The Effects of Stock Splits on Pricing and Bid-Ask Spread of Syndicated Loans

Transcription

1 The Effects of Stock Splits on Pricing and Bid-Ask Spread of Syndicated Loans Bill B. Francis Rensselaer Polytechnic Institute - Lally School of Management and Technology th Street, Troy, NY 12180, USA Iftekhar Hasan hasan@rpi.edu Rensselaer Polytechnic Institute - Lally School of Management & Technology th Street, Troy, NY 12180, USA Mingming Zhou m.zhou@uaf.edu University of Alaska Fairbanks School of Management 219B Bunnell Building, PO Box , Fairbanks, AK Working Paper January

2 The Effects of Stock Splits on Pricing and Bid-Ask Spread of Syndicated Loans ABSTRACT Stock splits have long challenged the standard textbook analysis (e.g., Brealey and Myers (1991, p ) that regards splits as a purely cosmetic change, as it is well documented since the seminal article by Fama et al. (1969) that splits are associated with real economic effects. However, existing studies are still inconclusive on whether stock splits signal private information as the signaling hypothesis predicts. Studies that test signaling hypothesis based on equity-based data may have contaminated results because the changes in the share price mandated by splits may cloud the signaling effects. In addition, there are no existing studies that document the effect of stock splits on the debt of the firm. Our study is motivated by the conjecture that stock splits signal private information that is firm-wide rather than equity-wide. Based on this conjecture, we attempt to achieve a cleaner test of signaling hypothesis by utilizing the loan pricing data. More specifically, we empirically examine the effect of stock splits on the bid-ask spread of the syndicated loans, which serves as a measure of information asymmetry (or opaqueness) of the borrowing firm. In addition, we empirically explore whether the private information signaled by splits (if there is any) has affected the prices of loans. Based on a sample of firms that announced stock splits during 1999 to 2004, we find that bid-ask spread of the loans significantly decreases around the splits event, implying that splits signal private information to the market, and such signaling reduces the opaqueness of the firm. Our results also show that loans experience significant negative abnormal returns around the splits, both wealth transfer and prepayment risk hypotheses are supported by further cross-sectional analysis. Keywords: Stock splits Bid-ask spread Secondary loan market 2

3 The Effects of Stock Splits on Pricing and Bid-Ask Spread of Syndicated Loans I. Introduction Stock splits have long puzzled researchers in finance. A split, which just subdivides the existing number of shares into a greater number of units, on the surface, is regarded by most standard textbook analysis (e.g., Brealey and Myers (1991, p )) as a purely cosmetic change that does not affect a firm s value. Yet many studies (see, e.g., Bar-Yosef and Brown (1977), Charest (1978), Grinblatt et al (1984), Lamoureux and Poon (1987), Muscarella and Vetsuypens (1996)) since the seminal article by Fama et al. (1969) document that splits are associated with real economic effects, i.e., positive valuation effects. One of the widely studied hypotheses that aim to explain the positive valuation effects associated with splits is the signaling hypothesis, which is formally proposed by Brennan and Copeland (1988a) to the splits literature. Based on the signaling hypothesis, stock splits signal private information to investors. However, inconsistent and even contradictory results are found regarding whether and what private information is signaled by splits. 1 In addition, to validate the hypothesis that splits are costly signal of favorable future prospects for the firm, one must show that there are some associated costs with splits that prevent those inferior quality firms from mimicking. One of these costs, as suggested by Brennan and Copeland (1988a), among others, are the increased liquidity costs by stock splits. When liquidity is defined as the ease or rapidity with which a financial instrument can be exchanged for currency (see Hicks (1946)), the bid-ask spread is a more direct measure of the cost of the liquidity than the trading volume, and still evidence is inconclusive. 2 Besides, the existing studies which seek evidence of signaling hypothesis unanimously examine only equity-based data, and because an split event itself, even without any real economic effects, automatically implies the changing of the share price and number of shares outstanding of the stocks, the results of signaling effects of splits documented by 1 For example, Lakonishok and Lev (1987), Conroy and Harris (1999), Ikenberry and Ramnath (2002), Louis and Robinson (2005), among others, empirically document that there are abnormal increases in earnings and/or dividends around and after splits, while some studies (see, e.g., Nayak and Prabhala (2001), Huang et al. (2002)) find that stock splits do not necessary imply future increase of dividends or profitability. 2 For example, Conroy et al. (1990), based on a sample of 133 NYSE firms that voted a stock split between January 1981 and April 1983, find that percentage bid-ask spreads increase after splits, and such increases are directly related to decreases in share prices following splits. Contradictorily, Goyenko et al. (2006) track the liquidity of thousands of stock splits taking place from 1963 through 2003, and find that percent bid-ask spread of splitting firms only temporarily increase and then return to even with control firms in 5 to 12 months. 3

