# A market microstructure explanation of ex-day abnormal returns.

David A. Dubofsky is an Associate Professor at Texas A&M University, College Station, Texas.Considerable evidence exists that "something unusual" occurs on the ex-distribution days of U.S. listed common stocks. For example, most empirical evidence suggests positive average abnormal returns on the ex-days of taxable cash distributions (Elton and Gruber |7~, Kalay |12~, Eades, Hess, and Kim |6~, Lakonishok and Vermaelen |16~, and Barclay |1~).(1) Positive excess returns have also been found to exist on the ex-days of nontaxable stock distributions, such as stock splits and stock dividends. This has been documented by Woolridge |28~, Eades, Hess, and Kim |6~, Grinblatt, Masulis, and Titman |9~, and Lamoureux and Poon |17~.(2)

Two explanations for the ex-cash dividend day abnormal return have been hypothesized. In the first theory, supported by Elton and Gruber |7~, Elton, Gruber, and Rentzler |8~, and Barclay |1~, the marginal investor, who faces a higher tax burden on dividend income relative to capital gains, decides to trade for reasons unrelated to the dividend. This long-term marginal investor must be indifferent to accelerating or delaying the planned trade around the ex-date. If dividend/tax clienteles exist, long-term investors in tax brackets higher (lower) than the marginal investor will generally own low (high) dividend yielding stocks. In addition, the ex-dividend day rate of return will be inversely related to the dividend yield. The ex-day price declines of stocks with high yields will approach the dividend amount. However, if corporations, who face tax deductions on capital losses which exceed the tax liabilities on dividend income, are the marginal investors for high-yielding stocks, then ex-cash dividend day returns on these stocks are predicted to be negative.

A second hypothesized determinant of the ex-cash dividend day abnormal return is the existence of short-term traders who trade because of the differential tax treatment of dividends and capital losses. This view is favored by Kalay |12~ and |13~, Lakonishok and Vermaelen |16~, Karpoff and Walkling |14~ and |15~, Michaely |18~, and Stickel |22~. The marginal investors are short-term traders who generally practice dividend capture by buying high-dividend-yielding stocks cum-dividend and selling them ex-dividend.(3) The size of any ex-day abnormal returns created by dividend clienteles will be arbitraged away by short-term traders up to (or down to) these traders' marginal transactions costs.

Empirical evidence consistent with both hypotheses has been found. It is likely that both long-term investors and short-term traders affect the magnitude of ex-cash dividend day abnormal returns, and the arguments and evidence provided by this paper add another explanatory factor: market microstructure influences.

In contrast, no satisfactory theory has explained the positive abnormal returns found on ex-stock distribution days.(4) Because of this and other results, Eades, Hess, and Kim |6~ conclude that "ex-dividend period returns remain an anomaly" |6, p. 33~. Grinblatt, Masulis, and Titman |9~ also caution against the tax explanation, suggesting that "perhaps, whatever underlies the abnormal stock dividend and split ex-date return drives the abnormal cash dividend ex-date return, too" |9, p. 489~.

This paper proposes such an explanation, which is based on market microstructure practices. The abnormal returns are argued to be the result of New York Stock Exchange (NYSE) Rule 118 and American Stock Exchange (AMEX) Rule 132, which dictate how open ("good-till-canceled") limit orders to buy and sell stock are handled on ex-days. On both exchanges, open limit orders to buy stock must be reduced by the cash dividend amount on ex-cash dividend days. If the resulting price is not a multiple of an eighth of a dollar, then an ex-cash dividend limit buy order price is reduced to the next lower eighth. Limit orders to sell are not changed by the ex-cash dividend event.(5) In addition, patterns in ex-day abnormal returns are shown to arise because trading is done at discrete price intervals (normally $1/8 for stocks selling above one dollar per share), while cash and stock dividends paid per share are usually not multiples of eighths.

On ex-stock distribution days, NYSE Rule 118 and AMEX Rule 132 require a specialist to reduce all outstanding limit buy orders. The new limit order price equals the original limit order price divided by 100% plus the percentage value of the stock distribution. For example, a limit order price will be divided by 101% when there is a one percent stock dividend. If the resulting price is not a multiple of an eighth of a dollar, then the limit buy price is reduced to the next lower eighth. Outstanding limit sell orders for NYSE stocks are not reduced; however, AMEX limit sell order prices are reduced by 100% plus the percentage value of the stock distribution, and rounded down if necessary, when a stock trades ex-distribution.(6)

These rules for handling open limit orders on ex-days are sufficient to create ex-day abnormal returns under the assumption that the closing ex-dividend day bid-asked spread is constrained by the good-till-canceled limit orders. Under this assumption, the ex-day bid-asked spread is wider(7) than "normal," and it is not symmetric around the "expected" ex-day adjusted price (the cum-dividend day closing price less the dividend). These conditions would most likely exist, and the impact of the rules observed, for thinly traded stocks that pay small dividends. For large cash dividends and stock dividends, the ex-day bid-asked spread created by the open limit orders will be too wide for the specialist to avoid participating in the execution of ex-day market orders; otherwise, he would fail in his task of maintaining an orderly market. Similarly, active trading and newly entered limit orders on the ex-day will likely lead to a narrower spread than the one implied by the adjusted limit orders in conjunction with NYSE Rule 118 and AMEX Rule 132.

