View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Treasury-Federal Reserve Study of
the U. S. Government Securities Market




GOVERNMENT SECURITIES MARKET PERFORMANCE IN THE WAKE
OF OFFICIAL OPERATIONS IN COUPON ISSUES
DAY-TO-DAY PERFORMANCE

Staff study prepared by
Louise Ereemaa- Ahearn
Economist
Federal Reserve Bank of New York
April 16, 1965




THE
FEDERAL
RESERVE
BANK of
ST LOUIS

Research Library

Table of Contents
Page
A.

Introduction and Summary

1

B.

Statistical Indicators of Daily Market Performance

3

1.
2.

b

3-

C.

Review of Criteria of Market Performance
The Dealers' Criticisms of Official Operations in Coupon
Securities
a. Achievement of the Objectives of "Operation Nudge"
b. Market Performance
Statistical Indicators Selected
a. Description of the Indicators
b. Use of the Indicators

Relationship Between Dealer Sales to Official Accounts and Jfexket
Indicators on Days When There Were No Financings, 8/22/62-12/31/63
1.
2.

5
6
7
8
9
10

11

Dealer Sales of Coupon Issues to Official Accounts
Sales to Official Accounts and Sales to Private Customers
a. Chi-square Tests
b. Multiple Regressions
3. Sales to Official Accounts and Purchases from Private Customers
a. Chi-square Tests
b. Multiple Regressions
Sales to Official Accounts and Changes in Prices
a. Chi-square Tests
b. Multiple Regressions
5. Sales to Official Accounts and Changes in Dealer Positions
a. Chi-square Tests
b. Multiple Regressions
6. Sales to Official Accounts and Offerings to the Trading Desk
a. Chi-square Tests
b. Multiple Regressions
7. Conclusions
D.

lb
19
19
2b
25
25
27
29
29
30
32
32
32
3^
3^
3^
37

Relationship Between Dealer Sales to Official Accounts and Retail
Trading During Treasury Rights Financings

40

1.
2.
3.
4.
5.

6.
E.

Dealer Sales of Coupon Issues to Official Accounts
Relative Magnitudes of Sales to Official Accounts and Retail
Trading
Hypotheses
Simple Correlation
Multiple Regressions
a. Variables Included
b. Results for Long Financing Periods
c. Results for Short Financing Periods
Conclusions

Appendix




b2

46
48
51
51
53
57
60
63

A.

Introduction and Summary
The objective of this study has been to analyze the impact

of official operations in coupon issues on the daily performance of the
Government securities market. The first section belov selects statistical
indicators of daily market performance, mainly on the basis of a review of
criteria of market performance proposed in the past and the dealers1 criticisms
of official operations in coupon securities. The second section studies the
relationship between these market indicators and dealer sales to official
accounts on days when there were no Treasury financings, during the period
from August 22, 1962 through December 31, 1963* As a supplement, the final
section considers the relationship between dealer sales to official accounts
and retail sales and purchases (two of the more important market indicators)
during Treasury rights financings from March 1961 through July 1964.
The major findings of the study are summarized below:
(1) Dealer sales to official accounts were not significantly
related to dealer sales to private customers on days when Treasury
financings were not in progress.

During financings, however, purchases

by official accounts apparently encoura.ged a higher level of sales to
retail customers of securities maturing in 5«10 years and after 10 years.
Thus, there was little or no evidence in the daily data to confirm the
dealers' contention that private buyers were discouraged by official
operations.
(2) Dealer sales to official accounts were positively correlated
with dealer purchases from private customers of securities in the 5-10 and
over-10 year maturity classes, both during financings and on other days.




2
These results were consistent with the dealers1 argument that buying by
official accounts led other customers to take the opportunity of dumping
securities.

They could also be explained, however, by the practice of

official accounts of buying securities when they were available—
availability presumably being increased by large dealer purchases from
private customers.
(3) Dealer sales to Treasury accounts on the current day were
negatively related to the average change in prices of coupon securities
on the same day in all three maturity classes studied ( > 1 ^ 5 ^ ^>5^.10,
> 1 0 years) on days without financings.

Doubtless this reflected the

Treasury's practice of concentrating investment orders in weak markets.
This relationship appeared to have been temporary and slight.

The possi-

bility that purchases by official accounts led to price increases should
be tested for periods longer than two days and with lags of more than one
.day.
(4) Dealer sales to official accounts were negatively associated
with the change in the dealers1 gross long positions and net positions in
all three maturity classes on days without financings.

This relationship,

however, was probably mainly an immediate technical reaction and so would
not necessarily reflect a lessened dealer willingness to take positions
over a slightly longer period.
(5) Dealer sales to official accounts on the current day were
also associated with a rise in offerings of securities in all three maturity
classes to the Trading Desk at the New York Federal Reserve Bank, on days
when there were no Treasury financings.
These findings are subject to a number of qualifications.

First,

the analysis was confined to the market's daily response to official operations
in the period after the Federal Reserve System began to buy coupon securities.



3
In most cases a complete appraisal of the arguments about official operations in coupon securities also requires analyses of weekly or monthly
data and comparison with market behavior in the period before 1961 when
there was no possibility of System intervention.

Second, the frequency

of days with large official operations was not great.

The findings might

have been different if operations had been larger or more frequent.

In

addition, the statistical reliability of the results, particularly during
financings, would have been greater if operations had been more frequent.
Third, even in the cases where official operations were significantly
related to market indicators, such operations usually did not explain
much of the variation in the indicator.

Finally, it would seem preferable

to approach the problem of the impact of official operations in coupon
securities on market performance indirectly, by the construction of a
statistical model of the Government securities market that would explain
positions, trading, prices and official operations at the same time.

3.

Statistical Indicators of Dally Market Performance
The first step in this study was to select statistical indicators

of the daily performance of the U. S. Government securities market. The
indicators selected should facilitate an appraisal of the dealers1 statements about the impact of official operations in coupon securities on
market performance, and they should also measure the technical criteria
used to judge market performance.

Therefore, the first two sections below

briefly review the criteria of market performance proposed in the past and
summarize the dealers' criticisms of official operations in coupon securities.
The third section describes the statistical indicators selected and the
hypotheses that can be tested through a study of these indicators.




b

1.

Review of Criteria of Market Performance
In general terms it is usually agreed that an adequately

functioning Government securities market would have the capacity to
accommodate Treasury financings, Federal Reserve open market operations,
and private investment transactions.

Such a market would be character-

ized by continuity in trading at prices which reflect demand and supply
and would not exhibit the sustained sharp price movements that might
reflect investor or dealer unwillingness to maintain an active, functioning market.

More technical criteria implicit in this definition have been

developed along two lines, one defining ideal markets and the other defining
disorderly markets.
The Ad Hoc Subcommittee on the Government Securities Market in
1952 characterized an efficiently functioning market as one possessing
"depth, breadth, and resiliency.11

These characteristics were defined in

terms of the orders on the dealers' books.

The market "possesses depth

when there are orders, either actual orders or orders that can be readily
uncovered, both above and below the market.

The market has breadth when

these orders are in volume and come from widely divergent investor groups.
It is resilient when new orders pour promptly into the market to take
advantage of sharp and unexpected fluctuations in prices."^

Adopting a

somewhat different approach, Chairman Martin, in his 1959 reply to a
questionnaire of the Joint Economic Committee, stated that the Government
securities market should be characterized by a relatively large volume
of continuous trading and moderate day-to-day price

changes.2

^ U. S. Congress, Joint Committee on the Econimic Report,
Subcommittee on Economic Stabilization (Flanders Committee), United
States Monetary Policy: Recent Thinking and Experience, Hearings,
83d Cong., 2d Sess., 195]+, p. 265.
2 U. S. Congress, Joint Economic Committee, Employment, Growth,
and Price Levels, Hearings, Part 6C, 36th Cong., 1st Sess., 1959> V* 1801.




5
The Ad Hoc Subcommittee defined a declining market as disorderly
"when selling feeds on itself so rapidly and menacingly that it discourages
both short covering and the placement of offsetting new orders by investors
who ordinarily would seek to profit from purchases made in weak markets."^
A similar definition of disorderly conditions was applied to rising markets.
Most of these characteristics, particularly,those referring to
orders on the dealers' books, cannot be measured directly.

Nevertheless,

several statistical series have been suggested or used to measure "depth,
breadth and resiliency."

For example, Mr. Sproul in the Flanders Committee

Hearings in December 195^ suggested that "depth, breadth, and resiliency"
might be measured by dealers' positions, volume of trading, or erratic
price movements.

The Fact Finding Staff Committee, which studied the

performance of the market from 1950-57.? used data on trading volume, dealers'
positions, and spreads between the dealers' quoted bid and asked prices.
Economists outside the Federal Reserve System, who have not had access to
data on positions and trading until recently, have studied ownership distribution of the Treasury debt, price movements and spreads between bid
and asked prices in order to appraise market performance.
2.

The Dealers' Criticisms of Official Operations in Coupon Securities
This section summarizes only the opinions of those dealers who

have criticized purchases of coupon securities by official accounts and
does not necessarily reflect the attitudes of all the dealers or even of
2
the majority.

The criticisms can be divided into two groups, one per-

taining to the role of what the market has dubbed "operation nudge" in
1 Flanders Committee, Hearings, p. 268.
2 The summary was based on the weekly letters of Aubrey G.
Lanston & Co., Inc., Robert Van Cleave!s article "Operation Nudge"
(Banking, April 1962), and a column by Paul Hefferman, "Stability in
Government Bond Market," (reprinted in American Banker, July
1964).



6
achieving the objectives of the Federal Reserve and the Treasury and the
other pertaining to the impact of "operation nudge" on market performance.
a.

Achievement of the Objectives of "Operation Nudge." The'

dealers admitted to some uncertainty about the actual objectives of
official accounts in buying coupon securities.

They thought, however,

that any recognized objectives of "operation nudge" that were achieved
could also have been accomplished without such purchases.

In addition,

"operation nudge" contributed to a shortening of the publicly held debt
during 1961—a development which might complicate future attempts to
combat inflation.
One publicized objective of "operation nudge," namely to supply
reserves with a minimum of downward pressure on bill rates, was achieved.
Treasury bill rates did not fall to as low levels in the 1960-61 recession
as in earlier recessions, but according to the dealers this was not a
result of "operation nudge."

They thought that the decline in bill rates

was relatively small because the Treasury sold a large volume of short-term
securities and. because the discount rate was held at higher levels than in
previous recessions.

In a recession banks tend to invest surplus reserves

in Treasury bills thus driving down rates, but this bank demand for bills
is in part determined by rates on other short-term assets. Thus since in
I96O-6I the discount rate was relatively high and the Federal funds rate
was frequently near the discount rate, banks hesitated to buy bills at
rates much lower than those on Federal funds.

"Operation nudge," on the

other hand, produced little reduction in net buying pressure on bill rates.
If a decrease in official purchases of bills led to relatively high bill
rates, they would be largely offset by an increase in private purchases or
a decrease in private sales.




7
Very few of the investors who sold intermediate and long-term
securities to the Federal Reserve reinvested
long-term credit markets.

the funds in the private

The large flow of funds into these markets

reflected a high level of savings, and so this second assumed objective
of "operation nudge" would probably also have been accomplished anyway.
Similarly, the reduction in the spread between long-term and short-term
yields could have occurred without the help of "operation nudge," as
happened in 1959•
b.

Market Performance.

The dealers also claimed that official

operations in coupon securities impaired the functioning of the market
for intermediate and long-term Government securities.

This impairment

resulted from the possibility of official operations as well as from the
actual operations themselves, but the detrimental effects on the market
were said to be most pronounced in periods when operations were largest.
These dealers alleged first that price levels were artificially
high, since they reflected actual and potential purchases by official
accounts rather than supply and demand from private investors.

The impact

of aqtual purchases by official accounts on prices was larger than was
warranted by their size.

Moreover, even in periods when official accounts

were not in the market, prices were artificially high because the possibility of official buying was always in the background.

The resultant

artificiality of price levels made it extremely difficult for the Treasury
to plan financings, first because substantial price concessions were
necessary to market new Treasury securities and second because price
quotations provided poor guidance to the size of necessary concessions
or even to maturity areas where demand existed.

Private borrowers were

also sometimes misled about the strength of market demand for securities




8
by artificial market prices
thus causing congestion.

and offered excessive amounts of new issues,

In addition, private investors found it difficult

to sell large blocks of Government securities because price quotations
were not firm on the bid side.
Second, retail trading was said to have been smaller because of
official transactions in coupon securities.-1-

Potential buyers hesitated

to buy because they thought price levels were too high or because they
were uncertain about the Federal Reserve1s goals and their future purcnases.
Sellers sometimes rushed in to take advantage of the chance to sell to
official accounts at high prices, but at other times they postponed sales
in the expectation that official buying would push prices still higher in
the near future.

This decline in retail activity may not have been evident

in the data, according to the dealers, because of an increase in trading
among dealers and brokers that also stemmed from official purchases, since
several dealers or brokers may handle securities ultimately sold to the
Federal Reserve.
Third, it was claimed that the dealers were less willing to take
positions in intermediate and long-term Government securities, especially
short positions, because of uncertainty about future official purchases.
3.

Statistical Indicators Selected
The statistical indicators selected for analysis of the market's

day-to-day response to official operations were the following:
(1) Dealer sales to private customers by maturity (retail sales).
(2) Dealer purchases from private customers by maturity
(retail purchases).
(3) Average daily price changes of securities in each maturity
class.
^ Advance refundings were also given some credit by the dealers
for the decline in trading-




9
(k) Dealer offerings to the Trading Desk by maturity.
(5) Daily change in dealers gross long positions, gross short
positions and net positions by maturity.
The maturity categories selected for all indicators in all series were
securities maturing after one year through five years, those maturing
after five years through ten years, and those maturing after ten years.
a.

Description of the Indicators. • Dealer sales to private

customers (also referred to as retail sales) were defined as dealer sales
to all customers except U. S* Government securities brokers and dealers,
other brokers and dealers, and official accounts (i.e., System and Treasury).
Similarly, dealer purchases from private customers (also referred to as
retail purchases) were defined as dealer purchases from all customers
except U. S. Government securities brokers and dealers, other brokers and
dealers, and official accounts.

The source of these statistics was the

daily reports submitted by the dealers reporting to the Market Statistics
Division of the Federal Reserve Bank of New York. Repurchase agreements
were not included in private trading or in official transactions.
The series on average daily price changes were calculated especially for this study, in order to obtain daily data for price changes in
each desired maturity class.

The basictsource of the data was the daily

closing bid prices of U. S. Government securities as published by the
Federal Reserve Bank of I F w York. Every Government security maturing
\e
after one year (except the 1 l/2!s) was classified daily in the appropriate
maturity class (>1

5 ~ 10, >10)."*" The dayfs price change for each

issue in the given maturity class was calculated, and then a simple arithmetic average was obtained.

The resulting averages were expressed in

points and decimal fractions, rather than in 32nds.
1 Issues moving from one maturity class to another were arbitrarily
shifted on the 10th trading day of the appropriate month, in order to simplify
programming.




10

Data on daily dealer offerings to the Trading Desk were calculated
by the Securities Department of the Federal Reserve Bank of New York from
the daily tabulations of all unsolicited offerings.1

Again, offerings were

classified into three desired maturity classes.
Daily changes in gross long positions, gross short positions,
and net positions of dealers came from the daily reports submitted by the
dealers to the Market Statistics Division of. the Federal Reserve Bank of
Mew York.

