage over all 224 issues increases from 17.7% at the end of the first post-issue quarter to27.7% at the end of the first post-issue year and to 32.8% at the end of the second post-issueyear. The previous analysis of the relation between quarterly returns and quarterly changes in40Table XIII : Regression of RKPINCH 1 (= rank of the change in institutional holdings over thefirst quarter of the second post-issue year) on the Amex-NYSE Performance Dummies over ThePast Year.PINAV is the ratio of the aggregate institutional ownership percentage to the number ofinstitutions holding the stock at the beginning of the quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.(P-values are in parentheses)Independ Var Coefficient EstimatesIntercept 104.85(0.000 1)PINAV -6.26(0.2329)SHRCH 56.13(0.0617)DUMAJ2 16.28(0.1828)DUMAJ3 11.55(0.3407)DUMAJ4 37.24(0.0027)R2 0.07P-value (*) 0.0866(*) for the difference in the dummy coefficientsinstitutional holdings suggests that, by the end of the first (second) year, institutions hold asubstantial stake of firms that consistently outperform their benchmarks over the first four (eight)post-issue quarters, and that they hold substantially less shares of firms that consistentlyunderperform their benchmarks. To provide evidence of this conjecture, I calculate averages41of the level of institutional ownership at the end of the first post-issue year (PINI) by quartilesof index-adjusted performance over the first post-issue year. I calculate similar averages of thelevel of institutional ownership at the end of the second post-issue year (PIN2) conditional ontwo-year index-adjusted IPO returns. Table XIV reports these averages. They suggest that thereis a strong monotonic relation between past IPO performance and the institutional ownershippercentage, both at the end of the first post-issue year and at the end of the second post-issueyear. At both moments, institutional holdings of IPOs in the highest quartile of Amex-NYSEadjusted past performance are more than twice as high, on average, as holdings of IPOs in thelowest quartile of Amex-NYSE adjusted past performance. It is interesting to note that, by theend of the second post-issue year, institutions have built up an ownership level of 50.82%, onaverage, in IPOs of the highest quartile of Amex-NYSE adjusted past performance, a figure thatis very near to the 53% that they hold of the average U.S. stock.Table XIV Average Level of Percentage Ownership Held by Institutions at The End of TheFirst (PIN1) and Second (PIN2) Post-Issue Years by Past Performance Quartiles. The P-valuefor all averages is 0.0001.Quartiles of one-year Quartiles of two-yearIndex-adjusted performance PINI Index-adjusted performance PIN21 (lowest) 17.13 1 16.892 21.49 2 27.193 33.41 3 36.194 (highest) 39.07 4 50.8242The following figures try to shed some light on the question whether there is any rationality inholding an IPO portfolio24,at the end of the first post-issue year, tilted towards past winners.The IPOs in the top quartile of institutional ownership at the end of the first post-issue year,experience an Amex-NYSE adjusted return of 17.63% on average over the first year, while theTPOs in the bottom quartile of institutional ownership at the end of the first post-issue year,experience an adjusted return of, on average, -38.76%. The corresponding medians are 11.28%and -46.18%, respectively. These findings are consistent with the previously documentedstrategy of buying past winners and selling past losers. The average adjusted returns for thesubsequent year are as follows 2.25% for the top quartile of institutional holdings at the endof the first post-issue year and 31.63 % for the bottom quartile. The medians are -0.09% forthe top quartile and -25.27% for the bottom quartile. The medians show that, for investors whoare concerned about the left tail of the crossectional distribution of the returns of the stocks intheir portfolio - and the previous analysis suggests that this is the case for institutions withrespect to their IPO investments - it may indeed be rational to hold a portfolio tilted towards pastyear’s winners. Such a strategy does not result in superior returns but avoids negative adjustedreturns more frequently. The high mean value of 31.63% adjusted return for the bottom quartileis driven by a highly (ex-post) undervalued company at the end of the first year. The retailer,KOO KOO ROO Inc experiences a raw return of -92.77% over the first post-issue year. Thelevel of institutional ownership in this company at the end of the first year is 0%. Over thesubsequent year, however, this firm experiences a raw return of 1563%. If we drop this outlier,24 The term ‘portfolio’ is a bit inappropriate here. The measure of