The effects of splitting on the ex: a microstructure reconciliation.
32. R.C. Lease, R.M. Masulis, and J.R. Page, "An Investigation of Market Microstructure Impacts on Event Study Returns," Journal of Finance (September 1991), pp. 1523-1536.
33. J. Lintner, "Distribution of Incomes of Corporations Among Dividends, Retained Earnings, and Taxes," American Economic Review (May 1956), pp. 97-113.
34. C.F. Loderer and D.P. Sheehan, "Seasoned Stock Offerings Michael T. Maloney is a Professor of Economics in the College of Commerce and Industry at Clemson University, Clemson, South Carolina. J. Harold Mulherin is a Visiting Associate Professor of Business Administration in the Tuck School of Business at Dartmouth College, Hanover, New Hampshire.
In spite of their algebraic simplicity, stock splits have been tied to significant changes in corporate value. Writing in 1928, Charles Amos Dice |13, p. 38~ notes that although a stock split merely divides share ownership into finer units, "practically always the price per share of the old stock rises on the news that a stock dividend is probable." Over the past 60 years, research of increasing sophistication has confirmed that the announcement of stock splits is associated with positive increases in shareholder wealth. The positive return is said to represent a verification of prior improvements and/or a signal of future improvements in splitting firms' fundamentals such as dividends and earnings.
Somewhat more surprisingly, Grinblatt, Masulis, and Titman |20~ report a positive abnormal return on the actual execution date of the stock split. Since these authors document that stock splits are rarely canceled after announcement, there is no apparent information-based explanation for this finding. They speculate that the effects of the stock split on the ex-day return may be due to factors tied to transaction costs.
In this paper, we support such conjectures by relating the anomalous ex-day return for stock splits to market microstructure phenomena. In the eleven trading days following execution, trades of splitting stocks tend to congregate at ask prices and the ex-day return is dominated by an asymmetric increase in the ask price compared to the bid. This causes an apparent abnormal return using observed prices that is not available to the average trader.
We trace the source of the microstructure bias in the execution day returns to conventional explanations as to why stock splits are implemented in the first place. Surveys of corporate managers (Baker and Gallagher |2~), statements of accounting standards boards (Chottinger and Young |8~), and general market folklore (Teweles and Bradley |46~) all suggest that the increase in round lots enabled by stock splits is used to increase a firm's shareholder base and thereby improve market liquidity. Academic analysis has often downplayed this role of stock splits because of no clear link between the number of shareholders and market liquidity. Indeed, prior research has reported mixed support for the view that stock splits improve liquidity, in part, due to differences in methods of analysis and in measures of liquidity.
For a sample of NASDAQ firms, we report evidence consistent with a variant of the conventional wisdom regarding stock splits, finding that stock splits lead to a greater number of shareholders, higher dollar volume, a greater number of trades, and narrower absolute bid-ask spreads, while returning share price and average trade size to a target range. We also find an increase in the number of institutional holders and the percentage institutional ownership for the sample firms subsequent to the split. These results suggest that the increase in round lots facilitated by the split benefits not prospective shareholders but instead preexisting shareholders who can sell off, in round lots, a portion of their holdings in the splitting firms. Pre-split shareholders desire to sell off a portion of their holdings in order to rebalance their portfolios because of the especially strong equity performance enjoyed by splitting firms.
We link the process of expanding the shareholder base and increasing institutional ownership to a temporary increase in buy orders around the ex-date. This order flow imbalance induces an above average number of reported closing prices to be at the ask and an asymmetric increase in the ask price compared to the bid in the period immediately surrounding the execution of the split, and is responsible for a major fraction of the ex-day abnormal returns. These findings for stock splits resemble recent microstructure analyses that relate order flow imbalances to observed stock returns at the execution of seasoned equity offerings (Lease, Masulis, and Page |32~), at the turn of the year (Keim |28~), across days of the week (Porter |42~), and around dividend capture trading (Choe and Masulis |6~), that more generally show the significant effects of bid-ask spreads on the behavior of stock prices (Phillips and Smith |41~, and Kaul and Nimalendran |27~), and that show the relevance of transactions-cost analysis to corporate financing decisions (Hansen |22~).
The first section reviews the prior research of stock splits and, in this way, provides the structure for our own analysis. The second section describes our data, comparing our sample of splitting firms to a more general NASDAQ sample. Section III replicates the announcement-day and ex-day returns for stock splits found in previous studies, which generally focus on NYSE securities, and presents evidence on the microstructure effects of stock splits on ex-day returns. Section IV reports evidence on the effects of stock splits on the stockholder base, institutional investors and other liquidity measures and relates the change in institutional investors to the ex-day anomaly. The final section summarizes and offers concluding comments.
I. Prior Research on Stock Splits
Research on stock splits dates at least to a series of papers by Dolley |14~, |15~ and also includes the seminal event study by Fama, Fisher, Jensen and Roll |19~. The research examines the change in the equity value of splitting firms in extended periods surrounding splits, computes the stock returns at the announcement and execution of splits, and measures changes in liquidity following splits.
A. Price Appreciation Over Extended Periods
Much of the prior research focuses on changes in equity value during extended windows surrounding stock splits. Three early studies by Myers and Bakay |39~, Barker |3~, and Johnson |26~ report positive stock price appreciation, net of market price indexes, in the period up to and including split execution. These studies also report that the price appreciation is related to contemporaneous increases in earnings and dividends.
In a more systematic analysis of 940 stock splits over the 1927-1959 period, Fama, Fisher, Jensen, and Roll |19~ find that splitting firms experience a 34% increase in stock price, net of market, in the 30 months up to and including split execution. The price appreciation is greater for firms that subsequently increase their dividends. Fama, Fisher, Jensen, and Roll |19, pp. 9-11~ conclude that
Companies tend to split their shares during "abnormally" good times -- that is during periods of time when the prices of their shares have increased much more than would be implied by the normal relationship between the shares and general market price behavior.
