.113: ,. ..
V ) . 1.3:: .54

 

 

 

 

 

 

> I

v95... 7
. a}?

5,115

 

A I.
; .\:..¢‘p(rl
: , .I:

 

:r; r

A . 1:!!5 -
L...

{tiltfrel

1.)) 4 .. ‘ I

villi

)

 

.51): . r.
l.fl,!l11
9):; .L r..n.\cl.)
it. :f.-\lr . L.

1 I: {5.1.} .4

 

 

 

 

a Iii. v1.3 (If:
H... 11.)) 1...: .Iavkaib): (It. 33$ 59 (I
15.12.21. (rt. I. {5.352. .;\ .rfr. :. .2.
ix)..§b c, r
...Ill)f}!t!. up».
.5. ii”... I
; .,(vr£r.¢;ln.!)3.c.’)))§f ((3 .
Av . .2. n 'l. .I) ((1')!!! a... by
f f. t) . .td 1.). .5; :7:
77.1.] k .
c5.i..¥.i+
(31:: ox
226..

ii. ..E
:zz...$.€i; 3
. grit?»
. 35......si ..
.,;.;..;,.§.. . ,
:a ”a. . , ,
S. EH. gin. -Ifn
i? :55! rd... pv'fl
iii: sills)... {(1:22
: 52.352, 1

.5.

x.:..wéw.¢

drags. ‘
. Fauna . .,

.: 9 .:

 

 

RABIES
MICHIGAN STATE UN

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\l\\)\\\\\ _~

This is to certify that the

thesis entitled

RECONSIDERING ECONOMIC SANCTIONS--
WHEN DO THEY WORK?:
A QUANTITATIVE ANALYSIS

presented by

Alfred Cooper Drury "

has been accepted towards fulfillment
of the requirements for

M.A. International Relations

 

 

degree in

/5f/fi(\

Major p fessor

7 A t 1992
Date ugus

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

  
    
    

 
 

LIBRARY
M'Ch'flan State
University

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE

 

 

,r
L

igpw. L G) ‘fi'l

 

 

2: E ‘35".

 

 

in» v

 

 

 

“‘5‘ Ta; 3\

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Institution
c:\clrc\datedm.nm3~p.1

RECONSIDERING ECONOMIC SANCTIONS—WHEN DO THEY WORK?:
A QUANTITATIVE ANALYSIS

By
Alfred Cooper Drury

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Multi-Disciplinary Program

1992

ABSRTACT

RECONSIDERING ECONOMIC SANCTIONS—WHEN DO THEY WORK?:
A QUANTITATIVE ANALYSIS

By
Alfred Cooper Drury

The purpose of this thesis is to examine the work done by Gary Hufbauer, Jeffery Schott
and Kimberly Elliott in Economic Sanctions Reconsidered. Using the data from their 115 case
studies of economic sanctions published in their books I test their results, and determine
when economic sanctions are effective. There are many problems with the data and with
some of the evaluations within their work, these will also be addressed. The plan of the
thesis is to address the relevant issues and definitions involved with economic sanctions,
including an evaluation of Hufbauer Schott and Elliott. Next, the quantitative analysis
and results are presented followed by examination of the faults of the study and some
possible solutions to them. Lastly, I will discuss the implications and problems presented

in the study and the direction of future research.

Copyright by
ALFRED COOPER DRURY
1992

To all my teachers who taught me
to love learning, of which my
parents top the list.

ACKNOWLEDGEMENTS

I would like to thank my committee, Scott Gates,
Jim Granato and Gretchen Hower, for all their
patience, help and support. I would also like to
thank Charlie Curtis for his help through this
last year.