The first section demonstrates how NYSE Rule 118 and AMEX Rule 132 lead to abnormal returns, and trading in eighths creates ex-day returns patterns that are a function of the difference between the dividend amount and adjacent multiples of 1/8. Specific hypotheses are presented. Section II contains the results of several empirical tests. Section III discusses the results and provides a conclusion.

I. Ex-Day Returns Under NYSE Rule 118 and AMEX Rule 132

Let |P.sub.c~ denote the mean closing cum-dividend price and |P.sub.e~ denote the mean closing ex-dividend price. Also, let P|B.sub.c~, and P|A.sub.c~ be the bid and ask prices at the close of trading on the last cum-day. Each is equidistant from |P.sub.c~; i.e., P|B.sub.c~ + X = P|A.sub.c~ - X = |P.sub.c~, where X is one-half of the bid-asked spread. D denotes the cash dividend per share. The asked and bid quotes on the ex-day are P|A.sub.e~ and P|B.sub.e~, respectively.

If the bid and asked quotes remain unchanged during intra-ex-day trading, and ex-day market orders are equally likely to be a buy or a sell order filled at P|A.sub.e~ or P|B.sub.e~, respectively, then the average ex-day closing price is (P|A.sub.e~ + P|B.sub.e~)/2. It follows that the stock's mean return is:

|Mathematical Expression Omitted~

In a perfect, frictionless market, both P|A.sub.e~ and P|B.sub.e~ would be reduced from their cum-dividend levels by exactly the dividend amount; thus, R would equal zero. However, stocks trade at 1/8 increments, and dividend amounts are typically not in increments of eighths. In addition, NYSE Rule 118 and AMEX Rule 132 dictate how specialists must handle the prices specified by good-till-canceled buy and sell limit orders.

NYSE Rule 118 and AMEX Rule 132 lead to nonzero abnormal ex-day returns under the following conditions: (i) the mean last cum-dividend day closing price is the mean of the inside quotes specified by outstanding good-till-canceled buy and sell limit orders, (ii) on the ex-dividend day, the bid and asked quotes are those of outstanding limit orders previously placed by investors during the cum-dividend period and adjusted as specified by NYSE Rule 118 and AMEX Rule 132, and (iii) the closing trade on the ex-day is at one of these two quotes, with equal probabilities.(8)

Under the two exchange rules, outstanding limit orders to sell remain unchanged on ex-cash dividend days. Outstanding limit orders to buy are reduced by the cash dividend amount; if the resulting limit buy price is not a multiple of an eighth, the price is further reduced to the next lower multiple of $0.125. On ex-stock distribution days, the NYSE reduces limit buy orders, but not limit sell orders; the AMEX reduces both limit orders to buy and limit orders to sell. If the limit price is reduced, the new ex-day limit price becomes the original limit price divided by 100% plus the percentage value of the stock distribution; the resulting price (if it does not equal a multiple of $0.125) is rounded down to the next lower multiple of $0.125.

Exhibit 1 illustrates how the practice of leaving limit orders to sell unchanged, and reducing limit orders to buy and then rounding the resulting prices down, leads to abnormal returns. The example in the exhibit assumes that the average closing cum-day bid is 49 7/8, and the ask is 50 1/8. The ex-cash dividend day return is a function of the proximity of the dividend amount to surrounding 1/8 multiples. Each segment of the function attains a local maximum when the dividend equals a 1/8 dollar multiple. The slope of each segment is 1/|P.sub.c~. Three testable predictions that exist under the conditions stated just above, and depicted in Exhibit 1, are that:

H1: The return for a cash dividend amount just below and equal to a multiple of an eighth exceeds the return for a dividend amount just above the same 1/8 multiple.

H2: Ex-cash dividend day returns are negative if the dividend is less than $0.0625.

H3: Except for the discontinuities around the 1/8 variations, returns increase as the cash dividend amount increases.

It is unlikely, however, that the high returns predicted for large cash dividends will be observed. As the size of a cash dividend (or the value of a stock distribution) increases, the model predicts that the ex-day bid-ask spread will widen. However, a specialist will likely intervene as a dealer on ex-dividend days at prices between the open limit order bid-ask quotes when spreads get large, thereby maintaining an orderly market and avoiding possible penalties imposed by the exchange. The specialist can act as a dealer by quoting his own bid at a price 1/8 above the highest limit order buy price, and his own ask at a price 1/8 below the lowest limit order sell price, yielding an ex-day pattern that is similar to, but less extreme than, the one depicted in Exhibit 1.

Active ex-day trading will lessen the likelihood that these predictions will be observed when returns are measured using closing prices. New ex-day limit orders and high trading volume will further narrow the spread. Lakonishok and Vermaelen |16~ and Karpoff and Walkling |14~ conclude that active short-term trading is concentrated in high-yield stocks. Thus, it is likely that the abnormal returns created by the limit order rules will be evident primarily in low-yielding, thinly traded stocks.

Abnormal returns on ex-days of small stock distributions are also created by the exchange rules. Large stock distributions create large effective dividends, and increase the likelihood that the specialist will act as a dealer; thus, market microstructure effects are likely to be observed only for small stock dividends.