In a related analysis of the 1963-1982 period, Lakonishok and Lev |29~ confirm that firms electing to split their stock have had above average performance prior to the split decision. For their sample of 1,015 splits, they find a 68% abnormal return, net of matched control firms, in the five years preceding the split announcement, but no continued abnormal price performance subsequent to the split. Correspondingly, the splitting firms enjoy above average earnings performance and dividend growth in the period prior to the split, with some continued, but smaller, improvement in earnings and dividends immediately following the split.
B. Price Appreciation Surrounding Announcement Day
Because of the association of stock splits with an overall improvement in performance, some of the more recent research addresses whether the announcement of stock splits has any information content or instead is confounded by other corporate announcements. Grinblatt, Masulis, and Titman's |20~ study of 1,380 stock splits during the 1967-1976 period finds an average three percent return in the two days surrounding split announcement. The result holds for subsamples of firms having no other confounding announcements when the split becomes public, for firms with no prior dividends and for firms with no subsequent dividend increases. These and similar results by Hausman, Largay, and West |25~, Brennan and Copeland |4~, Asquith, Healy, and Palepu |1~, and McNichols and Dravid |35~ indicate that split announcements convey information to the market.
C. Price Appreciation on Execution Day
Prior research has also studied price behavior on the actual date of split execution. Dolley |14~, |15~ analyzes 95 NYSE splits occurring during the 1921-1930 period and finds that a large majority of the splitting firms experience a positive return on the execution date. More recent papers by Chottinger and Young |8~, Woolridge |47~, Grinblatt, Masulis, and Titman |20~, Eades, Hess, and Kim |18~, and Lamoureux and Poon |30~ provide similar findings.
The ex-day results are not easily explained with the information rationale given for split announcements. The date of split execution is known well in advance. Moreover, splits are rarely canceled -- Grinblatt, Masulis, and Titman |20, p. 485~ report that only four of the 1,380 split announcements in their full sample are not followed by actual split execution. Not surprisingly, therefore, the results of Lamoureux and Poon |30~ indicate that the price response on the ex-day is eliminated within ten days of execution.
Further evidence against an information rationale for the ex-day effect are the interesting results regarding "when-issued" stock. Shares in when-issued stock trade prior to the execution but represent claims on post-split shares. Choi and Strong |7~ find that when-issued stock trades at a premium to the concurrently trading pre-split shares even after adjusting for the cost of carrying. This premium closely parallels the ex-day abnormal return. Since the premium between the pre-split shares and the when-issued shares is not arbitraged away, it must reflect a bid-ask spread phenomenon. Since the when-issued premium closely reflects the ex-day abnormal return, the ex-day premium is also likely to be a bid-ask spread effect. Lamoureux and Wansley |31~ reach similar conclusions. This implies that there is no information content to the ex-day price appreciation.
Because of the lack of an information rationale for the ex-day results, several authors have hypothesized that the price movements surrounding split execution stem from transaction costs. Grinblatt, Masulis, and Titman |20~ wager that the ex-day price movements are tied to market microstructure phenomenon. They offer a story |20, pp. 485-488~ in which specialists' inventory practices lead to an above average number of trades at bid prices on the day preceding the split and result in apparent abnormal returns on the ex-day that merely represent movement from bid to ask prices. They note, however, that their story does not explain the continuance of the abnormal return for several days following execution.
Lamoureux and Poon |30~ provide a story and some supporting evidence that the positive return following the ex-day stems from price pressures related to an increase in the number of shareholders subsequent to the split. The authors report that the number of shareholders of splitting firms increases by 35% in the year following the split, and that the percentage change in shareholders is positively related to the abnormal ex-day return. Unfortunately, Lamoureux and Poon |30~ relate the shareholder increase to tax-timing strategies that the authors predict |30, p. 1369~ would be "greatly diminished" by the 1986 Tax Bill. Since, as we report below, the positive execution return is maintained in more recent data, the tax-timing story cannot be the source of the ex-day result.
Yet, a modification of the Lamoureux and Poon |30~ argument that incorporates Grinblatt, Masulis, and Titman's |20~ conjectures on market microstructure can resolve the ex-day anomaly. The increase in shareholders may lead to temporary order imbalances that induce a concentration of orders at ask prices and creates illusory abnormal returns. This modified argument not only provides an explanation for the ex-day return but is pertinent to the more general issue of whether stock splits affect the liquidity of splitting firms. This latter issue is a subject of dispute in the literature.
D. Stock Splits, Shareholder Base and Liquidity
Lamoureux and Poon's |30~ finding of an increase in the number of shareholders following stock splits is consistent with earlier empirical research by Dolley |14~, |15~ and Barker |3~. The empirical findings also support the survey by Baker and Gallagher |2~ that finds that corporate managers use stock splits to expand firms' shareholder base. The mechanism by which splits expand the number of shareholders is said to be the reduction in the investment required to buy a round lot of 100 shares that is facilitated by the split. The underlying reason that an increased number of shareholders is desirable may rest on the belief, consistent with the analysis of Demsetz |12~, that more shareholders translate into greater overall liquidity as measured by variables such as trading volume and the bid-ask spread.
Prior research is mixed as to whether an enhanced shareholder base maps into greater overall liquidity following stock splits. Copeland's |11~ analysis of 25 NYSE splits during the 1963-1974 period and Lamoureux and Poon's |30~ study of 215 NYSE and AMEX firms during the 1963-1982 period report a decline in split-adjusted trading volume following stock splits. Similarly, Copeland's |11~ analysis of 162 OTC stock splits during the 1968-1976 period and Conroy, Harris, and Benet's |10~ study of 147 stock splits during the 1981-1983 period find a significant widening in the relative bid-ask spread following stock splits.(1) These results of lower volume and wider spreads suggest that splits reduce, rather than increase, liquidity.
Research by Lakonishok and Lev |29~, however, points out flaws in the finding of reduced liquidity following splits. They note that the above average equity performance that firms experience prior to splits induces abnormal trading activity in the year preceding execution. Comparing post-split volume to this heavy trading volume associated with factors that arguably induce the split in the first place unduly biases the results toward a finding of reduced liquidity following the split. By contrast, Lakonishok and Lev |29~ and Conroy, Harris, and Benet |10~ find no significant change in trading volume when comparing longer windows before and after stock splits.