 

iii

TABLE OF CONTENTS

List of Tables
List of Figures
Chapter I: Introduction

Chapter II: Liturature Review
Idealists and Realists
Idealism
. Realism
The Cold War Influence
Margaret Doxey
Robin Renwick
Johan Galtung
Jerrold Green
David Deese
David Baldwin
Problems defining economic sanctions
Definition of economic sanctions

Chapter III: Hufbauer, Schott and Elliott
Chapter IV: Quantitative Model

List and definition of variables

Description of model

Expected results
Chapter V: Statistical results and problems encountered
Chapter VI: Policy implications and conclusions
Appendix A
Appendix B

Bibliography

iv

11
ll
13
17
19
21
23
25
29
31

33
38
38
4O
41
43
52
54
74

79

iv

LIST OF TABLES

Recoded Success Values
First Round of Reductions
Second Round of Reductions
Third Round of Reductions
Fourth Round of Reductions
Final Model

Diagnostic Test Results

38

44

46

47

48

49

50

LIST OF FIGURES

Figure 1.1: Concurrent Sanctions
Figure 1.2: Enacted Sanctions
Figure 1.3: Success of Sanctions
Figure 2.1: Galtung’s Theory
Figure 4: Scaled Residuals
Figure 5: One Step Chow

Figure 6: Decreasing Horizon
Figure 7: Increasing Horizon

Figure 8

74

75

76

77

78

Chapter I:
Introduction

Economic sanctions have been used throughout history for different reasons and with
different levels of success. However, the study of sanctions has been rather sporadic. Most
scholars concentrate on the question of why sanctions fail so often (almost twice as many fail as
succeed) and yet are still frequently used.1 Before leaping into this thesis, a short example of a
sanction follows. Although there is no formula for a sanction, an example of one may prove to
be useful as a point of reference which can be used when sanctions are discussed at the world
level. What follows is a description of the US sanctions against the Dominican Republic in 1960.

In 1930 Rafael Trujillo assumed power of the Dominican Republic and began a reign of
oppression. Between 1959-1960 he initiated a campaign to exterminate all opposition ex—patriots.
During this same time the world recession was particularly severe to Latin America, damaging
the Dominican Republic’s earnings. Trujillo maintained military spending of $80 to $100 million,
approximately one third of the national budget, which intensified the damage. On 24 June 1960,
the Trujillo family attempted to assassinate President Betancourt of Venezuela, after 20 years of
mutual enmity. In response to this the US pushed the Organization of American States (OAS) to
break diplomatic relations, suspend all arms trade and begin consideration of further economic
sanctions. The sanctions were to be lifted only when the Dominican Republic ceased its threats
to the exiles residing abroad. In the beginning of 1961, the OAS extended the sanctions to include
petroleum, trucks and spare parts. The increase is only passed by a two thirds majority with
Cuba and Brazil being the largest dissenters.

Almost five months later Trujillo was assassinated with suspected CIA involvement.
However, his son maintained control of the country with the aid of his uncles who were generals
on the joint chiefs of staff which he headed. They escalated the campaign to purge the country
of all opposition. Six months later a US fleet appears off the Dominican Republic coast in support

for Joaquin Balaguer who ousted the Trujillo family and took control of the country. In the

 

‘ Data are from the Economic Sanctions Data Set.

2
beginning of January 1962 the OAS sanctions were lifted when a provisional government was
established in the Dominican Republic.

This case illustrates several aspects of an economic sanction. The basic cost to the
Dominican Republic was 1.9% of its annual GNP. The US enjoyed a size superiority of 596 times
that of the Dominican Republic. The sanctions are considered to have had a significant
contribution to the full success in this case. Other contributions include the suspected use of
covert operations by the CIA: Rafael Trujillo assassination; and a quasi-military operation: the
presence of the US fleet off the Dominican Republic coast. Thus the case includes measures other
than economic sanctions. International cooperation is apparent through the participation of the
OAS while there was no other aid to the Dominican Republic from another nation with the
intention of disrupting the sanctioning effort.2 This example should not be used as a blueprint
for sanctions, for as mentioned above, there is no real standard for sanctions. It should be used,
however, as a reference point or an aid to understanding sanctions while reading this thesis.

Economic sanctions

 

have a long history, one that
goes beyond what will be
discussed here. Figure 1.1
shows how many sanctions
were concurrently in effect

annually (The horizontal axis

 

 

 

shows years in the 1900’s).

 

4D 50 SO 7

1S 2 25 35 45 55 SS 75 E10 85
There has been an almost +NLNBER OF SANCTIONS OCCURING EACH YEAR

constant increase in FIGURELlConcurrentSanctions

sanctioning.