4 these studies may be potentially contaminated by the underlying changes in the share price and outstanding shares also. In other words, the compounded effects of the private information signaling and changes in the price and outstanding shares in the event of stock splits make the equity-based studies of stock splits questionable. In addition, despite of the abundant existing studies on stock splits, there is little effort or documentation on the effects of stock splits on the firm s debt side, such as loans that are borrowed from the banks. The spillover effect of stock splits on debt is an important issue to examine for two reasons. First, it is reasonable to assume that stock splits can have firm-wide impact rather than equity-wide impact if splits really signal private information of the firm. Second, as debt is also an important component of a firm s assets, it is interesting to see whether there is any real economic effects of stock splits on the firm s debt, such as the loans. Indeed, debt such as bank loans has received tremendous attention from both practitioners and academicians regarding a broad spectrum of issues, but (probably) due to data availability constraints, there are no existing studies that examine the effect of stock splits on the debt of the firms to our best knowledge. The secondary market data of the syndicated loans, available in Loan Trade Database (provided by Reuters Loan Pricing Corporation, i.e., Reuters LPC), has made empirical research on the daily pricing of loans possible. More specifically, Loan Trade Database provides loan trading data including the borrower s name, indicative loan bid and ask price quotes aggregated across dealers, and the number of market makers. According to Reuters LPC estimates, the Loan Trade Database covers 80% of the trading volume in the secondary loan market in the U.S. Such features of the Loan Trade Database, along with the associated DealScan database also provided by Reuters LTC, make the syndicated loan market a promising setting to test issues such as debt contracting and debt pricing. In this paper, we utilize the pricing information from Loan Trade Database to test the economic effects of splits on loans. Loan trading data offers a methodological advantage in evaluating the signaling hypothesis of stock splits: we utilize the bid-ask spread of the loans in the secondary market as a measure of information asymmetry associated with the borrowing firm (see more discussions in Moerman (2005)), and because such measure is not contaminated by the share price change mandated by stock splits, any changes of the bid-ask spread of the loans can be attributed to the private information signaling effects of stock splits. Further, if stock splits serve as a consistent signal of private information in the extent 4

5 that the information opaqueness of the firm is reduced, we should find a significant decrease in bid-ask spread of loans of splitting firms around the split announcement window. Alternatively, we should observe no significant changes or even an increase in the spread if stock splits act as a noisy signal that does not give any private information or even increases the opaqueness of the firm. Based on a sample of firms which announce stock splits during 1999 to 2004, we find that the bid-ask spread of the loans experience a significant reduction during the splits event window, which provides support for the hypothesis that splits signal private information to the market. In addition, we utilize the loan trading data to examine the effect of stock splits on the returns of the loans. We assume that if stock splits are associated with firm-wide information (e.g., growth of future earnings), then we should observe abnormal returns of loans around the splits window. Alternatively, the abnormal returns of loans should be zero if there is no firm-wide information signaled by stock splits. Our results show that loans, on average, experience significant negative abnormal returns during the splits window. We provide two non-exclusive hypotheses regarding the negative returns on loans during the splits event. The first is the wealth transfer hypothesis. Wealth transfer might happen in the event of splits because the splits event may imply changes of firm-wide riskiness, and when the riskiness of the firm changes, there will be a wealth transfer between stockholders and debtholders, due to the embedded call option in the stocks (see, e.g, Saunders et al., (1990)). Indeed, increases in firm risk around stock splits are supported by studies of beta changes around stock splits (e.g., Brennan and Copeland (1988)), though it is arguable that beta stands for the firm-wide or stock-wide riskiness. Such changes in the firm-wide risk may be caused by reduced holdings and monitoring of institutional investors (Mukherji et al. (1997), among others, provide evidence that splits can induce changes in the splitting firm s shareholder structure or ownership concentration). In this paper, we conjecture that stock splits induce increases in the firm s riskiness, which leads to wealth transfer between stockholders and loan-holders. Our second explanation regarding the negative abnormal return on loans is the prepayment risk hypothesis. When borrowers have better growth potentials or improved financial conditions, probabilities of loan prepayment from these borrowers increase. The increase of expected prepayment risk will lead to decrease of loan prices. In case of stock splits, the favorable information signaled by splits can hurt the loan prices because the favorable information signal increases the expected prepayment risk. Our 5

6 cross-sectional analysis empirically supports both wealth transfer and prepayment hypotheses. Taken as a whole, our study contributes to the literature in the following ways. First, we find support for the signaling hypothesis of stock splits by constructing a cleaner test and utilizes the bid-ask spread of secondary market loans as the measure of the information asymmetry of the splitting firms. Second, we document a negative spill-over effect of stock splits on the pricing of loans, which is a strong indication that splits have firm-wise impact rather than stock-wide impact. Third, we provide evidence that there is wealth transfer between stockholders and loan-holders, as the abnormal returns on loans are negatively and significantly correlated. We hope that the findings of this study can offer more insights into the puzzles around stock splits. The remainder of this paper is organized as follows. Section II presents existing theory and findings. Section III briefly describes the growth of secondary market for bank loans. Section IV presents our sample collection procedure and methodology. Section V summarizes our empirical results and Section VI concludes. II. Literature Review The notion that financial decisions convey information about firm value was proposed by Ross (1977), Leland and Pyle (1977), and Bhattacharya (1979) in adaptations of the Spence (1973) signaling model. This framework assumes that (a) asymmetric information between managers and investors, (b) managers have an incentive to convey favorable information to investors, and (c) it is prohibitively costly for low-value firms to mimic the financial decisions of high-value firms. However, unlike most cash dividends and capital structure changes, stock splits do not directly affect the future cash flows of the firm, and since there are no apparent costs to stock splits, the traditional signaling argument implies that these events should have no information content. However, numerous studies have provided empirical evidence that there are, in general, positive stock returns associated with stock splits. For example, the early seminal work on splits by Fama, et al. (1969), and then later by Bar-Yosef and Brown (1977), and Charest (1978), among others, presents evidence which is interpreted as a split announcement effect. However, these studies have some important limitations as suggested by Fama (1976). As an 6