Before proposing testable propositions concerning ex-stock dividend day returns, it is important to note that the current versions of NYSE Rule 118 and AMEX Rule 132 regarding stock distributions became effective on February 3, 1975, and November 22, 1976, respectively. Prior to those dates, both exchanges handled open orders to buy on ex-stock distribution days in the same way that they were handled on ex-cash dividend days, namely, neither exchange reduced open limit orders to sell stock. Thus, it is predicted that:

H4: For any given distribution size, cum-stock dividend price, and cum-stock dividend bid-ask spread, the ex-distribution day return for AMEX stocks will be less than for NYSE stocks, after November 22, 1976.

H5: There will be no difference between AMEX and NYSE ex-stock dividend day returns before February 3, 1975.

Because both limit orders to buy and limit orders to sell are now reduced and rounded down on the AMEX, it is predicted that:

H6: Ex-stock dividend day returns for AMEX stocks are predicted to be nonpositive after November 22, 1976.

H7: AMEX ex-stock dividend day returns should be lower after November 22, 1976, than before that date.

Finally, it is predicted that:

H8: Returns will be negative for one percent stock dividends, when the open limit buy price is less than $6.25 per share, because these conditions create an effective dividend less than $0.0625 per share.

II. Empirical Evidence

A. Data

The CRSP Daily Master File is the source of stock prices, ex-dividend dates, and cash dividend and stock distribution amounts. The sample constructed to test the cash dividend hypotheses uses the following criteria:

(i) Only cash dividends are considered (CRSP distribution codes less than 1300).

(ii) On an ex-cash dividend date, there are no other ex-distribution events.

(iii) The first (last) day of CRSP price data for the company is more than thirty days before (after) the ex-cash dividend date, and at least two "comparison period" returns exist during the sixty days surrounding the ex-cash dividend date.

Two abnormal return measures for an ex-cash dividend day event are computed. The comparison period approach defines AR1 to equal the ex-cash dividend day return less the average daily return during the sixty days surrounding the ex-date. The market-adjusted approach defines AR2 as the ex-cash dividend day return less the CRSP equal weighted market return on that day. For this study's sample of NYSE and AMEX ex-cash dividend days between July 2, 1962 and December 31, 1987, that meet the criteria stated above, the average AR1 (|Mathematical Expression Omitted~) equals 0.119% and the average AR2 (|Mathematical Expression Omitted~) equals 0.152%. There are 146,341 ex-cash dividend events in this sample.(9)

Samples of stock dividends between one percent and five percent on each exchange were also formed. Only small stock distributions were analyzed in order to lessen the chance of specialist intervention, which will likely be necessary for large stock distributions. The only qualification for inclusion in the sample for calculating raw returns, was that the stock did not also trade ex-cash dividend on the ex-stock dividend date. To form the sample for calculating abnormal returns, the first (last) stock price had to be more than thirty days before (after) the ex-cash dividend date, and at least two returns had to exist during the sixty-day comparison period.

B. Cash Dividends: Discontinuities Around 1/8 Variations

The first hypothesis (H1) predicts that market microstructure effects will cause the ex-cash dividend day abnormal return to be higher for a dividend just below or equal to a 1/8 multiple than for a dividend just above that 1/8 multiple. A local maximum return is realized if the cash dividend equals a multiple of $0.125. To test this, let D equal the dividend amount, and |B.sub.i~ denote a 1/8 variation for i = 1,2,...15. |B.sub.1~ = $0.125, and |B.sub.15~ = $1.875. If |B.sub.i~ - 0.025 |is less than~ D |is less than or equal to~ |B.sub.i~, then the ex-day event is placed in a portfolio labeled |BELOW.sub.i~. If |B.sub.i~ |is less than~ D |is less than or equal to~ |B.sub.i~ + 0.025, then it is placed in portfolio |ABOVE.sub.i~. Thus, for each dividend amount that is a multiple of $0.125, the average abnormal return (|Mathematical Expression Omitted~ and |Mathematical Expression Omitted~) is computed for cash dividends just below or equal to that multiple, and also for cash dividends just above that multiple.(10)

The results of this classification are presented in Exhibits 2 and 3 (NYSE and AMEX, respectively). The first column presents the cash dividend amounts, each of which is a multiple of $0.125. Below that are the number of ex-dividend day events for amounts just below or equal to that dividend amount (|N.sub.B~), and the number just above that amount (|N.sub.A~). The next columns are the average abnormal stock returns using the comparison period approach for the dividends just below or equal to the 1/8 multiple (|Mathematical Expression Omitted~) and just above that multiple (|Mathematical Expression Omitted~). The average abnormal stock returns for the market-adjusted returns method (|Mathematical Expression Omitted~ and |Mathematical Expression Omitted~) follow. Below each abnormal return is the T-statistic that tests the null hypothesis that the abnormal return equals zero. The next two columns are the T-statistics for testing the null hypothesis that there is no difference between |Mathematical Expression Omitted~ and |Mathematical Expression Omitted~ (TDIFF1), and that there is no difference between |Mathematical Expression Omitted~ and |Mathematical Expression Omitted~ (TDIFF2). Finally, the last two columns are the average dividend yield for the stocks below or equal to the 1/8 multiple, D|Y.sub.B~, and for those above it, D|Y.sub.A~.

