sales on the Exchange exceeded 1.5 billion shares, and members accounted for 23.6 to 25.7 percent of these shares.

Specialists have always accounted for more trading than any other class of members. Over the period 1937 to 1954 they accounted for between 7.8 and 11 percent of all round-lot purchases and sales on thie NYSE each year. This participation rate increased to within a range of 12.3 to 14.9 percent for each of the years 1955 to 1961.

Odd-lot dealers have consistently accounted for approximately 3 percent of all round-lot purchases and sales each year since 1937, ranging above 3.6 percent only once (4.0 percent in 1947) and never falling below 2.5 percent.

Floor traders accounted for 6.8 percent of all round-lot purchases and sales in 1937, but this figure has dropped over the years and since 1945 has fluctuated in the 2 to 3 percent range.

All other member purchases and sales, accounted for by members' off-floor trading, have ranged between 2.9 percent and 6.1 percent of total round-lot purchases and sales over the years 1937 to 1961. Prior to 1955 this trading exceeded 4.5 percent only once (6.1 percent in 1949), but it has since ranged from 4.8 percent to 5.3 percent.

4. CONCENTRATION OF MEMBER TRADING

a. General measures of member concentration

The above percentage figures, although they provide a sound general measure of member participation in the market, fail to reflect variations in member participation from day to day or stock to stock. Because the impact of any given group's trading is more discernible with respect to a given stock than to the market as a whole, and is more manifest over a shorter period of time (such as a day) than_a longer period of time (such as a year), much of the Special Study data is designed to reflect trading in each stock each day over the periods studied. That is, Exchange activity has been analyzed on a "stock day" basis.63

Data covering the 3 weeks studied indicate that all members tend to concentrate their activity in certain stock days, with some classes of members concentrating more prominently than others. Over the 3 weeks there were 16,174 stock days on the NYSE. One or more members participated in 14,970 or 92.6 percent of this total. The

63 The "stock day" concept is analogous to the man-hour or man-day concepts utilized in economic statistics. Thus, just as 10 man-days of labor may represent the work of 10 men on 1 day, 2 men on each 5 days, etc., 10 "stock days" may represent 10 stocks traded on 1 day, 2 stocks traded on each of 5 days, etc. That is, any stock which trades on more than 1 day is counted as one for each day it trades; e.g., General Motors trading on 3 days during a period under study would be counted as 3 stock days. If 900 stocks are traded on the Exchange on Monday, 1,100 on Tuesday, 1,050 on Wednesday, 950 on Thursday, and 1,100 on Friday, the total number of "stock days" for the week would be 5,100.

The primary merit of the "stock day" approach is that it allows a study, by days, of all trading over the period studied, but at the same time breaks down this trading according to the daily performance of each stock. Although stocks cannot be identified by name, the characteristics of each stock for each day it trades (its price, price range, and volume. etc.) are preserved. Thus it is possible to examine the trading of any particular group (public, members, or any class of members) with respect to stocks classified according to those characteristics; for example, public trading in low-priced stocks, member trading in volatile stocks, floor trader trading in active stocks, etc., as each stock displayed such characteristics in each day it traded-in other words, on each "stock day."

A complete explanation of the process of analyzing these data is set forth in app. VI-A.

following table sets forth data indicating the degree to which each member class tended to concentrate its trading over the 3 weeks:

TABLE VI-8.-Concentration of member trading

1 Averages noted in the text are slightly higher for each class of members than in fact obtained, due to the fact that stock days include only common stock days while total shares traded include preferred stocks. Single specialists traded in 12,825 stock days, while competing specialists traded in 644. The computation of total shares traded and average number of shares traded per stock day includes the data for both single and competing specialists. Other data on specialists presented in this part, however, are based on single specialists only.

As this table shows, floor traders tend to concentrate their trading to a greater degree than other members. For example, although floor traders traded about 25 percent fewer shares over the 3 weeks than odd-lot dealers, the odd-lot dealers' average participation per stock day was only 401 shares while the floor traders' average was 1,359. This average floor trader participation, it should be noted, was only 87 shares less than the specialists' average, despite the fact that specialists traded more than 6 times as many shares as floor traders over the 3 weeks.

