Printer Friendly

Trading costs and ex-day behavior: an examination of primes and stocks.

* The study of security price behavior around ex-dividend days is of interest because it can help us understand investors' valuation of cash dividends. The nature of taxation of dividends and capital gains has been a central issue in this research. In the U.S. individual investors have generally faced a tax penalty on dividends (relative to capital gains), while corporate investors have enjoyed a tax advantage on dividends, and entities such as securities dealers have been taxed equally on both forms of income.(1) Modelling the effect of the tax code on equilibrium ex-day behavior therefore requires an analysis of the interplay between these different groups.

Elton and Gruber[6] proposed that, in equilibrium, investors might sort themselves into tax-induced dividend clienteles, such that high-tax investors hold low-yield securities and vice-versa.(2) Consistent with this argument, they find that the ex-day relative price drop, ([P.sub.cum] - [P.sub.ex) / DIV, is noticeably smaller than 1.0 for securities with low dividend yield, and is closer to or greater than 1.0 for high-yield securities. This suggests a dividend tax penalty on lower yield securities and a smaller penalty or even a dividend preference for higher yield securities.(3) Lakonishok and Vermaelen [15] point out two distinctive features of this framework: (i) trading by "long-term investors," who have decided to buy or sell the security, for reasons unrelated to the dividend determines ex-day behavior, and (ii) trading costs are not an incremental concern, and therefore do not affect equilibrium. Equilibrium ex-day behavior is presumed to be such that investors within each clientele are indifferent between trading before or after the dividend.

However, Kalay [13] noted that agents may not behave so passively. Within the bounds imposed by trading costs, securities dealers (who face equal effective tax rates on dividends and capital gains) and incorporated entities (who have a tax advantage on dividends) may be able to profit from positive ex-day returns by buying cum-dividend and selling ex-dividend; securities dealers may profit from negative ex-day returns by shortselling. If securities dealers have the lowest transaction costs, the no-profit bounds for the ex-day relative price drop will be symmetric around unity. Within these no-profit bounds, either taxable individuals or corporations may be the marginal traders. Thus, the empirical relevance of this argument depends on the "width" of the no-profit bounds for realistic transaction costs. Elton, Gruber and Rentzler [7] argue that, at least for 1966 and 1967, transaction costs (even for broker dealers) were sufficiently large that ex-day behavior was well within the no-profit region, so that tax-related interpretations of the results might be appropriate, after all.

In more recent times, it is believed that trading costs may have declined sufficiently to make cum-ex trading "profitable" for incorporated entities. Several recent papers (e.g., Grammatikos [8], Karpoff and Walkling [14], and Lakonishok and Vermaelen [15]), are aimed at empirically documenting the prevalence and impact of corporate cum-ex trading. It should be noted at the very outset that these authors (e.g., Lakonishok and Vermaelen [15], and Karpoff and Walkling [14]), assume that shortselling around ex-days is not a common practice, and therefore consider only corporate cum-ex trading in their empirical analysis. Although not directly observable, they note that cum-ex trading is likely to be positively related to dividend yield, and negatively related to trading costs. Corporate cum-ex traders have to contend with not only direct transaction costs (bid-ask spread and commissions), but also costs arising from the price-risk and illiquidity of the trading positions taken.(4,5) Evidence indicating differential ex-day price drops for securities that differ only in trading costs would suggest discernible cum-ex trading activity. In particular, securities with lower trading costs, ceteris paribus, would be expected to exhibit larger ex-day relative drops.(6)

This paper seeks to add to this literature by studying the ex-day behavior of a set of stocks and of their newly created derivative securities known as Americus Trust Primes. The underlying stocks for which primes have been created are generally those of large firms with high dividend yields and listed options, making them good candidates for cum-ex trading, especially dividend capture strategies by incorporated entities. By construction, each prime offers virtually the same dollar dividend (and slightly higher dividend yield) as the underlying stock. However, the direct trading costs for the primes are far greater: their supply is limited (at most, to 5% of the outstanding shares of the underlying stock), trading frequency and daily volume are very low, and their bid-ask spreads are much higher than those of the stocks. The holding costs for the primes are also likely to be greater because hedged positions are not as easy to construct.(7)

From the standpoint of cum-ex traders, the benefits of the prime and the stock would be approximately equal, but the disadvantages of the prime (trading costs) are substantially greater. Cum-ex activity in the stock but not in the prime, would be reflected in larger relative price drops for the stock. Also, unusually great ex-period volume is expected in the stock, but not in the prime. These are the main predictions tested in the paper.