holdings used in theanalysis is the percentage of shares outstanding of a particular firm that institutions in theaggregate hold. Weights do not sum to 1.43the mean adjusted return for the bottom quartile falls to 1.13%, which is slightly lower than the2.31% for the top quartile.IV.4 Concentration versus Multiple Institutions HypothesisSufficient data are available to examine whether changes in aggregate institutional holdings arecaused by changes in holdings per institution or changes in the number of institutions holdinga particular stock. To distinguish between the two possibilities, regressions are estimated withpast and current quarter performance dummies as independent variables and with the rank of thechange in the number of institutions holding a positive stake over a given quarter as thedependent variable. Additional control variables are the number of institutions at the beginningof the quarter that hold shares of the firm (NUMNIV) and the percentage change in the numberof shares outstanding, adjusted for stock splits and stock dividends (SHRCH). The results ofthese regressions are given in Tables XV to XVII. They provide some interesting additionalinsights. First, the results are similar to the ones in the first series of regressions with the rankof the percentage change in aggregate holdings as dependent variable: the dummy coefficientsin most regressions have the same monotonically increasing pattern as in the first series ofregressions. This implies that an increase (decrease) in the percentage of a firm held byinstitutions in the aggregate is really caused by many institutions buying (selling) a small shareof the same stock at the same time25. It is evidence of parallel trading by institutions. Thefinding that additional money managers get in when the current price goes up makes the25 The correlation coefficient between the change in aggregate institutional ownershippercentage and the change in the number of institutions having a positive stake in the firm rangesfrom 0.40 to 0.65 over the eight quarters.44interpretation of fundamental feedback trading on price signals actually stronger. If managerswho already hold the stock simply expand their holdings when the stock has a high adjustedreturn, one could object to this interpretation because the manager already held the stock beforethe price run-up. The analysis, however, suggests that it is additional, probably less informed,managers who want to get in at a positive price signal. The strong positive relation betweenadjusted returns and the change over the subsequent quarter in the number of institutions holdinga positive stake in the stock confirms the cascading hypothesis in money managers’ tradesadditional managers, possibly reassured by or infering information from their colleagues’positions at the end of a given quarter, jump on the bandwagon the next quarter. The followingfigures are consistent with a number of institutions obtaining a stake in the firm during the firsttwo post-issue years (as opposed to one or a few institutions increasing their holdings) : whilethe average percentage of aggregate intitutional ownership over all 224 issues increases from17.7% at the end of the first post-issue quarter to 27.7% at the end of the first post-issue yearand to 32.8% at the end of the second post-issue year, the average over all 224 issues of thepercentage held by a ‘typical’ institution (i.e. the aggregate percentage divided by the numberof institutions) remains approximately constant: 0.96% at the end of the first post-issue quarter,1.33% at the end of the first post-issue year and 1.27% at the end of the second post-issue year.The small stake in IPO firms held by the average institution could be due to rules at the levelof individual institutions that restrict the amount invested in certain categories of firms. It couldalso be due to the Prudent Man Rule, which may lower the willingness of a money manager toinvest heavily in small, risky firms. Another interesting finding is that the variable NUMNIV,which represents the number of institutions with positive holdings in the firm at the beginning45of the quarter, is positively and in most regressions significantly related to the change in thenumber of institutions holding a positive stake. This is consistent with the conjecture that thenumber of institutions already holding shares of an IPO acts as a stimulant for other institutionsto also acquire a stake. This could be because of managers’ belief that colleagues who boughtshares previously must have superior information about the quality of the stock. Or it could beto a ‘sharing-the-blame’ effect : worldly wisdom teaches that is is better for reputation to failconventionally than to succeed unconventionally.26 