The results show considerable support for the market microstructure hypothesis. Consider Exhibit 2 (NYSE). When the cash dividend amount is $0.10 |is less than~ D |is less than or equal to~ $0.125, the ex-day average abnormal return using the comparison TABULAR DATA OMITTED period benchmark, |Mathematical Expression Omitted~, is 0.195% (T = 7.40). For dividends just above that 1/8 multiple, $0.125 |is less than~ D |is less than or equal to~ $0.15, |Mathematical Expression Omitted~ = 0.079% (T = 4.13). The T-statistic for the difference in means is TDIFF1 = 3.06.

The same pattern holds for |Mathematical Expression Omitted~ and |Mathematical Expression Omitted~ for cash dividends of $0.25, $0.375, and $0.50. Ex-cash dividend day returns for dividends just below or equal to a 1/8 multiple are greater than the returns if the cash dividend amount is just above the same multiple of $0.125.

The AMEX results (Exhibit 3) are consistent with the NYSE results. The differences between the mean abnormal returns (TDIFF1 and TDIFF2) are not always significant, but Exhibit 3 shows that the abnormal returns are greater for dividends just below and equal to 1/8 multiples TABULAR DATA OMITTED than for dividends just above the multiples, for dividend amounts up to and including $0.75. Indeed, for five of the six 1/8 multiples for AMEX ex-cash dividend days, |Mathematical Expression Omitted~ is negative, and for three of the six multiples, |Mathematical Expression Omitted~ is negative.

The last line of Exhibits 2 and 3 presents the average abnormal return for all cash dividends just below or equal to the 1/8 multiples, and for those just above 1/8 multiples. On the NYSE (Exhibit 2), |Mathematical Expression Omitted~ = 0.188% while |Mathematical Expression Omitted~ = 0.053%. There is a statistically significant difference between these abnormal returns (TDIFF1 = 8.61). A similar result is found for abnormal returns estimated using the market-adjusted returns method. On the AMEX (Exhibit 3), |Mathematical Expression Omitted~ = 0.297% and |Mathematical Expression Omitted~ = -0.028%, with TDIFF1 = 10.3. Similar results are found with |Mathematical Expression Omitted~.(11)

The return pattern is not observable for large dividends (above $0.50 on the NYSE and above $0.75 on the AMEX). There are several explanations for this. First, as previously discussed, it is likely that the specialist will participate in handling market orders if a stock's bid-asked spread implied by open limit orders is too wide. Other forces, such as the greater trading volume associated with high-dividend-yielding stocks on their ex-days (Lakonishok and Vermaelen |16~, Karpoff and Walkling |14~), participation in trades by nonspecialist individuals who are on the trading floor (e.g., floor brokers), and new ex-day limit traders, also work to temper or eliminate the effects of NYSE Rule 118 and AMEX Rule 132 when returns are measured close-to-close.

A second explanation lies in ex-cash dividend day selling pressure by the two investor groups that trade in high-dividend (yielding) stocks around the ex-day. These include low-taxed, long-term investors who prefer to buy high-dividend stocks cum-dividend, and sell them ex-dividend, as well as short-term traders, many of whom are involved in dividend capture. The specialist faces unusual inventory risk as the ex-cash dividend day nears. He knows that on the ex-day he will have to intervene more actively as a dealer when buy orders are received, because open limit orders to sell will be "away" from the market. If he expected a "normal" day on which equal numbers of orders to buy and to sell are received, then he would accumulate added inventory of shares cum-dividend to meet the ex-day requirement for his dealer services. On the other hand, he will also have to anticipate the ex-day selling pressure by short-term traders and the high-dividend stocks' normal clientele: low-taxed investors.(12) Thus, as the ex-day nears, the specialist must anticipate two forces: one requiring inventory build-up and one requiring inventory shrinkage.

A third possible reason that ARs are not higher for large dividends just below 1/8 multiples was indicated by written correspondence with two NYSE specialist firms. They claimed that about 10% of the orders received on a typical trading day for high-volume stocks (which are typically high-dividend stocks) are good-till-canceled limit orders, compared to 15-20% for low-volume stocks.(13) These statements are consistent with Hasbrouck's |10~ conclusion that limit orders appear to be more important in determining the trade prices and quotes for low-volume stocks than for high-volume stocks. The specialists also estimated that 80-90% of good-till-canceled orders are placed by individuals, rather than institutional investors. This anecdotal evidence is consistent with the decline of the predicted market microstructure effects shown in Exhibits 2 and 3 when dividend amounts are large.

C. Close-Open Returns Versus Close-Close Returns

For large dividends, the results of Exhibits 2 and 3 fail to support the market microstructure hypothesis; abnormal returns for dividends just below or equal to a 1/8 multiple are not significantly greater than those for dividends just above that multiple, when the dividend amount exceeds $0.50 on the NYSE and $0.75 on the AMEX.

However, the predictions of the market microstructure hypothesis may be dissipated by active ex-day trading of these large dividend stocks. It is possible that the predictions concerning returns for large dividends around 1/8 multiples will be detectable on a close-to-open basis since active trading could wash out the effect as bid and ask quotes are altered over the ex-day.

The New York Times reported opening stock prices prior to October 1, 1972. Our sample consists of all NYSE firms that traded ex-cash dividend in an amount equal to or greater than $0.625 per share, between August 14, 1962 and September 28, 1972, inclusively, and also met the sample criteria described in Section II.A. above.