The article employs a database which includes data on bid-ask quotes. Quote midpoints, rather than CRSP prices, are used to compute price changes, alleviating the bid-ask related effect present in CRSP returns (see Blume and Stambaugh [2]) and Venkatesh [18]). Also, because the price adjustment for the dividend takes place at the opening of trading, returns are calculated on a close-to-open basis. Close-to-close returns, which most ex-day studies use, will also impound the (unrelated) intraday movements and hence cannot isolate the dividend-related price changes as effectively.(8)

Overall, the results, although not especially strong, indicate that the ex-day price ratios are smaller and the ex-day returns are larger for the primes compared to the underlying stocks. Diagnostic checks confirm that these results are robust with respect to contaminating events and alternative specification of returns. Clear evidence of an increase in ex-period volume is found for the stocks but not for the primes. These results are consistent with the propositions that: (i) a security's trading costs affect its ex-day behavior, and (ii) cum-ex trading in stocks results in ex-day returns of stocks being smaller than they would otherwise be. That is, ex-day returns for stocks and primes may well be similar if their trading costs are similar.

I. Background and Hypotheses

A. Primes and Scores

The Americus Trust Corporation formed a unit investment trust and created new securities based on the stock securities of 26 large corporations; the trust has a life of five from inception.(9) The new securities, in effect, split the underlying stock's returns into a capital appreciation component ("score") and a dividend-oriented component ("prime"). A holder of the prime receives all the dividend payments by the underlying stock (less an annual trust fee of, at most, five cents per share) and any increase in the stock price up to a prespecified termination claim. The score component is entitled to all the stock price appreciation above the termination claim. The eventual exchanges are treated as nontaxable events. The primes and scores trade independently on the American Stock Exchange.(10)

Jarrow and O'Hara [12] examine the pricing of primes and scores relatives to the underlying stock. Their data set consists of daily closing prices for the prime, the score and the stock of five stock securities from the inception of the corresponding prime and score to June 1987. They conclude that the combination of the prime and the scores sells at a premium relative to the underlying stock, even after allowing for estimated transaction costs and using conservative statistical tests. They suggest that the premium might exist because the score, being similar to a long-term call option on the stock, can save on the costs of dynamic hedging.(11)

B. Ex-Day Behavior

Theory. The literature on the conceptual issues pertaining to ex-day effects is quite extensive and well-developed. Detailed discussions of the issues and further citations can be found in, for example, Copeland and Weston [4], Lakonishok and Vermaelen [15], Kalay [13], and Karpoff and Walkling [14]. The main relations are the following.

First, if equilibrium is determined by long-term investors (as per the tax-clientele argument), prices are set such that the marginal shareholder is indifferent between selling before or after the stock goes ex-dividend, so that(12)

[P.sub.cum] - ([P.sub.cum]-[P.sub.o])[t.sub.g]=([P.sub.ex]-[P.sub.o])[t.sub.g]+D(1-[t.sub.d]) (1)

where [P.sub.cum] is the cum-dividend price, [P.sub.ex] is the ex-dividend price; [P.sub.o] is the original purchase price (tax basis), [t.sub.g] is the tax rate on capital gains/losses, and [t.sub.d.] is the tax rate on ordinary / dividend income. This can be arranged to yield:

[P.sub.cum] - [P.sub.ex] / D = 1 - [t.sub.d] / 1 - [t.sub.g] (2) and

[P.sub.ex]+D-[P.sub.cum] / [P.sub.cum = ([t.sub.d]-[t.sub.g])D / (1-[t.sub.g])[P.sub.cum] (3) which are the ex-day relative price drop and return respectively. Two inferences can be drawn from the above. First, if [t.sub.d]>[t.sub.g], then the relative price drop is smaller than unity, or ex-day returns are positive. Also, a positive relationship would obtain between ex-day returns and dividend yield.(13) Second, if tax clienteles exist (high-tax investors hold low-yield stocks and vice-versa), then the relative price drop will be larger for high-yield stocks. Note that because the investor wishes to sell (or buy) the stock for reasons unrelated to the dividend, transaction costs are not an incremental concern, and therefore do not appear in the equation.

Theory. The literature on the conceptual issues pertaining to ex-day effects is quite extensive and well-developed. Detailed discussions of the issues and further citations can be found in, for example, Copeland and Weston [4], Lakonishok and Vermaelen [15], Kalay [13], and Karpoff and Walkling [14]. The main relations are the following.

First, if equilibrium is determined by long-term investors (as per the tax-clientele argument), prices are set such that the marginal shareholder is indifferent between selling before or after the stock goes ex-dividend, so that(12)

[P.sub.cum] - ([P.sub.cum]-[P.sub.o])[t.sub.g]=([P.sub.ex]-[P.sub.o])[t.sub.g+D(1-[t.sub.d]) (1)

where [P.sub.cum] is the cum-dividend price, [P.sub.ex] is the ex-dividend price; [P.sub.o] is the original purchase price (tax basis), [t.sub.g] is the tax rate on capital gains/losses, and [t.sub.d.] is the tax rate on ordinary / dividend income. This can be arranged to yield:

[Mathematical Expression Omitted] (2)


[Mathematical Expression Omitted] (3)

which are the ex-day relative price drop and return respectively. Two inferences can be drawn from the consistent with cum-ex traders affecting ex-day returns.