The third interesting finding relates to thesignificant positive coefficient on SHRCH for the fourth quarter. This suggests that the numberof institutions holding a positive stake is significantly affected through secondary equityofferings. It is possible that institutions get a priority to acquire new shares during thesecondary offering. This could be the case if the offering is done through a private placement.If the secondary offering is large, and individual institutions acquire only small fractions, thenumber of institutions acquiring new shares could indeed be large.26 The results that many institutions are involved in a certain increase or decrease ofinstitutional holdings and that the number of institutions holding a particular IPO acts as astimulant for others to become involved, leads to an alternative interpretation of the significantlynegative coefficient on the variable PINAV (the aggregate institutional ownership percentagedivided by the number of institutions with a positive investment in a particular stock) in the firstseries of regressions. In those regressions, the rank of the change in the percentage ownershipof a company held by institutions in the aggregate is the dependent variable. Previously, Iinterpreted the coefficient on PINAV as a price sensitivity factor for a ‘typical’ institution thatholds stock in the firm. If a typical institution already holds a lot of stock, I conjectured, itwould buy relatively less when the firm performs well than if it held less stock. However,because the evidence in this section indicates that new trades involve new institutions, thisinterpretation is incorrect. Rather, the coefficient on PINAV is significantly negative becausePINAV is a ratio, with the number of institutions holding a positive amount of the stock in thedenominator, and we now know that this latter variable is positively related to the increase ininstitutional holdings in a given stock over a given quarter.46Table XV: Regression of RKNUM21 (= the rank of the change in the number of institutionsholding a positive stake in the firm) on Contemporaneous Performance Dummies.NUMNIV is the number of institutions holding a positive stake in the firm at the beginning ofthe quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.DUMAJ is a dummy related to performance adjusted by the Amex-NYSE index; DUMAJMAis a dummy related to performance adjusted by matched firm (P-values are in parentheses).Regression CoefficientsIndepend Var Concurrent PerformanceIntercept 77.82 85.29(0.0001) (0.0001)NUMNIV 0.38 0.61(0.2645) (0.0844)SHRCH 32.84 42.75(0.2459) (0.1438)DUMAJ2 11.20(0.3292)DUMAJ3 36.59(0.0017)DUMAJ4 59.21(0.0001)DUMAJMA2 12.69(0.2840)DUMAJMA3 3.77(0.75 17)DUMAJMA4 41.14(0.0006)R2 0.15 0.09P-value (*) 0.0003 0.0058(*) for the difference in dummy coefficients47Table XVI Regression of RKNUM43 (= the rank of the change in the number of institutionsholding a positive stake in the firm) on Contemporaneous and Lagged Performance Dummies.NUMNIV is the number of institutions holding a positive stake in the firm at the beginning ofthe quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.DUMAJ is a dummy related to performance adjusted by the Amex-NYSE index; DUMAJMAis a dummy related to performance adjusted by matched firm (P-values are in parentheses).Regression CoefficientsIndepend Var Concurrent Performance Lagged PerformanceIntercept 59.05 86.28 86.60 80.76(0.000 1) (0.000 1) (0.0001) (0.0001)NUMNTV 0.50 0.62 0.37 0.44(0.0349) (0.0155) (0.1627) (0.0800)SHRCH 74.02 80.02 93.96 81.79(0.0278) (0.0270) (0.0100) (0.0220)DUMAJ2 37.83 11.51(0.0005) (0.3400)DUMAJ3 50.53 17.12(0.0001) (0. 1666)DUMAJ4 75.98 32.46(0.0001) (0.0078)DUMAJMA2 -3.87 18.05(0.7378) (0.1133)DUMAJMA3 7.51 12.57(0.5 125) (0.2755)DUMAJMA4 37.80 50.07(0.0012) (0.0001)0.24 0.12 0.09 0.14P-value (*) 0.0026 0.0020 0.1872 0.0029(*) for the difference in dummy coefficients48Table XVII : Regression of RKNUM65 ( the rank of the change in the number of institutionsholding a positive stake in the firm) on Contemporaneous and Lagged Performance Dummies.NUMNIV is the number of institutions holding a positive stake in the firm at the beginning ofthe quarter.SHRCH is the percentage change in the number of shares outstanding during the quarter.DUMAJ is a dummy related to performance adjusted by the Amex-NYSE index; DUMAJMAis a dummy related to performance adjusted by matched firm (P-values are in parentheses).Regression CoefficientsIndepend Var Concurrent Performance Lagged PerformanceIntercept 69.09 83.25 77.09 83.43(0.0001) (0.0001) (0.0001) (0.0001)NUMNIV 0.61 0.68 0.63 0.67(0.0030) (0.0018) (0.0041) (0.0025)SHRCH 7.97 12.43 -7.68 9.53(0.7810) (0.6799) (0.8022) (0.7587)DUMAJ2 11.40 9.45(0.3133) (0.4186)DUMAJ3 34.78 40.42(0.0014) (0.0007)DUMAJ4 65.30 31.30(0.0001) (0.0084)DUMAJMA2 1.51 6.04(0.8968) (0.6081)DUMAJMA3 5.90 15.46(0.6061) (0. 