A similar issue arises when measuring changes in spreads around splits. The pre-split base used to compute changes in spreads occurs in the midst of a significant run-up in stock price (Fama, Fisher, Jensen, and Roll |19~). This price run-up induces a corresponding decline in the relative spread and may significantly influence the value of the relative spread prior to the split. As in the analysis of trading volume, one way to address this issue is to use longer windows when comparing pre-split and post-split spreads. Indeed, Murray |38~ uses longer windows than Copeland |11~ and finds no significant change in relative spread for a sample of OTC splits during the 1972-1976 period.
A more general critique of the analysis of bid-ask spreads and stock splits is the emphasis on relative spreads in the reported results. While the relative spread is one method of comparing liquidity cross-sectionally, the same measure can create estimation problems in time series analysis when the event being studied causes significant changes in stock prices. In such cases, results can be driven by changes in the denominator of the relative spread rather than being due to any real changes in the bid-ask differential. Because of this problem with relative spreads, several authors have chosen to emphasize absolute bid-ask spreads in studying corporate finance phenomena such as earnings announcements (Morse and Ushman |37~), acquisitions (Conrad, Mandelker, Niden, Rosenfeld, and Shastri |9~), seasoned stock offerings (Loderer and Sheehan |34~), and open market stock repurchase programs (Miller |36~). In the existing work on stock splits, Conroy, Harris, and Benet |10~ report that the absolute spread narrows by nine percent following the ex-day, indicating that the finding of wider spreads following splits is being driven, in some sense artificially, by the sharp reduction in price brought about by the split.
E. Summary of Prior Research
The consensus view of stock splits is that:
* Stock splits follow a period of abnormal improvements in firm value.
* Stock split announcements convey information to the market about the value of the company.
* On the execution of the split, there is stock price appreciation that must be linked to some microstructure phenomenon since the execution date is known well in advance and is almost never canceled. The exact nature and source of this abnormal return has not been identified.
* Stock splits are linked to expansion of shareholder base but there is uncertainty regarding their overall liquidity effects.
In the remainder of the paper, we address these points to consolidate and solidify our understanding of this most simple of all capital structure choices of the firm.
II. The Sample of NASDAQ Stock Splits
Our research design is twofold. First, we relate the ex-day return to market microstructure phenomena. Next, we more generally address the effect of stock splits on shareholder base and other liquidity measures. The variables used and the event windows analyzed take into account the research discussed in Section I and clarify many of the ambiguities found in the prior studies.
We construct a stock split data set of NASDAQ firms from the CRSP NMS file. Our focus on NASDAQ firms stems, in part, from the practical fact that most of the variables we use to measure liquidity such as trading volume and bid-ask quotes are readily available from CRSP for the nonexchange firms. Moreover, any liquidity effects associated with stock splits should be more apparent for the NASDAQ stocks vis-a-vis the larger, more actively traded NYSE securities that have been the focus of most of the prior analyses.
The sample is comprised of all splits executed between January 1, 1985, and December 31, 1989, that are reported on the NASDAQ daily master file. There are 1,882 of these. From this set of splits, we delete all companies with multiple occurrences of splits in the period from January 1, 1984, to December 31, 1990. This allows one year of trading before and after each split in our sample that is uncontaminated by another split. This reduces the sample to 1,076. For the remaining sample, we delete the firms on the CRSP NMS file that have less than 160 days of non-missing data on closing bid and ask quotes over the periods 250 days to 51 days prior to the execution of the split and 51 to 250 days following the execution. We cross-checked the split factor reported by CRSP with that reported in the Wall Street Journal and in the Standard & Poor's Daily Stock Price Record and two splits were deleted because of discrepancies. The final sample contains 446 splits.
The average split in our sample resulted in an 81.6% increase in the number of shares, which is comparable to studies of NYSE stock splits.(2) One hundred percent increases (i.e., 2-for-1 splits) are both the median and the mode. At the low end is one five percent split, which may be a mislabeled stock dividend in the CRSP database. Our sample also has some 6-for-5 exchanges, that is, 20% share increases. At the high end, the largest split is an 8-for-1 swap. The sample firms are a diverse lot that includes notable growth companies such as Apple Computer, Intel, Lotus Development, and Microsoft, as well as a number of banks, savings and loans, and insurance companies.
As an overview of the sample, Panel A of Exhibit 1 shows summary statistics for these 446 firms for the period 250 to 51 days prior to split execution. At this reference point, the representative firm in the sample had $210 million in equity with a stock price of $26.83. The dollar volume of shares traded each day averaged $892,487 with an average value per trade of $28,921. (Average value per trade is daily dollar volume divided by the number of trades.)
As a comparison, Panel B gives the median values of the same variables for all NASDAQ firms broken into equity size quintiles. Splitting firms are large relative to all NASDAQ issues. The median equity for splitting firms is $81.383 million, which is almost exactly on the line breaking the fourth and fifth quintiles for all NASDAQ firms. The median values for dollar volume and value per trade for the split sample also fall on the borderline of the highest NASDAQ quintile.
One fact that stands out in Exhibit 1 is that the stock price for splitting firms is at the high end of the full NASDAQ sample. Prior to splitting, the sample firms have prices in the highest ten percent of all NASDAQ firms. Correspondingly, the bid-ask spreads of the splitting firms are relatively large, even though relative spread (not shown) is low. The amount of equity per shareholder is relatively large for the sample firms prior to the split. This latter result suggests that the conventional wisdom may be on to something. Expansion of the shareholder base, which lowers the equity per shareholder, may be an important consideration in the decision to undertake a stock split.
Exhibit 2 shows the classic pattern of stock price movement for splitting companies reminiscent of Fama, Fisher, Jensen, and Roll |19~. The plot reports average stock price across the 446 firms in the stock-split sample for 50-day intervals starting two years prior to the split. The picture is consistent with the prior research as well as the adage that the purpose of the split is to return the stock price to its prior level following a run of strong performance. It is clear from Exhibit 2 that firms split their stock following a period of price appreciation. After the split, these firms experience normal performance. That is, some have further price appreciation that is offset by others that underperform. On average, price is relatively flat following the split.