Evidence. The empirical research on ex-day behavior is quite extensive. I concentrate here on recent papers whose main concern is with cum-ex trading. Lakonishok and Vermaelen[15] and Grammatikos[8] present evidence consistent with cum-ex trading activity. Lakonishok and Vermaelen[15] argue that cum-ex trading would lead to abnormally large ex-period volume, whereas trading by long-term investors (the key players in the tax-clientele argument) would not lead to any net abnormal volume. They find evidence of unusually great ex-period volume.

Grammatikos studies the effect of the 1984 Tax Reform Act, which sought to increase the risks and/or reduce the rewards to cum-ex trading.(16) Conducting a matched pre- and post-analysis, he finds that the post-Act ex-day returns are larger. This is consistent with a withdrawal of cum-ex traders leading to larger ex-day returns. He finds these effects to be strongest for (i) high-yield stocks, and (ii) for stocks without listed options. The first relation is consistent with cum-ex trading being generally greater in high-yield securities, as expected. Since listed options provide cum-ex traders with a means of reducing the risk of their trading position (see Brown and Lummer [3] and Zivney and Alderson[19]), the risks of dividend capture could still be curtailed for optioned securities but not for nonoptioned securities. Thus, the second relation strengthens the notion that cum-ex traders do affect ex-day behavior and that holding costs are an important component of the total costs of a dividend capture strategy.

Eades, Hess and Kim [5], in a careful and comprehensive study, find that ex-day behavior for taxable distributions is largely consistent with tax arguments. However, they also document a number of other perplexing results with respect to nontaxable distributions, stock splits, and behavior before and after ex-days, which the tax hypothesis cannot explain. Research continues on these "anomalies." While Lakonishok and Vermaelen[15] find similar results for returns around ex-days, Grammatikos does not.

Trading Costs and Ex-Day Behavior. In a recent paper, Karpoff and Walkling[14] examine the relationship between trading costs and ex-day behavior for a large sample of ex-day events on NASDAQ securities. While all other studies have used relatively weak proxies for trading costs, they use the bid-ask spread. They find a positive cross-sectional relationship between ex-day returns and bid-ask spreads. The strength of the relationship increases across dividend yield groups and is most significant in high-yield stocks. They conclude from this evidence that dividend capture does affect the ex-day returns of at least some, especially high-yield, NASDAQ stocks.(17)

The present study and the paper by Karpoff and Walkling have similar research objectives: namely, to discern whether ex-day behavior is related to trading costs. However, there are sufficient differences in the samples studied and the methodologies that the results are complementary rather than competing. Their paper provides valuable cross-sectional results and methodological aspects not present in this paper. A brief comparison of the two papers follows.

My sample, although considerably smaller, has the advantage that it permits a more controlled experiment in which one is reasonably sure that all other relevant factors are very similar in design. Karpoff and Walkling must use grouping and regression procedures to control for the other factors, which is unavoidable in large samples. For the sake of argument, one may note the following potential problems with the latter approach. First, dividend yield and the bid-risk spread are both likely to be related to the total variability of the security and also to each other; after conditioning or "grouping" on dividend yield, it is conceivable that there is not a great deal of cross-sectional variation in trading costs. Second, in academic studies, a security would be placed into a dividend yield group or a risk group generally on the basis of historical averages. The actions of traders, however, will depend on current information, which may indicate a risk measure or yield value different from the historical average. It is not inconceivable that such misclassification would generate "outliers" that affect results even in large samples. Because the prime is a derivative security and the analysis uses matched differences, such problems would be minimized. In sum, although the small sample size precludes generalization, it does guard against such problems.

C. Hypotheses

Ex-day behavior will be evaluated using the ex-day price ratio, defined as ([P.sub.cum] - [P.sub.ex])/DIV, and the ex-day close-to-open return, ([P.sub.ex] - [P.sub.cum] + DIV) X 100/[P.sub.cum]. The results will be interpreted in the spirit of the following (alternative) hypotheses.(18)

(i) If the ex-day relative price drops of the primes are

smaller than those of the stocks, that would support

that argument that trading costs do affect

ex-day returns. It would be consistent with greater

cum-ex trading in the stock, leading to larger relative

price drops. The absence of a significant difference

would suggest that trading costs do not

noticeably affect ex-day behavior.