1955)DUMAJMA4 40.89 27.95(0.0004) (0.0182)R2 0.20 0.12 0.11 0.08Pvalue(*) 0.0001 0.0015 0.0315 0.1939(*) for the difference in dummy coefficients49Chapter V : ConclusionIn the first part of the empirical analysis, the focus was on the long-term performance of twoIPO samples with issue years 1990 and 1991, respectively. For the 1990 sample, I found anaverage cumulative 3-year underperformance of approximately 23% : the 3-year buy-and-holdreturn on a portfolio of IPOs formed by investing $1 in each IPO at the end of the first listingdate is 23.2%, compared to an equivalent return for the index of 45.8% and to an equivalentreturn for the matched firm portfolio of 48.6%. For the 1991 sample, I found no evidence ofunderperformance during the first two post-issue years. Interesting, though, is the possibilitythat the much documented IPO underperformance is, at least partly, the result of anotherphenomenon known as mean reversion, possibly driven by investor overreaction, which affectsIPO returns differently than matched firm returns or the index. This possibility is suggested bythe finding that the average compounded yearly return on the portfolio consisting of the 1990IPOs rises from 7.2% for a buy-and-hold strategy to 13.0% for a rebalancing strategy, whererebalancing is done at each anniversary of the issue date. Rebalancing doesn’t affect the averageyearly compounded return on the index, and raises the equivalent return on the matched firmportfolio only by 0.2%, compared to a buy-and-hold strategy. Rebalancing to equal weights ata given moment is equivalent to putting more weight (relative to a buy-and-hold strategy) in pastlosers and less weight in past winners. If past extreme winners and/or past extreme losers tendto revert to the mean, a strategy of putting more weight at a given moment in past losers andless weight in past winners (relative to a buy-and-hold strategy) would indeed result in superiorreturns (relative to a buy-and-hold strategy).50In the second part of the empirical analysis, I examined the process of building up aninstitutional ownership percentage in IPOs from 17.7% to 27.7%, on average, over the firstpost-issue year and from 27.7% to 32.8%, on average, over the second post-issue year. Inparticular, I examined how the process relates to the adjusted performance of the IPOs. Theanalysis suggests that many money managers are involved in the stock acquisition process.Some managers, having no access to superior information on future IPO payoffs, update theirexpectations when current prices move and follow a positive fundamental feedback tradingstrategy. A concern about reputation and/or over the fiduciary duty may strengthen the ‘buywinners-sell losers’ pattern. Other managers react to overperformance with a one-quarter lag.This behavior may be driven by a belief that past trends are likely to continue. Alternatively,they may be reassured by the positions of their colleagues as presented in the quarterlystatements, and engage in similar trades during the subsequent quarter. Given that the managerswho trade on current price movements follow a positive feedback trading strategy, mimickingthem during the subsequent quarter induces a positive relation between changes in institutionalholdings and past quarter’s adjusted returns. The relation between changes in holdings andlagged or contemporaneous returns for institutions as a whole is stronger than the relationdocumented by Lakonishok, Shleifer and Vishny (1991) for pension funds. Also, the evidenceof herding presented here is stronger than in their study. As mentioned previously, their findingof a positive relation between holdings and returns and their evidence of herding are limited tofirms in the two smallest size quinti]es of their sample. A factor that could explain part of thestronger results is size. As shown in Table II, the firms in my sample are relatively small73.2% of the firms studied have a market value at the end of the first post-issue quarter51pertaining to the five lower size deciles as defined by the Amex and NYSE stocks. I don’tthink, however, that the size factor is the whole explanation. The stronger result could be partlydue to the particular nature of IPOs : First, there is great uncertainty as to what their exactvaluation is. Less informed investors update their expectations strongly by current pricemovements. Parallel trading and intertemporally correlated trades could occur becauseinstitutions try to infer information about a certain IPO from other institutions’ holdings andtrades. Second, because IPOs are relatively unknown and small