Exhibit 4 presents the raw returns for the entire sample of 1,581 events, and (line 2) those for large dividends just below or equal to a 1/8 multiple, and (line 3) those for large dividends just above a 1/8 multiple. Abnormal returns cannot be computed since the expected returns (comparison period return or the ex-day market return) cannot be dichotomized into cum-day close to ex-day open, and ex-day open to ex-day close components.

Exhibit 4 indicates that the ex-cash dividend day effect is entirely a cum-day close to ex-day open phenomenon for large-dividend stocks.(14) |Mathematical Expression Omitted~, the average overnight raw return, is 0.218%. During the ex-day itself, stock prices decline on average, since the average close-to-close return of 0.142% is lower than |Mathematical Expression Omitted~. In 757 cases, |Mathematical Expression Omitted~ exceeds |Mathematical Expression Omitted~, compared to only 636 cases in which prices actually rise during ex-day trading. The tendency for prices to decline on the ex-day is consistent with the existence of TABULAR DATA OMITTED short-term traders who capture dividends and sell the stock on the ex-day.

The market microstructure hypothesis not only predicts positive cum-day close to ex-day open returns for these large dividend stocks, but also predicts that returns will be greater for dividends just below or equal to a 1/8 multiple than for dividends just above that multiple. When abnormal returns are computed on a close-to-close basis, Exhibits 2 and 3 fail to support this hypothesis for high-dividend stocks. However, lines 2 and 3 of Exhibit 4 provide some evidence that is consistent with it. |Mathematical Expression Omitted~, the mean close-to-close return of 0.253%, is significantly greater for large dividends just above 1/8 multiples than the |Mathematical Expression Omitted~ of 0.079% for dividends just below or equal to 1/8 multiples. A test of the difference between these means yields a T-statistic of 2.08. These results for |Mathematical Expression Omitted~ are the opposite of market microstructure predictions. However, the difference between mean returns for dividends just above 1/8 multiples versus those for dividends just below 1/8 multiples is insignificant on a close-to-open interval. |Mathematical Expression Omitted~ is 0.262% for dividends just above 1/8 multiples, and 0.223% for dividends just below or equal to multiples of $0.125; the T-statistic for the difference between means test is 0.68. This is consistent with market microstructure effects appearing on a cum-day close to ex-day open interval, with the impact dissipating during active trading on the ex-day.

D. Very Small Dividends

Market microstructure hypothesis H2 predicts that returns are negative for small cash dividend amounts less than $0.0625 per share. Ex-day abnormal returns for dividends of $0.06 or less were examined during the period July 2, 1962 to December 31, 1987. For 6,540 small dividends paid by NYSE-listed firms, |Mathematical Expression Omitted~ = 0.048% and |Mathematical Expression Omitted~ = 0.070%. These abnormal returns, though not negative, are smaller in magnitude than those presented on the last line of Exhibit 2. For the sample of 7,638 AMEX-listed stocks, |Mathematical Expression Omitted~ = -0.114% and |Mathematical Expression Omitted~ = -0.066%. These negative returns are consistent with the hypothesis H2. These results sharply contrast with the dividend clientele hypothesis, which predicts that the actions of long-term investors in high tax brackets will result in the smallest dividend (yielding) stocks having the largest abnormal returns. Thus, it would seem that market microstructure is having an impact.

E. Stock Dividends

Exhibit 5 presents the raw and abnormal returns for small stock distributions on each exchange before and after the effective dates of the limit order rules as they currently exist. Panel A contains returns measures for NYSE stock distributions. The columns under the |Mathematical Expression Omitted~ headings list the mean raw returns for samples of size N1. The columns for |Mathematical Expression Omitted~ and |Mathematical Expression Omitted~ are the mean abnormal returns using the comparison period approach and market-adjusted approach, respectively, for samples of size N2. The last three columns present the T-statistics for standard differences between means test. T1, T2 and T3 test whether there are differences between |Mathematical Expression Omitted~, |Mathematical Expression Omitted~ and |Mathematical Expression Omitted~, respectively. Panel A results indicate that NYSE raw and abnormal returns have generally been higher since the rule change, though only the abnormal returns (|Mathematical Expression Omitted~, |Mathematical Expression Omitted~) for five percent distributions have been significantly higher (T2 = -2.20, T3 = -2.17).

On the AMEX, however (Panel B), all returns measures have been lower since the rule change, as predicted by H7, for stock distributions of two percent, three percent, four percent and five percent. All AMEX returns measures for three percent stock distributions, and raw returns for five percent stock distributions, have been significantly lower. Under the model's assumptions, negative returns are predicted after November 22, 1976 for AMEX stock distributions (H6); empirically, |Mathematical Expression Omitted~, |Mathematical Expression Omitted~, and |Mathematical Expression Omitted~ for the three percent stock distributions and the abnormal returns for the two percent stock distributions are negative. None of the NYSE stock distributions produce negative returns.