(ii) Unusually great ex-period trading volume in the

stock, but not in the prime would similarly support

the idea that trading costs affect ex-day behavior,

and that cum-ex traders are active in the stock, but

not in the prime.

II. Data

The primary data source is the database from the Institute for the Study of Securities Markets (ISSM) [10]. This database provides information on all securities on the NYSE and AMEX, for the year 1988. For each trading day, the file provides the entire time-stamped sequence of transaction (trade price and shares traded) and bid-ask quotes data, along with codes to indicate any special aspect of a trade or the quotes.(19) The files also contain data on the number of shares outstanding and all the dividend-related information necessary for this study. These data were cross-checked against the dividend information from the CRSP master files and the Standard & Poor's Daily Stock Price Record, and corrections made where necessary. The Standard & Poor's Stock Guide was used to obtain data on institutional shareholdings and the termination claims and dates for the units (the same data apply for the prime and the score).

Exhibit 1 presents some characteristics of the sample. For each firm, the time-series mean (over 253 trading days in 1988) of a variable is used as its representative value. The exhibit presents summary statistics from the cross-sectional distribution of these individual representative values. The stock sample is composed of firms which are actively trades, as reflected in the large values of average transactions per day, shares traded per day and blocks traded per day.(20) The mean (annualized) dividend yield of 4.19% would put these stocks in the medium- to high-yield group, relative to the population of stocks (see below). They also have large equity values, and institutions hold about half the outstanding shares, for a typical firm. The low percentage bid-ask spreads, coupled with active trading, indicate that trading costs would be quite low in terms of the bid-ask spread and liquidity.


The primes have slightly higher dividend yields than the stocks and substantially lower price variability - the average standard deviation of daily returns is 1.24%, which is almost half that of the stocks.(21) Because the prime is a derivative security, its price behavior relative to the stock will be quite complicated.(22) However, they are clearly inactively traded securities, averaging only about three trades and 3,200 shares per day. The mean percentage spread of 1.52% is three times the average spread for the stocks.(23) Overall, these data indicate that trading costs would be substantially greater for the primes than for the stocks, in terms of the spread and illiquidity. About a third of the outstanding shares are held by institutions.

Finally, we note that the scores enjoy somewhat greater trading activity than the primes, but show large price variability, with standard deviation of daily returns averaging 13.5%! This is not surprising in view of the option characteristics and the inherent leverage in the scores. While the dollar spread for the scores averages about half that of the primes, the percentage spread is about three times as large. The scores are generally "out-of-the money" since, on average, the stock price is only 75% of the termination claim (which is akin to the exercise price).

It is apparent that the dollar bid-ask spreads on the prime and the score are quite large, averaging about eighty cents and forty-two cents respectively, and the question arises whether the systematic price discrepancy noted by Jarrow and O'Hara [12] persists after the spread is accounted for. Using the opening quote midpoint for each security, a time-series of the daily difference between the stock price and the sum of the prices of the prime and the score was obtained.(24) For virtually every firm, the mean value indicated that the prime plus score combination was worth more than the stock; the cross-sectional mean of the individual means was thirty-two cents. Although the figure is smaller than that reported by Jarrow and O'Hara, this analysis strongly confirms their finding.

III. Empirical Analysis

A. Basic Results

Two standard measures are used to examine ex-day price behavior. The first is the ex-day relative price drop, defined as ([P.sub.cum] - [P.sub.ex])/DIV, and the second is the ex-day close-to-open return, ([P.sub.ex] - [P.sub.cum + DIV) X 100/[P.sub.cum]. is taken to be the midpoint of the closing quotes on the day prior to the ex-day, and [P.sub.ex] is taken to be the midpoint of the opening quotes on the ex-day. Quotes from the security's primary trading exchange (NYSE for the stocks and AMEX for the primes) are used for these calculations.(25) Alternative definitions of [P.sub.cum] and [P.sub.ex] are discussed later.

The ex-day price drop and the ex-day return are computed for each security. Separate cross-sectional distributions are generated for the stock sample and the prime sample to indicate the typical values of these variables. However, the key hypotheses of differential ex-day behavior are based on the matched difference (stock value less the corresponding prime value) of the ex-day relative price drop and the ex-day return.

The results are presented in Exhibit 2 for the final sample of 98 ex-day events.(26) For both the stocks and the primes, the ex-day relative price drop is significantly less than 1.0, and ex-day returns are significantly positive, as reflected in the means and medians; however, the interquartile ranges and the standard deviations do indicate substantial cross sectional variation.