firms, and generally perceivedas risky investment, money managers may be particularly concerned about the prudent investorrule and/or the bad impression given when they present holdings of an underperforming IPO.Parallel trading and intertemporally correlated trades could also occur because of this factorif a certain IPO turns out to be a bad investment, at least, one will not be perceived as a lonefool. Or, if a lawsuit occurs, at least, one has the excuse that other institutions were holdingthe same stock.52BibliographyAdmati, A. R., P. Pfeiderer and J. Zechner, 1993, Large Shareholder Activism, Risk Sharingand Financial Market Equilibrium, Working Paper, University of British Columbia.Agrawal, A. and G.N. Mandelker, 1990, Large Shareholders and the Monitoring of Managers:The Case of Antitakeover Charter Amendernents, Journal of Financial and Quantitative Analysis,143-161.Beatty, R.P. and J. Ritter, 1986, Investment Banking, Reputation, and The Underpricing ofInitial Public Offerings, Journal of Financial Economics, 213-232.Black, B., 1992, Agents Watching Agents: The Promise of Institutional Investor Voice, UCLALaw Review 39, 811-893.Buser, S. A. and Chan, K.C., 1987, NASDAQ/NMS qualification standards, Ohio RegistrationExperience and The Price Performance of Initial Public Offerings, Columbus, Ohio Departmentof Commerce and National Association of Dealers, Inc.Brennan, M. and H.H. Cao, 1994, Information, Trade and Derivative Securities, WorkingPaper, University of California.Brickley, J., R. C. Lease and C. W. Smith, 1988, Ownership Structure and Voting onAntitakeover Amendments, Journal of Financial Economics, 267-291.De Bondt, W. and R. Thaler, 1985, Does the Stock Market Overreact, Journal of Finance, 793-808.Degeorge, F. and R. Zeckhauser, 1993, The Reverse LBO Decision and Firm PerformanceTheory and Evidence, Journal of Finance, 1323-1348.De Long, J.B., A. Shleifer, L.H. Summers and R.J. Waldmann, 1990, Positive FeedbackInvestment Strategies and Destabilizing Rational Speculation, Journal of Finance, 379-395.Drorns, W., 1992, Fiduciary Responsibilities of Investment Managers and Trustees, FinancialAnalysts Journal, 58-64.Heliwig, M. F., 1980, On The Aggregation of Information in Competitive Markets, Journal ofEconomic Theory, 477-498.Hendricks, D., J. Patel and R. Zeckhauser, 1993, Hot Hands in Mutual Funds : Short-RunPersistence of Relative Performance, 1974-1988, Journal of Finance, 93-129.53Ibbotson, R. G., 1975, Price Performance of Common Stock New Issues, Journal of FinancialEconomics, 235-272.Jensen, M. C., 1968, The Performance of Mutual Funds in The Period 1945-1964, Journal ofFinance, 389-416.Keloharju, M., 1993, The Winner’s Curse, Legal Liability, and The Long-Run PricePerformance of Initial Public Offerings in Finland, Journal of Financial Economics, 251-277.Kraus, A. and H. Stoll, 1972, Parallel Trading by Institutional Investors, Journal of Financialand Quantitative Analysis, 2107-2137.Lakonishok, J., A. Shleifer and R. Vishny, 1992, The Impact of Institutional Trading on StockPrices, Journal of Financial Economics, 23-43.Lakonishok, J., A. Shleifer and R. Vishny, 1992, The Structure and Performance of the MoneyManagement Industry, Brookings Papers on Economic Activity: Microeconomics, 339-391.Lehmann B. and D. M. Modest, 1987, Mutual Fund Performance Evaluation : A Comparisonof Benchmarks and Benchmark Comparisons, Journal of Finance, 233-265.Levis, M., 1993, The Long-Run Performance of IPOs: The UK, Financial Management, SpringEdition, 28-41.Mikkelson, W.H. and K. Shah, 1994, Performance of Companies Around Initial PublicOfferings, Working Paper, University of Oregon.Ritter, J., 1984, The ‘Hot Issue’ Market of 1980, Journal of Business, 2 15-240.Ritter, J., 1991, The Long-Run Performance of Initial Public Offerings, Journal of Finance, 3-27.Scharfstein, D.S. and J.C. Stein, 1990, Herd Behaviour and Investment, American EconomicReview, 465-479.Sharpe, W., 1966, Mutual Fund Performance, Journal of Business, 119-138.Shleifer, A. and R. Vishny, 1986, Large Shareholders and Corporate Control, Journal ofPolitical Economy, 46 1-488.Smith, J.W., 1986, Investment Banking and The Capital Acquisition Process, Journal ofFinancial Economics, 3-25.Stern, R. L. and P. Bornstein, 1985, Why Issues Are Lousy Investments, Forbes, 152-190.54Stoll, H.R. and A.J. Curley, 1970, Small Business and The New Issues Market for Equities,Journal of Financial and Quantitative Analysis, 309-322.Truernan, B., 1988, A Theory of Noise Trading in Securities Markets, Journal of Finance, 83-95.Treynor, J., 1965, How to Rate Management of Investment Funds, Harvard Business Review,63-75.55

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0087461/manifest

Comment

Related Items