The next two exhibits show the bid-ask spread over this same interval where the spread is reported in unadjusted dollar value, as well as in relative form, that is, weighted by stock price. As shown in Exhibit 3, the relative spread follows a pattern that reflects the inverse of stock price pictured in Exhibit 2. There is a run-down in relative spread prior to the split. Since relative spread has price in the denominator, this is not surprising. By contrast, as shown in Exhibit 4, the unadjusted bid-ask spread rises over the period of price run-up. Following the split, the unadjusted spread falls back to a level roughly the same as it was in TABULAR DATA OMITTED the pre-price-appreciation period, and it seems to stabilize there. From these data, we conclude that the event of a stock split definitely changes the bid-ask spread in both unadjusted dollar value and in relative terms.
With these broad patterns in mind, we now turn to a discussion of the stock price reactions in the days immediately surrounding the announcement and execution of the split.
III. Stock Returns on Announcement and Execution Dates
A substantial body of research reports significant increases in shareholder wealth on the announcement of stock splits of NYSE firms. There is also some evidence of positive price movements on the subsequent execution of the stock splits. We use our data on NASDAQ stock splits to replicate the prior analysis of the NYSE and to offer an explanation.
A. Stock Return Evidence From NASDAQ
Exhibit 5 reports data on stock returns around the announcement and the actual execution of the splits in our sample. As in the previous research, stock returns around the announcement date are significantly positive. There is appreciable price run-up in the ten days prior to the announcement date, with a further 3.3% increase in the three-day period beginning with the announcement date.
Consistent with prior work such as Grinblatt, Masulis, and Titman |20~, the positive announcement effect in our sample is not restricted to firms that have experienced or will experience improvements in fundamentals such as dividends or earnings. For our sample, 114 out of 446 firms announced a dividend increase along with the split. Another 71 announced a dividend increase within the year following the split. We broke the three-day cumulative announcement returns into groups along these lines to see if the cumulative performance differs. In results available upon request, we find that it does not. The firms not undertaking dividend increases at the time of the split or in the next year experience announcement returns comparable to those that do.
As reported in Exhibit 5, the execution date also evinces positive stock returns, although of a smaller magnitude and in a more discrete fashion than the announcement date. There is less evidence of run-up preceding the execution date. On the execution day, stock prices increase by an average of 1.6%.(3) There is some continuance in the positive returns following the split; the three-day stock return beginning with the execution date is 2.2% and the cumulative return by day 10 is 3.9%.(4)
B. Microstructure Analysis
Since stock splits are rarely canceled, the execution date is not likely to be an information event. Indeed, a number of authors question the reality of the ex-day price movements. Grinblatt, Masulis, and Titman |20~, for example, conjecture that the ex-day price movements are tied to market microstructure rather than representing actual wealth changes. They speculate (|20, pp. 485-488~) that there are an above average number of trades at bid prices on the day preceding the split. In contrast, we argue that the positive ex-day stock returns can also arise from an above average number of closing prices recorded at the ask on the execution date and the days following. The preponderance of ask prices stems from temporary order imbalances as stock splits accomplish their intended purpose, that is, increasing the number of shareholders of the splitting firms.
Exhibit 5. Daily and Cumulative Returns for Announcement and Execution Events Panel A. Event Time Announcement Event Execution Event Event Daily Cumulative Daily Cumulative Time Return Return Return Return -10 0.0032 0.0032 -0.0016 -0.0016 -9 0.0039 0.0071 -0.004 -0.0020 -8 0.0048 0.0119 0.0004 -0.0016 -7 0.0028 0.0147 0.0015 -0.0001 -6 0.0037 0.0183 -0.0006 -0.0007 -5 0.0027 0.0210 0.0011 0.0004 -4 0.0025 0.0236 -0.0002 0.0003 -3 0.0025 0.0261 0.0010 0.0013 -2 0.0039 0.0299 0.0023 0.0036 -1 0.0008 0.0308 0.0047 0.0083 0 0.0136 0.0443 0.0164 0.0247 1 0.0114 0.0557 0.0036 0.0283 2 0.0079 0.0637 0.0034 0.0317 3 0.0003 0.0640 0.0023 0.0339 4 0.0047 0.0686 0.0005 0.0344 5 0.0015 0.0701 0.0002 0.0345 6 0.0006 0.0707 0.0044 0.0389 7 -0.0001 0.0707 -0.0010 0.0379 8 -0.0005 0.0702 0.0029 0.0408 9 0.0006 0.0696 -0.0017 0.0391 10 -0.0002 0.0693 0.0005 0.0396 Panel B. Three-Day Cumulative Return CR (0,+2) 0.0329 0.0224 t-statistic 11.25 7.41 Note: Announcement event returns exclude returns for firms within five days of execution event. Returns are unadjusted based on daily stock prices which are the average of the bid-ask spread when no trades occur, but closing spreads are reported. Cumulative returns for both series are staable after ten days. The t-statistic testing three-day cumulative return is based on the pooled standard error from the regression of returns across all firms over the entire two-year period on dummy variables for the three-day announcement and execution windows.
Exhibit 6 provides evidence in support of our story. Panel A reports the average occurrence of closing trades at or above the ask, at or below the bid or somewhere in between on the days surrounding the execution of stock splits.(5) In the 10-day period prior to the execution date, 42% of the closing prices occur at or above the ask, 37% at or below the bid, and the remainder in between. On the day of the actual split, the fraction at the ask jumps to over 50% and stays at roughly this level for at least ten days following the split. This suggests an above average number of buy orders in the period immediately following the stock split. Clearly, the number of trades in the post-execution window increases dramatically. In the ten days prior to the execution, the number of trades is relatively flat compared to the average number from 151 to 250 days following the execution. On and after the execution, the number of trades jumps by 50% and stays up for 20 days.