 

2 Hufbauer, Gary Clyde Jeffrey]. Schott and Kimberly Ann Elliott annmir “ " n J 4
‘Case Histories 1990, Washington, DC. Institute for lntemational Economics, pp. 182- 186.

C

 

3

The attempt to use economic sanctions as an instrument of policy to induce an

offending government to change its conduct and thereby resolve or help to

resolve the international problems created by it is an essentially twentieth century

phenomenon. States since time immemorial had interrupted commercial

relations...[But] It was not imagined that economic measures falling short of a

blockade could have decisive effects.3
The only effective way to sanction before the Twentieth Century was to impose a blockade, which
required a great deal of force and was rarely successful. Also, blockades are typically thought of
as military actions; certainly if a nation is attempting to cut off the life blood of another nation,
it is considered a militant act. This changed as the world began to move from a state of autarky
to one of trade and eventual interdependence. This turnaround appeared about the time of the
industrial revolution.‘

After the First and Second World Wars the interdependence among nations increased,
especially within the East-West blocs.

Growing international economic interdependence has been a second circumstance

encouraging the resort to external economic coercion. Foreign trade and the flow

of capital among nations have expanded significantly in the post-war period and

have generated vulnerabilities that would have been absent in a world of greater

autarky.5
This growth, as defined by Nincic and Wallensteen, has had a positive effect on the increase of
sanctions over the years. What has become apparent since 1914 is both the increase in the number
of sanctions running concurrently (Figure ONE), and the number of sanctions enacted in a given
year, (Figure 1.2). As one can see there is a significant increase in sanctions since the end of
WWII. The use of sanctions during the Cold War obviously increased due to the dangerous

nature of the world arena. The US and the USSR found it safer to impose sanctions on each other

than to engage in militarized conflict. Also, during this time the lines between allies and enemies

 

3 Renwick, Robin Economic Sanctions, 1981, Rensselaer: Hamilton Printing Company, p. 4.

‘ Doxey, Margaret R, '" ' “ " and International " ’ ‘ 1971, London: Oxford
University Press, p. 15.

5 Nincic, Miroslav and Peter Wallensteen eds. Dilemmas of " ' Coercinn' Sanctions in World
Politics, 1983, New York: Praeger, p. 2.

were rather well marked,

 

a _ allowing for clear definitions
of who should be sanctioned.
The sanctions against
COMECON are an example

of the Cold War effect.6

 

One would then

  

 

 

 

 

a l l' ' ' ' "" ‘l I III [I 1111 1 ll llllll Ll Ullllllllllllllllll
.- ”I” -— w I. It

152°253°354045505580837073808590

+NWBER 0F SANCTIONS ENACTED EACH YEAR the Cold War we should see
FIGURE 1.2 Enacted Sanctions

expect that with the ending of

 

 

a significant decrease in the
number of sanctions being levied. Although it is too early to tell, there has been somewhat of a
decrease in concurrent sanctions world wide. There is a rather convincing counterpoint to this,
raised by Fukuyama’s paper The End of History. Simply, he argues that the end of the Cold War
has ushered in an era of military peace and no new ideologies, only to be replaced by economic
warfare and democracy. This certainly would not predict a decline in the number of sanctions
imposed, in fact the opposite would be true. Along with the arrogance required to believe that
in this century we have created the highest form of government possible, democracy, the main
tenets of The End of History argument are wrong. This can be seen in the eruption‘of the Gulf
War, and all the little wars that are beginning to divide nations into ethnic blocs, i.e. Yugoslavia
etc. Violence is not on its way to extinction, it is strong and thriving.

The one point of Fukuyama’s that does have some validity is that there is an increase of
economic warfare. The world is quickly dividing into economic blocs, while the established
military blocs are quickly disbanding. Blocs like the EC have already opened their markets

internally, but have closed much of them off to the outside. It seems that economic sanctions as

 

6 Nincic and Wallensteen, 1983, p. 2.

 

5
a part of this economic warfare could soon follow. Since economic sanctions may increase in the
future, their study is an important one.