( 1) The specifics of the tax code applicable to the period under study are discussed in the next section. ( 2) Of course, portfolio diversification considerations will also play a role in this process. See Long [16]. ( 3) Elton and Gruber [6] also infer the marginal tax bracket of each clientele from the price drop. Also, the ex-day return, (([P.sub.ex] - [P.sub.cum] + DIV) / [P.sub.cum]) is positive for low-yield securities and close to zero or negative for high-yield securities. Elton and Gruber suggest that the dividend preference implied by the negative returns could be due to corporate investors. ( 4) Brown and Lummer [3], Zivney and Alderson [19], and Grammatikos [8] discuss alternative hedging strategies for dividend capture. In particular, for incorporated traders, the minimum holding period (46 days at present) creates a nontrivial exposure to risk. The funds invested in the dividend capture position are also illiquid during this period. Short-term traders, who have no holding period requirement, would face liquidity concerns regarding trading frequency and volume. These determine the ability to reverse the position when desired. Poor liquidity in this sense would also increase risk exposure by increasing the expected holding period. ( 5) While the bid-ask spread itself reflects the security's risk and liquidity characteristics to some extent, a cum-ex trader must not only compensate the market-maker for these costs, but additionally suffer them on his own account. ( 6) That is, given no shortselling, corporate cum-ex traders would drive the relative price drop to larger values than taxable individuals would. As Lakonishok and Vermaelen [15] point out, if corporate cum-ex trading is concentrated in high-yield securities, this will lead to larger price drops for high-yield securities - an outcome very similar to the tax clientele effect. Hence, they argue that an examination of ex-day price behavior alone cannot distinguish between the two hypotheses; they employ ex-period volume for that purpose. ( 7) Position in the stocks can be hedged by using listed options. Because of the prime's somewhat unusual payoff structure, finding an appropriate hedge may not be as easy. ( 8) Lakonishok and Vermaelen [15] use close-to-open transaction prices, which will not eliminate the bid-ask effect. Eades, Hess and Kim [5] use close-to-open returns for the 30 stocks in the Dow Jones Index. ( 9) Most of the trusts expire in 1992. (10) Due to an adverse tax ruling than in effect leads to double taxation, it is expected that no more new trusts will be created. See Jarrow and O'Hara [12, footnote 2,p. 1265] for details. (11) Ingersoll [11] provides a theoretical and empirical examinations of dual-purpose funds, which are similar in some respects to the primes and scores. He demonstrates that the combination of the income and capital shares may sell at a discount to the asset value, because they cannot be redeemed prior to maturity. The Americus Trust Corporation, however, stands ready to exchange a prime plus score for the unit, or equivalent, the underlying stock. Hence, a discount would present a very straightforward arbitrate opportunity.

The scores are not particularly useful for hedged dividend capture, because for the time period studied, they are generally out-of-the-money, and they are very long-term call options. Hedged dividend capture is most effective with short-maturity, deep-in-the-money options (Brown and Lummer [3]). Following the 1984 Tax Reform Act, only the least-in-the-money option can be used for this purpose. (12) A similar argument can be made for a buyer. The same final relation obtains if all parameters (e.g., tax rates) are the same. (13) These equations predict that the relationship would disappear after 1986, when [t.sum.d] = [t.sub.g]. The tax timing option associated with capital gains is not reflected here, because the derivation of Equation (1) assumes that the investor has already decided to sell and will incur the capital gains tax. See also Equations (3) and (4) in Lakonishok and Vermaelen [15]. (14) Under the present tax law, a corporation that holds up to 20% of the common stock of the dividend-paying firm can deduct 70% of the dividend income for tax purpose. To qualify for this, the firm must hold the stock for at least 46 days, and be at