A convenient summary statistic for this order imbalance is the order flow ratio used by Lease, Masulis, and Page |32~. The order flow ratio for firm i at time t is computed as:
Closing Ask |Quote.sub.it~ - Closing Transaction |Price.sub.it~ /Closing Ask |Quote.sub.it~ - Closing Bid |Quote.sub.it~
For a given firm i on date t, when the closing price is at the ask, the order flow ratio equals zero; for a closing price at the bid, the ratio equals one. Since the closing price and bid-ask quotes are not completely synchronous, that is, there are times when the last transactions price is outside the closing bid or ask quotes, the order flow ratio is not bounded by zero and one. But when averaged over many firms, the ratio does succinctly exposit the relation between closing prices and bid-ask quotes.
Panel A of Exhibit 6 presents the statistic at its sample mean for our firms for the ten days preceding through the 20 days following the execution date. Panel B presents the averages over windows of these days and over the entire nonevent period. Note that the average over all nonevent days in our sample is 0.511, which is not statistically different from one-half. This says that, on average, stocks trade at the midpoint of the bid and ask quotes. During the ten days preceding the execution date, the average of the order flow ratio across firms is 0.479, slightly lower than the midpoint. The ratio drops to 0.396 on day 0, indicating an above average number of closing prices at the ask on the execution date. The order flow ratio averages 0.423 for the period from day 0 through day +10; this is significantly lower than the average over the nonevent periods and the ten days preceding the split as shown by the t-statistics of 7.7 and 3.46 in Panel B. This tendency to trade at the ask persists TABULAR DATA OMITTED up to 20 days following the split; for the (+11, +20) period, the ratio remains at 0.443.
The order imbalance associated with stock splits appears to be a major reason for the positive return around the execution date. But this is only the first stage of the story. We expect that prices congregate at the ask whenever positive information enters the market.(6) Informed traders enter the market to buy the underpriced stock at the asking quote. The nature of the CRSP data should minimize this phenomena in reported returns because the closing quotes should follow the closing prices and hence, impound their information content. When a transaction causes price to increase, the bid and ask TABULAR DATA OMITTED should move up to account for the higher transaction price. Indeed, in an inventory framework, as analyzed by Stoll |45~, the ask and bid should exaggerate the price movement in order to induce uninformed traders to move to the other side of the market. Even so, there is a tendency for congregation of prices at the ask any time prices are rising. Indeed, we find that trading congregates at the ask around the declaration event as well as the execution.
Importantly in this regard, our analysis of execution day trading reveals a second stage to the story that goes beyond clustering at the ask. Exhibit 7 shows event returns for the execution computed from the bid and ask quotes rather than the closing transaction prices. As a comparison event, the same bid and ask returns are computed for the announcement event. The two events are strikingly dissimilar. Note the asymmetry of the bid and ask quotes around the execution event compared to the symmetry characterizing the announcement period. During the 21-day announcement period, the bid and ask quotes move step for step; the difference between the cumulative bid return and the cumulative ask return is never more than 0.1% and, as shown in Panel B, the difference in the three-day event window cumulative announcement return between the bid and the ask is less than 0.1%. In contrast, the movement in the bid and ask around the execution is significantly different. Up to the execution day, they move together just like the announcement returns. However, on the execution day, the ask quote jumps two percent while the bid only increases 0.6%. The difference over the three-day event window is 1.1%, significant at the 0.0001 level, and the difference is still one percent after ten days.
Note that this means the transaction-induced movement in the bid and ask quotes is not a profit opportunity to the average trader. The holding period return for a trader that buys at the ask ten days before the execution and trades at the bid ten days after averages 0.07%, that is, nearly zero.
The asymmetry between the bid and the ask returns resembles microstructure analysis of cash dividend ex-days (Choe and Masulis |6~ and Dubofsky |17~) and is consistent with more general empirical results on the response of quotes to order imbalances (Handa |21~). In the next section, we tie the asymmetric movement to changes in the shareholder base that accompany stock splits.
IV. Liquidity, Shareholder Base and Price Pressure
So far, we have shown that stock splits are the occasion for several discrete events. One is a change in the bid-ask spread. Exhibits 3 and 4 show that the spread changes abruptly around the execution of split. The dollar spread falls while the relative spread rises. Another event is the pattern of trading. At the execution, there is an asymmetric abnormal jump in the ask price quote relative to the bid. However, none of these events motivate the split. That is, it is still not clear why the stock split is a profit opportunity for the firm. In this section, we offer such a motivation by relating the increase in the shareholder base to the microstructure events. First, we look more broadly at the pattern of trading before and after the split.
A. Liquidity Measures
Exhibit 8 shows the dollar volume of each security traded each day averaged over the stock-split sample in 50-day intervals. There is a decided increase in daily trading activity starting around 200 days prior to the split. The increase in trading activity is much more precipitous than the run-up in stock price and the level of trading activity stays high following the split execution. Even so, there is nothing in Exhibit 8 that indicates the split is the cause of an increase in trading volume. The dollar volume of trading more or less tracks the run-up in firm value and then stabilizes as does firm value.
Exhibit 9 shows the average value of each security transaction labeled "value per trade". As before, this measure is computed as the total value of stock traded in a day divided by the number of trades. Value per trade increases in the period prior to the split. After execution, value per trade returns to a level roughly the same as it was two years prior to the split. In contrast to Exhibit 8, Exhibit 9 shows that the split is the occasion of a discrete change in the pattern of trading activity. Based on the timing of this change, we conclude that the act of splitting is the cause. After the split, there is a marked increase in the number of trades per day and this causes the average value of each trade to return to some pre-split benchmark value.
Next, we examine the change in the shareholder base surrounding the split. Exhibit 10 shows several characteristics. The number of shareholders increases from an average of 6,817 before the split to 8,288 following. The percentage increase for each firm averages 23.4; the t-statistic testing the hypothesis that this percentage change is zero is 15.56. Not only does the number of shareholders increase, so does the number of institutional shareholders. Before the split, there are approximately 32 institutional holders of each stock (on average), while after the split, this increases to 55. Firm by firm the percentage change averages 68.7. Moreover, the percentage of shares held by institutions increases; it goes from 23.8% to 31.2%, on average. The average percentage change across firms is 118.9, while the median percentage change is 29.2.
The data on the number and type of shareholders is not observed as periodically as the number and value of trades, so we cannot link the changes in shareholder base directly to the split. Maybe the number of shareholders would have increased without the split just as a consequence of the run-up in firm value. However, the expansion of the shareholder base is theoretically consistent with the increase in TABULAR DATA OMITTED the number of trades per day, which is directly linked to the split.