Positions were on a commitment basis, and securities placed with

customers under repurchase agreements were included in gross long positions
and net positions.
b.

Use of the Indicators.

These series were selected because

they were measures of market performance, because they could be used to
test the dealers' allegations about the market performance, and because
they were available en a daily .basis.

If these indicators responded to

official purchases in the ways suggested by the dealers, official purchases
of coupon securities generally should have led to a low level of dealer
sales to private customers, a high level of dealer purchases from private
customers (unless prices were expected to rise further, when purchases
-should also have been low), a rise in security prices, a rise in dealers'
short positions or a decline in long positions, and a decline in net
positions.

In addition, official buying should have caused dealers zo

offer still more securities to the Trading Desk.
Unfortunately, however, official purchases were also partly a
response to the market indicators.

This might lead one to expect different

relationships between official operations and market indicators and complicate the interpretation of the relationships.

For example, any negative

relationship between official purchases and dealer sales to private customers
1 Offerings made in response to System requests are not included.
quantity) revisions are made in the original
not occur.

 If a dealer changes his price (or
http://fraser.stlouisfed.org/
offering, so double.counting does
Federal Reserve Bank of St. Louis

11
and any positive relationship between official purchases and dealer purchases from private customers might also have reflected the tendency for
official accounts to buy when securities were available.

To take another

example, Treasury accounts tended to buy when prices were falling—a fact
which should have caused a negative relationship between price changes and
dealer sales to official accounts rather than the positive relationship
suggested earlier.

These hypotheses are discussed further when the tests

are presented and interpreted.
This study was confined to the impact of official operations on
the market indicators on the day when these accounts were in the market
and, in some cases, on the following day*

Effects attributable simply

to the possibility of official transactions during the entire period
would have to be assessed by comparisons with pre-1961 statisitcs.
C.

Relationship Between Dealer Sales to Official Accounts and Market

Indicators on Days Fnen Thore Were No Financings, 8/22/62-12/31/63
Preliminary examination of the data showed that the behavior of
most market indicators was much different during Treasury financings than
at other times, even when there were no official transactions.

For example,

trading was, of course^, much higher during financings than at other times.
Therefore, this section of the study analyzes the response of market
indicators to official transactions on days when Treasury financings were
not in progress.

(Section D considers the response of selected market

indicators during financings.) Moreover, because of the time-consuming
nature of the task of keypunching' daily data, this study of days without
financing covered only August 22, 19622 through December 31*




12
not the entire period for which data were available."*-

Official operations

in coupon securities were relatively heavy in this period, however; so
that the sample provided a good basis for analysis.
Financing periods were selected independently for each of the
three maturity classes (>l-£l-5>

10, *i>10), so that the number of

days without financings in the samples differed for each maturity class.
For example, if a financing involved only issues maturing in 1-5 an(i
5-10 years, days during this financing were included in the sample of
days with financings for the 1-5

5-10 year maturity classes and in

the sample of days without financings for bonds maturing after 10 years.
A financing was defined as extending from the day after the announcement
day through the payment date if the securities in the maturity class included new issues.

If the issues in the given maturity class that were

involved in the financing included only rights, the financing was defined
as covering the day after the announcement through the day the books
closed.

The samples of days without financings, of course, excluded all

days falling within financing periods so

d e f i n e d . 2

After a description of the frequency and size of official
operations, the remainder of this section discusses the results of
statistical tests of the day-to-day response of each market indicator
to dealer sales of coupon securities to official accounts.

The first

1 This period was selected arbitrarily because official operations were relatively frequent. August 22, 1962 was the starting day
because it was the beginning of a period of operations.
2 The days when reinvestment of the proceeds of municipal
advance refundings took place were also excluded (i.e., 2/14/63 and
IO/I0/63 for 1-5 year issues and 7/13/63 f o r 5-10 year issues). Days
of competitive bond auctions were not excluded.




13
type of test, the chi-square test, utilized frequency distributions of
the days in the sample classified according to dealer sales to official
accounts and also according to each of the market indicators.

For example,

the frequencies of days with high and low dealer sales to private customers
when official accounts were in the market and when they were out of the
market were compared to the frequency distributions expected on the basis
of the distribution of such sales on all days in the sample.

This test

should indicate whether there was a relationship between sales to official
accounts and the given market indicator.

(An example of how the test was

applied is given below,)
Chi-square tests were applied to the relationship between each
market indicator and each of the following measures of official transactions
on the current day:

l) the existence or non-existence of sales to the

System and Treasury accounts,
to the System,
accounts,

2) the existence or non-existence of sales

3) the existence or non-existence of sales to Treasury

k) large and small sales to the System and Treasury accounts,

5) large and small sales to the System and
Treasury accounts.

6) large and small sales to

The tests were also applied to the relationship between

market indicators and the same measures of sales to official accounts on
Previous day, in order to test for the existence of a lagged relationship.

In all tests the frequency distributions were compressed to four

cells to increase the number of frequencies in each cell.

Even so, not

all the tests could be calculated for all three maturities because there
were not enough theoretical frequencies in some cells.^

1 The original data from the computer showed more classes for
the market indicators and also the mean" and standard deviation of the
market indicators for each frequency distribution.




14

The chi-square test does not indicate the form or magnitude of the
relationships between sales to official accounts and the market indicators.
Thus, as a second test, multiple regressions were also calculated to see
how much of the variations in the market indicators could be "explained"
by official operations.

In each regression five measures of official

operations were used as independent variables, namely, dealer sales to
Treasury accounts on the current day

sales to the System on the

current day (Sst), sales to the Treasury on the previous day (S^-i)* sales
to the System on the previous day (S^.^), and a measure of the duration of
sales to official accounts (D) which was the number of days out of the most
recent five days on which dealers made sales to either Treasury accounts
or the System.

These measures permitted analysis of the current response

of the market indicators to System and Treasury operations separately, as
well as analysis of the lagged response and of the response to the duration
of operations.

Obviously, the market indicators also responded to each

other and to other factors (in ways discussed below); and so, where reasonable, market indicators were also included as independent variables in
the multiple regressions.

1.

Dealer Sales of Coupon Issues to Official Accounts
Tables I through III describe the frequency and size of the

dealers1 daily sales to official accounts of coupon securities in each of
the three maturity classes, when financings were not taking place.

In

each of the three maturity classes sales to either the System or the
Treasury or both occurred on slightly over one-fourth of the days—63 days
out of 2b2 days for 1-5 year issues, 71 days out of 265 days for 5-10, and
91 days out of 311 days for over-10 year issues.
tions were much less frequent.



Relatively large opera-

For example, sales of over $20 million

15

Table I
Frequency Distribution of Trading Days Without
Financings Classified According to Level of Dealer Sales
to Official Accounts of Securities Maturing In > 1
Years
(August 22, 1962-December 31, 1963)

Volume of Sales to
Official Accounts
(in millions of dollars)

Type of Official Account
SOMA
Total
Treasury
(number of days)
212

207

179

(30)

(35)

(63)

Greater than 0 but less than or equal to 1

11

0

11

Greater than 1 but less than or equal to 20

12

11

22

( 7)

(2t)

(30)

Greater than 20 but less than or equal to 40

h

10

12

Greater than 40 buo less than or equal to 60

l

10

12

Greater than 60

2

h

6

2k2

2k2

2k2

Subtotal:

Subtotal:

Greater than 0

Greater than 20

Total:

(in millions of dollars)
Highest Daily Sales
Average Daily Sales*

-- Based on entire number of days in the sample.
x




73-5

78.7

78.7

1.7

4.9

6.6

16

Table II
Frequency Distribution of Trading Days Without
Financings Classified According to Level of Dealer Sales
to Official Accounts of Securities Maturing in
5 ^ 1 0 Years
(August 22, 1962-December 31, 1963)

Volume of Sales to
Official Accounts
(in millions of dollars)

TVpe of Official Account
SOMA
Total
Treasury
(number of days)
222

229

194

to)

(36)

(71)

Greater than 0 "but less than or equal to 1

IT

0

11

Greater than 1 but less than or equal to 10

11

7

17

(15)

(29)

Greater than 10 but less than or equal to 20

8

10

17

Greater than 20 but less than or equal to 30

3

14

15

Greater than 30 but less than or equal to 40

3

3

8

Greater than 40

1

2

3

265

265

265

Subtotal:

Subtotal:

Greater than 0

Greater than 10

Total:

to)

(in millions of dollars)
Highest Daily Sales
Average Daily Sales*

*

Based on entire number of days in the sample.




47.5

45-7

47-:

1.6

2.7

h.:

17

Table III
Frequency Distribution of Trading Days
Without Financings Classified According to Level of
Dealer Sales to Official Accounts of Securities Maturing After 10 Years
(August 22, 1962-December 31, 1963)

Volume of Sales to
Official Accounts
(in millions of dollars)

Type of Official Account
SOMA
Treasury
Total
(number of days)
23^

288

220

(77)

(23)

(91)

Greater than 0 but less than or equal to 1

21

1

18

Greater than 1 but less than or equal to 5

27

17

ko

Subtotal:

Subtotal:

Greater than 0

Greater than 5

(29)

( 5)

(33)

Greater than 5 "but less than or equal to 10

11

5

14

Greater than 10 bu'c less than or equal to 15

7

0

8

Greater than 15 but less than or equal to 20

5

0

Greater than 20

6

0

7

311

311

311

Total:

(in millions of dollars)
Highest Daily Sales
Average Daily Sales*

*

Based on entire number of days in the sample.




61.3

10.0

61.3

2.0

0-3

2.3

18

occurred on 30 days in 1-5 year issues.
executed on

Sales exceeding $10 million were

days in the 5 - 10 year class; and sales of more than $5 niillion

were made on 33

for securities maturing after 10 years.

In addition, sales to official accounts were not divided equally
between the Treasury and the System.

The Treasury accounted for most of

the activity in the longest maturity class, while the System was more
important in the two shorter maturity classes.

Thus, sales to Treasury

accounts of securities maturing after 10 years were made on 77 days while
sales to the System were made on only 23 days.

Similarly, sales to the

Treasury of this maturity exceeded $5 million on 29 days, whereas sales to
the System exceeded
the i-5 year class.

million on only 5 days.

The opposite was true in

Large sales—over $20 million—were made on only J days,

to the Treasury and on 2k days to the System.

A somewhat more balanced

situation occurred in the 5-10 year class, although more large daily sales
were made to the System (over $10 million) than to the Treasury--29 days
compared to 15 days.
The contrast in the relative importance of System and Treasury
accounts was also evident in the daily average volume of dealer sales to
these accounts.

In the over-10 year class, sales to the Treasury averaged

$2.0 million a day in this period, while sales to the System averaged only
$.3 million.

In contrast sales of 1-5 year and 5-10 year securities to

the Treasury averaged $1.7 million and $1.6 million a day respectively,
against $4.9 million and $2.7 million to the System.
The largest concentration of sales to official accounts.during
this period (excluding financings, of course) occurred in the Fall of 1962,
May 1963>

late June and early July 1963*

Relatively heavy daily sales

were also scattered through the rest of the period.



19
2.

Sales to Official Accounts and Sales to Private Customers
a.

Chi-square Tests.

The comments of the dealers would lead one

to expect a lower level of dealer sales to private customers when dealers
were also selling to official accounts.

Analysis of the frequency dis-

tributions of daily sales to private customers for August 22, 1962 through
December 31> 19^3 revealed no such relationship either with sales to
official accounts on the current day or on the previous day in any of the
three maturity classes.

This analysis was based on the chi-square test

of the significance (or lack of significance) of the relationship between
two principles of classification, in this case two characteristics of the
Government securities market.
This test is described in detail for retail sales of securities
maturing after 10 years and large sales (over $5 million) of such securities
to official accounts.

The first row of Table IV shows the actual frequency

distribution of daily dealer sales to private customers.

In other words,

each of the 3 H days is classified according to the level of retail sales.
Thus, on 207 days retail sales were less than or equal to $10 million,
while on 10b days retail sales were greater than ^10 million.

The second

row shows the percentage of the total days falling in each class—i.e., on
66.6 per cent of the 3 H

sales were $10 million or less.

The third

and fourth rows show the same absolute frequency distributions for days
when dealer sales to official accounts were less than or equal to $5 million
(278) and days when such sales exceeded $5 million (33)The question to be answered by the chi-square test is whether
the frequency distributions in rows three and four are what would be
expected, if there had been no relationship between large sales to
official accounts and retail sales.




If there had been no relationship,

20

Table IV
Example of Calculation of Chi-Square
Relating Retail Sales to Large Dealer Sales to
Official Accounts of Securities Maturing After 10 Years
(Dollar amounts in millions)

Sales to Official Accounts

Dealer Sales to Private Customers
^ 10
>-10
Total

All Days Without Financings
(l) Number
(2 ) Per cent

207

104

66.6

33A

311
100

Days When Volume of Sales to
SOMA + Treasury were
Actual Distribution
(3)

189

89

278

GO

18

15

33

Theoretical Distribution
(5) ^ 5

(6)

>5




185

93

278

22

11

33

21
66.6 per cent of the 278 days with small sales to official accounts and
66.6 per cent of the 33 days with large sales to official accounts would
have had retail sales of $10 million or less, since 66.6 per cent of all
days in the sample had retail sales of $10 million or less (row two).
The so-called "theoretical" frequency distributions in rows five and six
were calculated on the basis of this assumption of the same percentage
distribution for the two sub-groups as existed for the total sample.
In fact, the actual and theoretical distributions were not exactly the
same, but these differences may have been due to chance.
The chi-square statistic is calculated from these theoretical
and actual frequency distributions by a formula-'- which gives a chi-square
of zero if the distributions are exactly the same and increasingly large
chi-squares as the difference between the distributions increases.

Then

prepared taoxes are consulted to see if the resulting value of chi-square
is significantly different from zero—in other words, if the differences
between the actual and theoretical frequency distributions were too great
to be due to chance, under the hypothesis that large sales to official
accounts and retail sales were really unrelated.
In this case, chi-square equaled 2.4^01 and was not significantly
different from zero at the 5 per cent confidence level.

In other words, a

value this large could occur more often than 5 times in 100 if the true
value were zero.

Thus, the data were consistent with the hypothesis of no

relationship between retail sales and large sales to official accounts.

1 Chi-square equals the sum of the squared differences between
the numbers in corresponding cells of the theoretical (f) and actual
frequency (f0) distributions divided by the theoretical frequency.
Chi-square =
(f0 - f)2|




22
If chi-square had been 3-8^1 or greater in this example, the hypothesis' of
no relationship would have been rejected, since such a large value of chisquare would occur by chance only 5 times in 100 if there were in fact no
relationship.

In fact, none of the chi-squares relating retail sales and

sales to official accounts were significant at the 5 V e r cent level.
Where there were enough observations, the same chi-square tests
were also calculated for days when prices declined (the average price decline exceeded l/32), when prices were unchanged (the average change was
between + and - l/32) and when prices rose (the average price increase was
greater than 1/32)., on the possibility that a reaction between retail sales
and prices had obscured a relationship between retail sales and official
transactions.

Again there were no significant chi-squares.