7 effort to control for the other information releases on the returns at the split announcement dates that potentially contaminate previous studies, Grinblatt et al (1984) examine three different samples and the findings lead them to conclude that the valuation effect of split announcement, for the most part, cannot be explained by forecasts of near term increases in cash dividends. Studies based on more recent samples also provide consistent findings that there are positive abnormal returns around stock splits announcement (e.g., Muscarella and Vetsuypens (1996)). With regard to the persistent empirical findings of positive valuation effects of stock splits, existing literature has proposed many theories and explanations, but little consensus is achieved till now. Among others, some studies interpret the positive stock market reaction to split announcements as a response to managers signaling favorable insider information (see, e.g., Brennan and Copeland (1988a)). The basis of Brennan and Copeland (1988a) s argument is that it is costly for a firm to reduce its share price by splitting because the structure of brokerage commissions makes it more costly to trade in low priced shares. The relation between splits and trading costs has been acknowledged in the business press (see, e.g., Wall Street Journal, October 13, 1989). Signaling hypothesis is supported by studies that empirically document the abnormal increases in earnings and/or dividends around and after splits (see, e.g., Lakonishok and Lev (1987), Doran and Nachtmann (1988), Asquith et al. (1989), McNichols and Dravid (1990), Conroy and Harris (1999), Ikenberry and Ramnath (2002), Louis and Robinson (2005)). Lakonishok and Lev (1987) use a sample of 1,015 stock splits that occurred from 1963 to 1982, and find that earnings of splitting firms in the first quarter following the splits is significantly larger than non-splitting firms, suggesting that managers of splitting firms signal to investors that earnings increase prior to split is not temporary. Doran and Nachtmann (1988) examine the relationship between stock splits and earnings realized after a split compared to the expected earnings before the split, and they find that earnings realized after splits are greater than expected. Asquith et al. (1989), based on a sample of 121 NYSE and AMEX firms that issued splits of at least 1.25-for-1 during 1970 to 1980, report that earnings of splitting firms increases significantly in the four years before splits announcement, and the splitting firms are generally the best performers in the wellperforming industries. McNichols and Dravid (1990) construct a sample consisting of 1,376 7

8 stock splits and stock dividends during 1976 to 1983 and provide evidence that firms signal their private information about future earnings by their choice of split factor, and investors revise their beliefs about firm value accordingly. Conroy and Harris (1999) provide further evidence that in cases that firms make larger than expected split, analysts tend to upward adjust their earnings forecasts, and such adjustment is a direct confirmation of an information effect linked to stock split. Ikenberry and Ramnath (2002) find that splitting firms have an unusually low propensity to experience a contraction in future earnings as compared to non-splitting firms. Louis and Robinson (2005) find evidence that the stock splits signal lends credibility to the accrual signal whereas the accrual signal reinforces the split signal, and the market construes the pre-split abnormal accruals as a signal of managerial optimism rather than managerial opportunism. Consistent with the hypothesis that future fundamental financial performance is a source for the positive valuation effect of stock splits, some studies (see, e.g., Ikenberry et al. (1996), Desai and Jain (1997)) further find positive abnormal returns in years subsequent to stock splits. Ikenberry et al. (1996) observe significant post-split excess returns of 7.93% in the first year and 12.15% in the first three years for a sample of 1,275 two-for-one splits initiated from 1975 through 1990, which suggests that managers self-select by conditioning the splits decision on expected future performance. Similarly, Desai and Jain (1997) based on a sample of 5,596 stock splits announcements made during , find that on average, the 1-and 3-year buy-and-hold abnormal returns after the announcement month are 7.05% and 11.87%, respectively. In addition, they find that both abnormal returns during the announcement period and the long-run window are positively associated with an increase in dividends. However, in contrast of the evidence on favorable future financial performance of splitting firms, some studies (see, e.g., Nayak and Prabhala (2001), Huang et al. (2002)) find that stock splits do not necessary imply future increase of dividends or profitability. Nayak and Prabhala (2001) find that splits and dividends are informational substitutes and a significant portion of the split valuation effects, 46% according to their estimates, cannot be attributed to dividend information in splits. Huang et al. (2002), based on a sample includes 635 split announcements during 1982 through 1997 that have both a not-close-to-themedian post split share price and split factors equal to or less than 0.5, find little evidence that stock splits are positively related to future profitability. Instead, they find that stock 8