Another reason for
20

 

studying sanctions is their

 

 

 

 

 

 

poor success record. 1,_ ' I W" I. n n n "
Although sanctions have N S' H H II! "n
increased over time they have 10 - i I I
not become more successful. I I l

a m 11
Instead there is a randomness s _ . . . .

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

of the success of sanctions k U l. "

ll I! I! l I! I! I” I II“ I II I l
O llllllllllllllllllllllllllllllllllllllllllllllIllllllllllllllllllllllllllllllllllllllllllllllllllllllIlllllllllllll

when viewed over time, as +SUCCESS MEASURE OVER ALL CASES

 

 

 

shown in Figure 13. Which FIGURE 1.3 Success of Sanctions

depicts the measure of success as determined by Hufbauer, Schott and Elliott. In this data
sanctions are rated on two separate four point scales, one for the success towards the goal and
one for the overall contribution of the sanctions (1—8 equates with failure while 9-16 with
success).7 Although the success of sanctions is seemingly random, over time there are many other
factors that can be studied to determine what could be causing success. This is exactly what
Hufbauer et al have done in Economic Sanctiog Recogsidered: History and Current Policy. They
use a case analysis format to attempt to understand what makes economic sanctions tick. They
conclude with what they call nine commandments, or results of their study.8

1. "Don’t Bite Off More That You Can Chew": Sanctions that attempt to make great
changes in the target’ 3 policies often fail, less ambitious plans succeed more often.

 

7This operationalization will be discussed later, see Hufbauer, Gary Clyde Jeffrey J. Schott and Kimberly
Ann Elliott Economic Sanctions Reconsidered: History and Currant Policy, 1990, Washington, DC: Institute
for International Economics, pp. 41-42.

8 All of the nine commandments can be found in: Hufbauer, Schott and Elliott, 1990, pp. 94-105. A full
discussion of Hufbauer et al’s work and results will be presented later.

 

6

2. ”More 13 Not Necessarily Merrier”: International cooperation is not that important
for success, "In most cases, multilateral sanctions are not associated with
success."9

3. "The Weakest Go To The Wall "2 Weak nations, especially those under distress,
succumb to sanctions much more often. This includes the idea that the sender
to target ratio has a strong effect, that is, the more skewed the fight, the more
likely the sender will win.

4. "Attack Your Allies, Not Your Adversaries": Sanctions levied against friendly
nations are much more likely to succeed.

S. "If It Were Done, When ’Tis Done, Then ’Twere Well It Were Done Quickly":
Overall, failure is associated with longer periods, and vice verse.

6. "In For a Penny, In For a Pound ": The more intense the cost to the target, the
more likely the sanctions will succeed.

7. "If you Need to Ask the Price, You Can’t Afford the Yacht": Sanctions that are
costly to the sender often fail.

8. "Choose the Right Tool For the Job”: Supplemental policies, (military, quasi-
military and covert) can often enhance the probability for success.

9. "Look Before You Leap ": Sender nations should consider the consequences and
expected outcomes before they impose sanctions.

This thesis quantifies the work of Hufbauer et a1 by creating the Economic Sanctions Data
Set from their published case histories.10 The format of this thesis is straight forward. Following
this introduction, a literary review of 7 different scholars is undertaken, showing the development
of theory about sanctions since the pre-WWI era. The authors are chosen for their ability to
characterize the different types of theory as well as having an interesting focus. Following the
literary review is the analysis and critique of Hufbauer et al’s work. In this work, Chapter IV
consists of the quantitative model. Here the model is formalized, the variables defined and a list
of expected outcomes set out. Next, the statistical results are presented and problems with the

model discussed. Lastly, in Chapter VI, there is a discussion of the policy implications and

 

9 Hufbauer, Schott and Elliott, 1990, p. 95.

1° The case histories are found in Hufbauer, Schott and Elliott, 1990, and Hufbauer, Schott and Elliott,
1990, Supplemental Case Histories.

7
conclusions. This also includes a discussion of possible improvements on the research as well as

other paths that can be taken.

 

8

Chapter 11:
Literature Review
Idealists and Realists

The purpose of this review is to give the reader a sense of the background of economic
sanctions; their legacy, if you will. During this process different scholars are examined for their
theoretical tradition pertaining to economic sanctions and their specific focus on sanctions. This
should serve as an introduction to Hufbauer, Schott and Elliott’s work, allowing for a better
understanding of it. It is not an attempt to formulate a separate comprehensive theory, although
in the conclusion a suggestion is given for further work, and possible theories.