TABULAR DATA OMITTED

The negative AMEX returns for one percent stock distributions before November 22, 1976, are consistent with hypothesis H8. Prior to that date, open limit orders to buy were handled in the same way for both stock distributions and cash dividends. Open orders to buy a stock with a last cum-day closing price of $6.25 or below would be handled on the one percent stock dividend ex-day as if a $0.0625 (or smaller) cash dividend were paid. Exhibit 1 illustrates that the market microstructure hypothesis predicts that these small cash dividends (or similarly treated stock dividends) will result in negative returns. Of the 32 one percent stock dividends in the AMEX raw returns sample, 15 were stocks selling below $6.25 per share. Ten of these 15 low-priced one percent stock dividends produced negative raw returns, and the |Mathematical Expression Omitted~ for these was -1.484%. Of the 17 one percent stock dividends paid by stocks priced above $6.25 per share, only six had negative raw returns, and the |Mathematical Expression Omitted~ for these was +0.002%. Similar results were found for the abnormal returns sample. In contrast, only one of the NYSE one percent stock distributions before February 3, 1975, was made by a stock selling below $6.25. Thus, negative returns would not be expected, and this expectation is supported by the data.

Finally, Panel C of Exhibit 5 presents T-statistics that test whether there are differences between mean returns on the two exchanges. Hypothesis H5 predicts that there will be no difference before the rules changed (open orders were handled in the same way), but returns will be lower on the AMEX after the rules changed (H4). Panel C shows that abnormal returns on the two exchanges were not significantly different before the rules changed, except for one percent stock distributions (T2 = 2.05 for |Mathematical Expression Omitted~ and T3 = 1.83 for |Mathematical Expression Omitted~). This latter result is expected since hypothesis H8 explains that there will be low AMEX returns for one percent distributions prior to November 22, 1976. Subsequent to the rule change, all NYSE returns measures have been greater than AMEX returns for every stock distribution size. However, only T2 for three percent distributions, and T2 and T3 for five percent distributions are statistically significant at the five percent level in one-tailed tests.

III. Discussion and Conclusion

The evidence presented in this paper is consistent with two market microstructure details: trading in eighths, and NYSE Rule 118 and AMEX Rule 132. Abnormal ex-day returns are induced by the rules, which dictate that specialists lower all outstanding limit buy orders by the dividend; if the resulting price is not a multiple of $1/8, then the limit order price is rounded down. Outstanding limit orders to sell remain unchanged on the ex-dividend day. Ex-dividend day abnormal returns will be an increasing saw-toothed function of the dividend amount because of these rules, and because stocks trade in eighths while dividends are relatively continuous. Abnormal returns will be positive when the dividend exceeds $0.0625 per share. The return for a dividend just below or equal to a multiple of an eighth will exceed the return for a dividend just above the same 1/8 multiple. Ex-day returns are predicted to be negative if the dividend is less than $0.0625 per share.

Ex-stock dividend day returns should be affected by the rules, too. Prior to February 3, 1975, the NYSE and the AMEX handled outstanding limit orders on ex-stock distribution days in the same way. More recently, NYSE specialists have been reducing limit orders to buy on ex-distribution days, but not reducing limit orders to sell. Both limit orders to buy and to sell AMEX-listed stocks are now reduced on ex-distribution days. Thus, this paper makes specific predictions concerning differential ex-stock dividend day returns before and after the rules were changed, and cross-sectionally for NYSE stocks relative to AMEX stocks.

Evidence supporting the market microstructure hypothesis is provided by the result that abnormal returns are significantly greater for ex-cash dividend amounts just below and equal to $0.125, $0.25, $0.375, and $0.50 than for dividends just above those amounts. Abnormal returns for very small cash dividends, less than $0.06 per share, are low for NYSE stocks, and negative for AMEX stocks. The lack of support for the market microstructure predictions when dividends are large can be explained by active ex-day trading for these stocks.

Results also support the hypothesis that market microstructure affects ex-stock dividend day returns. AMEX abnormal returns are found to be lower on ex-stock dividend days subsequent to the date on which the rules for handling open limit orders changed. AMEX ex-stock dividend day returns have been below NYSE returns since the rules were changed. AMEX stock distributions of two percent and three percent have realized negative abnormal returns after November 22, 1976. The negative abnormal returns for one percent stock dividends by AMEX-listed stocks before November 22, 1976, can be explained by the cases in which the effective dividend was below $0.0625 per share.

A normative question to be asked is: What should be the rule for handling open orders on the ex-day? Do NYSE Rule 118 and AMEX Rule 135 increase efficiency, or would some other rule work better? The rules may create thinner markets because limit sell orders are not reduced on ex-cash dividend days. It is possible that the rules lead to a wider and biased bid-ask spread, and increased demand for specialists' dealer services.

Because the competition provided by the public's limit orders is lessened, the specialist has greater market power on the ex-day, particularly at the opening. Even on a non-ex-dividend day, the specialist has considerable market power at the opening, relative to the remainder of the day (Stoll |23~ and |24~). Stoll and Whaley |26~ find that opening price volatility exceeds close-to-close volatility. With wider open limit order prices, opening ex-day volatility may be even greater than that of a "normal" opening. The range of possible opening prices (hence, the degree of pricing power) is greater on an ex-distribution day, as limit buy orders are reduced by an amount greater than or equal to the distribution, and limit orders to sell are not reduced at all (with the exception of AMEX stock distributions).

It cannot, however, be concluded that any resulting profits are excessive. As argued earlier, the specialist faces added risk (some of which is nondiversifiable) on the ex-day. He must anticipate the need for greater inventory to satisfy ex-day market buy orders. The specialist's buying may add upward price pressure during the cum-distribution period. The profits that the specialist might earn on the ex-day may only be compensation for the added price/inventory risk he faces.