If judged solely on the basis of the average number of shares traded per stock day, specialists and floor traders would appear to have approximately equal importance in the stock days in which they trade. Such is not the case, however. Since floor traders tend to concentrate their trading in very active stock days, they tend to account for smaller percentages of total stock day volume than specialists (table VI-2). This fact is reflected in the following table, which indicates the approximate percent of total stock day purchases and sales accounted for by the median stock day (ranked by percent of total stock day purchases and sales) of each member class for the 3 weeks: 64

4 See table VI-2 for more complete presentation of data.

TABLE VI-b.-Trading by members of various classes relative to total trading per stock day

1 The amount of trading on an exchange may be measured by computing the total number of shares sold, which is the conventional measure ("total volume"), or by computing the total number of shares purchased plus the total number of shares sold ("total purchases and sales"). The latter measure will always be exactly twice as large as the former, since each share sold must have a matching purchase.

Because the trading of any group (e.g., all members, specialists, etc.) will not ordinarily consist of an equal uumber of purchases and sales, the trading of a given group must be measured by adding its total purchases and its total sales together. This total may then be expressed as a percent of either total exchange volume (sales) or total exchange purchases and sales, depending on the purpose for which it is to be used. If the purpose is to establish the group's participation in total purchases and sales the participation is best expressed as a percentage of total purchases plus sales. If, however, the purpose is to determine the percent of "total volume" in which a group member participated as buyer or seller, the group's purchases and sales should be expressed as a percent of "total volume.' The resulting percentage may overstate somewhat the percentage of volume in which the group participated if members of the group were on both sides of any of the transactions included in the data, but such instances are generally limited and therefore do not destroy the usefulness of this measure.

In any event the market participation of a particular group expressed as a percent of total purchases and sales will always be one-half as large as when expressed as a percent of total volume since total purchases and sales are always equivalent to "twice total volume" (a phrase used interchangeably with "total purchases and sales").

Specialists, therefore, accounted for approximately 35 percent or more of total stock day volume in half of the stock days in which they traded, while floor traders accounted for only about 6.75 percent or more of total stock day volume in half of the stock days in which they traded.

b. Concentration of member trading by stock days

The fact that specialists and odd-lot dealers trade in a greater number of stock days (i.e., have a broader dispersion of their trading) than other members is attributable in large part to their market functions, which foster broader participation in the market. An effort was made to determine whether other factors-the price, price range, or trading volume of stock on a given day exhibit any relationship to the trading patterns of these members, and whether such factors play a determinative role in the trading of floor traders or members off floor. The first step in this effort was to plot the distribution of stock days in which each member class participated against total stock days over the 3 weeks. Thus, a series of line charts were prepared,65 the bases of which represented every stock day over the 3 weeks, arranged from the stock day of lowest price (or price range, or total volume) to the stock day of highest price (or price range, or total volume). The vertical axis of each chart was marked off from zero to 100 percent to indicate the cumulative percent of each member class' stock days falling at or below each price (or price range, or total volume) level. As the bases of the charts are also divided into percent rankings (lowest 10 percent, next 10 percent, etc. of stock days as ranked by price, price range, or total volume, as appropriate), it is

Charts VI-1 through VI-4.

66

6 The actual percent rankings are set forth across the top of each chart, but the 10 percent, 20 percent, 30 percent, etc., lines are carried to the bases of the charts to facilitate reading of the data.

96-746-63-pt. 2--5

possible to read from each chart the percent of any given member class' stock days which fall into the lowest 10 percent (or 20 percent, 30 percent, etc.) of all stock days over the 3 weeks, as ranked by price, price range, or total volume.

As measured by the number of stock days in which each class of members participated, it does not appear that any class of members has a notable preference for either high- or low-priced stocks (chart VI-1). Floor traders tend to favor very slightly the lower priced stocks, as 22 percent of the stock days in which they participated fell within the lowest 20 percent of all stock days, as ranked by stock price. The other member classes show a slight preference for the higher priced stocks. Thus approximately 46.5 percent of the stock days in which specialists participated, 44.5 percent in which members off floor participated, and 44 percent in which odd-lot dealers participated fell in the lowest 50 percent of all stock days over the 3 weeks as ranked by stock price (i.e., fell below the median stock day as ranked by stock price).