Most of the literature on economic sanctions falls into two broad categories of
international relations: the Idealists and the Realists. Like international relations, the Idealists
form the foundation for the other paradigm and as such have the longest history. In addition to
these two traditions the Cold War is discussed due to its profound influence on the study of
economic sanctions.

One important point to watch for is the focus of each scholar. All the literature on
sanctions, regardless of its theoretical foundation, explores the success/ failure of sanctions. That
is, the focal point is why sanctions succeed and fail, not why they occur in the first place. This
shall become apparent below.

Idealism

The Idealists are the oldest of the two theoretical traditions, dating back to the 1880’s.”
They focus on morality, legality and the universality of international organizations. "...there are
important differences between economic measures used as techniques of warfare [economic

warfare]... and economic sanctions employed by an international organization as part of a

 

‘1 Sanctions actually go back to Thucydides’ times as he looked at economic sanctions. However only
recent history will be discussed here.

 

9
constitutionally authorized enforcement process."12 This typifies the Idealist. The distinction
between economic warfare and sanctions is made by others subscribing to Idealist theory. The
focus on international organizations is not a derivative of Neo-Institutionalism, instead it is the
qualifier for legality and morality. This can be seen in Idealist studies where the cases used as
evidence consist purely of UN and League sanctions.13

In the modern era, Idealist theory has a long history dating back to the end of the First
World War. In pre—WWI historyxhe conceptualization of sanctions was one that saw them as a
useful tool to be instituted by a group of nations in the world arena§§There was one main codicil
to this: neutral nations could withdraw from being considered as part of a sanctioning coalition.”
That is, if the European nations decided to sanction a country, there was no obligation for other
nations, including allies, to support or even respect the sanction. Notice the morality or idealism
that is implicit but obvious here: only wronged nations would participate, others remained
neutral. The concepts of Realism have no place in this setting.

World War I was considered so devastating in both battle deaths and the failure of the
balance of power system that a modified theoretical perspective on sanctions arose. This new
"paradigm", as Daoudi and Dajani call it, is one that saw sanctions as a response to one evil nation
victimizing a benevolent nation, the lines between good guy and bad guy were clear. In addition
to this previously accepted view all nations were expected to act selflessly in an effort of collective
security. This was promoted primarily by the Geneva School of Thought.15 They argue that the
Geneva School’s idea of full responsibility for world security made a distinct imprint on the

forming of the League of Nations. But the old concept of neutrality had not yet died, so when

 

12 Doxey, 1971, p. 14.

‘3 For example see Doxey, 1971, and Doxey, Margaret P. International Sanction: in (‘
Persgtive, 1987, London: Macmillan Press Ltd. and Daoudi, M. S, and M. S. Dajani Economic Sanctions.
[deals and Experience, 1983, London: Routledge 8: Kegan Paul.

1‘ Daoudi and Dajani, 1983, p. 18.

‘5 Daoudi and Dajani (1983), pp. 20-28.

 

 

 

10
the US did not join or support the League, it failed and subsequently WWII transpired. This
failure set the Geneva School’s paradigm in motion: universality was pushed from the theoretical
into real policy implementation. The UN’s ideals reflect this as does the high enrollment in it.

Thus, the Idealistic concept that economic sanctions were an instrument to maintain the
world order emerged as the dominant theory. It was held that their use, which was not as
destructive as war, could keep all nations within the bounds of the international norms, the idea
that sanctions truly worked. Around the 1960’s this began to be questioned, and a new theory
was being developed by the Realists. This group started to dominate in the 1980’ s as the literature
points out. Although there are many questions within it which are discussed below, the Realist
paradigm is presently dominant.