9 splits are in general negatively related to future profitability in subsequent years after the announcement. Consistent with studies that find no favorable dividends or earnings after splits, a number of studies find negligible evidence of positive long-run performance of splitting firms, or even evidence of overreaction or mean reversion in the prices of splitting firms. For example, DeBondt and Thaler (1985, 1987) show that firms that experience large returns over 3-5 years display mean reverting returns in subsequent years (see DeBondt and Thaler (1989) for a summary and discussion of overreaction and mean reversion in stock prices after stock splits announcements). A more recent study, Byun and Rozeff (2003), measures the post-split performance of 12,747 stock splits from 1927 to 1996 and do not find splitting firms enjoy significantly superior performance than those non-splitting firms. Another version of the information asymmetry hypothesis is the attention-getting hypothesis. Brennan and Hughes (1991) develop a model in which the managers with favorable private information about their firms have an incentive to split their firm s shares in order to attract the attention of security analysts, given that more analysts will produce earnings forecasts based on the lower share price. Brennan and Hughes (1991) find empirical evidence that supports the prediction of their model that the number of analysts following a firm is inversely related to its share price. Some other studies (e.g., Grinblatt, et al. (1984), Brennan and Hughes (1991)) provide additional evidence that splits draw increased attention to a stock, which triggers positive revaluation of the splitting stock. Alternative explanations regarding to abnormal reaction of stock prices to splits, among others, include: (a) Heinkel (1984) s reputation model. Heinkel (1984) suggests that firms tend to maintain honest signaling reputations in order to have the opportunities to signal favorable information in the future. Consistent with this hypothesis, some empirical studies (e.g., Fama et al. (1969)) attribute split valuation effects mainly to an implied promise of higher dividends. (b) Trading range hypothesis. Trading range hypothesis suggests that managers split to keep the price of their shares within a customary trading range. Given the costs associated with splits and reversals, managers with unfavorable inside information might decide not to split, even if their firm s stock price is high, because they expect that future events will force the price of the splits shares to fall below the customary trading range. Investors, observing the correlation between splits and subsequent stock performance, could then use the split announcement to draw inferences about this information. Some 9

10 studies (e.g., Baker and Gallagher (1980), Lakonishok and Lev (1987), Baker and Powell (1993)) provide evidence that suggest that splits return stock prices to an optimal trading range. (c) Private communications. Grinblatt et al. (1984) propose that the valuation effect of stock splits come from the communications between securities analysts, fund managers and corporate managers at the announcement of a split or stock dividend in order to ascertain the announcement s implications for future cash dividends payouts. (d) Taxoption hypothesis. Lamoureux and Poon (1987) argue that the increase in idiosyncratic volatility of a stock after a split increases the tax-option value and hence the price of a stock. (e) Liquidity hypothesis. Most recently, Muscarella and Vetsuypens (1996), who analyze ADR solo splits, suggest that improved liquidity explains positive split valuation effects. For a comprehensive review of stock splits literature, please see Baker et al. (1995). To validate the hypothesis that splits are costly signal of favorable future prospects for the firm, one must show that there are some associated costs with splits that prevent those inferior quality firms from mimicking. Brennan and Copeland (1988a), among others, argue that if stock splits increase liquidity costs to investors, such splits can serve as valid signals. The existing literature provides abundant studies on the liquidity costs associated with splits. Most empirical studies of the liquidity effects of splits use trading volume as liquidity proxy. The use of volume to measure liquidity is largely motivated by empirical findings of a negative relationship between volume and the bid-ask spread (see, e.g., Demsetz (1968), Benston and Hagerman (1974), and Tinic (1972)). Among others, Copeland (1979), Lamoureux and Poon (1987), and Murray (1985) conclude that shareholder liquidity decreases after a split based on finding dollar trading volume decreases after splits. Contradictorily, Lakonishok and Lev (1987) find abnormally high volume in the year preceding a split and, abstracting from this abnormal volume, that splits are not associated with a change in trading volume. When liquidity is defined as the ease or rapidity with which a financial instrument can be exchanged for currency (see Hicks (1946)), the bid-ask spread is a more direct measure of the cost of the liquidity than the trading volume. Some studies provide evidence of liquidity costs of splits based on stock bid-ask spread measures. For example, Conroy, Harris, and Benet (1990), based on a sample of 133 NYSE firms that voted a stock split between January 1981 and April 1983, find that percentage bid-ask spreads increase after splits, and such increases are directly related to decreases in share prices following splits. These spreads 10

11 increase, as they argue, represent a liquidity cost to investors, and thus validate that stock splits are a signal of favorable information about the firms. Contradictorily, Goyenko et al. (2006) track the liquidity of thousands of stock splits taking place from 1963 through 2003, and find that percent bid-ask spread of splitting firms only temporarily increase and then return to even with control firms in 5 to 12 months, while splitting firms have significantly lower spread than that of the control firms in 12 to 39 months. Goyenko et al. (2006) s evidence of a net benefit of liquidity for splitting firms seems to suggest a missing link for the signaling hypothesis. As we discuss above, existing studies provide inconclusive evidence on the signaling hypothesis and its assumptions of liquidity costs. In addition, these studies raise serious concern about their results when their tests of signaling effects of splits are equity-based. For example, in studies that test the effects of stock splits on bid-ask spreads (e.g., Copeland (1979), Murray (1985), Conroy et al. (1990), among others), they use the bid-ask spreads of the exactly same stocks that splits, and because the changes in the bid-ask spreads can resulted from a mixture of both liquidity effects and signaling effects, and it is possible that the liquidity effects are so overwhelming that they dominate the empirical findings, these studies can barely yield convincing conclusions of signaling effects of stock splits. Our study is motivated by the conjecture that stock splits signal private information that is firm-wide rather than equity-wide. If this is true, then we should also expect that the splits affected value of the debt of the firm. Because debt is not splitted as the stocks are, we can achieve a cleaner test of signaling hypothesis based on the debt pricing data. The secondary market data of the syndicated loans, available in Loan Trade Database (provided by Reuters Loan Pricing Corporation, i.e., Reuters LPC), has made empirical research on the daily pricing of loans possible. In fact, the bank loans have received tremendous attention from academicians over decades. Banks, who lend to corporations, are considered special for several reasons, including reducing the agency costs of monitoring borrowers (e.g., see Saunders (2002) for a comprehensive review of why banks are considered special). Some more recent studies further examine the role of bank monitoring in the presence of loan sales in the secondary market. Kroszner and Strahan (2001) find that the lead bank, which typically holds the largest share of a syndicated loan, rarely sells its share of a loan in order to preserve its banking relationship with the borrower. Taylor and Yang (2003) document that the 11