risk. The deductible amounts are 80% for holders of up to 80% of the stock, and 100% for holders of greater than 80% of the stock. (15) Karpoff and Walkling use this expression; Lakonishok and Vermaelen [15] omit the holding cost, H. [P.sub.ex] should be viewed as an expected price. The holding cost incorporates the after-tax transaction costs of hedging plus the risk premium for any remaining exposure to unsystematic risk under the optimal hedging strategy. (16) First, the minimum holding period was extended from 16 days to 46 days. The provides no additional gains, but adds to the risks. Second, whereas dividend captures would like to hedge by writing deep-in-the-money call options, the law decreed that only the least-in-the-money option could be used. See Brown and Lummer [13] for further discussion and analysis. (17) Less direct evidence on this point is also provided by Lakonishok and Vermaelen [15] and Eades, Hess and Kim [15]. They find that ex-day returns are generally smaller after May 1975, when transaction costs fell as commission rates became negotiable. (18) Another testable implication, for which I do not have sufficient data, is to examine whether the ex-day behavior of stocks is altered by the introduction of the primes. (19) A detailed description of the database can be found in the ISSM manual. The procedure by which the data are recorded is also discussed in Blume, Mackinlay and Terker [1]. (20) Comparable figures (cross-sectional means) for the population of NYSE and AMEX stocks are as follows: 74,6000 shares traded per day and 33 trades per day. The customary definition of a block, namely, trade size of greater than 10,000 shares, is used. (21) Annualized dividend yield decile value were computed for the entire sample of NYSE and AMEX securities on the 1988 ISSM tape. The decile values were: 1.26% (10th percentile), 1.82%, 2.27%, 2.80%, 3.22%, 3.75%, 4.52%, 6.12%, 8.08%, 64% (maximum). The stocks and primes are concentrated roughly between the 65th and the 75th percentiles, so in an overall sense, their yields are very similar. (22) It will depend a good deal on the ratio of the stock price to the prime's termination claim. A regression of daily prime returns on stock returns (based on quote midpoint changes) was run for each prime-stock pair. The cross-sectional mean [R.sup.2] was 51% and the mean "[beta]" coefficient was 0.52. Naturally, these regressions have to the interpreted with caution because there is an upper bound to the prime's price (much like a firm's risky debt). The prime's predicted response to a stock price change would be a function of the ratio of the stock price to the prime's termination claim, so that the theoretical "regression coefficient" is not expected to be a fixed number. However, the relatively small [R.sup.2] does suggest that hedged dividend capture with the primes would not be as effective as with the stocks. (23) This relatively large percentage spread highlights the improvement in measurement accuracy by using quote midpoint returns. Even if the closing and opening quotes were identical, transaction returns may be quite large; e.g., when a trade at the bid is followed by a trade at the ask, or vice-versa. (24) The data are adjusted for splits. Opening quotes were used because they were better synchronized (in time) than closing quotes. (25) "Composite" closing quotes, which may be from the Pacific Stock Exchange, are considered subsequently. (26) The necessary data is available for 102 ex-day events. However, because the minimum allowable change in bid and ask prices is 1/8th, observation with dividend [is less than or equal to] 12.5 cents were deleted.

These magnitudes are similar to those found in other studies.(27) The results for the matched difference indicate that the relative price drop of the stock is generally greater than that of the associated prime; the median and the mean of the matched difference are 0.10 and 0.08, respectively. This implies that the price drop ([P.sub.cum] - [P.sub.ex]) relative to the dividend, is greater for the stock than for the prime. Likewise, the matched difference of ex-day returns shows that the stock's return is smaller than the prime's return; that is, the stock's return, although positive, is closer to zero.(28)

These results, although not overwhelming statistical significance, are generally consistent with the basic thesis put forth earlier. That is, although the prime offers the same dividend and the same risk characteristics of the stock, the higher trading costs deter cum-ex trading in the prime. This absence (or lower level) of cum-ex trading results in the average ex-day return of the prime being greater than the stock's average ex-day return. Or, alternatively, cum-ex trading in the stock lowers the stock's ex-day return below what it would otherwise have been. In that sense, the evidence here suggests that: (i) cum-ex traders do affect ex-day returns, especially for high-yield, low-trading-cost securities, and (ii) a security's ex-day behavior is affected by its trading costs.(29)

B. Robustness of the Results

A number of diagnostic checks were carried out to examine the robustness of the results.

Contaminating Events. The Wall Street Journal Index was scanned for "major" announcements/events that took place on the ex-day or the day before. Five ex-day events were dropped after this screening.(30)

The results for the reduced sample of 93 ex-day events are presented in Exhibit 3. The mean return for the stocks and the primes is slightly greater than that for the full sample, suggesting that the announcements on average, had a negative impact. The results are essentially the same as those in Exhibit 2, except that the matched difference price drop is now statistically significant. Thus, the cleaner sample corroborates and reinforces the previous results.


Mondays Excluded. Harris [9] concludes that, on average, large firms accrue a negative return from the close on Friday to the opening on Monday. For smaller firms, the negative return takes place during the trading hours on Monday, rather than at the open itself. Hence, the presence of ex-days that fall on Mondays may impart a downward bias to the results.(31) Excluding the contaminating events and Mondays reduces the sample to 69 ex-day events.

Exhibit 4 contains the results. There is some evidence of a negative Monday effect for the stocks, because the stocks' mean return goes up to 0.15% versus 0.12% in Exhibit 3. There is no visible effect on the prime returns. The statistical significance of the matched differences falls slightly.


Nonsynchronous Quotes. If stocks and primes close and/or open at widely different times, the returns may be measured over different time intervals and may incorporate price movements other than just the overnight ex-day effect. Exhibit 5 provides data on the opening and closing times. Stocks open within a few minutes of 9:30 a.m., and close within a few minutes of 4:00 p.m. Interestingly, opening quotes on primes are available by around 9:40 a.m. But primes do tend to close as much one hour before 4:00 p.m. Given that the primes are generally inactive securities, this correspondence is quite good, and nonsynchronocity should not be too serious a problem.