B. A Role for Stock Splits
Our interpretation of this relation between trading patterns and expanded shareholder base is that the split allows small investors to diversify out of a portion of the wealth they hold in the security without completely divesting. The splitting firm is an asset whose value has appreciated significantly. The firm splits its shares to return the stock price in approximate value to some prior level. Investors can sell their new shares in round lots and effectively return their portfolio wealth held in this asset to the prior level.(7) The split allows them to more easily diversify their portfolio, while still maintaining a position in the firm.
Consistent with this interpretation is the fact that even though institutional ownership increases, equity per shareholder returns to pre-split levels. Exhibit 10 shows that the mean and median equity per shareholder stays the same from before to after the split.(8) The equity value of the sample firms is rising over the pre-split period and flat thereafter. The number of shareholders increases pre- to post-split. The net effect on equity per shareholder is essentially nil.
In sum, the new shares created by the split are soaked up by new investors, principally institutions, whose interest in the security has grown with the appreciation in value.(9) The fact that institutions, who can be assumed to not suffer significant round lot constraints, increase their holdings following the split suggests that the greater number of round lots enabled by the split does not generally aid prospective shareholders. Instead, the split allows pre-existing shareholders to sell, in round lots, a portion of their holdings in a firm that has experienced substantial equity appreciation.
This argument, linking trading patterns to the expansion of the shareholder base, has the appealing feature that it is based on the assumption that the firm operates to maximize current stockholder wealth. Thus, we have a rational, logically consistent motivation for a firm choosing to split its shares. Moreover, the argument is consistent with the change in the spread caused by the split. Consider an investor holding 100 shares before the split. The spread falls from 81 cents to 58 cents as a result of the split.(10) If the stockholder liquidates his entire position, the split increases trading costs from half of 81 cents times 100 to half of 58 cents times 200. However, if the investor chooses to sell only 100 of the new shares, then the split causes trading costs to fall by half of the difference in the pre- and post-spread.
The argument assumes that investors hold portfolios that mimic the market without holding all traded assets. While we do not have information on small investors, data are available from SEC 13f filings detailing the holdings of portfolios of $100 million or more. From the filings for the third quarter 1989, across 826 portfolios, the average number of stocks held was 286. Hence, even institutional investors hold portfolios that are subsets of the market and must be rebalanced when assets change dramatically in value.
The argument also assumes that 100-share lots are a constraint. Again, we do not have information on small investors, but across the 826 institutions filing 13fs in 1989, 124 held stocks in lots of 100 shares or less. The small-lot stocks held by these 124 funds were predominately standard industrial companies. This says that even institutional investors have small-lot holdings and we expect the frequency to be substantially higher for small investors.
Assuming that the pre-split shareholders have a significant demand to diversify out of part, but not all, of the security, then the split reduces their transactions costs. Hence, the split is motivated by maximization of shareholder wealth by minimizing shareholder transactions costs.(11)
C. The Source of the Ex-Day Price Pressure
The entry of new investors documented above is a possible source of the ex-day return behavior. While we do not know why the new investors would necessarily rush in to purchase shares immediately, the data in Exhibit 6 show that trading activity is decidedly higher on the execution day and the days immediately following. These results parallel recent evidence on the effects of listing on the S&P 500. Harris and Gurel |24~ show that stocks added to the S&P 500 experience a temporary, positive stock return of roughly three percent around the listing date. Pruitt and Wei |43~ report that newly listed firms experience a subsequent increase in institutional ownership and that the increase in institutional ownership is directly related to the price pressure around the listing date.
If a similar phenomenon is operating here, it fits together with our rationale for the stock split to explain why the ex-day return is asymmetric. If new investors, principally institutions, flock to the stock on the ex-day, this creates buying pressure that drives up the ask price. If they were the only new force entering the market, we would expect the bid to increase also, as it does on the announcement day. However, from the selling side, our argument implies that there are ample reserves to be tapped. Splitting stocks have investors who are interested in selling some, but not all, of their shares. These investors are not interested in selling their shares for a discount, but the split does cause an increase in the number of shares offered at the current price. The net effect is an increase in the average of the bid and ask quotes driven by an increase in the ask while the bid remains relatively flat.
This argues that there should be a link between the execution stock price reaction, the entry of new investors, and the diversification effect. The execution-event phenomenon that we focus on is the asymmetric reaction of the spread during the event window. Recall from Exhibit 7 that the three-day return computed at the ask price is larger by 0.0108 than the same cumulative return computed at the bid.(12) The results of Pruitt and Wei |43~ suggest that the ask price increases because of new buyers who are likely to be institutions. On the other side of the market, our diversification argument says that the reaction of the bid price to this buying pressure is muted by an increased willingness to sell the split shares by the current holders.
We proxy the diversification effect by two different variables. One is the change in the value per trade around the split. If our argument is correct, the increase in the TABULAR DATA OMITTED number of trades per day relative to the dollar volume of shares transacted reflects the interest of pre-split shareholders to trade part, but not all, of their holdings. The other proxy is the long-term change in the spread. Here, we average the dollar spread over the period of the general price run-up and compare it to the average in the post-split period (where we exclude the execution-event period). Note that this does not generate a mechanical relation: the change in the long-term spread is computed as the average spread over the year before the split compared to the year following, omitting 50 days on each side of the execution day. As we argue above, the decline in the dollar spread measures the reduction in trading costs that the firm is trying to achieve by the split. If this reduction in trading costs is positively related to the demand by existing shareholders to sell their split shares, then the more the long-term spread declines the less upward pressure there should be on the bid price on the ex-day.