The above is laid out in Daoudi and Dajani who take a Kuhnian perspective of sanctions,
citing that there are two paradigms to be studied, the pre—WWI paradigm and the Geneva
School’s. This stretches the definition of a paradigm. Sanctions occupy a very small area of
study, and are part of a larger paradigm, that of international relations. To say that the study of
sanctions has undergone its own personal paradigmatic shift is giving the area too much
independence. A better study, one that would be far more accurate, would be to look at how the
shifts in international relations have affected the sub-field of sanctions. For example, it is far more
accurate to say Idealism, Realism or Neo-Institutionalism are paradigms, than it is to say sanctions
are. It would be better to call the different views of sanctions theories; I have done so here.

A second problem with Daoudi and Dajani’s paradigmatic shift is the lack of shift that
took place. When comparing the two paradigms (Pre— and Post- Geneva School) one is hard
pressed to see much of a difference; in fact the only one is the idea that neutrality/ universality.
The pre—Geneva School believed in neutrality while the Geneva School saw neutrality as the cause
of past failures. The consequence of this shift is the increase of importance on moralistic, legalistic

and especially universal sanctions. That is, as the importance for universal participation in the

 

 

ll

levying of sanctions grew, so did the need for moral grounds on which to levy them. The three
considerations interacted with each other, intensifying their importance.
Realism

The Realist sees sanctions as means of a nation-state to enforce its will and promote its
security. This view is diametrically opposite to the Idealist who takes legal and normative
questions into account while the Realist only looks at what is the best method for a country to
survive and prosper. This mode of operations is certainly more scientific than the traditionalist
because questions of what are just and unjust sanctions are ignored, leaving the researcher much
more room for objective analysis. Truly, one can argue that no view is objective, some are more
than others, and this is the case between the Realist and the Idealists. This is more fully argued
later.

The review of David Baldwin provides an example and full discussion of Realist theory.
But first, a few remarks must be made concerning it. There are two types of Realists with
reference to sanctions: first, those who believe sanctions are futile policies that cost the sender
country dearly and whose success rate is pathetic at best; secondly, those who View sanctions as
a viable option in international relations. The doubting types began discovering evidence in the
1960’s that found sanctions to be rather ineffective.16 This continues today. The basis for this
conflict of opinions centers around what the sanctions are expected to achieve. That is, what is
the goal of the sanctioning effort: punishment, policy change, domestic rallying etc. This debate
is Baldwin’s primary focus and as such is discussed then.
The Cold War Influence

As the Cold War locked its protagonists into two rival military blocs, and as the

potential costs of their confrontation grew, economic coercion became a less
perilous manner of fighting the East-West conflict.”

 

1‘ For a sample list and quotations of those scholars see Daoudi and Dajani (1983) pp. 43-48.

‘7 Nincic and Wallensteen, 1983, p. 2.

 

12

This statement fundamentally represents the Cold War influence on economic sanctions.
The development of the nuclear arsenals and possibility of nuclear holocaust made the actual
thought of an hot war more deadly than ever before. Fear of escalation through direct military
confrontation led to the use of less direct measures such as military aid to third world countries
and economic sanctions. The sanctions provided a manner in which the East and West could jab
at each other without risking escalation to full scale war. Many of the theorists writing at this
time take this approach to sanctions, comparing them as an option to military warfare.

The consequences of the Cold War influence are twofold. First the scholars tend to view
sanctions as an all or nothing option, not something that can be imposed incrementally. Although
some of the scholars do not always take this to heart it is a common trend. The second influence
is to think of economic sanctions as a substitution for military actions, this is an inaccurate view
of sanctions. The only similarity between economic and military force is that they both are tools
of the nation-state, their likeness ends there. Economic sanctions have both the capability of
punishing, showing disapproval and garnering domestic and allied support. They are a less
drastic resort to action as compared to military actions, and more severe than diplomatic
sanctions. Because they are not on the same scale of severity, one can not say that the two,
military and economic sanctions, are substitutes for each other. This is not to say they cannot be
used in conjunction with each other, in fact slightly over a third of all sanctions were imposed in
combination with military action.“

One interesting incongruity must be noted in the Cold War theory. Sanctions were seen
as a substitute for military confrontation, one to be used to avoid escalation. But sanctions to a
theorist writing under the influence of the Cold War are a full scale operation, they aim to be as

destructive as possible.19 One would expect that full scale sanctions of this type are the most

 

‘8 Data are from the Economic Sanctions Data Set.