Alternative Computations. The analysis presented thus far has been based on the closing quotes from the security's primary exchange and the opening quotes. The main advantages of this approach are: (i) the ex-day phenomenon is a close-to-open effect, and (ii) the use of quote-based returns circumvents the bid-ask bounce in transaction returns. Nonetheless, at least three other alternative computations are worth investigating.

First, I examined whether using closing composite bid-ask quotes, rather than the primary exchange quotes had any effect, and found none.(32) Second, I used the average of the last two quotes, and the average of the first two quotes to check for possible data errors. The results were virtually unaffected.

The final alternative tried was the use of the opening auction price (rather than the opening bid-ask average), which may, under certain circumstances, be a good estimate of the "true price" (Stoll and Whaley [17]).(33) Exhibit 6 provides the results for the clean sample, using closing quotes and the opening auction price. If there is no opening auction price, that observation is dropped; this happens for 14 observations for the prime.(34) The mean return for both stocks and primes is slightly lower than those in Exhibit 3. Otherwise the results are essentially the same.


IV. Trading Around Ex-Days

To study unusual trading behavior around ex-days, two portfolios were formed, one for the stocks and one for the primes. Exhibit 7 provides data on trading patterns for the stock and the prime portfolios, for days -3 to +3, where day 0 is the ex-day. The data presented are the percentage deviations for that day from the normal (time-series mean over 253 trading days) portfolio values of those variables.


There is a sharp increase in stock trading volume in days -1 and 0, followed by slightly less than normal activity after the ex-day. Some of that increase in shares traded seems to come from block trades. The nonblock transaction size also appears to be larger, since the number of transactions is indistinguishable from normal levels. These results are quite similar to those in Lakonishok and Vermaelen [15] and is consistent with the activity of cum-ex traders. The evidence for the primes in weakly suggestive of a mild increase in trading before the ex-day, followed by a mild decrease in trading in the post ex-day period. However, trading in the primes is inherently erratic and unstable, so that percentage deviations have to be viewed cautiously; indeed, none of the deviations is statistically different from the normal value. Overall, there is no abnormal ex-period volume in the primes.

V. Conclusions

Evidence on the prevalence and the impact of corporate cum-ex trading can help in deciding whether observed ex-day behavior can be properly attributed to tax effects. This paper studies the ex-dividend day behavior of twenty-six stocks and their derivative, dividend-oriented securities, the Americus Trust Primes. From the standpoint of cum-ex traders, the prime and the stock offer very similar benefits (viz., very similar dividend yield) but the trading costs for the primes are substantially higher, in the form of larger bid-ask spreads, thinner trading and limited supply.

The results, although not especially strong, suggest cum-ex activity in the stock but not in the prime. Specifically, the ex-day relative price drop is larger for stocks (ex-day returns for the stocks are smaller) and while there is a visible increase in stock volume, there is hardly any unusual trading in the primes. These results are consistent with the propositions that: (i) the ex-day behavior of these securities is related to their trading costs, and (ii) cum-ex trading in the stocks makes their ex-day price drops larger (or, returns smaller) than they would otherwise be.

(27) Whether this should be interpreted as a "tax effect" is debatable. Give [t.sub.d] = [t.sub.g], Equation (1) therefore predicts that individuals who have decided to sell (buy) would "profit" by selling at [P.sub.ex] (buying at [P.sub.cum]). On the other hand, investors do face a tax penalty on dividends because of the tax deferral aspect of capital gains, which Equation (1) does not reflect. Investors with long expected holding periods would prefer low-dividend payouts, ceteris paribus. If they are marginal, a price drop less than 1.0 may still be interpretable as a dividend tax penalty. Further research is clearly necessary. This paper's focus is on the differences in price drops (between primes and stocks), rather than the levels. (28) The difference in stock and prime returns may be due, in part, to differences in price levels and/or differences in yield levels. I tested for this, rather simply, by regressing the difference in returns on the differences in yield and price level. Neither the regression nor the individual coefficients were statistically significant. (29) Because trading costs and holding costs are not insubstantial, ex-day returns can be positive even if cum-ex traders are the marginal traders. (30) Two points should be noted. First, because these are large firms, they tend to have a fair number of entries in the Index; judgment must be used in determining which announcements are "major." The following were dropped: (i) American Home Products, 2/8/88 - firm was engaged in a bidding contest; (ii) Eastman Kodak, 2/24/88 - a lawsuit was filed against the firm; (iii) Johnson & Johnson, 2/17/88 - firm announced plans for a major acquisition; (iv) Johnson & Johnson, 11/16/88 - a lawsuit was decided in the firm's favor; (v) Mobil Corp., 2/2/88 - firm announces sale of Montgomery Ward. Second, although it is not necessarily the case that these announcements were made after the close or before the opening of trading, the price adjustment may take some time, and hence also affect opening and closing prices. (31) Eades, Hess and Kim [5] test for a day-of-the-week effect in ex-day returns and conclude that there is none. (32) All the regional exchanges close along with the NYSE at 4:00 p.m., with the exception of the Pacific Stock Exchange which closes at 4:30 p.m., eastern time. (33) All market orders that have accumulated overnight, limit orders to buy at or above the opening price, and limit orders to sell at or below the opening price are all executed at a single price - the opening price. The specialist may also participate to offset order imbalances. If the specialist does not intervene, the open would be more like a classic auction and the opening price closer to the "true" price. This interpretation does not hold if the specialist buys or sells for his own account to offset order imbalances. (34) Note that bid-ask quotes may be posted even if there is no batch open.


