-27

In the model, a broker always attempts to borrow stock to cover short sales and segregation requirements whenever his available resources are less than needed. In addition, there is the alternative

of borrowing to cover fails-to-deliver to customers or to brokers. Table 3 shows the results of enforcing alternative stock loan policies of covering fails-to-deliver to institutions or to brokers. In both alternatives, stock loans are mandatory when stock is available. Further, a completely efficient stock lending and borrowing system has been assumed, i.e., if stock is available and needed, it will be found and borrowed. The only restriction is that firms are allowed to lend

a maximum of 85 percent of the stock held in their free box on any given day. In time, this may result in the lending of over 85 percent of the free box.

Borrowing to Prevent Broker-to-Institution Fails

Table 3 shows that borrowing stock to prevent failing to deliver to institutional customers results in an annual cost reduction of $8 million, accounted for by reductions of $3 million in broker-to-broker fails, $2 million in broker-to-public cash customer fails, $9 million in broker-to-institution fails, and an offsetting $6 million increase in the cost of making stock loans. The objective of this policy, to effect a significant reduction in the cost of broker-to-institution

E

-28

fails, was not realized. An examination of some of the back-up information showed that, with institutions receiving first priority, the majority of the remaining broker-to-institution fails were caused by DKs. The availability of additional stock with which to attempt : deliveries to institutions had little impact. If the DK problem were solved first, significantly different results might be obtained.

Borrowing to Prevent Broker-to-Broker Fails

Table 3 also shows what happens when stock is borrowed to prevent broker-to-broker fails. As before, the hypothesis was that improving performance between brokers will, in turn, improve overall system performance. There may be sufficient stock in the aggregate system inventory to make all deliveries, but it may be in the wrong hands. The rationale was that, in such a case, an efficient stock lending system might overcome the difficulty. The results show that there is validity in this line of reasoning. Although there is an $11 million increase in the clerical cost of stock loans made to prevent broker-to-broker fails, the net cost drops by $16 million per year. The significant savings are a reduction of more than one-third in the cost of brokerto-broker fails and of one-fourth in broker-to-public cash customer fails.

With the previously-stated caveat regarding the imputed interest impact of stock loans on individual firms, the conclusion is that an efficient and enforced stock borrowing policy can improve the overall performance of the trade completion system.

A FAST TRANSFER ALTERNATIVE

Poor performance of the trade completion system is in part a problem of a limited supply or inventory of stock to be processed. There are certain operations that must be performed on stock certificates to make them transferable, and these take time. During the transfer process itself, certificates are removed from circulation and the available inventory is effectively reduced. It follows that any reduction in the time required to perform the transfer operations

-29

will have the effect of increasing the size of the available inventory and should, as a consequence, raise the level of satisfactory performance in trade completion. This possibility was investigated by reducing average transfer time from 8 days to 2 days. The industry incurred a cost of $10 million per year to achieve this reduction; but as is shown in Table 4, the result was a net saving for incomplete transactions of $24 million or approximately 20 percent. The costs associated with fails of all kinds were reduced: broker-to-broker fails by 26 percent, broker-to-institution fails by 11 percent (still held up by DKs and wrong denominations), and broker-to-public cash customer fails by 25 percent. Reducing transfer time shows encouraging results and should be high on the list of items considered for implementation.

Brokers frequently find themselves with only enough stock to make a partial delivery. NYSE, AMEX, and NASD rules take cognizance of this by providing that partial deliveries will be accepted if the delivery is offered in a multiple of the unit of trading. Although this is the rule, most brokers are reluctant to accept partial deliveries because they complicate bookkeeping. Consequently, partial deliveries from

-30

broker to broker are seldom made. This also is true of partial deliveries to banks, although it seems logical to assume that such deliveries would be of particular advantage to banks because of the large block trading in which they engage as agents for institutional investors. Banks, generally, are willing to accept deliveries of stock corresponding to the way in which an order was executed. For example, if

It

an order placed for 1000 shares resulted in trades of 500, 200, and 300 shares, delivery of any one of these units would be accepted. seems clear that full implementation of the partial delivery rule would reduce the level of fails and consequently the cost of these fails. The question is whether the reduction would be sufficient to offset the cost of the extra bookkeeping and the additional deliveries. Table 4 shows the results of the simulation undertaken to investigate these possibilities.

In the benchmark case, partial deliveries are not allowed, but they are allowed in the test case and in the form specified under the rules of the NASD, AMEX, and the NYSE. The results are very promising. The net cost of all fails could be reduced by $46 million or 37 percent per year. The greatest single impact is the $25 million reduction in the cost of broker-to-broker fails. The expectation was that the cost of broker-to-institution fails would go down considerably, but as before, the DKs and wrong denominations kept it up. The relatively slight reduction in the cost of broker-to-public cash customer fails is solely attributable to the enhancement of the broker's overall stock availability position. The additional delivery and bookkeeping cost associated with partial deliveries is estimated to cost $6 million per year, but the benefits were nine times the increased costs. It is evident, therefore, that taking full advantage of the partial delivery rules offers another promising way to improve the performance of the trade completion system.

REDUCING DKS, UN COMPARES, AND WRONG DENOMINATIONS

The simulation model is next used to examine the effect of reducing Ks, uncompares, and wrong denominations. Although the DK has its impact on fails-to-deliver to institutional customers and the wrong

-31

denomination is a problem common to deliveries of all kinds, the two have an impact on system performance different from that of the uncompared trade. Each of the first two results in stock being removed from the system until the difficulty can be resolved. When a DK occurs, the stock used in the attempted delivery is held in limbo until either the receiving bank obtains the necessary instructions or until the irregularities preventing the delivery are ironed out. When a wrong denomination occurs, stock must be sent to transfer for change to be made. During the time it is at transfer, it also is effectively removed from the available inventory.

An uncompared trade has a somewhat different impact. Until the two parties to the transaction agree to its terms, no delivery can be made and none is attempted. Obviously, there will be a failure to satisfy either side of the transaction until resolution can be accomplished, but no stock will be removed from inventory as in the other

cases.

Based on empirical evidence, the following rates have been specified for the existing systems: DKs occur in 3 out of 10 deliveries to banks, uncompared trades occur in 6 out of 100 cases, and wrong denominations are a problem in 5 out of 100 deliveries. DKs are relevant only to deliveries to banks; uncompares affect only broker-to-broker deliveries; and wrong denominations can be a problem with any type of delivery.

Table 4 shows what happens when the DK rate is lowered to 10 percent, the uncompare rate is lowered to 1 percent, and the wrong denomination rate is reduced to 2 percent. The estimated net result of these combined changes is a reduction in cost of $59 million per year, from $125 million to $66 million. The reduction in broker-to-institutional fails was primarily due to the reduction in the DK rate, with a small effect from the change in the wrong denomination rate. The broker-to-broker fails reduction was due to the combined effect of the reductions in the uncompared trade and wrong denomination rates, while the reduction in broker-to-public cash customer fails was almost entirely due to reducing the wrong denomination rate.

Although each of these factors has an impact on system performance,