Thus, the

tests were consistent with the hypothesis that there was no relationship
on a day-to-day basis between dealer sales to private customers and dealer
sales to official accounts.
b. Multiple Regressions. As a further test for a relationship
between dealer sales to private customers (S^t) anci sales to official
accounts, mulitiple regressions were calculated. The independent variables
included the five measures of official operations mentioned earlier, dealer
sales to Treasury accounts on the current day (STt)> dealer sales to the
System on the current day (SSt), dealer sales to Treasury accounts on the
previous day

dealer sales to the System on the previous day

(Ss-t-l)> and the number of days out of the last five on which either
dealers made sales to the System or Treasury accounts ( D ) .

In addition,

the dealers' net position on the previous day (NP-fc-l)* "the average change
in prices on the current day (/\pt)>

an(3

-

average change in prices on

the previous day (Z\P+_-,) were used as independent variables.



23
Even with the addition of these other variables, however, the
equations were far from expressing the complex interrelationships in the
Government securities market.

Obviously, a number of factors on both the

supply and demand sides of the market, other than official activity, may
influence dealer sales to private customers.

The supply of securities

available for sale depends partly on the dealers* inventories, partly on
the dealers1 willingness to sell their inventories, and partly on what the
dealers expect to be able to buy in the market.

The dealers1 willingness

to sell their inventories would, in turn, depend on expected price changes
and, on the cost of financing the securities relative to the coupon.

The

dealers* net position on the previous day was included in the regressions;
and the actual change in prices on the past two days may have reflected the
dealers1 expectations; but there undoubtedly were a number of influences
on the supply side that were left out.
variables were even more numerous.

On the demand side the omitted

Presumably, private customers1 demand

for Governments would depend on relative or absolute yields, their own
cash flows, and expected price changes.
The implication of this brief discussion is that the regressions
calculated would probably not explain much of the variation in retail
sales on days without financing.

Furthermore, it raises the possibility

that the true relationship between retail sales and official operations
may not be uncovered.

One possible solution, a detailed model of the

Government securities market, is beyond the scope of this study.
The final equations1 were as follows:
(1-5 yrs.)

S p t = 25.737 - .3^9 S T t
(.3^2)

S S t + .613
(.148)

S

(.336)

T
t

_ l

+

-

2 1 8

s

V

(.147)

1 In each equation all variables referred only to the given
maturity class. In these equations and those in the following sections,
dollar variables were in millions and changes in prices were in points.
 Price changes less than 1 point were expressed as decimals, not 32nds.
http://fraser.stlouisfed.org/
See Appendix Table XIV for other results.
Federal Reserve Bank of St. Louis

l

24
- 2.777 D + .085 MPt.! + 108.115 A Pt + 18.475 A
(1.16k)

(.017)

(55.525)

Pfl

(55.307)

S p t = 3^.6^9 - .094 S T t + .128 S s t - .126 S T t -l + .113 S s t -1

(5-10 yrs.)

(.231)

(.158)

(.233)

(.162)

- 2.641 D + ^033 NP-t-i + 6.629 A ? t " 8-785 A Pt-i
(1.073)

(-009)

(16.867)

(16.858)

S p t = 8.280 + .112 S T t - .452 S s t - .118 S T t - 1 + .191 S s t - 1

(>10 yrs.)

(.058)

(.274)

(.068)

+ .176 D + .016 NP t - 1 + .231 A Pt
(.287)

(-005)

(4.802)

+

(.092)
-501 A P t - 1

(4.843)

The numbers in parenthesis are the standard errors of the regression
coefficients.

The underlined regression coefficients are those that are

significantly different from zero at the 5 per cent level or better—i.e.,
a coefficient that large could occur less than 5 times in 100 if the true
coefficient were zero.
As can be seen from the equations, no significant relationship
was found between the measures of official activity in most maturity
classes, when all the variables were included.

One exception was sales to

the System on the previous day in the over-10 year maturity class. which
was positively related to retail sales.^ Another exception occurred in
the 5"10 year maturity class, where there was a significant negative relationship between retail sales and the duration of sales to official
accounts.

(An increase of one day in the number of days out of the previous

five on which official accounts bought 5-10 year Governments led to a
decrease of $2.6 million in retail sales, when the other variables were
1 As Appendix Table XIV shows, the simple correlation coefficient
between retail sales and sales to the Treasury on the current day was positive and significant at the 1 per cent level, but it was not significant


in
http://fraser.stlouisfed.org/ multiple
Federal Reserve Bank of St. Louis

regressions because of multicollinearity with net position.

25
held constant.) Even the regression coefficients for these two variables,
however, were on the borderline of significance:

they were significant

at the 5 per cent level but not the 1 per cent level.
implications of the two coefficients are contradictory.

Moreover, the
In the 5-10 year

class official activity led to lower retail sales, as the dealers suggested;
but in the over-10 year class, it led to higher retail sales.
Of the three other variables, only net positions on the previous
day was significant in all three maturity classes.

Sales to private cus-

tomers were somewhat higher when positions were large.

Moreover, the

total variation in retail sales explained by all the variables was very
small indeed—12 per cent in the 1-5 year maturity class, 7 per cent in
the 5-10 year maturity class, and 6 per cent in the over-10 year class.
As was the case with the chi-square tests, the multiple regressions provided almost no confirmation of the dealers1 contention that
buying by official accounts led other buyers to withdraw from the market
on days without Treasury financings in late 1962 and 19^3-

Of course,

these results might not hold if the magnitude or frequency of official
purchases were greater than in this period.

Moreover, as was explained

above an integrated and more complete model of the market might reveal
some more pronounced relationships.

3.

Sales to Official Accounts and Purchases from Private Customers
a.

Chi-square Tests.

Chi-square tests were calculated on the

relationship between dealer purchases from private customers and the various
measures of dealer sales to official accounts.

In all three maturity classes,

some of the chi-squares based on the current dayfs measures of official
activity were significant at the 5 per cent level, suggesting that there
was a relationship between retail purchases and sales to official accounts




26
on the current day.

Evidence of a lagged relationship also existed In the

over-10 year maturity class, but not in the other two maturity classes.
Examination of the frequency distributions showed that these relationships
were in the direction suggested by the-most frequent dealer criticisms.
Retail purchases were higher on days when the dealers were also selling
to official accounts.
In the 1-5 year maturity class, a significant chi-square of
4.07^5 vas found between the size of retail purchases and the existence
or non-existence of dealer sales to official accounts.
Table I.)
class.

(See Appendix,

None of the other chi-squares were significant in this maturity

In the 5-10 year class, more of the chi-squares were significant,

as Appendix, Table II shows.

Significant relationships were found between

retail purchases and the two measures of sales to the System and the two
measures of sales to the System and Treasury accounts together.

No

significant relationship was evident between retail purchases and sales
to Treasury accounts.
In contrast, for the longest maturity class, there were no
significant relationships between retail purchases and sales to the System,
but there were significant relationships between the size of retail purchases and large sales to Treasury accounts on the current day and between
retail purchases and large sales to Treasury accounts and the System,
taken together. [See Appendix, Table III).

Classification of the current

day's sales to official accounts on a zero and greater than zero basis did
not result in a significant chi-square.

Probably this was because there

were so many days when sales to Treasury accounts, although greater than
zero, were very small (less than $1 million).

On the other hand, there

were significant chi-squares between existence or non-existence of sales




27
to Treasury accounts on the previous day and the current day's retail
purchases (7-U200) and also between the existence or non-existence of
sales to Treasury accounts and the System together on the previous day
and retail purchases ( * . 1077)*
1-

Possibly the dealers were aware of such

small sales only with a lag and so also reacted to them by increasing
purchases from private customers with a one day lag.
b.

Multiple Regressions.

Hie multiple regressions calculated
same

to explain dealer purchases from private customers ( F ^ ) used

eight independent variables as the equations explaining retail sales,
namely, S T t , S s t , S T t _ l f S s t .i, D, MPt^if A p t > ^

A p t-1-

the

A* ^

case with retail sales, retail purchases were undoubtedly influenced by
many other variables not included in the regressions, such as investors
cash needs, relative yields, price changes expected by the dealers and by
customers, the cost of financing positions, and volume of sales to private
customers expected by the dealers.
The final equations1 were:
(1-5 yrs.)

pP t = 39.906- .ikk s T t + .159 S s t + .193 S T t _l + .170
(.396)

(.171)

(.390)

s3tml

(.170)

- 2.268 D + .052 NP t - 1 + 62.^6 A P t + 129*553 Ap-t-i
(2.0lf5)

(5-10 yrs.)




(.019)

T^t m 31.656 -- .361 S T t +
t
(.238)

(6U.356)

S s t + .077 S
(.163)

(.009)

1

(.21*1)

- 1.U02 D + .001 NP t - 1 + U6.162 A
(1.106)

(61*.1<*)

(17.too)

1 See Appendix Table XV for other results.

p

t

^ 4 .020 S s t - 1
(.168)

+

^.662 A
(17.391)

p

t-1

28

(>10 yrs.)

^

= 8.448 + .204 S T t - .118 S s t + .109
(.060)

(.284)

(.070)

- .o4f d - .001 NP t . x + 3j^48 A p t
(.297)

(.005)

- .006 S s t - 1
(.095)
+

(4.971)

n * ^

Apt-i

(5-013)

These variables, however, explained only a very small part of the variation
in retail purchases--4 per cent in 1-5 year issues, 8 per cent in 5-10 year
issues, and 10 per cent in issues maturing after 10 years.
In the equations for the 5-10 and over-10 year maturity classes,
dealer sales to official accounts on the current day were positively
correlated with retail purchases.

Reflecting differences in activity dis-

cussed earlier, sales to Treasury accounts was the significant variable in
the over-10 year class, while sales to the System was significant in the
5-10 year class.

In the latter case retail purchases rose $510 thousand

for every $1 million increase in sales to the System, assuming the other
variables were constant.

In the over-10 year maturity class retail pur-

chases rose |204 thousand for a $1 million increase in sales to Treasury
accounts.

Both of these regression coefficients were significant at the

1 per cent level.
The other measures of official activity were not significant in
the 5-10 and over-10 year class, while in the 1-5 year class none of the
measures of official activity were significant.
As the equations above indicated, price changes were positively
related to retail purchases in all maturity classes.

In the two longest

maturity classes, the regression coefficients for both the current day's
change in prices and the previous day's change in prices were significant
at the 5 per cent level or better, while in the 1-5 year class the only
significant coefficient was that for price changes on the previous day.



29

For example, an increase of 1 point in the average price of securities
maturing after 10 years on the current day was associated with an increase
of $17.5 million in dealer purchases from retail customers, when there was
no change in other variables.
In conclusion, the multiple regressions and the chi-square tests
suggested that sales to official accounts have been associated with higher
dealer purchases from private customers on a day-to-day basis, particularly
in the case of longer-term securities.

These results were consistent with

the dealers J assertion that buying by official accounts led other customers
to sell while good prices could be obtained.

On the surface they also

seemed consistent with the contention that official accounts bought when
securities were available and that large dealer purchases implied greater
availability.

k.

Sales to Official Accounts and Changes in Prices
a.

The dealers1 comments that purchases by

Chi-square Tests.

official accounts in the coupon market result in artificially high prices
would lead one to expect the average daily change in prices of Government
securities to have been positively related to dealer sales of coupon
securities to official accounts.

On the other hand, the fact that the

Treasury tended to buy in a weak market suggested that there might have
been a negative relationship between sales to Treasury accounts and the
daily change in prices.
Oil-square tests provided no evidence of any relationship between
sales to Treasury accounts and the price change on the current day or
between sales to Treasury accounts and the two-day price change (the
current day and the next day).




In the shortest maturity class, 1-5 years,

30

the chi-square test was significant when sales to the System were subdivided
into zero and non-zero groups.

For price changes on the current day chi-

square was 7-521^; and for the two-day price change it was 3*9^03.

The

smaller figure for the two-day price change implies that any price effect
was quickly dissipated.

As suggested by the hypothesis, price increases

were more frequent than normal on days when dealers made sales to the
System.

No other significant relationships were revealed.
b.

Multiple Regressions.

The interrelationships between the

change in prices and other market variables were also complex.

While

changes in prices over several months may have been mainly a reaction to
actual supply and demand, on a daily basis price changes often did not
reflect customer activity but were a technical reaction to some news item
or development that might affect demand and supply in the future.

In

addition, while changes in retail sales or retail purchases sometimes may
have caused price changes, price changes may also have reacted on retail
sales and purchases.

Thus it seems futile to try to explain daily price

changes with the data available.
the change in prices (A^t) a s

Nevertheless multiple regressions with
dependent variable were calculated,

using net positions on the previous day (NB^i),
measures of official transactions (S^., S ^ ,

as w e

H

as

five
as inde-

pendent variables.
In all three maturity classes, sales to the Treasury on the
current day was the variable entered first, which means it produced the
greatest reduction in unexplained variation in the change in prices.
In all three cases this reduction in variation was significantly different
from zero at the 5 V e r cent level.

As was expected the correlation was

negative, indicating that the Treasury bought when prices were falling.




31
After the other variables were added, however, the regression coefficient
for sales to the Treasury on the current day was no longer significant in
the two longest maturity classes, because of multicollinearity with the
other variables.
This negative price relationship, however, appears to have been
relatively short-lived.

For example, there was no significant simple

correlation between the change in prices on the current day and Treasury
(or System) operations on the previous day.

On the other hand there was

still no evidence that System or Treasury operations had a stimulating
impact on (or positive correlation) with the change in prices.

As a

further test of the possibility that the negative correlation between the
change in prices and sales to Treasury accounts was soon offset by the
stimulating impact of sales to Treasury accounts, the two-day price change
ft + (t - lJ7 w a s also used as the dependent variable.

The two-day price

change was also negatively correlated with sales to the Treasury on the
current day.

These simple correlation coefficients were about the same

as those between the current day's price change and sales to the Treasury.
The two-day change in prices, however, was negatively correlated with sales
to the Treasury on the previous day in the two shorter maturity classes,
but not in the longest maturity class where Treasury activity was heaviest.
Moreover, in the shortest maturity class there was a positive simple
correlation between the two-day price change and sales to the System on
the previous day (.123).

These results seem to suggest that the negative

relationship between sales to the Treasury and the change in prices was
temporary.

They also raise the possibility that studies based on longer

time periods or lags might turn up a positive relationship between official
operations and prices.




32

None of the other independent variables were significant at the
5 per cent level.

Moreover, the explained variation in the change in

prices was exceptionally small—in the 1-5 year class 3 per cent, in the
5-10 year class l/2 of 1 per cent, and in the over-10 year class actually
negative.^

In view of the fact that almost no regression coefficients

were significant the equations are not given here, but are shown in
Appendix, Table XVI.

5#

Sales to Official Accounts and Changes in Dealer Positions
a.

Chi-square Tests.

Comparing frequency distributions through

the chi-square test showed that the daily change in gross long positions
and net positions were related to some measure of sales to official accounts
in all maturity classes.

As was expected, increases in long and net posi-

tions were less frequent on days when dealers were selling to official
accounts.

In the over-10 year maturity class there was also a significant

relationship between sales to official accounts and the change in short
positions—with a rise in short positions being more frequent when official
accounts were buying.

Appendix Tables IV through X show the actual and

theoretical frequencies used to calculate the values of chi-square.
b.

Multiple Regressions.

Multiple regressions were also

calculated to explain the daily change in the dealers 1 gross long position,
gross short position, and net position, although it did not seem that
daily data were the most suitable for testing the dealers 1 allegations
that official operations had caused dealers to carry lower positions.
Even if the dealers* willingness to carry.positions over a longer period of
time were not affected by official operations, some of any day's sales to
official accounts would normally come out of positions.

In the next week

or so, however, positions could be rebuilt.

http://fraser.stlouisfed.org/
freedom
Federal Reserve Bank of St. Louis

1
-* It can be negative because R2 is adjusted for the degrees of
by a subtractive adjustment.

33
The results for the regressions explaining the change in gross
long position and net position were similar.1

The proportion of the

variation explained by all eight variables was about 33 per cent in the
1-5 year class, about 25 per cent in the 5-10 year class, and kl per cent
in the over-10 year class.

As expected, dealer sales to both the System

and the Treasury on the current day were negatively correlated with the
changes in both long and net positions in all maturity classes.

In the

over-10 year class, however, sales to the Treasury on the previous day was
positively related to the change in long and net positions.

This implies

that the dealers had already started to rebuild positions on the day after
official operations—a fact not indicative of a reduced willingness to
hold positions over the longer run.
The dealers1 net position on the previous day was negatively
related to the change in long and net positions in some maturity classes,
while the change in prices on the current and previous day was positively
related to the change in positions.
The results of the regressions on short positions showed much
smaller R 2 s and lower correlations (if any) between the change in positions
and official activity.

Only 7-9 per cent of the variation in the change

in short positions was explained by the eight variables.(See Appendix,
Table XVIII).

In the 1-5 year class, both sales to the System on the

current day and sales to the Treasury on the previous day were positively
related to the change in short positions, when all variables were included.
In the over-10 year class, the duration of official sales was positively
correlated with the change in short positions.

Multicollinearity prevented

sales to the Treasury on the current day from being significant in this




1

See Appendix, Tables XVII and XIX.

regression when all variables were added, although the simple correlation
coefficient was significant.

No other measures of official activity were

significant; but the change in prices had significant negative coefficients
in several cases.
The fact that the few significant correlation coefficients between the change in short positions and the measures of official activity
were positive implied that dealers sometimes went short in order to sell to
official accounts-

These results were thus contrary to dealer criticisms,

which suggested that dealers were less willing to carry short positions
when official accounts were buying, because of fear that prices would rise
and make it expensive to cover short positions.

This would appear to be

further evidence that correlations of daily data measured mainly immediate
technical reactions and did not really reflect changes in dealers1 willingness to carry positions even for the short-run, such as a week.

6.

Sales to Official Accounts and Dealer Offerings to the Trading Desk
a.

Chi-square Tests.

Chi-square tests also showed significant

relationships between daily offerings to the Trading Desk and some measures
of dealer sales to official accounts in all three maturity classes.
Appendix, Tables XI-XIII.)

(See

In the two shorter maturity classes, there

was not a significant relationship between offerings and sales to Treasury
accounts alone, while in the longest class there was not a significant
relationship with sales to the System.

This appears to have been another

example of the actual division of labor between the System and the Treasury,
with the Treasury concentrating on long maturities and the System concentrating on short maturities.
b.

Multiple Regressions.

In contrast to the other multiple

regressions, a relatively high proportion of the variation in offerings



35
to the Trading Desk (0 t ) was explained by the eight independent variables
included, which were S T t , S s t , STt-i, SSt-i, D, NPt-l> A

p

t

and

A Pfl#

In the 1-5 year maturity class 35 per cent of the variation in offerings was
explained; in the 5-10 year class, 59 per cent; and in the over-10 year
class 64 per cent.

This probably reflected a less complex relationship and

the inclusion of more of the determinants.

Presumably, offerings should

vary with the availability of securities to the dealers, their willingness
to sell these securities, and the strength of their expectations of being
able to sell to official accounts.

Of the variables included in the re-

gression net positiont-Bl reflects availability, A Pt

an(i

A Pfc-i influence

the willingness to sell, and the measures of official operations influence
their expectations of being able to sell to official accounts.

Other

variables not included—such as financing costs—may still have played a
role, of course.

There is still the problem, however, that offerings may

also have influenced official purchases.
Bie equations1 were:
(1-5 yrs.)

0 t = 42.630 + .684 S T t + 1.059 S s t + .116 S T t - 1 + ,380 S s t - 1
(.409)

(.177)

(.403)

(.176)

- .216 D + .107 NP t - 1 - 363.259 A P t - 60.729 A P t - 1
(2.111)

(5-10 yrs.)




(.020)

(66.444)

(66.183)

Ot = 35.678 + .907 S T t + .931 S s t + .121 S T t - 1 + .262 S s t - 1

(.258)

(.177)

(.261)

(.182)

+ .610 D + .134 NP t - 1 - 169.136 A P t - 51.632 A P t - i
(1.200)

(.010)

(18.869)

1 See Appendix, Table XX for other results.

(18.859)

36
(>10 yrs.)

0 t = 11.384 + .655 S T t + .896 S s t - .102 S T t - 1 + .132 S S t - 1
(.090)

(.427)

(.106)

(.143)

+ 1.462 D + .114 NP t - 1 - 46.622 A P t - 4.398 A P t -i
(.447)

(.008)

(7-467)

(7.531)

As was suggested by the chi-square test, sales to official
accounts on the current day were definitely significant in explaining
offerings to the Trading Desk.

In the shortest maturity class, sales to

the System were more important, while in the longest class sales to Treasury
accounts were the more important variable.

In the 5-10 year class, sales

to both accounts were significant at the 1 per cent level, although the
partial correlation coefficient for sales to the System (.313) "was somewhat higher ^han that for sales zo Treasury accounts (.215).
In the 1-5 year class sales to the System on the previous day
were also significant at the 5 per cent level; and in the over-10 year
class the duration of sales to official accounts was significant at the
1 per cent level, when all eight independent variables were included.
No other measures of official activity were significant in the multiple
regressions because of multicollinearity.

However, simple correlation

coefficients between offerings and almost all measures of official activity
were significant and positive.
Another indication of the importance of official operations was
that the five variables reflecting official operations together explained
18 per cent of the variation in offerings in the 1-5 year class, 12 per
cent in the 5-10 year class, and 33 per cent in the over-10 year class.
The net position on the previous day was significant at the 1 per
cent level in all three regressions.

In fact, net positions alone ex-

plained 33 per cent of the variation in offerings in the 5-10 year class



3T
and k6 per cent in the over-10 year class.

The change in prices on the

current day was significant at the 1 per cent level in all three maturity
classes, while the change in prices on the previous day was significant
in the 5-10 year class.
Clearly both tests show that official purchases of coupon
securities led to higher offerings to the Trading Desk.

7.

Conclusions
This section has presented the results of a statistical analysis

of the relationship between dealer sales of coupon securities to official
accounts and selected indicators of market performance on days when there
were no Treasury financings from August 22, 1962 through December 31f 1963*
Relationships in each of three maturity classes (">1:^5*

^ 1 0 , and

> 1 0 years) were studied separately, with all variables in each study
applying only to the given maturity class.

In this analysis several

measures of dealer sales to official accounts were studied, namely, the
size of sales on the current day, the size of sales on the preceding day,
the official account involved (i.e., Treasury or System), and the duration
of the official activity.

Unless otherwise noted, the relationships

summarized below were based on multiple regressions that included all
five measures of official activity and some other variables.

The con-

clusions were:
(l) There were almost no significant relationships between
dealer sales to official accounts and dealer sales to private customers.
The two exceptions—a positive correlation between dealer sales of securities
maturing after 10 years to private customers and sales to the System on the
previous day and a negative relationship between dealer sales to private
customers of securities maturing in 5-10 years and the duration of dealer
sales to official accounts—were borderline cases, that were significant at

the 5 per


cent level but not at the 1 per cent level.

38

(2) Dealer sales to Treasury accounts on the current day were
positively related to dealer purchases from private customers in the
over-10 year maturity class, while sales to the System were positively
related to purchases from private customers in the 5-10 year maturity class.
The difference in the relative importance of System and Treasury accounts
in the two maturity classes undoubtedly reflected the fact that Treasury
activity was concentrated in longer-term issues and System activity in
the shorter- and intermediate-term issues.
The lagged value of sales to official accounts and the duration
of sales to official accounts had no significant relationship to retail
purchases in these two maturity classes, when the above variables were
included.

Suggestions in the chi-square tests of a positive relationship

between sales to official accounts and retail purchases in the 1-5 year
maturity class were not confirmed by the multiple regressions.
This positive relationship between dealer sales to official
accounts and dealer purchases from private customers could mean that
official activity stimulated private investors to sell their securities,
possibly because they thought that prices were artificially high.

It

could also mean that official accounts bought securities when dealers were
buying them from other customers, since securities were more available at
such times.
(3) The daily average change in prices on the current day was
negatively related to dealer sales to Treasury accounts on that day in all
three maturity classes, when no other variables were included.

When other

variables were included, this relationship was no longer significant in
the 5-10 and over-10 year maturity classes.

Moreover, the current day's

price change was not significantly related to the previous day's sales
to the Treasury.



These results probably imply that Treasury accounts

39
tended to buy Governments on days when prices were weak.

Some of the data

did suggest, however, that a positive impact between official operations
and prices might show up if longer time periods or lags were used.
(4) Both the change in gross long positions and net positions
were negatively related to dealer sales to Treasury accounts and also to
sales to System accounts on the current day in all maturity classes.

In

the two shorter maturity classes sales to the System were more important
than sales to the Treasury, but in the longest maturity class the opposite
was true.
In the 1-5 year maturity, the change in dealers1 gross short
position was positively related to sales to the System on the current day
and to sales to the Treasury on the previous day, at the 5 per cent significance level.

In the over-10 year maturity class, the change in short

positions was significantly related to the duration of sales to official
accounts, rising as the frequency of official activity in recent days
increased.

In fact, this relationship, in a sense, overpowered the positive

correlation between the size of sales to the Treasury and the change in
short positions when both variables were included.

No correlation between

changes in short positions and any measure of official activity was found
in the 5-10 year class.
It seems likely that these relationships between dealer positions
and dealer sales to official accounts were mainly short-run and technical,
since on the actual day of the sale at least part of the securities sold
would normally have come from positions.

Thus, the results do not neces-

sarily imply that the dealers were less willing to hold positions over a
longer period of time.

This view gained some confirmation from the positive

correlation with the change in short positions, since if dealers were less
willing to take positions they should have reduced short positions too.




40

(5) Sales to official accounts also "were associated with a rise
in offerings to the Trading Desk in all maturity classes on the same day.
Again sales to the System appeared to be more important in the two shorter
maturity classes, with sales to Treasury accounts assuming the dominant
role in the over-10 year class.

There also appeared to be a significant

lagged reaction to sales to the System in the 1-5 year class.
In appraising these conclusions, several qualifications should be
kept in mind.

First, the conclusions were based on official operations of

the frequency and magnitude in the sample period.

The impact might be

different if operations were larger or more frequent (or smaller and less
frequent) or if other economic conditions changed.

One possible test of

this would be to run similar tests for 1961-62 or 1964.

Secondly, no

attempt was made to develop a complete model of the Government securities
market.

Further work along the lines of developing a model that would

simultaneously explain positions, prices, trading, and perhaps official
operations, might also result in different conclusions about the impact of
official operations.

D.

Relationship Between Sales to Official Accounts and Retail Trading

During Treasury Rights Financings
This section supplements the analysis of the previous section by
studying the day-to-day relationships between sales of coupon securities to
official accounts and retail trading during rights financings (i.e., exchanges and advance refundings) from March 1961 through July 1964.

The

study concentrated on retail purchases and sales and excluded other market
indicators because the criticisms or hypotheses about the impact of official
operations on retail trading seemed more testable than the hypotheses concerning other indicators.



Moreover, the allegations that official purchases

11
+
led private buyers to withdraw from the marker, and private sellers to dump
securities on the market would represent serious departures from an ideal
marke t perf ormance.
Only advance refundings and rights exchanges were covered in
this study, because the pattern of trading was different during cash
financings.

In other words, in cash offerings no trading in new issues

could take place until the books closed, whereas in rights financings the
heaviest trading occurred before the books closed.

Cash financings were

not analyzed separately because the samples would have been too small.
(See footnote 1 , p. k k . )

Even in the samples on rights financings, the

results would have been more conclusive if there had been a greater number
of days with large official transactions.
Both a long definition and a short definition of financing
periods were used in this section.

Under the long definition (also the

one used in the previous section) the financing period extended from the
day after the announcement through the payment day when the issues in the
given maturity class included new issues.

If the issues included only

rights, the financing period ended the day the books closed.

Under the

short definition, the financing period extended from the day after the
announcement through the day the books closed, regardless of whether the
issues involved in the financing were rights or new issues.
This section first provides background information on the size
and frequency of sales of coupon securities to official accounts and the
average volume of retail trading.

Then the hypotheses to be tested are

reviewed; and finally the results of the simple and multiple correlations
are presented and interpreted.




42
1.

Dealer Sales of Coupon Issues to Official Accounts
Table V classifies all days during Treasury rights financings

according to the size of the dealers 1 sales to official accounts on that
day.

During long rights financings, dealers made sales of securities

maturing after 10 years to official accounts (the System and Treasury
together) on 32 days out of 92.
sales did not exceed $5 million.

On about half of these days, however,
In short financings dealers made sales

of long-term Governments on Ik days, and on 11 of these days such sales
exceeded $10 million.
Treasury accounts:

Almost all sales of long-term Governments were to

the System bought on only k days in long financings

and did not buy at all in short financings.
In the intermediate-term maturity class,

5 — 10 years, sales to

official accounts were made on 60 days in- long financings and 29 days in
short financings.

These sales exceeded $5 million on 33 days in long

financings and 15 days in short financings.

In short financings all such

sales were to Treasury accounts, but in long financings relatively large
sales were made to System accounts almost as often as to Treasury accounts.
Sales of short securities (>>li*=5) to official accounts were
made on 53 days in long financings but only on 16 days in short financings.
In long financings, such sales exceeded $5 million on 32 days, but in short
financings sales of this size occurred on only 6 days.
The table indicates the tentative nature of conclusions based
on the following analysis.

Although the total number of days in each

sample seems adequate, the number of days on which there were large sales
to official accounts was not as large as would be desired.

The number of

such days was particularly small for securities maturing in > 1 ^ 5 years
in short financings.



Sales to the System and Treasury accounts were not

Table V
Days During Treasury Financings,* Classified
According to the Size of Sales to Official Accounts

Maturity
class

Total
days

0

Sales to Official Accounts (in millions of dollars)
10 > 1 0 ^ 2 0 > 2 0 * ^ 4 0 > 1 * 0 ^ 6 0
> 0 ^ 5
(number of days)
Long Financing Periods

All Accounts
> 1 ^ 5

166

113

21

6

10

8

5

3

>5

158

98

27

8

16

8

1

0

92

60

17

3

3

3

3

3

> 1 < 5

166

Ikl

k

2

7

5

k

3

> 5 2s 10

158

142

2

5

5

h

0

0

92

88

3

1

0

0

0

0

> 1 ^ 5

166

138

17

h

3

3

1

0

>5 ^10

158

107

30

7

10

3

1

0

92

6k

lh

2

3

3

3

3

^10

>10
SCMA

>10
Treasury

>10

Short Financing Periods
All Accounts
> 1 ^ 5

78

62

10

0

h

1

1

0

>5^10

67

38

14

5

6

3

1

0

>10

39

25

3

0

2

3

3

3

*

Includes rights exchanges and advance refundings from March l£6l through July I96U.




44

considered separately in the following analysis, because this would have
further reduced the number of days with large official transactions.1

2.

Relative Magnitudes of Sales to Official Accounts and Retail Trading
Table VI provides some indication of the magnitude of dealer

transactions with retail customers compared to dealer sales to official
accounts during rights financings from March 1961 through July 1964.
Dealer sales to retail customers of bonds maturing after 10 years averaged
$28 million a day during long financing periods while retail purchases
were $15 million.

The actual daily highs, of course, were much larger,

$108 million for sales and $8l million for purchases.

In the 5-10 year

maturity range retail sales averaged $57 million and purchases averaged
$38 million, with daily peaks of $196 million and $145 million,respectively.
Trading in 1-5 year issues was still heavier, with both retail sales and
purchases averaging slightly over $90 million a day in long financings and
with the record highs in such periods reaching $334 million and $445 million,
respectively.

Daily average retail sales and purchases were, of course,

much larger in all maturity classes in the short financing periods, although
the actual daily highs were unchanged.
Average dealer sales to official accounts were not large during
long financing periods, equaling $6 million for securities maturing after
10 years,$4 million for securities maturing in 5~10 years, and $6 million
for 1-5 year issues.

Even in short financing periods, average daily

sales to official accounts were moderate.

These averages, however, were

pulled down by the many days on which there were no sales to official
accounts (see Table V).

On certain days such sales were substantial,

1 This problem was even greater in the case of cash financings.
There were 43 days that fell in cash financings for > 1 ^ 5 year issues
from March 196l-June 1963* and sales to official accounts were made on
15 days. In the 5-10 year class, there were 4 such days out of 42; in the

over-10 year class, only 2 out of 14.


^5
Table VI
Daily Average, High and Low Levels of Dealer Trading with Retail
Customers and Dealer Sales to Official Accounts During Treasury Rights Financings*
(in millions of dollars)

Daily Trading
I.

Securities Maturing in
> 1 ^ 5 yrs. > 5 ^ 1 0 yrs.
> 1 0 yrs.

Long Financing Period (From Day After Announcement Through Payment Date)
Average
Dealer Sales to Official Accounts
Dealer Sales to Retail Customers
Dealer Purchases from Retail Customers

5-9
92.0
91.2

4.0
57-0
38*3

6,0
27,5
15-4

Dealer Sales to Official Accounts
Dealer Sales to Retail Customers
Dealer Purchases from Retail Customers

95-6
334.4
445.3

40.2
196.1
145.4

75.0
108.1
81.4

Dealer Sales to Official Accounts
Dealer Sales to Retail Customers
Dealer Purchases from Retail Customers

0.0
19-7
14.3

0.0
5-j+
3-7

0.0
2.4
1.6

High

Low

II.

Short Financing Period (From Day After Announcement Through Day Books Closed)
Average
Dealer Sales to Official Accounts
Dealer Sales to Retail Customers
Dealer Purchases from Retail Customers

2.1
119.4
124.2

4,5
79.5
49,3

12,5
45.4
23,0

58.0
334,4
445.3

40*2
196.1
145.4

75.0
108,1
8l.4

HijSh
Dealer Sales to Official Accounts
Dealer Sales to Retail Customers
Dealer Purchases from Retail Customers
Low
Dealer Sales to Official Accounts
Dealer Sales to Retail Customers
Dealer Purchases from Retail Customers

*

0.0
35 - 3
32.1

0.0
5-6
4.5

0.0
11.6
1.6

Includes advance refundings and rights exchanges for jNSarch 1961 through July 1964.




46

For example, the record high in long financings was $75 million for securities
maturing after 10 years, $40 million for 5-10 year issues and $96 million
for 1-5 year securities.
As Table VI makes clear, average dealer sales to official accounts
and the highest dealer sales to official accounts were much larger relative
to retail sales and purchases in the longest maturity class than in the
other two classes.

For example^ average sales to official accounts of

securities maturing after 10 years were 22 per cent of average daily sales
to retail customers and 39 per cent of retail purchases in long financings.
In the 5-10 year maturity class sales to official accounts represented only
7 per cent of retail sales and 11 per cent 6f retail purchases, while in
the 1-5 year class sales to official accounts averaged just over 6 per cent
of both retail sales and retail purchases.
were similar.

Results for short financings

The relative frequency of heavy sales to official accounts

was also much greater for longer-term securities, as Table I showed.
These comparisons raise the possibility that an impact of official transactions in trading would be more likely to be found in the longest maturity
clelss.

3.

Hypotheses
As was discussed earlier, the major hypotheses about the impact

of sales of coupon issues to official accounts on retail trading were
derived from the dealers1 criticisms.

According to these criticisms,

dealer sales to retail customers were lower than otherwise when official
accounts were buying because price levels were artificially high.

At such

times, dealer purchases from retail customers were higher than otherwise,
particularly if prices were expected to fall in the future, since purchases
by official accounts offered investors a chance to sell at relatively
 high prices.


vr
The dealers, however, did not apply their criticisms specifically
to sales to official accounts during Treasury financings, and it seems
probable that the impact of official purchases on retail sales may have
been different at such times.

Purchases by official accounts during

financings probably made the Treasury's offerings look more attractive and
thus may have led to larger dealer sales to retail customers.

The potential

market at such times probably consisted of a larger number of less sophisticated investors who would not be aware of official purchases and would
be favorably impressed by strong prices.

Even if investors were aware of

official buying, they might be encouraged to buy themselves by the thought
that the Treasury was going to keep prices from falling.
It may,also be argued that the causal mechanism underlying any
relationships between retail trading and official operations ran from
retail trading to official operations as well as from official operations
to retail trading.

In this case, there should have been negative correla-

tion between retail sales and sales to official accounts and a positive
correlation between retail purchases and sales to official accounts.

In

other words, official accounts bought when the dealers1 purchases from
retail customers were high or sales to retail customers were low, because
at such times securities were available.

An even more important reason

for such relationships during financings was that high retail purchases
or low retail sales may have led to price weakness which the Treasury
especially wanted to avoid at such times.
Thus, the hypotheses to be tested were as follows:
(l) Daily retail purchases during Treasury financings were
higher when official accounts were buying, because retail customers took
advantage of these opportunities to sell securities at relatively attractive



48

prices and/or because official accounts were buying securities purchased
by the dealers from retail customers in order to maintain a favorable
atmosphere for the financing.
(2) Daily retail sales were lower when official accounts were
buying, because retail customers were discouraged from buying by artificially high prices or because official accounts were buying securities
which dealers could not sell in order to maintain a favorable atmosphere.
(3) Daily retail sales were higher when official accounts were
buying because retail customers were encouraged to purchase securities
by the strong price performance and the implied attractiveness of the
financing.
These hypotheses were tested by simple correlations and multiple
regressions.

In both cases three variables were included as measures of

official transactions.

They were dealer sales to official accounts of

securities in the given maturity class on the current day (DS°^), dealer
sales to official accounts on the previous day (DS0-^!), and the cumulative
volume of dealer sales to official accounts from the first day during the
financing to the current day (2DS°t).

The latter two variables were

included on the possibility that the influence of official accounts was
felt with a lag.

In addition, if the lagged variables prove to be signifi-

cant, it suggests that the line of causation ran from sales to official
accounts to retail trading rather than in the other direction, since
official transactions cannot be influenced by retail trading on days in
the future.

4.

Simple Correlati on
Table VII shows the simple correlation coefficients between

retail trading and dealer sales to official accounts.



For securities

49
Table VII
Simple Correlation Coefficients between Retail Trading and
Official Purchases of Coupon Issues During Treasury Rights Financings
(March 1961-July 1964)

Retail Trading on Current Day
X.

Dealer Sales to Official Accounts
Cumulated for
Current
Previous
financing
day
day
to date

Long Financing Period
Dealer Purchases from Retail Customers
of Securities Maturing in
> 1 ^ 5 years
> 5 "^.10 years
> 1 0 years

.041

.086

.121

.018

.470**

.324**

.183*
.027
.192

Dealer Sales to Retail Customers
of Securities Maturing in
> 1 ^ 5 years
> 5 ^s10 years
> 1 0 years
II.

.049
• 199*
.336**

.025
.080

.296**

.145
.020
.064

Short Financing Pferiod
Dealer Purchases from Retail Customers
of Securities Maturing in
> 1
5 years
> 5
10 years
> 1 0 years

.018
.324**
.403*

.024
.256*
.364*

.026
.287*
.466**

Dealer Sales to Retail Customers
of Securities Maturing in
> 1 * ^ 5 years
> 5 "SrlO years
> 1 0 years

*
**

Significant at 5 per cent level.
Significant at 1 per cent level.




.091
.335**

.168

.178
.245*

.282

.183
.370**
.310

50
maturing after 10 years there was significant positive correlation "between
dealer sales to official accounts on the current day and dealer purchases
from retail customers in both long and short financing periods.

In long

financing periods there was also significant positive correlation in this
maturity class between sales to official accounts and sales to retail
customers.
These relationships are illustrated in Charts I and II.

Chart I

is a scatter diagram relating daily retail sales to daily sales to official
accounts of securities maturing after 10 years.

Plottings for the day

after the announcement day through the day the books closed are stars;
plottings for days after the books closed through the payment day are dots.
Chart II is a scatter diagram relating daily retail purchases to daily
sales to official accounts.

When all the plottings are considered (i.e.,

in long financings) there seems to have been some tendency for both retail
sales and retail purchases to be higher on days when official accounts
were large buyers.

The simple correlation coefficients of .336 between

retail sales and official sales and of .470 between retail purchases and
official sales for long financings confirmed this impression.

When only

the stars (short financings) were studied, however, this relationship
vanished in the case of retail sales.
In contrast to longer term securities, the positive correlation
between official purchases of 5-10 year securities on the current day and
retail trading was more consistent in short financings.

In the latter

financings both retail purchases and retail sales showed significant
positive correlation with sales to official accounts on the current day.
In long financings positive correlation showed up only with retail sales.




Chart I

SCATTER DIAGRAM. RETAIL SALES AND SALES TO OFFICIAL ACCOUNTS
OF GOVERNMENT SECURITIES MATURING AFTER TEN YEARS DURING FINANCINGS
Retail sales

MilliHS of dollars-

100

Key:
x - D a y after the anioattceient threagh day the beaks dosed.
• - D a y after heeks closed threagh payieat day.

60

50
40
1?

30

<>

•:
a

20
•

•

10
•

•

1

0

10

20

30

40

50

Millions of dollars (Sales to Official Accounts)

60

M t i : l i s t ! n tailf lata fir Traasirj ri(kts fiiaiciiis t u n March 1911-ill; 1SS4.
Sam lira talus if retail salts w i n p l i t t i i t i t i l lift if tki i i r i lilt kitaist if lack if spaci.



70

Chart II

SCATTER DIAGRAM. RETAIL PURCHASES AND SALES TO OFFICIAL ACCOUNTS
OF GOVERNMENT SECURITIES MATURING AFTER TEN YEARS DURING FINANCINGS
Retail purchases
Millieas of hilars

Key:
i
x—Day after the a i N i i c e i e a t tbreifh day the heeks c osed.

100

• - D a y after heeks closed thronfh payieat l a y .

90
80

70
60

SO

40

0

10

20

30

40

50

Millions of dollars (Sales to Official Accounts)
far Triasirj rifkts liiaiciiis f r u March I S Jl 1964.
S I if

60

data: lasii 11 daily lata

Sana m a Mlaas af ratail pirclisis vara plallai ta tka laft af tka w a iiaa kacusa af lack af spaca.


70

51
In most cases in the 5-10 year and over-10 year maturity classes
where sales to official accounts on the current day were positively correlated with retail trading, sales to official accounts on the following day
were also positively correlated with retail trading.

In short financings

positive correlation also showed up between cumulated sales to official
accounts and retail trading.

In fact, the correlation coefficient for

cumulated sales to official accounts was usually the highest of the three
for each maturity class.
Table VII also indicates that there was no significant correlation between sales to official accounts on the current or previous day
and retail trading in 1-5 year securities.

It is possible, however, that

some relationship was obscured by large variations in trading caused by
technical factors, such as the size of issues in the financing or whether
the books were open.

On the other hand, the simple correlations indicated

in Table VII for longer securities may really have reflected these technical
factors.

In order to hold such technical factors constant, multiple

regressions were calculated.

5.

Multiple Regre s s i ons
a.

Variables Included.

The dependent variables in the multiple

regressions were dealer sales to retail customers of securities maturing
in over one through five years, over five through ten years, and over
ten years.

Dealer purchases from retail customers of securities in the

same three maturity classes were also used as dependent variables.
When the financing period was defined as lasting through the payment date, nine independent variables were included to take account of certain
technical differences among days within financings and between financings.




52

For example, trading during rights financings was usually higher when the
hooks were open.

Therefore, one variable (A) was a dummy variable equal

to +1 on the days the books were open and zero otherwise.

Alternatively,

another dummy variable (B), which equaled +1 from the day after the
announcement day through the day the books closed and zero otherwise,
was included.

A second technical consideration which might have influenced

the volume of trading is whether the financing was an advance refunding
or a rights exchange, so variable C was equal to +1 in advance refundings
and zero otherwise.

Another set of variables took account of the size

and type of issues in the given maturity class that were involved in the
financings.

Thus, variable D measured the size of rights in the maturity

class held by the public and was the same for each day during the financings;
and variable E equaled the size of the allotments to the public of new
issues in the maturity class and was also the same value throughout each
-financing.

Variable F was +1 on the payment date, and variable G was +1

on the day before the payment date, on the assumption that dealer sales and
purchases might have been higher on these days as customers prepared to
make payment.

Variable H was zero from the announcement day through the

day the books closed and thereafter increased by one on each day, reflecting
a tendency for trading to fall gradually to normal levels after the books
close.

Variable I was constant throughout each financing and equaled the

number of days from .the day after the announcement day through the day
the books closed, since daily trading might have been lower if investors
had a longer time to act before the books closed.
When regressions were calculated for the short financing period
(i.e., from the day after the announcement date through the day the books
closed) five of the above independent structural variables were included,
namely A, C, D, E, and I.




53
The independent variables included in these regressions were not
necessarily the only determinants of the daily volume of retail trading.
For example, the expectational element was omitted, except to the extent
that it was reflected in the size of public allotments of the new issues.
Dealers1positions and price behavior were also not considered, although
they probably had an influence on trading.
b.

Results for Long Financing Periods.

When technical factors,

as well as sales to official accounts were taken into account, a "significant"
positive relationship showed up between some measure of sales to official
accounts and retail sales and between sales to official accounts and retail
purchases in both the 5-10 year and over-10 year maturity classes during
rights financings.

There was still no significant relationship, however,

between official purchases and retail trading in the 1-5 year maturities.
Table VIII shows the net regression coefficient and the corresponding
values of Student's t for each of the independent variables for the long
financing periods.
In the regression explaining dealer purchases from retail customers of securities maturing in more than 10 years, dealer sales to
official accounts on the current day (DS0^) was the significant measure of
official activity.

An increase in dealer sales to official accounts of

$1 million of bonds maturing after 10 years was associated with an increase
of $211 thousand in dealer purchases from retail customers, when the other
variables were held constant.

This independent variable was the first one

added to the equation^ and by itself explained 21 per cent of the variation
in retail purchases.

The other measures of activity by official accounts

In the program used, the variable reducing variance the most
was added to the equation first.



Table VTII
Results of Multiple Regressions for Long Financing Periods
Dependent Variables
Dealer Purchases from Retail Customers
Dealer Sales to Retail Customers
of Securities Maturing in
of Securities Maturing in
> 10 yrs.
> 1 0 yrs.
>
5 yrs.
> 5
10 yrs.
5 ^ 1 0 yrs.
1 ~ 5 yrs.
t
t
b
b
b
t
b
t
b
t
b
t

Independent
variables
DS°t

.172

.583

DS

°t-l

.015

.046

DS°

.02 4

.261

A

14.620

1.281

B

^

U
r\

.112
.268

1.004

.071

.703

.188* 2.347

-

.446

.032

1.343

.098

4.544

38.168* 2.536

20.002* 2.737

7-112

1.340

.708

-12.666* 2.613

15.452

I.636

.000
.004

-

-

.005* 4.036

E

-

.010* 3.088

.001

• 739

.002* 2.014

-

.826

.068

•395

• 293

.815

26.839* 2.606

1.544

D

• 975

.078
.241

.125

-10.7^3

.605

.211* 2.072

-

1.112

.228

.604

.352* 3.103
.214

- 1.864

46.189* 3-400

46.941* 4.545

.525

.038

-20.825* 3.039

.001

1.074

-

-

.034

.261

.271* 2.119
.003

.0y2

- 8.021

1-353

-

32.969* 4.88p
61.567*

5.126

-

.005* 3-937

.007* 4.487

.013* 4.558

.008* 2.767

-

.009* 3.156

F

22.088

1.246

- 8.912

1.043

3.501

.609

7.955

.497

-12.881

1.066

8.658

1.184

G

31.025

1.850

- 5-644

.709

1.184

.217

20.306

1.342

- 2.928

.260

4.233

.611

H

- 3.645

1.494

-

.247

.588

.828

-

• 273

- 1.243

1-375

I

.736

.111

- 3.132

1.195

R 2 adj. for DOF
DOF

.301

4.655* 1-979

-

.848

.385

- 6.980

1.171

.469

6.812* 2.049

-15•210* 4.562

.382

.197

.325

.402

.436

•571

153

145

79

153

145

79

Note: All dollar variables were measured in millions of dollars.
* Significantly different from zero at 5 per cent level.




-

55
were not significant, in part because of multicollinearity with sales to
official accounts on the current day.

Dealer sales to retail customers

of securities maturing after 10 years were significantly related to dealer
sales to official accounts on the previous day.

A rise of $1 million in

sales to official accounts on the previous day was accompanied by a rise
of $271 thousand in retail sales on the current day.
In the 5-10 year maturity category, the measure of official
activity which was significantly related to retail sales and purchases
was the volume of sales to official accounts cumulated from the beginning
of the financing through the given day.

Holding the other variables

constant, dealer sales to retail customers were $352 thousand higher for
every $1 million increase in sales to official accounts, while dealer
purchases from retail customers were $188 thousand higher.

As is clear

from comparing simple correlation coefficients in Table VII with Table VIII,
this positive relationship between " E DS° and retail trading depends on
S
talking account of technical factors, particularly the fact that average
trading at the beginning of the financing period is much higher than at
the end, which in itself would lead to a negative relationship between
cumulated sales to official accounts and retail sales or purchases.

When

this was allowed for (through variable B), dealer sales and purchases with
retail customers rose as the cumulated volume of official purchases increased.
These results are consistent with the hypothesis that retail
purchases were higher when official accounts were buying.

Moreover, the

fact that retail purchases were positively correlated with the cumulated
values of sales to official accounts, as well as with the current day's
sales, suggests that large sales to official accounts caused the high level
of retail purchases.



The results are also consistent with the hypothesis

56

that retail sales were higher when sales to official accounts were high
because investors were favorably impressed by price performance.

They are

not consistent with the hypothesis suggesting negative correlation between
retail sales and sales to official accounts.
Although this study was mainly concerned with the relationship
between official transactions and retail trading, a few comments on the
relationships between the technical variables and retail trading follow.
Variable B, a dummy variable which was +1 for days from the day after the
announcement day through the day the books closed was positively related
to retail purchases and retail sales at the 5 per cent significance level
in five out of the six regressions.

Such a relationship is evident just

from inspection of the daily figures, since trading was generally higher
before the books closed than afterwards.
The signs of other technical variables that were significant
differed among maturity classes.

Of particular interest, the coefficient

for C, the dummy variable that was +1 in advance refundings and zero in
rights exchanges, was negative for the 5-10 year maturity range.

In other

words, daily retail sales of securities maturing in 5-10 years were
$21 million, lower in advance refundings than in rights exchanges, and
daily retail purchases were $13 million lower, when other variables
(in particular the size of the new issues in the maturity class) were held
constant.

In this maturity range, the size of the new issues alloted to

the public were positively correlated with retail trading.

This result

suggests that advance refundings led to less, rather than more, daily
activity in the retail market in 5-10 year securities than when an equivalent amount of new securities was sold through an exchange offering."*"
1 In contrast, the coefficient for C, the advance refunding dummy,
was positive in the case of retail sales of securities maturing after 10 years,
but there was only one of the financings in this maturity class which was an

exchange in the sample so that this result is not very convincing.


57
Possibly this reflected the fact that there was usually also a longer new
issue in which to trade in advance refundings but in exchanges the 5-10
year issue was the longest option*
In the regression for securities maturing after 10 years, the
negative coefficient for E, which measures the size of the new issues
sold to the public, was unexpected.

Possibly it can be explained by the

fact that the smaller public allotments of new long-term bonds occurred
in those advance refundings and exchanges where the rights were very
short-term.
market,

Presumably these holders of rights sold them in the secondary

and the dealers then sold new issues to other investors, thus

leading to greater activity in the secondary market when public allotment
of new issues were small.
c.

Results for Short Financing Periods.

Table IX shows the

regression coefficients and the associated values of Student's t for those
variables that were significant at the 5 per cent level in the multiple
regressions for the short financing periods.

Although the other variables

mentioned on page 52 were tried in the regressions, the results in the
table are for equations including only the variables listed in each column*
This procedure was adopted because of the small size of the samples and
because adding other variables frequently resulted in insignificant regression coefficients for all variables.
In the short financing period, retail sales of securities maturing
in 5-10 years and after 10 years were positively related to the cumulated
volume of sales to official accounts at the accepted significance level.
Retail sales in the 1-5 year and over-10 year maturity classes, however,
were also negatively related to the volume of sales to official accounts
on the current day, when the seven other independent variables were taken
into consideration.




Table IX
Results of Multiple Regressions for Short Financing Periods

Independent
variables

Dependent Variables
Deiiler Purchases from Retail Customers
Dealer Sales to Retail Customers
of Securities Maturing in
of Securities Maturing in
> 10 yrs.
> 10 yrs.
> 5 ^ 10 yrs.
> 1 < 5 yrs.
^>5 ^ 1 0 yrs.
> 1 ^ 5 yrs.
t
b
b
t
b
t
b
t
b
t
t
b_
1.030* 2.759

- 4.842* 2.072

-

^717* 2.702

- 2.548

l.l6l

-

.309

1,129

2.239

1.617

.463*

3.136

A

22.545

1.323

C

53.385* 2.086

u

.169*" 3*587

D

.007*

5*377

E

.017*

3.895

.003* 2.215

DOF

-13 A O ?

.259

.091

.273

75

65

36

Note: All dollar variables were measured in millions of dollars,
* Significantly different from zero at 5 per cent level.




.001

4.173

- 9.351
-24.731*

2.238

390

1.291

HI.750*

5.302

.014* 6.177

1.681

.3 6k

• J+37

69

63

.011* 4.607

-

.018*

3,222

-31.407*

•397

.019*

I

R^ adj. for DOF

o
n
1.066*

5.131

.U68
30

00

59
Retail purchases of securities maturing after 10 years were positively correlated with the cumulated volume of sales to official accounts,
while retail purchases in the 5-10 year class were positively correlated
with the current day's sales to official accounts.

No significant corre-

lation between sales to official accounts and retail purchases existed in
the 1-5 year maturity class.
The relationships (or lack of relationship) between retail sales
of 1-5 year securities (or retail purchases) and sales to official accounts
are questionable and should not be accepted without further confirmation,
because there were only 6 days out of 78 when sales to official accounts
exceeded $10 million.

Even in the other maturity classes it would have

been desirable to have more days with sizable sales to official accounts.
In the 5-10 year maturity class there were 10 days out of 67 when sales to
official accounts were greater than $10 million, and in the over-10 year
class there were 11 such days out of 39-

(See Table V. )

It should also be pointed out that the variables included in
these regressions did not explain much of the variation in trading either
in short financing periods or long financing periods.

For example, only

57 per cent and kj per cent of the variation in daily retail sales of
securities maturing after 10 years in long financings and short financings,
respectively, was explained.

In all other cases the explained variance

was lower.
Chart III illustrates this point by plotting the actual level of
daily retail sales of securities maturing after 10 years against the level
calculated by the regression equation for short financing periods.

If there

were no sales to official accounts during a financing, the calculated level
of retail sales would drop when the books opened and be constant thereafter, as in the September 19&1 and



July 196^ advance refundings.

60

Any other day-to-day movement in the calculated level of retail sales during
a given financing was a result of sales to official accounts.

The average

level of calculated retail sales during any financing (which depended
largely on the size of public allotments of new issues and the advance
refunding dummy) was generally close to the actual.

In addition, the

actual day-to-day variation in retail sales was similar to the variation
in calculated retail sales in some of the financings—see, for example, the
financings in February 1962, February 1963 and September 1963*

Of course,

there was also considerable day-to-day movement in retail sales during
financings that was not explained.

6.

Conclusions
(l) Ihe results of the multiple regressions for 5-10 year and

over-10 year Government securities were consistent with the hypothesis
that the dealers ? daily purchases from retail customers during Treasury
financings were larger than otherwise when official accounts were buying
such issues.

Both the current day's sales to official accounts and the

cumulated volume of sales to official accounts for the financing to date
were positively related to retail purchases, although because of multicollinearity only one of these variables was significant in any multiple
regression.

Hiis correlation may be explained both by the behavior of

customers and by the behavior of official accounts:

customers took

advantage of the opportunity to sell at artificially high prices created
by official operations, and official accounts bought because the dealers
had taken on a large volume of securities from retail customers which
would depress the market if not purchased.

The fact that the positive

correlation was between retail purchases and both the current day's sales
to official accounts and the cumulated volume of sales to official accounts



Chart III
A C T U A L A N D C A L C U L A T E D R E T A I L SALES OF G O V E R N M E N T S E C U R I T I E S M A T U R I N G A F T E R 10 Y E A R S D U R I N G SHORT F I N A N C I N G P E R I O D S




61

suggests that the former interpretation may have been the dominant one,
since retail activity on the current day could not have affected sales
to official accounts on earlier days in any financing.

(Obviously the

cumulated volume of sales to official accounts is heavily weighted by
sales on earlier days.)
(2)

In contrast, there was no significant correlation between

retail purchases of 1-5 year Governments and dealer sales to official
accounts.

Perhaps this reflected the relatively low level of sales to

official accounts of 1-5 year securities.
(3) Daily sales to retail customers of Governments maturing
in 5-10 years were positively related to the cumulated (for the financing
to date) volume of sales to official accounts in both long and short
financing periods, when certain other variables were held constant.
(k) The dealers1 daily sales to retail customers of Governments maturing after 10 years were positively related to sales to official
accounts on the previous day in long financing periods and to the cumulated
volume (for the financing to date) of sales to official accounts in short
financing periods, when other variables were held constant.

In short

financing periods retail sales were also negatively related to the current
day's sales to official accounts, again holding constant the other" independent variables.

The positive correlation of retail sales of long

bonds with the cumulated volume of sales to official accounts during short
financing periods suggests that official buying did stimulate greater
retail sales probably by maintaining a favorable atmosphere and by preventing or reducing price declines while the books were open.

In view of

this positive correlation between retail sales and cumulated sales to
official accounts, the negative correlation between retail sales and the



62

current day's sales to official accounts should probably be interpreted
as an indication that official accounts bought when retail sales were low
in order to prevent an unfavorable atmosphere from developing rather than
as an indication that sales to official accounts discouraged sales to
retail customers.
(5) Retail sales of Governments maturing in 1-5 years were also
negatively correlated with the current day's sales to official accounts,
when other variables were held constant.

These results were somewhat

dubious, however, because of the small number of days when official accounts
were buying.
While these statistical results did suggest that official purchases of coupon securities (particularly longer-term securities) had an
influence on retail trading, a number of precautionary comments are in
order.

First, although according to the t test, the regression coefficients

discussed above were significantly different from zero at the 5 per cent
level, the t values were not especially high*

Second, official sales

accounted for only a small part of the variation in retail trading.

Third,

more financings, particularly financing with heavy official operations,
would be desirable in order to test the relationships further.




66
Appendix, Table I
Comparison of Actual and, theoretical Frsquisnci.es« Days
Classified According to the Current Day's Sales to Official
Accounts and Retail Purchases of Securities Maturing in 1-5 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

Betail Purchases
(in millions of dollars)
50.0
>50.0
Total

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent
Days When Sales to SOMA. + Treasury Were
0 Theoretical
> 0 Numbers

2U2
100

Days When Sales to SOMA. Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers

80
28

179
63

79
29

179
63

117
17

95
13

212
30

92
l6

212
30

115
19

16

92

207
35

121

20 Actual.
20 Numbers

99
35

120
14

•20 Theoretical
•20 Numbers

86

13

~ 20 Theoretical
> 20 Numbers

22

121
13

97
11

126

^ 2 0 Actual
> 2 0 Numbers

92

4.0745
207
35

218
24

218

8

16

117
17

95
13

212
30

113
21

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers

Note:

108
44.6

100
34

0 Actual
> 0 Numbers

0 Actual
> 0 Numbers

134
55.4

99
9

212
30

2k

Treasury ^ 20, > 20 not calculated because of insufficient frequencies of > 2 0 .


* If significant


at 5 per cent level. Calculated using Yates' correction.

64
Appendix, Table II
Comparison of Actml and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official
Accounts and Retail Purchases of Securities Maturing in 5-10 Years
(August 22, 1962-Deceraber 31> 1963)

Sales to Official Accounts
(in millions of dollars)

Retail Purchases
(in millions of dollars)
0-30.0
>30.0
Total

Chi-Square*

(in number of days)
All Days Without Financings
Number
Bar cent

141
53-2

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

265

124
46.8

100

103
38

91
33

194
71

0 Actual
> 0 Numbers

114
27

80
44

194
71

^ 10 Theoretical
>• 10 numbers

118
23

104
20

222
43

tSlO Actual
> 1 0 Numbers

131
10

91
33

222
43

8.5241

17.4326

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers

122
19

107
17

229
36

229

0 Actual
> 0 Numbers

132
9

97
27

— 10 Theoretical
>»10 Numbers

126
15

110
14

236
29

< 10 Actual
> 1 0 Numbers

136
5

100
24

236
29

118
23

104
20

222

121
20

101
23

11.6421

222
^3

36

13-9999

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:

Treasury -^.10, > 1 0 not calculated because of insufficient frequencies of ">10.

Digitized * FRASER
for If significant


at 5 per cent level.

Calculated using Yates' correction.

65
Appendix, Table III
Comparison of Actual and Theoretical. Frequencies, Days
Classified According to the Current Day's Sales to Official
Accounts and Retail Purchases of Securities Maturing After 10 Years
(August 22, 1962-December 31> 1963)

Sales to Official Accounts
(in millions of dollars)

Retail Purchases
(in millions of dollars)
^ 10
Total
>10

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent

217
69.8

30.2

311
100

Days When Sales to SOMA. + Treasury Were
0 Theoretical
> 0 Numbers

15U
63

66
28

220
91

^ 0 Actual
> 0 Numbers

155
62

65
29

220
91

^ 5 Theoretical
> 5 Numbers

194
23

84
10

278
33

- 5 Actual
;>5 Numbers

200
17

78
16

278
33

201
16

87
7

288
23

198
19

90
4

288
23

163
54

71
23

234
77

^ 0 Actual
^>0 Numbers

168
49

66
28

234
77

Theoretical
> 5 Numbers

197
20

85
9

282
29

$ 5 Actual
5 Numbers

203
14

79
15

282
29

I+.8562

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers
S 0 Actual
> 0 Numbers

Days When Sales to Treasury Were
<•0 Theoretical
> 0 Numbers

5.3831

Note:

SOMA.

> 5 not calculated because of insufficient frequencies for > 5 *


* If
http://fraser.stlouisfed.org/ significant
Federal Reserve Bank of St. Louis

at 5 per cent level.

Calculated using Yates1 correction.

66
Appendix, Table IV
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official Accounts
and Change in Gross Long Positions for Securities Maturing in 1 - 5 Years
(August 22, 1962-December 31> 1963)

Sales to Official Accounts
(in millions of dollars)

Change in Long Position
(in millions of dollars)
Total
>0
= 0

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent

137
56.6

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

105
43.4

242
100

101
36

78
27

179
63

89
48

90
15

179
63

^ 20 Theoretical
> 20 Numbers

120
17

92
13

212
30

§ 20 Actual
> 2 0 Numbers

109
28

103
2

212
30

117
20

90
15

207
35

0 Actual
> 0 Numbers

106
31

101
4

207
35

— 20 Theoretical
> 2 0 Numbers

123 '
14

95
10

218
24

— 20 Actual
>•20 Numbers

115
22

103
2

218
24

120
17

92
13

212
30

118
19

94
11

212
30

11.5766
0 Actual
•>0 Numbers

17.0833

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers

15.0298

IO.6923

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:

Treasury ^ 2 0 , > 2 0 not calculated because of insufficient frequencies of > 2 0 .


* If significant at


5 per cent level. Calculated using Yates' correction.

68
Appendix, Table V
Comparison of Actual and Theoretical Frequencies9 Days
Classified According to the Current Day's Sales to Official
Accounts and Change in Net Position for Securities Maturing in 1-5 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

Change in Net Position
(in millions of dollars)
% 0
Total
>0

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent

137
56.6

Days When Sales to SOMA. + Treasury Were
0 Theoretical
> 0 Numbers

101
36

78
27

89
48

90
15

179
63

^ 20 Theoretical
> 20 Numbers

120
17

92
13

212
30

20 Actual
> 2 0 Numbers

109
28

103
2

212
30

117
20

90
15

207
35

0 Actual
> 0 Numbers

106
31

101
k

207
35

^ 20 Theoretical
20 Numbers

123
Ik

95
10

218
2k

^ 20 Actual
> 2 0 Numbers

115
22

103
2

218
2k

120
• 17

92
13

212
30

118
19

94
11

212
30

105
kl.k

242
100

179
63
11.5766

0 Actual
> 0 Numbers

17.0833

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers

15.0298

IO.6923

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:

Treasury


* If significant


20, > 2 0 not calculated because of insufficient frequencies of > 2 0 .
at 5 per cent level. Calculated using Yates' correction.

68
Appendix, Table VI
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Dayfs Sales to Official Accounts
and Change in Gross Long Position for Securities Maturing in 5 - 1 0 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

Change in Long Position
(in millions of dollars)
<0
> 0
Total

Chi-Square*

(in number of days)
All Days Without Financings
Number
Bar cent

177
66.8

88
33.2

265
100

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

130
47

64
24

194
71

0 Actual
> 0 Numbers

123

71
17

194
71

"<10 Theoretical
> 1 0 Numbers

148
29

74
14

222
43

< 1 0 Actual
> 1 0 Numbers

142
35

80
8

222
43

153
24

76
12

229
36

0 Actual
> 0 Numbers

147
30

82
6

229
36

^ 1 0 Theoretical
> 1 0 Numbers

158
19

78
10

236
29

152
25

84
4

236
29

l48
29

74
14

222
43

147
30

75
13

222
43

Days When Sales to SOMA. Were
0 Theoretical
> 0 Numbers

4.3769

5.1964
^•10 Actual
> 1 0 Numbers

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:

Treasury ^ 10, > 1 0 not calculated because of insufficient frequencies.. Of > 1 0 .


* If significant


at 5 per cent level.

Calculated using Yates' correction.

69
Appendix, Table VII
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official
Accounts and Change in Net Position for Securities Maturing in 5-10 Years
(August 22, 1962-December 31> 1963)

Sales to Official Accounts
(in millions of dollars)

Change1 in Net Position
(in millions of dollars)
Total
> 0
^ 0

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent

172
64.9

93
35.1

265
100

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

126
46

68
25

194
71

0 Actual
> 0 Numbers

119
53

75
18

194
71

— 10 Theoretical
> 1 0 Numbers

144
28

78
15

222
43

10 Actual
> 1 0 Numbers

138
34

84
9

222
43

149
23

80
13

229
36

0 Actual
> 0 Numbers

143
29

86
7

229
36

~ 1 0 Theoretical
> 1 0 Numbers

153
19

83
10

236
29

^ 10 Actual
> 1 0 Numbers

148
24

88
5

236
29

• ! .
> •
j.TL; «
Ci
f
—

78
15

222
43

142
30

80
13

222
43

Days When Sales to SOMA. Were
0 Theoretical
> 0 Numbers

4.2232

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:

Treasury ^ 1 0 , > 1 0 not calculated because of insufficient frequencies of > 1 0 .


http://fraser.stlouisfed.org/ significant
* If
Federal Reserve Bank of St. Louis

at 5 per cent level.

Calculated using Yates' correction.

70
Appendix, Table VIII
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official Accounts
and Change in Gross Long Position for Securities Maturing After 10 Years
(August 22, 1962-December 31> 1963)

Sales to Official Accounts
(in millions of dollars)

Change in Long Position
(in millions of dollars)
Total
> 0
^ 0

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent

193
62.1

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

118
37.9

311
100

137
56

83
35

.220
91

0 Actual
> 0 Numbers

125
68

95
23

220
91

^ 5 Theoretical
> 5 Numbers

173
20

105
13

278
33

8.6989

8.0002
165
28

113
5

278
33

179
14

109
9

288
23

174
19

114
4

288
23

1*5
48

89
29

0 Actual
> 0 Numbers

137
56

97
21

234
77

~ 5 Theoretical
> 5 Numbers

175

107

282

^ 5 Actual
> 5 Numbers

Days When Sales to SOMA. Were
0 Theoretical
> 0 Numbers

M953
0 Actual
> 0 Numbers

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers

'

234
77
4.1315

.18

168

^ 5 Actual
> 5 Numbers
Note:

25

SQMA-^5> > 5


* If significant


11

114
4

29
282
29

not calculated because of insufficient frequencies of > 5 .

at 5 per cent level.

Calculated using Yates' correction.

6.8244

70
Appendix, Table IX
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official Accounts
and Change in Gross Long Position for Securities Maturing After 10 Years
(August 22, 1962-December 31> 1963)

Sales to Official Accounts
(in millions of dollars)

Change in Short Position
(in millions of dollars)
^0
> 0
Total

Chl-Square*

(in number of days)
All Days Without Financings
Number
Bar cent

150
48.2

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

106
44

114
47

220
91

0 Actual
> 0 Numbers

119
31

101
60

220
91

^ 5 Theoretibal
> 5 Numbers

134
16

144
17

278
33

~ 5 Actual
> 5 Numbers

139
11

139
22

278
33

139
11

149
12

288
23

141
9

147
14

288
23

113
37

121
'40

234
77

124
26

110
51

234
77

136
Ik

lk6
15

282

iko

lk2
19

282

161
51.8

311
100

9.7203

Days When Sales to SOMA. Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers

7.6229
0 Actual
^-0 Numbers
5 Theoretical
Numbers
^ 5 Actual
5 Numbers
Note:

10

29

29

S 0 M A ^ 5 > > 5 not calculated because of insufficient frequencies of


http://fraser.stlouisfed.org/
* If significant
Federal Reserve Bank of St. Louis

at 5 per cent level.

Calculated using Yates1 correction.

72
Appendix, Table X
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official
Accounts and Change In Net Position for Securities Maturing After 10 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

Change in Net Position
(in millions of dollars)
<= 0
> 0
Total

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per cent

191
6l.4

Days When Sales to SOMA + Treasury Were
0 Theoretical
^ > 0 Numbers

120

38.6

311
100

135
56

85
35

220
91

0 Actual
> 0 Numbers

117
74

103
17

220
91

^ 5 Theoretical
> 5 Numbers

171
20

107
13

278
33

5 Actual
> 5 Numbers

161
30

117
3

278
33

111

288

9

23

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers

5 Theoretical
> 5 Numbers
5 Actual
Numbers
Note:

SOMA'S 5,




288

171
20

117
3

144
47

90
30
105
15

234
77

173
18

109
11

282
29

164
27

118
2

282
29

12.8261

5.9652

234
77

62

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers

177
14

20.0902

129

23

1.5224

not calculated because of insufficient frequencies of

75
Appendix, Table XI
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official Accounts
and Offerings to the Trading Desk of Securities Maturing After 10 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

uiienngs
(in millions of dollars)
Total
>7?

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per Cent

l4i
58.3

242
100

75
26

104
37

179
63

0 Actual
> 0 Numbers

82
19

97
44

179
63

= 20 Theoretical
> 2 0 Numbers

88
13

124
17

212
30

99
2

113
28

212
30

86
15

121
20

207
35

0 Actual
> 0 Numbers

94
7

113
28

207
35

~ 20 Theoretical
> 20 Numbers

91
10

127
14

218
24

100
1

118
23

218
24

88
13

124
17

212
30

89
13

123
17

212
3C

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

101
41.7

17.1080
= 20 Actual
> 2 0 Numbers

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers

7.6815

13.7486
tS 20 Actual
> 2 0 Numbers

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:

Treasury ^


http://fraser.stlouisfed.org/ significant
* If
Federal Reserve Bank of St. Louis

20, > 2 0 not calculated because of insufficient frequencies of > 2 0 .
at 5 per cent level.

Calculated using Yates1 correction.

Appendix, Table XII
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official Accounts
and Offerings to the Trading Desk of Securities Maturing in 5-10 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

uiienngs
(in millions of dollars)
^50

>50

Total

Chi-Square*

(in number of days)
All Days Without Financings
Number
Per Cent

143
54.0

Days When Sales to SOMA + Treasury Were
0 Theoretical
> 0 Numbers

105
38

89
33

194
71

0 Actual
> 0 Numbers

114
29

80
42

19^
71

< 10 Theoretical
> 1 0 Numbers

120
23

102
20

222

< 10 Actual
> 1 0 Numbers

136
7

86
36

222

124
19

105
17

229
36

122
46.0

265
100

5.5906

26.8157

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers

11.6461
0 Actual
> 0 Numbers

13^
9

95
27

229
36

^ 10 Theoretical
> 10 Numbers

127
16

109
13

236
29

* 10 Actual
> 1 0 Numbers

139
4

97
25

236
29

120
23

102
20

222
43

120
23

102
20

222
^3

20.6933

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers
Note:
*

Treasury^ 10, > 1 0 not calculated because of insufficient frequencies of

If significant at 5 per cent level.




Calculated using Yates1 correction.

>10.

75
Appendix, Table XIII
Comparison of Actual and Theoretical Frequencies, Days
Classified According to the Current Day's Sales to Official Accounts
and Offerings to the Trading Desk of Securities Maturing After 10 Years
(August 22, 1962-December 31, 1963)

Sales to Official Accounts
(in millions of dollars)

Offerings
(in millions of dollars)
^ 25
>25
Total

Chl-Square*

(in number of days)
All Days Without Financings
Number
Bar cent

225
72.3

86
27.7

311
100

Days When Sales to SOMA. + Treasury Were
0 Theoretical
> 0 Numbers

159
66

61
25

220
91

171
54

k9
37

220
91

201
24

77
9

278
33

215
10

63
23

278
33

208
17

80
6

288
23

208
17

80
6

288
23

169
56

65
21

234
77

182

52
34

234
77

5 Theoretical
]> 5 Numbers

204
21

78
8

282
29

^
>

217
8

65
21

282
29

10.2936
>

0 Actual
0 Numbers

< 5 Theoretical
> 5 Numbers

31.1174
< 5 Actual
> 5 Numbers

Days When Sales to SOMA Were
0 Theoretical
> 0 Numbers
0 Actual
> 0 Numbers

Days When Sales to Treasury Were
0 Theoretical
> 0 Numbers

13.5591
0 Actual
> 0 Numbers

29.7409

Note:
*

5 Actual
5 Numbers
S0MA<- 5» >

5 not calculated becauses of insufficient frequencies of > 5 -

If significant at 5 per cent level.




Calculated using Yates1 correction.

Appendix, Table XIV
Results of Multiple Regressions Explaining Dealer Sales to Private
Customers on Days Without Financings, August 22, 1962-Pecember 31, 1963

Treasury^
Simple r
1-5
5-10
>10

6
6
3

F Level, When Added
1-5
1.315
5-10
.196
>10
2.810

.081
.033
- .061

.061
- -115
.001

.062
.008
.121*

5
4
4

3
5
2

1.007
1.274
2.481

1.673
• 772
2.584

Adjusted for Degrees of Freedom, When Added
.126
1-5
.126
.125
5-10
.072
.077
.077
>10
.056
.066
.061
Net Regression
1-5
5-10
>10

Independent Variables
Duration of sales
'
S0MA t _ 1
to SOMA + Treasury

7
3
5

- .009
- .109
.159**

Order Added
1-5
5-10
>10

Dealer Sales to
S0MAt
Treasury

1.291
• 512
3.622
.121
• 075
.051

Coefficient at Final Stage
.148
.613
.349
.128
- .126
.094
- .452
- .118
.112

.218
.113
.191*
Standard Error of Regression Coefficient at Final Stage
.148
.336
.147
1-5
-342
.158
.162
5-10
.231
•233
.068
.092
.274
>10
.058
Partial Correlation Coefficient at Final Stage
1-5
- .067
.065
.119
.097
- .034
.051
.044
5-10
- .026
- .094
>10
.111
- .099
.119
At Final Stage:
1-5
5-10
>10

Degrees of freedom
233
256
302

Note: Dollar variables in millions; price changes in points.
* Significant at 5 per cent level.
** Significant at 1 per cent level.



- .060
- .170**
.017
4
2
6

Net
positiooj.^

A Prices^

.323**
.228**
.214**

.137*
.042
- .020

1
1
1

2
8
8

£ Prices.,.,
.036
.006
.000
8
7
7
0.112
.192
.016

1.258
9-045**
.383

28.027**
14.386**
14.802**

6.310*
.154
.002

.122
.077
.064

.101
.048
.043

.120
.066
.058

.123
.069
.061

108.115
6.629
.231

18.475
-8.785
.501

55.525
16.867
4.802

55.307
16.858
4.843

-2.777
-2.641*
.176

.085**
.033**
.016**

1.764
1.073
.287

.017
.009
.005

- .103
- .152
• 035
Constant
25.737
34.649
8.280

.318
.236
.178

.127
.025
.003
r 2 adjusted
.123
.066
.058

.022
- -033
.006

Appendix, Table XV
Results of Multiple Regressions Explaining Dealer Purchases from
Private Customers on Days Without Financings, August 22, 1962-December 31, 1963

Treasuryt
Simple r
1-5
5-10
>10

- .042
.041
.203**

Order Added
1-5
5-10
>10

8
4
2

F Level When Added
1-5
.132
5-10
1.941
>10
18.521**

Dealer Sales to
SQMAt
Treasury

Independent Variables
Duration of sales
S0MA t _ 1
to SOMA + Treasury

.078
.181**
- .001

- .031
.005
.138*

.061
.043
- .005

- .048
- .015
• 079

6
2
5

7
6
4

4
7
8

5
5
6

.124
.097
2.912
Hr Adjusted for Degrees of Freedom, When Added
1-5
.041
.048
.045
5-10
.088
.065
.087
>10
.096
.109
.111
.882
7-870**
.204

.669
.014
.004

1.036
1.628
.037

.048
.083
.100

.049
.090
.106

Net Regression Coefficient at Final Stage
.170
1-5
- .144
.159
• 193
5-10
.361
.510**
.020
.077
>10
.204** - .118
- .006
.109
Standard Error of Regression Coefficient at Final Stage
.170
• 390
1-5
.396
.171
5-10
.238
.163
.241
.168
>10
. 060
. 284
.070
• 095
Partial Correlation Coefficient at Final Stage
.032
1-5
- .024
.061
.020
5-10
.094
.192
>10
.193
- -024
.089
At Final Stage:
Degrees of freedom
1-5
233
5-10
256
>10
302

.065
.007
- .004

Note: Dollar variables in mill ions; price changes in points.
* Significant at 5 per cent level.

http://fraser.stlouisfed.org/
** Significant at 1 per cent level.
Federal Reserve Bank of St. Louis

-2.268
-1.402
- .047

Net
positional

^ Prices^

.176**
.025
.092

.109
.210**
.218**

1
8
7

3
1
1

.151*
.197**
.170**
2
3
3

7.639**
.008
.010

1.568
12.095**
15.385**

6.224*
6.696**
4.545*

.027
.080
.103

.050
.040
.044

.048
.085
.106

.052**
.001
.001

62.446
46.162**
17.5W**

2.045
1.106
.297

.019
.009
.005

64.356
17.400
4-971

- .072
- .079
- .009

.175
.006
.006

Constant
39.906
31.656
8.448

^Prices t _j

R2

.063
.164
.199
adjusted
.041
.080
.100

129.553*
44.662*
11.646*
64.104
17.391
5-013
.131
.158
•133

Appendix, Table XVI
Results of Multiple Regressions Explaining Change in
Prices on Days Without Financings, August 22, 1962-December 31, 1963

Treasury^
Simple r
1-5
5-10
>10
Order Added
1-5
5-10
>10

- .188**
- .132*
- .128*
1
1
1

F Level When Added
1-5
8.780**
5-10
4.639*
>10
5.185*

Independent Variables
Dealer Sales to
SOMAj.
Treasuryt_1
SCMA t _ 1
.090
.067
.03.0

- .083
- .102
- .028

4
2
5
.909
1.006
.108

R 2 Adjusted for Degrees of Freedom, When Added
.0348
1-5
.0313
5-10
.0136
.0136
>10
.0133
.0029
Net Regression Coefficient at Final Stage
.0002
1-5
- .0011**
5-10
- .0010
.0006
.0012
>10
- .0014

3
5
6
• 995
.200
.042

Note:
*
**

- .046
- .009
- .074

6
3
2

5
6
3

.003
.777
.265

.029
.027
.192

.0267
.0128
.0110

.0308
.0053
.0084

.0004
.0004
.0002

.0002
.0005
.0005

.0001
- .0038
.0009

.0000
.0000
.0000

.0002
.0006
.0012

.0021
.0041
.0036

.0000
.0000
.0001

Dollar variables in millions; price changes in points.




1.967
.823
.199

- .010
- .077
- .009

.0352
.0121
.0058

Degrees of freedom
235
258
304

Significant at 5 per cent level.
Significant at 1 per cent level.

2
4
4

Net
positiont_^

.0351
.0091
.0003

Standard Error of Regression Coefficient at Final Stage
1-5
.0004
.0002
.0004
5-10
.0009
.0006
.0009
>10
.0007
.0009
.0035
Partial Correlation Coefficient at Final Stage
.061
1-5
- .168
.057
- .028
5-10
- .07^
.067
.020
.012
>10
- .110
At Final Stage:
1-5
5-10
>10

.098
.054
.028

Duration of sales
to SOMA. + Treasury

.066
.051
.025
Constant
- .0047
.0030
- .0028

.oo4
- .058
.014

- .010 .
- .010
- .030
R 2 adjusted
.0267
.0053
- .0003

Appendix, Table XVII
Multiple Regressions Explaining Change in Dealers' Gross Long
Position on Days When There Were No Financings, August 22, 1962-December 31, 1963

Treasury^
Simple r
1-5
5-10
>10

- .269**
- .112
- .610**

Order Added
1-5
5-10
>10

Defiller Sales to
SOMAfc
Treasuryt-1
- .469**
- .289**
- .152**

2
4
1

- .219**
- .033
- .069

1
2
5

P Level When Added
1-5
28.889**
5-10
3.445
10
182.697**

67.796**
22.409**
4.158*

7
6
3
.214
1.673
8.188**

Independent Variables
Duration of sales
SOMlf..! to SOMA. + Treasury
- .060
- .102
.012
6
8
8

- .253**
- .082
- .117*
8
7
7

Net
positional
- .224**
- .354**
- .301**
3
1
4

£ Pricest
- .043
.119
.206**
5
5
2
2.294
2.029
8.316**'

• 453
1.177
.000

.009
• 737
.011

7.149**
37.752**
5.914*

.322
•233
.409

.317
.233
.411

.316
.122
.408

.060
- .109
.000

.116
.781
.032

- .030*
- .033**
- .013*

-57.778
15.859
12.658*

Standard Error of Regression Coefficient at Final Stage
.104
.236
.103
1-5
.240
5-10
.143
.098
.144
.100
.292
.072
>10
.061
.098

1.238
.664
.306

.012
.005
.005

38.966
10.438
5.118

.006
.073
.006

- .165
- .361
- .133

Adjusted for Degrees of Freedom, When Added
.217
.320
1-5
.299
5-10
.229
.188
.234
>10
.370
.414
.398
Met Regression
1-5
5-10
>10
-

Coefficient
.756**
.329*
.717**

at Final Stage
.842**
- .111
.476**
.122
.588*
.219**

Partial Correlation Coefficient at Final Stage
- .031
- .470
1-5
- .202
- .291
.053
5-10
- .143
- .115
.172
>10
- .558
At Final Stage:
1-5
5-10
>10

.038
- .068
.000

Degrees of freedom
233
256
302

Note: Dollar variables in millions;, price changes in points.
* Significant at 5 per cent level.

** Significant at 1 per cent level.


Constant
12.775
-I.678
- .121

.324
.232
.384

- .097
•095
.141
r 2 adjusted
.317
.233
.409

£

Prices^
.076
.181**
.108
4
3
6
2.307
12.113**
.509
.320
.221
.413
65.234
32.229**
3.655
38.813
IO.433
5.161
.109
.190
.04l

Appendix, Table XVIII
Multiple Regressions Explaining Change in Dealers' Gross Short
Position on Days When There Were Ho Financings^ August 22, 1962-December 31, 1963

Treasuryt
Simple r
1-5
5-10
>10
Order Added
1-5
5-10
>10

.154*
.052
.141*
8
6
4

Desiler Sales to
SOMA^.
Treasury t . .
.}
.109
.038
.071

.222**
- .007
.004

3
4
7

1
3
3

P Level When Added
12.450**
1-5
.076
4.346*
5-10
.950
.806
.839
>10
2.673
.311
1.926
R 2 Adjusted for Degrees of Freedom, When Added
1-5
.088
.086
.045
5-10
.090
.091
.092
>10
.072
.074
.067
Net Regression Coefficient at Final Stage
1-5
.035
.112*
.311*
- .088
5-10
.098
.080
- .074
>10
.056
.096

Independent Variables
Duration of sales
s(m fc . 1
to SOMA + Treasury
.082
- .022
.066

.073
- .015
.171**

.096
- .001
.015

4
7
6

5
5
2

7
8
8

1.657
.007
1.439
.089
.086
.076

.099
.006
.067
Standard Error of Regression Coefficient at Final Stage
.126
1-5
.128
.055
•055
.102
5-10
.101
.069
.071
.043
.058
>10
.036
.172

Partial Correlation Coefficient at Final Stage
.160
1-5
.018
.132
- .054
5-10
.061
.073
- .100
>10
.090
.032
At Final Stage:
1-5
5-10
>10

.118
.005
.067

Degrees of freedom
233
256
302

Note: Dollar variables 'n millions; price changes in points.
* Significant at 5 per cent level.
** Significant at 1 per cent level.



Net
positional

A Prices^
-

.102
.256**
.205**

& Prices-^
-

6
1
1

.216**
.243**
.137*
2
2
5

2.061
.595
9.459**

.675
.000
.013

• 931
18.462**
13.493**

8.269**
9.780**
1.900

.093
.090
.064

.091
.082
.071

• 093
. .062
.039

.073
,09?
.071;

- .905
- .442
.491**

.005
.000
.000

-17.880
-25.425**
- 8.704**

.659
.468
.180

.006
.004
.003

- .090
- .059
.155
Constant
-1.939
.845
- .779

.049
- .001
- .007

20.743
7.365
3.005
- .056
- .211
- .164
2
R adjusted
.088
.082
.071

-55,462**
-23.511**
- 4.052
^0.66l
7.361
3.030
-

.173
.196
.077

Appendix, Table XIX
Multiple Regressions Explaining Change in Dealers' Net Position
on Days When There Were No Financings, August 22, 1962-December 31, 1963

Dealer Sales to
SOMA^
Treasuryt_1

Treasury^
Simple r
1-5
5-10
>10

- .294**
- .125*
- .594**

Order Added
1-5
5-10
>10

- .450**
- .269**
- .164**

- .278**
- .02'!
- .063

Independent Variables
Duration of sales
SOMA^
to SOMA. + Treasury
- .084
- -075
- .017

- .248**
- .062
- .175**

8
8
8

7
7
7

.109
1.021
.400

.420
1.869
1.860

.33^
.279
.414

.337
.279
.415

- .039
- .115
- .070

1.021
1.223
- .456

Standard Error of Regression Coefficient at Final Stage
1-5
.274
.118
.269
.118
5-10
.161
.110
.163
.114
>10
.070
.332
.082
.111

2
5

1
3
1

4

5
6
3

F Level When Added
1-5
33.718**
60.804**
2.364
5-10
2.651
20.266**
3.554
>10
168.675**
5.653*
8.244**
R^ Adjusted for Degrees of Freedom, When Added
1-5
.295
.199
-339
5-10
.269
.232
.276
>10
.351
.^09
.400
Net Regression
1-5
5-10
>10
-

Coefficient at Final Stage
.791**
- .955**
- .422
.428**
- .556**
.211
.779**
- .691*
.289**

Partial Correlation Coefficient at Final Stage
1-5
- .186
- .467
- .102
5-10
- .163
- .300
.080
>10
- .541
- .119
.198
At Final Stage:
1-5
5-10
>10
Note:
*

- .022
- .063
- .036

Degrees of freedom
233
256
302

Dollar variables in millions; price changes in points.

Significant at 5 per cent level.
at 1 per cent level.


** Significant


Net
positional
- .232**
- .303**
- .263**
4
l
5

APrices-t
.004
.244**
.267**

^Prices^
.152*
.290**
.154**

6
4
2

3
2
6

7.55^**
26.634**
2.433

.843
12.621**
18.637**

8.536**
28.706**
1.788

• 335
.089
.412

• 338
.265
.386

.316
.175
.413

- .035**
- .033**
- .011

-39.897
41.284**
21.586**

120.697**
55.740**
7.938

1.413
• 751
.347

.013
.006
.006

44.466
11.805
5.805

44.291
11.798
5.855

.047
.101
- .075
Constant
14.713
-2.523
.657

- .167
- .323
- .103

- .059
.214
.209
R? adjusted
• 334
.279
.414

.176
.283
.078

Appendix, Table XX
Results of Multiple Regressions Explaining Offerings to the
Trading Desk on Days Without Financings, August 22, 1962-December~l, 1963

Treasury^
Simple r
1-5
5-10
>10
Order Added
1-5
5-10
>10

.233**
.244**
.560**
4
4
2

Dealer Sales to
S0MA.t
Treasuryt _ 1
• 357**
.257**
.143*
1
3
5

.160*
.184**
.280**
7
8
7

F Level When Added
.074
1-5
6.702*
35.151**
5-10
21.299**
.217
33-999**
>10
79.630**
5-044*
.807
2
R Adjusted for Degrees of Freedom, When Added
.124
1-5
.3^6
• 355
5-10
.581
.549
• 592
>10
.571
.641
.641
Net Regression Coefficient at Final Stage
.116
1-5
.684
1.059**
5-10
.907**
.121
.931**
.102
>10
.655**
.896*

Independent Variables
Duration of sales
SOM/L^-L
to SOMA + Treasury
.150*
.110
.069

.345**
.576**
.681**

8
7
4

3
1
1

4.976*
2.854
.893

.010
.46l
12.724**

33.612**
130.505**
267.477**

• 357
.59*
.641

.352
• 593
.636

5
6
6

.380*
.262
.132
Standard Error of Regression Coefficient at Final Stage
.176
.403
1-5
.409
.177
.182
.261
5-10
.258
.177
.106
>10
.090
.427
.1*3
Partial Correlation Coefficient at Final Stage
.019
.l4o
1-5
.109
.365
.029
.090
5-10
.215
.313
.056
.120
>10
.388
.053
At Final Stage:
Degrees of freedom
233
1-5
256
5-10
302
>10
Note: Dollar variables in millions; price changes in points.
* Significant at 5 per cent level.
** Significant at 1 per cent level.



.204**
.229**
.202**

Net
position^._2

- .216
.610
1.462**

• 331
• 329
.462
.107**
.134**
.114**

£Prices t
- .310**
- .410**
- .312**
2
2
3

APrices t _2
- .119
- .206**
- .151**
6
5
8

37.380**
85.151**
42.596**

7.494**

.239
.492
.622

• 357
• 591
.640

-363.259**
-169.136**
- 46.622**

.988
.341

-60.729
-51.632**
- 4.398

2.111
1.200
.447

.020
.010
.008

66.444
18.869
7.467

66.183
18.859
7.531

- .007
.032
.185
Constant
42.630
35.678
11.384

• 330
.657
.644

- .337
- .489
- .338
R2 adjusted
• 352
.592
.640

- .060
- .169
- .034