[ 1.] M. Blume, C. Mackinlay, and B. Terker, "Order Imbalances and

Stock Price Movements on October 19 and 20, 1987," Journal of

Finance (September 1989), pp. 827-848. [ 2.] M. Blume and R. Stambaugh, "Biases in Computed Returns,"

Journal of Financial Economics (November 1983), pp. 387-404. [ 3.] K. Brown and S. Lummer, "A Reexamination of the Covered

Call Option Strategy for Corporate Cash Management," Financial

Management (Summer 1986), pp. 13-17. [ 4.] T. Copeland and F. Weston, Financial Theory and Corporate

Policy, Addison-Wesley, Reading, MA, 1988. [ 5.] K. Eades, P. Hess, and H. Kim, "On Interpreting Security

Returns During the Ex-Dividend Period," Journal of Financial

Economics (March 1984), pp. 3-34. [ 6.] E. Elton and M. Gruber, "Marginal Stockholder Tax Rates and

the Clientele Effect," Review of Economics and Statistics,

(February 1970), pp. 68-74. [ 7.] E Elton. M. Gruber, and J. Rentzler, "The Ex-Dividend Day

Behavior of Stock Prices: A Reexamination of the Clientele

Effect: A Comment," Journal of Finance (June 1984), pp. 551-556. [ 8.] T. Grammatikos, "Dividend Stripping, Risk Exposure and the

Effect of the 1984 Tax Reform Act on the Ex-Dividend Day

Behavior," Journal of Business (April 1989), pp. 157-173. [ 9.] L. Harris, "A Transaction Data Study of Weekly and Intradaily

Patterns in Stock Returns," Journal of Financial Economics

(May 1986), pp. 99-118. [10.] Institute for the Study of Securities Markets, c/o Professor

Robert Wood, Memphis State University. [11.] J. Ingersoll, "A Theoretical and Empirical Investigation of the

Dual Purpose Funds," Journal of Financial Economics (March

1976), pp. 83-124. [12.] R. Jarrow and M. O'Hara, "primes and Scores: An Essay on

Market Imperfections," Journal of Finance (December 1989),

pp. 1263-1288. [13.] A. Kalay, "A Reexamination of the Clientele Effect," Journal of

Finance (September 1982), pp. 1059-1086. [14.] J. Karpoff and R. Walking, "Dividend Capture in NASDAQ

Stocks," Working Paper, University of Washington, 1990. [15.] J. Lakonishok and T. Vermaelen, "Tax-Induced Trading Around

Ex-Dividend Days," Journal of Financial Economics (July 1986),

pp. 287-320. [16.] J. B. Long, Jr., "Efficient Portfolio Choice With Differential

Taxation of Dividends and Capital Gains," Journal of Financial

Economics (August 1977), pp. 25-53. [17.] H. Stoll and R. Whaley, "Stock Market Structure and Volatility,"

Review of Financial Studies (1990), pp. 37-71. [18.] P. Venkatesh, "Empirical Evidence on the Impact of the Bid-Ask

Spread on the Characteristics of CRSP Daily Returns," Working

Paper, University of Houston, 1991. [19.] T. Zivney and M. Alderson, "Hedged Dividend Capture With

Stock Index Options," Financial Management (Summer 1986),

pp. 5-12.
COPYRIGHT 1991 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Venkatesh, P.C.
Publication:Financial Management
Date:Sep 22, 1991
Previous Article:Ownership structure, leverage, and firm value: the case of leveraged recapitalizations.
Next Article:A survey of corporate leasing analysis.

Related Articles
A market microstructure explanation of ex-day abnormal returns.
The effects of splitting on the ex: a microstructure reconciliation.
Prime and score premia: evidence against the tax-clientele hypothesis.
Nonstandard-settlement transactions.
Singapore premier urges APEC members to attend summit.
Who trades around the ex-dividend day? Evidence from NYSE audit file data.
Limit order adjustment mechanisms and ex-dividend day stock price behavior.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters