THE HYPOTHESIS ABOUT BROKER/DEALER PERFORMANCE IN CLERICAL WORK

We were told that clerical errors are made too frequently and that these increase the probability of a fail. We were told, and observed first hand, that people in the operations offices of broker/dealers are working many overtime hours, and that this situation has continued for weeks and months. Many reasons are offered; good clerical workers are hard to find, and trading volume jumps up and down from day to day and continues to increase at rates that no one has forecasted. Broker/dealers want to keep their costs for operations as low as possible and therefore delay in increasing the size of their clerical staff. Mechanization of operations has been slower than required. The complex, largely clerical record systems are error prone. We were told that if the broker/dealer on the sell side of the trade were known that it would be possible with some assurance of being correct to forecast whether or not the delivery would be made simply from prior acquaintance with the accuracy and up-to-dateness of the broker/dealer's system of records.

THE HYPOTHESIS ABOUT EFFECTS OF BROKER/DEALER FISCAL

MANAGEMENT

We were told that in these days of high interest rates it is increasingly common to find a broker/dealer or a bank acting for a broker/dealer postponing payment by finding small technical inadequacies in things like signatures on certificates offered for delivery, or even "not knowing" the trade. We were told that fails of this kind may be accepted a day or two later even when the delivery is submitted exactly as it was submitted the first time. If the high price of money. and close fiscal management by a broker/dealer is contributing to fails, it is the fiscal condition of the receiving broker/dealer which should contribute to the probability of a fail.

THE DYNAMIC INSTABILITY HYPOTHESIS

Those acquainted with securities trading indicate that turnover in stock ownership is higher now than at any time before. This tendency, coupled with the fact that stock clearing is a significantly more time consuming process than is stock trading, may lead inevitably to a system which performs less and less well. Observers note that they cannot make deliveries on securities that they are owed by others. The characteristics of the present stock trading and clearing systems are such that as trading volume increases the fails problem may increase at an even greater rate. Stock clearing houses have offered fails clearing services for the first time in 1968 in an effort to reduce the fails caused by one fail leading to another and then in turn to yet others.

THE HYPOTHESIS ABOUT MARKET VOLATILITY EFFECTS

When a security is experiencing large price fluctuations, the delay in time between trade date and settlement date-especially when settlement is accomplished less and less often on the fifth business day and more and more often on still later days-means that price movements become increasingly important to buyers and sellers as well as buying brokers and selling brokers. This can increase the motivation to take advantage of any opportunity to DK (don't know) the trade. It could be argued that if stock "delivery" and title transfer could be accomplished within a very few minutes after the execution of the trade, fails which occur because of motivations prompted by market volatility would disappear and therefore some fails would be eliminated. It is our impression that everyone in the industry would prefer a faster stock clearing system to a slower one since the faster one would reduce the possibility of less than responsible action following changes in value between the time of the trade and the time of the settlement.

"Which are the most important causes of fails?" we asked. If answers were available, they could focus the effort in intermediate and long-range improvements in the stock clearing system on those features which would most reduce fails. There was little confidence shown by those answering our question that they really know which causes are most important. Since many of the hypotheses about causes seemed reasonable to us, we were willing to guess that there are multiple causes for fails. Diagnostic information about which causes are most important could indeed be useful in guiding modifications and improvements in the stock clearing system. For that reason we proposed that a significant part of our study effort be devoted to a diagnosis of the relative importance of the causes of fails.

IV. AN APPROACH TO ASSESSING THE CAUSES

With expert industry opinion pointing toward multiple causes of fails, yet unable to indicate the relative importance of the several causes, we urged that some time taken in a diagnosis of the major causes of fails in today's stock clearing system could lead to insights which are useful in guiding the allocation of effort to remedy the problem. We undertook this portion of our work after receiving support for that suggestion.

In order to do the diagnosis, we visited brokerage houses around the country and observed actual trades and balance orders from clearing houses, noting trade date and the time when they settled. In addition, we observed many conditions associated with the trade or balance order, such as the security being traded, the identity of the buying broker/dealer, the identity of the selling broker/dealer, the identity of the broker/dealer making the delivery of stock certificates if he was different from the selling broker/dealer (we called him the "settler," the broker on the other end of the settlement from a buyer), what exchange/market was used in making the trade, whether a clearing house was used in settling, the geographical locations of the offices of the buying, selling, and settling broker/ dealers, and still other conditions associated with the trade. Then we assembled from many sources information about the broker/dealers involved in our sample of trades and information about the securities traded in our sample of trades. We put this information together, then studied how the many conditions were related to each other. In particular, we studied how all these conditions were related to whether the settlement "failed" or not, and how long it was in a fail state.

This chapter describes where and how we observed trades, where we found the information we needed, and how our diagnostic effort serves-to_test_the hypotheses about causes of fails which had been developed from our talks with knowledgeable people in the securities industry and the stock clearing system in the United States.

SAMPLING ACTUAL TRADES

One place to discover how the stock clearing system is performing is at the level of the trade itself. With the help of broker/dealers' records, both the trade date and the date on which the trade was cleared can be observed. When the trade is cleared through a clearing house, the balance order-or one of the balance orders-through which that trade was cleared can be identified and the date on which the balance order was cleared can be established. Observations of this kind can be made in today's stock clearing system in most parts of the country. Observations of this kind cannot be made when the clearing house for the Pacific Coast Stock Exchange is involved in stock clearing, however, because of its practice of debiting and crediting a member's account in a particular security. The

identity of both particular trades and a particular day's net balance (balance order) is lost in this method of clearing. While some exceptions of this kind make inappropriate our method of observing the performance of the stock clearing system, the method selected was appropriate for a very large proportion of today's stock clearing work.

We observed trades in ten broker/dealers in New York City and in two houses each in Atlanta, Houston, San Francisco, Chicago, and Boston. This sampling of houses gave us a view of trading and stock clearing throughout the United States. In each city we visited large houses and small houses to make sure our observations included any patterns which happened to be related to the size of the broker/dealer we visited. All of the houses we visited were members of the National Association of Securities Dealers, Inc., but because of the overlapping memberships in the three exchanges/associations, many of them were also members of the New York Stock Exchange and the American Stock Exchange.

In each house we observed buys which had been made three to eight weeks before the date of our visit. By observing buys, we were able to see how a very large number of broker/dealers are meeting their stock deliveries, a much larger number than could be visited in the time and budget we planned for our work. Buys which were made three to eight weeks earlier had had at least two weeks beyond settlement day in which to clear. Thus we could observe the rate at which buys were settled on settlement day and on succeeding days through a 17 calendar day period following settlement day. Records of buys made during the two months prior to our visit were still a part of the active records in the operations offices, and did not need to be retrieved from record storage.

We sampled from a broker/dealer's buys on several exchanges/markets for a particular day. We sampled from records of compared trades, trades which both the buying broker and the selling broker recognized as having been a trade between them on that date. We were able to complete our observations on approximately 100 buys in a day's visit with the broker/dealer, and observed 60 buys in the over-the-counter market, 20 buys on the American Stock Exchange, and 20 buys on the New York Stock Exchange if the broker/dealer traded in all three markets. When he traded exclusively in over-the-counter securities, we observed only those buys. If the number of trades we wished to observe, such as 60 over-the-counter buys, equalled or exceeded the broker/dealer's business on one day, we took all of his buys on one day and continued making observations of buys from an earlier or a later day's business. When our sample was less than a day's business, as was usually the case, we observed every third buy, or every tenth buy, or one buy from every fifth page of a long list of buys from that day, or some other pattern which gave us essentially a random selection of buys from his entire list of buys for a trading day. Sampling of this kind was done from each

of the lists of over-the-counter buys, New York buys, American buys, 3rd market buys, and sometimes buys on regional exchanges-the latter observations being made only in cities outside New York when the broker/dealer made such trades.

At the conclusion of our observations, we had data on 2114 buys involving a total of 487 different broker/dealers as buying, selling, or settling broker and 1030 different securities. Of these buys, 1490 buys had been made in the over-the-counter market, 286 on the New York Stock Exchange, 233 on the American Stock Exchange, 72 in the third market, and 33 on regional exchanges.

OBSERVING EVENTS ASSOCIATED WITH THE TRADE

Having sampled from his lists of compared buys for a trading day, we next went to the broker/dealer's record of balance orders from a clearing house and located the balance order for the security and trade date on which the buy had been made. We noted the settling broker for that balance order, as well as the other information on the balance order. When two or more balance orders had been received from the clearing house for that security and that date, as occurs less than half the time, the one balance order which agreed most closely with the trade in number of shares and total dollars was selected.

Finally, we went to the broker/dealer's records of fails-to-receive and recorded when the trade (if cleared direct) or balance order actually cleared on a scale of calendar days including and subsequent to settlement day. The time scale is pictured in Figure 6. Observations were made in the ten categories shown as "coded number of calendar days in fail state" in Figure 6. Fails-to-receive records were traced through the 16th calendar day following trade date, and trades or balance orders not cleared by that date were assumed to clear in the time period defined as the 17th calendar day following settlement day, or later. This completed the observations about each trade. A list of the data observed about each trade is shown in Figure 7.

Having observed the trade date, it was possible for us to determine what the trading volume was on the New York and American exchanges on the days immediately prior to the trade, whether or not settlement day was followed by a trading day, how many days had elapsed since the last fails clearing date for the security being traded in our buy, and still other kinds of information related to the time of our trade. Having observed the identity of the buying, selling and settling brokers (the settling broker was the selling broker when clearing was direct), we could associate the characteristics of each of the several brokers with the other characteristics of the trade and settlement, including whether or not the settlement failed on settlement day. Having observed the identity of the security, we could associate its number of outstanding shares, whether or not it had had a stock split or a dividend date in recent history, and yet other information about the security and its transfer agent with the other characteristics of the buy we had observed.