To test this final link in the chain of argument, we regress the difference in the three-day ask return and the three-day bid return on the percentage change in the number of institutional holders, on the percentage change in value per trade, and on the long-term change in the spread. The results are shown in Exhibit 11. The regressions offer support for the hypothesis. As predicted, there is a positive relation across the sample of splitting firms between the increase in the number of institutional holders and the divergence of the bid and ask returns on the ex-day. The statistical significance varies across the various specifications from 0.10 to 0.025. This corroborates the results of Pruitt and Wei |43~. Also as predicted, the percentage change in value per trade and the percentage change in the long-term spread are both negatively associated with the asymmetric increase in the ask relative to the bid on the execution day. The significance levels for these diversification-effect proxies vary between 0.15 and 0.02. Thus, looked at either in terms of the cost of trading (the spread variable) or in terms of the frequency of small trades, the diversification effect is predictably linked to the microstructure phenomenon that occurs on the ex-day.
Finally, we include the long-term level of the spread as an additional control variable in the regressions. The bigger the dollar spread is, measured in the post period but distant from the execution event, the larger the separation of the bid and ask returns at the execution. This relation is very strong. The inclusion of the level of the long-term spread increases |R.sup.2~ from roughly 0.02 to 0.17. Presumably, the level of the long-term spread in some way measures price sensitivity. However, no matter the explanation for its significance, its inclusion improves the precision of the estimated relation between the divergence in returns computed at the ask versus the bid and the other independent variables -- the percentage change in the long-term spread, the percentage change in value per trade, and the percentage change in the number of institutional holders.(13)
While arguably the simplest of corporate finance procedures, stock splits have been the subject of substantial research that has produced anomalous evidence on ex-day returns and conflicting results regarding splits and market liquidity. The ambiguity in the prior research arises because stock splits can be viewed in either of two ways. On the one hand, stock splits appear much like Sarah Lee Bakeries changing the labelling on their 12-ounce cheesecake from six to eight servings -- calories per serving go down but not calories per bite. Another view, one that we find more appealing, is that stock splits are like stained-glass windows. The Cathedral of Chartres has windows of the same overall size but with many more pieces than those in the National Cathedral in Washington. More pieces allow for more diversity which, in turn, is valued more highly by window aficionados.
We resolve some of the ambiguity concerning stock splits through our analysis of 446 NASDAQ firms that split once in the 1984-1990 period. As in the prior research, we find that splitting firms experience substantial price run-up in the two years prior to the split and experience an additional positive return at the announcement of the split. The announcement effect is independent of any confounding dividend increases on the announcement day or in the year following split execution.
As in prior work, we also find a positive return on the execution day itself. We show that, rather than representing real wealth changes, the ex-day price movement stems from market microstructure phenomena. On and immediately following the ex-day, trades of the splitting stocks congregate at ask prices. Moreover, the order imbalances induced by the splits result in asymmetrically larger returns computed at ask prices relative to bid prices. While affecting observed stock returns, these microstructure effects do not represent a trading rule for investors.
We link the ex-day price movements to changes in measures of liquidity that are associated with the split. For the sample firms, stock splits are associated with a greater number of shareholders, higher dollar volume, and a greater number of trades while returning share price and average trade size to the range prior to the run-up in the splitting firms' prices. Of these, the increase in shareholders, trades per day and the decline in the value per trade are caused by the firm's decision to split its shares. Most importantly, we find that institutional ownership increases after the split and that the increase in institutions is directly related to the asymmetric behavior of the bid and ask quotes around the ex-day. In addition, this asymmetric divergence of the bid and ask is related to the long-term change in the bid-ask spread and in the dollar volume per transaction, both of which are caused by the split.
We summarize our interpretation of the weight of the evidence: stock splits are undertaken by firms experiencing abnormal equity appreciation. The split allows existing shareholders to sell off, in round lots, a portion of their holdings in the now more valuable firm thereby rebalancing their portfolios. The announcement of the split conveys information, in a fashion similar to Lintner's |33~ portrayal of dividend increases, that the firm will maintain its new level of performance. The actual execution conveys no new information, but does draw in new shareholders that, at least in our sample, are disproportionately institutional investors. The order imbalances created by the transfer of shares to the new investors causes an abnormal number of trades at the ask and an asymmetric increase in ask price returns. The microstructure nature of the ex-day price movements indicates the absence of any anomalous trading rules around the execution of stock splits.
Michael T. Maloney is a Professor of Economics in the College of Commerce and Industry at Clemson University, Clemson, South Carolina. J. Harold Mulherin is a Visiting Associate Professor of Business Administration in the Tuck School of Business at Dartmouth College, Hanover, New Hampshire.
1 Conroy, Harris, and Benet |10~ report that the increase in the relative spread is one source of the increase in volatility following splits that has been reported by authors including Ohlson and Penman |40~.
2 Lakonishok and Lev |29~, for example, report an average split factor of 85%.
3 We attempted to assess the relation between the execution-event return and the premium paid for when-issued shares. However, using the Standard & Poor's Daily Stock Price Record, we could find when-issued quotes reported for only eight of the sample firms and only one firm had actual trading in when-issued stock.
4 Lamoureux and Poon |30~ provide evidence that the ex-day phenomenon on the NYSE is eliminated after about ten trading days. The quicker resolution on the NYSE arguably stems from the greater liquidity of the stocks traded there.
5 The failure of closing prices to fall within the bid and ask quotes 100% of the time stems from a number of factors, including stale prices and the fact that reported quotes are inside quotes rather than reflecting those of the particular individual dealer that may have executed a trade.
6 Dopuch, Holthausen, and Leftwich |16~ make the symmetric observation regarding bid prices and negative information.
7 This is our interpretation of the notion of a "target price."
8 Equity is computed based on the average stock price over the period of 250 to 51 days prior to and following the split. Notice that the percentage change on a firm-by-firm basis averages 12%, even though the change in the means of the two groups is zero.
9 In a more general setting, Reinganum |44, p. 20~ reports that ". . . significant shifts in institutional ownership seem to be contemporaneously correlated with large price movements." The growth in institutional ownership subsequent to the split may also be associated with the finding of Brennan and Hughes |5~ that the number of analysts tracking a stock rises following a split.
10 These numbers are the averages in the 50-day periods before and after the execution day as shown in Exhibit 4. They are also similar to the data shown in Exhibit 6 where the average spread around the execution day is listed. The drop in the dollar spread at the execution is not quite this large because of the microstructure phenomenon that we describe in Exhibit 7 and Exhibit 11. Nonetheless, there is a decline on the execution day from 80 cents to 67 cents.
11 The argument here is similar to the explanation of the fact that rights offerings are rarely used (Hansen |22~). Rights offerings impose an extra transaction on shareholders who do not wish to hold more of the asset and, thus, increase their transactions costs.
12 We focus on the divergence of the ask and the bid rather than on the increase in the midpoint because the extent of the separation between the bid and ask accentuates the forces that we see at play. Empirically, we regressed the return computed at the midpoint on the variables analyzed below, but found no significant relations.
13 We estimated the reported regressions controlling for heteroscedasticity. There was little sign of this problem and no change in the significance levels of the independent variables when the standard fix was applied.
1. P. Asquith, P. Healy, and K. Palepu, "Earnings and Stock Splits," Accounting Review (July 1989), pp. 387-403.
2. H.K. Baker and P.L. Gallagher, "Management's View of Stock Splits," Financial Management (Summer 1980), pp. 73-77.
3. C.A. Barker, "Effective Stock Splits," Harvard Business Review (January/February 1956), pp. 101-106.
4. M.J. Brennan and T.E. Copeland, "Stock Splits, Stock Prices and Transaction Costs," Journal of Financial Economics (October 1988), pp. 83-101.
5. M.J. Brennan and P.J. Hughes, "Stock Prices and the Supply of Information," Journal of Finance (December 1991), pp. 1665-1691.
6. H. Choe and R.W. Masulis, "Measuring the Impacts of Dividend Capture Trading: A Market Microstructure Analysis," Working Paper, Vanderbilt University, November 26, 1991.
7. D. Choi and R.A. Strong, "The Pricing of When-Issued Common Stock: A Note," Journal of Finance (September 1983), pp. 1293-1298.
8. S. Chottinger and A. Young, "A Test of the AICPA Differentiation Between Stock Dividends and Stock Splits," Journal of Accounting Research (Autumn 1971), pp. 367-374.
9. J. Conrad, G. Mandelker, C.M. Niden, A. Rosenfeld, and K. Shastri, "The Impact of Corporate Acquisition Announcements on Bid-Ask Quotes in an Electronic Market with Competitive Market Makers: The NASDAQ System," Working Paper, University of North Carolina and University of Pittsburgh, February 28, 1990.
10. R.M. Conroy, R.S. Harris, and B.A. Benet, "The Effects of Stock Splits on Bid-Ask Spreads," Journal of Finance (September 1990), pp. 1285-1295.
11. T.E. Copeland, "Liquidity Changes Following Stock Splits," Journal of Finance (March 1979), pp. 115-141.
12. H. Demsetz, "The Cost of Transacting," Quarterly Journal of Economics (February 1968), pp. 33-53.
13. C.A. Dice, The Stock Market, Chicago, A.W. Shaw Company, 1928.
14. J.C. Dolley, "Characteristics and Procedure of Common Stock Split-Ups," Harvard Business Review (April 1933), pp. 316-326.
15. J.C. Dolley, "Common Stock Split-Ups -- Motives and Effects," Harvard Business Review (October 1933), pp. 70-81.
16. N. Dopuch, R.W. Holthausen, and R.W. Leftwich, "Abnormal Stock Returns Associated with Media Disclosures of 'Subject To' Qualified Audit Opinions," Journal of Accounting and Economics (June 1986), pp. 93-117.
17. D.A. Dubofsky, "A Market Microstructure Explanation of Ex-Day Abnormal Returns," Financial Management (Winter 1992), pp. 32-43.
18. K.M. Eades, P.J. Hess, and E.H. Kim, "On Interpreting Security Returns During the Ex-Dividend Period," Journal of Financial Economics (March 1984), pp. 3-34.
19. E.F. Fama, L. Fisher, M.C. Jensen, and R. Roll, "The Adjustment of Stock Prices to New Information," International Economic Review (February 1969), pp. 1-21.
20. M.S. Grinblatt, R.M. Masulis, and S. Titman, "The Valuation Effects of Stock Splits and Stock Dividends," Journal of Financial Economics (December 1984), pp. 461-490.
21. P. Handa, "Order Flow and Bid-Ask Dynamics: An Empirical Investigation," Working Paper, NYU Stern School, July 1991.
22. R.S. Hansen, "The Demise of the Rights Issue," Review of Financial Studies (Fall 1988), pp. 289-309.
23. L. Harris, "A Day-End Transaction Price Anomaly," Journal of Financial and Quantitative Analysis (March 1989), pp. 29-45.
24. L. Harris and E. Gurel, "Price and Volume Effects Associated with Changes in the S&P 500 List: New Evidence for the Existence of Price Pressure," Journal of Finance (September 1986), pp. 815-829.
25. W.H. Hausman, R.R. West, and J.A. Largay, "Stock Splits, Price Changes, and Trading Profits: A Synthesis," Journal of Finance (January 1971), pp. 69-77.
26. K.B. Johnson, "Stock Splits and Price Change," Journal of Finance (December 1966), pp. 675-686.
27. G. Kaul and M. Nimalendran, "Price Reversals: Bid-Ask Errors or Market Overreaction?" Journal of Financial Economics (November-December 1990), pp. 67-93.
28. D.B. Keim, "Trading Patterns, Bid-Ask Spreads and Estimated Security Returns: The Case of Common Stocks at Calendar Turning Points," Journal of Financial Economics (November 1989), pp. 75-97.
29. J. Lakonishok and B. Lev, "Stock Splits and Stock Dividends: Why, Who and When," Journal of Finance (September 1987), pp. 913-932.
30. C.G. Lamoureux and P. Poon, "The Market Reaction to Stock Splits," Journal of Finance (December 1987), pp. 1347-1370.
31. C.G. Lamoureux and J.W. Wansley, "The Pricing of
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Market Microstructure and Corporate Finance Special Issue; stock split execution date|
|Author:||Maloney, Michael T.; Mulherin, J. Harold|
|Date:||Dec 22, 1992|
|Previous Article:||A market microstructure explanation of ex-day abnormal returns.|
|Next Article:||AMEX-to-NYSE transfers, market microstructure, and shareholder wealth.|