Market microstructure has been used to explain other phenomena. Stoll and Whaley |25~ and Schultz |21~ consider the bid-ask spread as a possible cause of the size effect. Roll |20~ and Keim |11~ examine the spread's impact on the turn of the year effect. Ohlson and Penman |19~ find a persistent increase in ex-split returns variance for NYSE stocks. The largest increase is on the ex-stock distribution day itself (and on the next few days). Market microstructure may contribute to the explanation of this "empirical aberration". Finally, Karpoff and Walkling |14~ and |15~, Stickel |22~, and Venkatesh |27~ find ex-day returns to be positively correlated with transaction costs. Market microstructure may contribute to explain their result, though it is necessary to obtain further understanding of the role of specialist intervention and nonintervention in order execution.

The author would like to thank C. Barry, J. Karpoff, S. Lummer, A. Mahajan, J. Reising, T. Zivney, the participants of the Finance Workshop at Texas A&M University and the Texas Finance Symposium, and the editors of this issue of Financial Management for useful comments and suggestions.

1 Campbell and Beranek |3~ support this conclusion, and also cite many early papers that examined this issue.

2 See Woolridge |28~ for citations of earlier works which document ex-stock dividend day abnormal returns.

3 Other tax-related strategies for short-term traders exist. See Lakonishok and Vermaelen |16~.

4 The results of Choi and Strong |5~ suggest that it is more costly to buy shares prior to an ex-date of a distribution that creates an odd lot from a round lot, and less costly to sell the same old shares. The marginal investor will buy shares on the ex-stock distribution day, and prefer to sell them cum-stock distribution. The resulting price pressure (if any), bid-ask effect (see Grinblatt, Masulis, and Titman |9, p. 485~), and brokerage commission impacts (Brennan and Copeland |2~) can explain ex-day returns for distributions creating odd lots, such as small stock dividends. However, this transactions cost hypothesis does not explain abnormal ex-day returns for 2-for-1 and other distributions that create round lots.

5 Open stop orders to sell and stop limit orders to sell are reduced in the same way as open limit buy orders. Also, an investor may place a "do not reduce" limit order, though such orders are rare. Responding to my request, the NYSE found that on a randomly selected day (July 9, 1990), only 10% of all good-till-canceled limit orders were marked "do not reduce". In addition, for a random sample of twenty stocks trading ex-dividend between July 9, 1990 and July 13, 1990, the NYSE found that only 7.7% of previously placed good-till-canceled orders were canceled on their ex-dividend days. Thus, uncanceled limit orders are quite likely to have an impact on ex-day pricing.

6 The rationale behind these rules is that the exchanges do not want to assume what a customer wishes to do. Both exchanges believe that reducing (and rounding down) limit buy orders and leaving limit sell orders unchanged is the most conservative approach to handling these orders, as the rules force traders to change their orders if the new limit prices are not to their liking. When asked about why the AMEX also reduces limit sell orders on ex-stock distribution days, an AMEX official responded that he did not have the vaguest idea, and that he was surprised that there had been no complaints about the rule (to his knowledge).

7 Several other factors affect specialists' bid ask spreads. See Stoll |22~ for a review. Choe and Masulis |4~ find that spreads increase on ex-cash dividend days, in particular, at the open and for high-dividend-yielding stocks.

8 Bid and asked quotes are unlikely to remain stationary on any trading day. However, it is only necessary that the last trade occurs at either of the opening quotes, with equal probability, for the model's predictions to be observed empirically. This is most likely to occur for thinly traded stocks paying small cash or stock dividends.

9 Eades, Hess, and Kim |6~ and Lakonishok and Vermaelen |16~ equally weight all of the sample observations on a given day (ex-cash dividend days tend to cluster), and also standardize the ex-day abnormal returns in order to minimize heteroscedasticity. I replicated their method, and all of the conclusions presented below remained unchanged.

10 The specification of |B.sub.i~ |+ or -~ $0.025 was made after inspecting Exhibit 1; it demarcates the ranges for which the returns for dividends just below or equal to a one-eighth multiple strictly exceed the returns for dividends above that multiple.

11 All of the above results were confirmed when the sample was separated into the pre-negotiated commissions era and post-negotiated commissions era. I also examined the results for nontaxable cash distributions, and found |Mathematical Expression Omitted~ = 0.191%, |Mathematical Expression Omitted~ = -0.189% (TDIFF1 = 1.96), |Mathematical Expression Omitted~ = 0.180% and |Mathematical Expression Omitted~ = -0.159% (TDIFF2 = 1.64), for sample sizes of |N.sub.B~ = 271 and |N.sub.A~ = 245. Thus, market microstructure appears to be a determinant of ex-cash dividend day returns independent of commission rates, and the tax status of the dividend.

12 Choe and Masulis |4~ conclude that sellers dominate on the ex-dividend day, particularly in high-dividend-yielding stocks.

13 While 10-20% may seem insignificant, consider that unfilled good-till-canceled limit orders stay on the specialists' books until they are subsequently filled or canceled. Thus, the cumulative influence that these orders have on trades and quotes exceeds the impact that would be inferred by the daily 10-20% figures.