12 changing role of banks, from loan originations to loan dealers and traders, which has facilitated the development of a secondary market for loans, may provide additional channels of monitoring. Consistently, Altman et al. (2006) also find evidence that support the argument that the bank advantages and incentives to monitor are likely to be preserved in the secondary loan market. The monitoring role of banks in the secondary loan market makes the loan pricing more efficient than the bond market and even the stock market, or in other words, new information is incorporated into loan pricing more quickly (see, e.g., Altman et al. (2006)). Such features of the secondary loan pricing has provided another motivation of our study, which is first motivated by the conjecture that stock splits signal private information that is firm-wide rather than equity-wide. More specifically, we utilize the bid-ask spread on the firm s loans traded on the secondary loan market as a measure of information asymmetry associated with the borrowing firm (see more discussion on the validity of this measure in Moerman (2005)) and test the effects of stock splits on the bid-ask spread and prices of the loans of the firm. To our best knowledge, our paper is the first study that documents the effects of splits on the loans. III. The Secondary Market for Bank Loans: Background The U.S. loan market bridges the private and public debt market and provides borrowers with an alternative source of financing to high yield bonds and relationship-based bilateral bank loans. During the past decades, the market for loans has expanded rapidly. There are generally two broad categories typically included in the market for loans. The first is the primary or syndicated market, in which portions of a loan are placed with a number of banks, often in conjunction with, and as part of, the loan origination process (usually referred to as the sale of participations). After the close of the primary syndication, syndication debt instruments can be traded on the secondary loan market, which is the second category. We focus on the latter category of the loan sales market in this study. Even though banks and other financial institutions have sold loans among themselves for over 100 years, this market grew slowly until the early 1980s when it entered a period of spectacular growth, largely due to growth in highly leveraged transactions (HLT) loans to finance leverage buyouts (LBOs) and mergers and acquisitions (M&As). With the decline in 12

13 LBOs and M&As in the late 1980s after the stock market crash of 1987, the volume of loan sales fell to approximately $10 billion in 1990 (source: Thomson Financial Securities Data Corporation). However, since then the volume of loan sales has expanded rapidly. From a trading volume of $8 billion in 1991, the secondary loan market has increased to a trading volume of $144.6 billion in 2003 (source: Gold Sheets, Loan Pricing Corporation). Syndicated loans are floating rate debt issues, priced at a specified interest rate spread above a reference rate; the most frequently used pricing options include Prime, LIBOR and Certificate of Deposit. Syndicated loans are always senior debt instruments. Another distinct feature of syndicated loans is the inclusion of extensive financial covenant restrictions in the loan agreement. These covenants are typically calculated quarterly and provide syndicate lenders with considerable control over a borrowers actions (Assender, 2000). The secondary loan sales market is sometimes segmented based on the type of investors involved on the buyside, e.g., institutional loan market versus retail loan market. An alternative way of stratifying loan trades in the secondary market is to distinguish between the par loans (loans selling at 90% or more of face value) versus distressed loans (loans selling at below 90% of face value). IV. Sample and Methodology A. Sample Selection Our sample consists of U.S. firms that announced splits between January 1, 1999 to December 31, 2004 based on records from Center for Research in Securities Prices (CRSP). Multiple databases are utilized when we collect the information on equity, and loans of these firms. Due to the absence of a unique identifier that ties all these datasets together, these datasets had to be manually matched based on the name of the company and/or other identifying variables. The following presents detailed description of our sample construction procedures. We first collect information of loans for our sample firms. Two major datasets are utilizied here, and both are provided by Reuters Loan Pricing Corporation (Reuters LPC): Loan Trade Database and the DealScan database. Loan Trade Database provides loan trading data including indicative loan bid and ask price quotes aggregated across dealers, 13