Members show a more pronounced tendency to participate in those stock days with wider daily price ranges (chart VI-2). With respect to both specialists and odd-lot dealers, only 44 percent of the stock days in which they participated fell below the median of all stock days ranked by price range. For members off floor the percent dropped to 39, and for floor traders it plummeted to 22. That is, as measured by the number of stock days in which each member class participated, 78 percent of all floor trading, 61 percent of all off-floor trading, and 56 percent of all specialist and odd-lot dealer trading occurred in those stock days with daily price ranges wider than the median price range over the 3 weeks. The most notable concentration in stock days of wide price range occurred with respect to floor traders. For example, 80 percent of all stock days over the 3 weeks had a price range of 2.8 percent or less, yet only 56 percent of all stock days in which floor traders participated had a price range of 2.8 percent or less. Stated conversely, 44 percent of all stock days in which floor traders participated were among the top 20 percent of all stock days as ranked by price range.

67

Even more pronounced, when measured by the number of stock days in which members participated, is the tendency for members to trade in stocks experiencing high volume on any given day (chart VI-3). As in the previous instances, specialists reveal this propensity to a lesser extent than other members; approximately 44 percent of the stock days in which they participated fell below the median stock day as ranked by share volume. The percent of stock days of other members falling below the median were: odd-lot dealers 36 percent, members off floor 23 percent, and floor traders 10 percent. If market activity is measured by number of transactions per stock day rather than by share volume per stock day, the pattern of member concentration remains virtually the same (chart VI-4).

c. Concentration of member trading by share volume

Concentration of member trading may be studied not only by analyzing the types of stocks or stock days in which they have any trading

Price range is expressed as a percent of the closing price on each stock day.

at all but also by analyzing their volume participation in various types of stocks or stock days. For example, while as shown in the last section, members seem to have no strong preference between high- and low-price stock days as measured by the relative number of stock days in which they trade at all, when the share volume of their trading is analyzed there is a decided tendency to higher participation rates in high price stock days.

68

The data used in this analysis (member percentage participation rate data) and the presentation of these data differ fundamentally from the stock day data above in that they reflect aggregate member trading rather than the trading of each class of members separately. As a result, trading patterns of the membership as a whole that appear in the data may not hold true with respect to each member class individually. Since specialists accounted for a preponderance of all shares traded by members over the 3 weeks, the specialist data tend to dominate the aggregate data and may obscure trading patterns of other member classes.

69

The member trading patterns that appear in these data do not in all cases coincide with the patterns indicated in the stock day data noted above. Thus, although members tended to trade in an equal number of stock days of high-priced and low-priced stocks (chart VI-1), as noted above they showed a definite tendency to increase their percentage participation rate in higher priced stocks (app. VI-A, chart 1 and table 1). That is, within the stock days in which they traded, members accounted for a greater percent of total volume in the high-priced stocks than in the low-priced stocks.

With respect to the price range of stock days on the other hand, both the stock day data and the participation rate data reflect a member tendency to trade more heavily in stock days of wide price range. That is, members tend not only to participate more frequently in days of relatively wide price range (chart VI-2), but as well to increase the volume rate of their participation as the price range widens (app. VI-A, chart 2 and table 2).

No clear relationship appears in the participation rate data between total stock day volume and member-participation rates as a percent of that volume (app. VI-A, chart 3 and table 3). Thus, although most member classes show a strong tendency to trade in the more active stock days (chart VI-3), the percentage rate of their participation within such stock days does not increase as total volume increases.

5. MEMBER PURCHASE AND SALE BALANCES

A most significant question with respect to member trading is whether it tends to stabilize or destabilize prices. Although different measures of stabilization have been employed, as noted in subsequent parts of this chapter, the measure employed here is the degree to which members are buyers or sellers on balance in stock days of rising and declining prices. That is, a member trading pattern which tends to produce purchase balances on declining stock days and sale balances on rising stock days would indicate that members exert a stabilizing

These types of data are presented separately for specialists and members off floor in pts. D and G of this chapter.

Another respect in which these data differ is that only the stock days in which members traded are included in the analysis in this section.