14 Eades, Hess, and Kim |6, pp. 10, 11~ and Lakonishok and Vermaelen |16, p. 310~ also conclude that ex-cash dividend day abnormal returns occur overnight.

References

1. M.J. Barclay, "Dividends, Taxes and Common Stock Prices: The Ex-Dividend Day Behavior of Common Stock Prices Before the Income Tax," Journal of Financial Economics (September 1987), pp. 31-44.

2. M.J. Brennan and T.E. Copeland, "Stock Splits, Stock Prices and Transaction Costs," Journal of Financial Economics (October 1988), pp. 83-101.

3. J.A. Campbell and W. Beranek, "Stock Price Behavior on Ex-Dividend Dates," Journal of Finance (December 1955), pp. 425-429.

4. H. Choe and R.W. Masulis, "Measuring the Impacts of Dividend Capture Trading: A Market Microstructure Analysis," Working Paper No. 91-13, Pennsylvania State University, May 29, 1992.

5. D. Choi and R.A. Strong, "The Pricing of When-Issued Stock: A Note," Journal of Finance (September 1983), pp. 1293-1298.

6. 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.

7. E.J. Elton and M.J. Gruber, "Marginal Stockholders Tax Rates and the Clientele Effect," Review of Economics and Statistics (February 1970), pp. 68-74.

8. E.J. Elton, M.J. Gruber, and J. Rentzler, "The Ex-Dividend Day Behavior of Stock Prices; A Re-Examination of the Clientele Effect: A Comment," Journal of Finance (June 1984), pp. 551-556.

9. M.S. Grinblatt, R.W. Masulis, and S. Titman, "The Valuation Effects of Stock Splits and Stock Dividends," Journal of Financial Economics (December 1984), pp. 461-490.

10. J. Hasbrouck, "Trades, Quotes, Inventories, and Information," Journal of Financial Economics (December 1988), pp. 229-252.

11. 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. 75-98.

12. A. Kalay, "The Ex-Dividend Day Behavior of Stock Prices: A Re-Examination of the Clientele Effect," Journal of Finance (September 1982), pp. 1059-1070.

13. A. Kalay, "The Ex-Dividend Day Behavior of Stock Prices; A Re-Examination of the Clientele Effect: A Reply," Journal of Finance (June 1984), pp. 557-562.

14. J.M. Karpoff and R.A. Walkling, "Short-Term Trading Around Ex-Dividend Days," Journal of Financial Economics (September 1988), pp. 291-298.

15. J.M. Karpoff and R.A. Walkling, "Dividend Capture in NASDAQ Stocks," Journal of Financial Economics (November/December 1990), pp. 39-66.

16. J. Lakonishok and T. Vermaelen, "Tax-Induced Trading Around Ex-Dividend Days," Journal of Financial Economics (July 1986), pp. 287-320.

17. C.G. Lamoureux and P. Poon, "The Market Reaction to Stock Splits," Journal of Finance (December 1987), pp. 1347-1370.

18. R. Michaely, "Ex-Dividend Day Stock Price Behavior: The Case of the 1986 Tax Reform Act," Journal of Finance (July 1991), pp. 845-860.

19. J.A. Ohlson and S.H. Penman, "Volatility Increases Subsequent to Stock Splits," Journal of Financial Economics (June 1985), pp. 251-266.

20. R. Roll, "Vas Ist Das? The Turn of the Year Effect and the Return Premium of Small Firms," Journal of Portfolio Management (Summer 1983), pp. 18-28.

21. P. Schultz, "Transaction Costs and the Small Firm Effect: A Comment," Journal of Financial Economics (June 1983), pp. 81-88.

22. S. Stickel, "The Ex-Dividend Behavior of Nonconvertible Preferred Stock Returns and Trading Volume," Journal of Financial and Quantitative Analysis (March 1991), pp. 45-62.

23. H.R. Stoll, "Alternative Views of Market Making," in Market Making and the Changing Structure of the Securities Industries, Y. Amihud, T.S.Y. Ho, and R.A. Schwartz (eds.), Lexington, MA, Lexington Books, 1985, pp. 67-91.

24. H.R. Stoll, "The Stock Exchange Specialist System: An Economic Analysis," Salomon Brothers Center for the Study of Financial Institutions Monograph 1985-2, 1985.

25. H.R. Stoll and R.E. Whaley, "Transaction Costs and the Small Firm Effect," Journal of Financial Economics (June 1983), pp. 57-80.

26. H.R. Stoll and R.E. Whaley, "Stock Market Structure and Volatility," Review of Financial Studies (1, 1990), pp. 37-71.

27. P.C. Venkatesh, "Trading Costs and Ex-Day Behavior: An Examination of Primes and Scores," Financial Management (Autumn 1991), pp. 84-95.

28. J.R. Woolridge, "Ex-Date Stock Price Adjustment to Stock Dividends: A Note," Journal of Finance (March 1983), pp. 247-255.

Printer friendly Cite/link Email Feedback | |

Title Annotation: | Market Microstructure and Corporate Finance Special Issue |
---|---|

Author: | Dubofsky, David A. |

Publication: | Financial Management |

Date: | Dec 22, 1992 |

Words: | 7499 |

Previous Article: | Order flow, trading costs and corporate acquisition announcements. |

Next Article: | The effects of splitting on the ex: a microstructure reconciliation. |

Topics: |