14 reported to LPC by trading desks at institutions that make a market in these loans. In addition to the pricing information, the Loan Trading Database also provides the borrower s name, quote data and the number of market makers reporting indicative price quotes to LPC. According to LPC estimates, the Loan Trade Database covers 80% of the trading volume in the secondary loan market in the U.S. On the other hand, DealScan database provides historical database of detailed deal teams and conditions on loans and bonds from around the world, and according to LPC, DealScan covers over 155,000 loan and bond transactions, including loans from 6686 public U.S. companies and private U.S. companies. However, a notable feature of companies included in the DealScan and Loan Trading Database is that these companies are significantly larger in size than the average companies in CRSP database. Therefore, any generalization of the results based on DealScan and Loan Trading Database should be of caution because the sample firms in these databases are biased towards very large companies. We match the Loan Trade Database to the DealScan database, based on the Facility-ID and/or Loan Identification Number (LIN). 3 Connecting these two databases allows us to identify the deal characteristics of the traded loans, such as the loan size, maturity, seniority, securitization, covenant package and syndicate structure. Connecting Facility-IDs with LINs is possible because both DealScan database and Loan Trade Database provide lists of matched Facility-IDs and LINs. But a rough combination of these lists will give multiple Facility IDs associated with the same LINs. For this reason, we hand-match the list of Facility-IDs with LINs. The secondary loan market database provided by LTC has 2,742,272 total number of trading observations during period from January 1, 1999 to December 31, We drop those observations either with missing LIN/Facility-ID numbers (50,571 trading observations), or with LINs with less than 13 digits (97,618 trading observations). 4 We then match the LTC database with the DealScan database, and we are able to identify 2,580,577 trading observations connecting to 9638 LINs, where 866 LINs cannot find the corresponding Facility-IDs. These facilities are syndicated to 2,531 borrowers. We further 3 Facility is a tranch/whole of a type loan arrangement between company and a lender/syndicate of lenders. The different types of facilities in an umbrella of arrangement between a company and lender/group of lenders form a package. Facility- ID is a number assigned by LPC to each syndicated facility on the primary loan market. LIN is assigned to each syndicated facility that is traded on the secondary loan market. 4 According to LPC, observations missing Facility-ID and LIN identifiers on the Loan Trade Database belong to the period when LPC just started covering the secondary loan market, and LINs with less than 13 digits are assigned to the trading facilities which cannot directly matched with records in DealScan database. 14

15 limit our sample to firms which are denoted as USA country original and USA country of syndication, and now there remains 5,900 LINs (i.e., facilities) which are syndicated to 1,950 borrowers. We then match the borrowers with the list of stock splits events obtained from CRSP database. A split is included in the list if it was announced between January 1, 1999 to December 31, There are 1803 stock split announcements of 1371 stocks over the sample period, from 1348 firms. After we further match the CRSP split list with DealScan and LTC database, we are able to identify 140,176 trading observations of 303 LINs (and Facility-IDs) from 97 borrowers. We collect the stock daily prices data from CRSP database, stock bid-ask consolidated quotes from TAQ database for the 97 borrowers that mentioned above. We only extract the common stock observations from CRSP and TAQ and we further drop observations that are in the banking industry (two-digit SIC codes 60 and 61). Our source for loans and stock index returns are the S&P/LSTA Leverage Loan Index from Standard & Poors, and the NYSE/AMEX/NASDAQ Value-weighted Index from CRSP. The LSTA (Loan Syndications and Trading Associations), in conjunction with Standard & Poor's/LCD, has developed the S&P/LSTA Leveraged Loan Index (LLI) in order to provide investors with a performance benchmark. The LLI is a weekly total return index that uses LSTA/LPC Mark-to-Market Pricing to calculate market value change. On a real-time basis, the LLI tracks the current outstanding balance and spread over LIBOR for fully funded term loans. The facilities included in the LLI represent a broad cross section of leveraged loans syndicated in the United States, including dollar-denominated loans to overseas issuers. According to LSTA, by December 31, 2001, the LLI consisted of approximately 470 facilities and $104 billion in outstandings, which represents approximately 70% of the institutional universe, a coverage comparable to the S&P 500's coverage of the equity universe. Finally, we match our sample with Compustat database to identify more firm-level characteristics. B. Methodology We start with event study analysis to examine the impact of stock splits on price of stocks and secondary market loans. We measure return performance by cumulating daily abnormal returns during a pre-specified time period. Specifically, we present empirical evidence for three different windows surrounding the split event: 3-day window [-1,+1], 11-15

16 day window [-5,+5] and a 21-day window [-10,+10], and define the estimation period as [- 70,-11], where day 0 refers to the stock split announcement date. We use several different methods to compute daily abnormal returns. First, on an unadjusted basis, i.e., using the raw return, as a first-approximation of the magnitude of the return impact on a stock or loan of the same corporate around an event date. Three other return measures are also examined based on test methodologies described in Brown and Warner (1985). Specifically and secondly, a mean-adjusted return, i.e., average daily return during the 60 day estimation time period [-70,-11], is subtracted from a stock, or loan daily return. The third and fourth measures are based on a single-factor market index (we use the S&P/LSTA Leverage Loan Index as a market index for loans). More specifically, the third measure is a market-adjusted return, i.e., the return on a market index is subtracted from a stock, or loan daily return. The fourth measure is a market-model adjusted return, i.e., the predicted return based on a market-model regression is subtracted from a stock, loan return. More formally, we have A i,t = R i,t E[R i,t ] (1) Where A i,t is the abnormal return, R i,t is the observed arithmetic return (i.e., R i,t = P i,t /P i,t-1 1, where P i,t and P i,t-1 denote the price for security i at date t and t-1), and E[R i,t ] is the expected return for security i at date t. The four different methods of computing daily abnormal returns correspond to four different expressions for the expected return for security i at date t. That is, E[Ri, t] 0 R i = R MKT,t ˆ α + ˆ βi i R MKT, t unadjusted mean adjusted market - adjusted market - model adjusted Where R i is the simple average of security i s daily return during the 60-day estimation period (i.e., [-70,-11]: = 11 1 t i = R i, t 60 t= 70 R (2) R MKT,t is the return on a market index defined as below: R MKT, t R = R L, t S,t loan index stock index 16

17 Where R L, t is the return on the S&P/LSTA Leveraged Loan Index, on NYSE/AMEX/NASDAQ value-weighted index. The coefficients R, is the return S t ˆ α and ˆ β are Ordinary Least Squares (OLS) values from the market-model regression during the estimation period. That is, we regress security I s returns on market index returns and a constant term to obtain OLS estimates of ˆ α and ˆ β during the estimation period. The intercept and slope coefficients for the multi-factor models are defined analogously to the single-factor models. The test statistic under the null hypothesis of zero normal returns for any event day and for multi-day windows surrounding the stock splits announcement dates is described below, though more detailed can be found in Brown and Warner (1985). The test statistic for any day t is the ratio of the average abnormal return to its standard error, estimated from the time-series of average abnormal returns. More formally, At Sˆ( A ) t ~ N(0, 1) i i i i (3) Where A and S ˆ( ) are defined as t A t A = 1 N t t A i, t N i= 1 (4) t = = 11 1 S ˆ( A ( * ) 2 t ) At A (5) 59 t= 70 Where A * in (5) is defined as 1 * t= A = 11 ( A * ) 2 t A (6) 60 t= 70 Where N t is the number of securities whose abnormal returns are available at day t. For tests over multi-day intervals, e.g., [-5,+5], the test statistic is the ratio of the cumulative average abnormal return (which we simply refer to as CAR) to its estimated standard error, and is given by t=+ 5 At t= 5 ~ N(0, 1) (7) 5 Sˆ 2 ( A ) t=+ t= 5 t 17

18 The event study of bid-ask spread is conducted in the similar manner as describe above, and we construct the daily changes of bid-ask spread of our sample loans (stocks) similar to the daily returns on loans (stocks). The market index of bid-ask spread on loans is proxied by value-weighted average of all bid-ask spreads on all available loans of U.S. borrowers in Loan Trade Database during 1998 to The market index of bid-ask spread on stocks is proxied by value-weighted average of all closing bid-ask spreads on all available common stocks of U.S companies in CRSP database during 1998 to Another method of examining the changing bid-ask spread around the event window is employed by Espen et al. (1990), where basic OLS regressions are used to obtain the coefficients of a dummy variable which indicates the windows of stock splits announcements. The size of this dummy variable, along with its significance level, captures the abnormal changes and shifts of bid-ask spread. Though this dummy variable method appears to be more straightforward, the event study we use in the paper are equivalent or even superior in terms of econometric techniques employed to correct some potential biasness in the results. Once the cumulative abnormal returns (CAR) are obtained from the time-series regressions for each stock or loan s response to splits during the stock splits announcement window, we run the cross-sectional regressions of CAR on a number of stock or loan- and firm-level specific characteristics. Grinblatt et al (1984) provide some reasons why linear regression is used in the event study of stock splits: differences in returns might arise because the information being announced has been partially discounted or because the announcements provide different types of information that vary with the particular characteristics of the firm (e.g., firm size or the variability of its stock returns). These effects will vary across subsamples of events and even across individual securities. V. Results Table I presents the comparison of certain firm-level, stock-level, and loan-level characteristics between our selected sample of firms (and the associated stocks and loans) and all firms that announced splits during January 1, 1999 to December 31, The first section of Table I presents comparison of the firm-level characteristics in terms of the following dimensions: total assets (in millions USD), profitability (measured as is the ratio of 18

19 EBITDA to total assets, estimated in the end of fiscal year prior to stock splits announcement), Tobin s Q (the ratio of market price of stock to the book value of equity, estimated in the end of fiscal year prior to stock splits announcement), capital expenditure/sales (the ratio of capital expenditure to sales, estimated in the end of fiscal year prior to stock splits announcement), interest coverage (the ratio of EBITDA to interest expense, estimated in the end of fiscal year prior to stock splits announcement), and leverage (the ratio of the long-term debt to total assets, estimated in the end of fiscal year prior to stock splits announcement). The descriptive statistics and t-test of the difference show that our sample firms are different in several aspects. For example, it shows that the average firm in our sample is almost three times as big as the average firm in all splitting firms in terms of total assets (US$19.2 billion versus US$7.0 billion). In addition, the firms in our sample have a higher mean value of profitability (0.15 versus 0.12) and leverage (0.34 versus 0.14), and a significantly lower mean value of interest coverage (16.84 versus ). The second section in Table I presents the comparison of stock-level characteristics such as market value of common equity, number of shares outstanding, and the split factor. The split factor is defined as the number of additional shares per old share issued. More specifically, we have Split factor = (S (t) S (t') )/S (t') = (S (t) /S (t') ) 1 where S (t) is the number of shares outstanding, t is a date after or on the exercise date for the split, and t' is a date before the split. Consistent with the results in the first section of Table I, we notice that our sample stocks have higher market value (US$18.6 billion) and more shares outstanding ( million) on average than the population splitting firms. The difference in split factor are modest (0.93 versus 0.53), but is statistically significant. The third section of Table I compares the loan (i.e., facility)-level characteristics. Facility amount is the dollar size of the loan at the facility level. As mentioned earlier, facility designates a loan in the syndicated loan market. Usually, a number of facilities with different maturities, interest rates spreads and repayment schedules are structured and syndicated as one transaction (deal) with a borrower. Institutional term loan is defined as the term loan B, C and D, and is a loan commitment that does not allow the amounts repaid to be re-borrowed. Maturity is estimated by the number of months between the facility s issue date and the date when the facility matures. Secured is a dummy variable that takes the value of one is the facility is 19

20 backed by collateral, zero otherwise. Senior is a dummy variable that takes the value of one if the facility is senior, zero otherwise. Spread over default base is the amount the borrower pays in basis points over the default base. Dividends restrictions take the value of one if the loan covenants contain terms of restricting dividends payout, zero otherwise. This section shows that our sample loans, on average, are bigger, and have a higher proportion of loans that are institutional type, secured, senior, and having restrictions on dividends. Another interesting feature of our sample loans is that they have much less spread over default base than the population loans (i.e., all loans in the DealScan database) Insert Table I here Table II presents the cumulative abnormal changes of bid-ask spread of loans surrounding three different windows surrounding the split event: 3-day window [-1, +1], 7- day window [-3, +3], 11-day window [-5, +5] and a 21-day window [-10, +10]. Loan prices are measured as the mean of average daily bid quotes and average daily ask quotes (i.e., mean of the mean) in the secondary loan market. The loan returns are defined as the percentage difference between today s loan price and yesterday s loan price, i.e., R L,T = P L,T / P L, T-1 1. Loan bid price is the mean of bid quotes in each trading day, and loan ask price is the mean of ask quotes in each trading day. Loan bid-ask spread is measured in two ways. The absolute measure of loan bid-ask spread is the dollar difference between mean of bid quotes and mean of ask quotes, and the relative measure is the percentage difference between the two. As mentioned earlier, the loan pricing data are obtained from Loan Trade Database. Stock returns are defined as the percentage difference between today s closing price and yesterday s closing price, i.e., R S, T = P S, T / P S, T-1 1, and are obtained from CRSP database. Stock bid-ask spread in absolute terms are measured as the mean of bid-ask spread across all consolidated quotes in TAQ database, and stock bid-ask spread in relative terms are measured as the mean of percentage bid-ask spread across all consolidated quotes in TAQ database. In the event study, the estimation period is defined as [-70, -11], where day 0 refers to the stock split announcement date. We use four difference methods to compute daily abnormal returns (changes) and cumulative abnormal returns (changes) of bid-ask spread: raw, mean-adjusted, market-adjusted, and market model-adjusted method. The raw spreads are the unadjusted daily changes in spreads, and the mean-adjusted method estimates 20

21 the mean-adjusted daily changes in spread by subtracting the average daily changes in spread during the 60-day estimation time period [-70, -11] from the daily changes in spread in the event window. The market-adjusted method estimates the market-adjusted spread by subtracting the average daily changes in market bid-ask spread from the sample loan (stock) s daily spread in the event window. The market model estimates the market model-adjusted changes of the spreads, i.e., the predicted changes in spread based on a market-model regression is subtracted from the changes in spread during the event window. Table II shows that the cumulative abnormal changes of bid-ask spread of loans are negative and significant both in absolute and relative terms, and this result holds across different specification of event windows and specification of estimation models. For example, based on market model, the cumulative abnormal changes of absolute bid-ask spread is (US$) during [-1,+1] window, (US$) during [-3,+3] window, (US$) during [-5,+5] window, and (US$) during [-10,+10] window. The stock bid-ask spread shows different pattern of changes however. Cumulative abnormal changes of absolute bid-ask spread of stocks are positive and significant, while the cumulative abnormal changes of relative bid-ask spread of stocks are negative and significant most of the time Insert Table II here Table III presents the cumulative abnormal returns on the loans and stocks surrounding three different windows surrounding the split event: 3-day window [-1, +1], 7- day window [-3, +3], 11-day window [-5, +5] and a 21-day window [-10, +10]. The estimation period is defined as [-70, -11], where day 0 refers to the stock split announcement date. We use four difference methods to compute daily abnormal returns and they are raw, mean-adjusted, market-adjusted, and market model-adjusted methods. The raw returns are the unadjusted returns, and the mean-adjusted method estimates the mean-adjusted returns by subtracting the average daily returns during the 60-day estimation time period [-70, -11] from the loan (stock) daily returns in the event window. The market-adjusted returns are estimated by subtracting the return on a market index from the loan or stock daily returns. The market model estimates the market model-adjusted returns of the loan or stock, i.e., the predicted returns based on a market-model regression is subtracted from the returns during the event window. Table III shows that the cumulative abnormal returns on loans around 21