‘9 For a full discussion of this see: Deese, David A. The Vulnerability of Modern Nations: Economic Diplomacy
in East-West Relations, found in Nincic and Wallensteen, 1983; or the description given in the analysis below.

13

escalatory as it intends to devastate the target’ 5 economy. That is, punishment is applied in such
a manner that little room is left for a subtle reaction by the target. The only options left to the
target are capitulation, resistance and full scale retaliation. Of these, retaliation would be the be
expected if the target refused to capitulate since it had been backed against a wall. The only
explanation I imagine is during the Cold War neither East nor West actually intended to severely
damage the other, or more likely if the intention did exist the capability did not. For example,
the actual damage inflicted on the Soviet Union, as a percent of GNP, by the Western sanctioning
of COMECON was nil.

Margaret Doxey

Margaret Doxey, in both of her books, is exemplary of the Idealist scholar of economic
sanctions. The basic question she attempts to answer is not very clear as the books are a general
look at the history, types and composition of sanctions. She does not develop a theory as such,
but she does come to some conclusions in the books; these act as a surrogate theory, however,
they are really the results inquiry. The nation-state is the level of her analysis, and she sees the
international arena as one that was once in a state of anarchy, but is developing into one where
international organizations are in control.

There is a problem with using the nation-state as an unit of analysis: it is rational, and
quite possible for a government to enact sanctions against an opposing nation for the purpose of
bolstering internal support for their regime. An example is the standing declaration of war Libya
has maintained against the US for many years. It is totally unreasonable to believe that they
intended to actually attack the US, what is reasonable however, is the use of the declaration as
a method of rallying the masses around Qadaffi. Similarly, the sanctions the US imposed on the
Soviet Union over the Afghanistan invasion were not wholly to push the USSR out, but to gather
the popular support of the people and as proof of the resolve against the Warsaw Pact for the
sake of the US’s allies. Thus, there are many aspects to sanctions; the reasons for their

implementation are rarely as clear cut as Doxey theorizes.

14
This idea that a nation may engage in sanctions or other types of international activities
principally to affect the domestic politics of their nation is brought up by Ronald Rogowski.20 In
attempting to analyze what effects international trade has on domestic politics he used the idea
of comparative advantages and a theorem that states while protectionism benefits the scarce
factors it harms abundant ones (and vice verse). That is to say, while a scarce factor will benefit

"x N

from trade policyx ,abundant factors will suffer. Conflict will appear along these lines, which
are delineated into capital, labor and land. Regardless of its faults, Rogowski has brought an
interesting point to bear, international affairs can play an integral part in domestic politics.21 This
is the fundamental problem with taking the approach that Doxey does, the same approach that
most scholars take.

To play Devil’s Advocate I will now defend the nation-state as a unit of analysis. By
focusing on the nation-state one is allowed to disregard much of the often messy details involved
in each different nation, e.g. type of political system, strength of ruling regime etc. This allows
one to focus on the aspects of international economic sanctions common to all nations, such as
Hufbauer et al did, and I have when testing their results. Both levels of analysis are useful, the
decision is dependant on what the study is attempting to accomplish. A

By focusing as she does on universal sanctions, Doxey limits her ability to form any
conclusions that go beyond those sanctions imposed by the League, UN or a large multi-lateral
group such as NATO. That is, she is forced to focus on sanctions that were imposed under full
international cooperation. She points out there are several different reasons that international

cooperation is difficult to attain, and as such how sanctions are difficult to attain:

Collective action which involves the commitment and diversion of national
resources will be easier to organize if the overriding policy goals of a group of

 

2° Rogowski, Ronald Commerce and Coalitions: How Trade Affects Domestic Political Alignments, 1989,
Princeton: Princeton University Press.

2‘ Rogowski’s analysis 18 flawed in many ways, primarily the idea that people will organize along factors
lines, 1 e capital, labor and land, and not along industrial lines.

 

15

nations happen to coincide. Much more common are cases when collective
action appears to conflict with national goals...22

This is the problem of consensus, as Doxey defines it; otherwise known as the collective action
problem. She does not discuss the specific problems of collective action, namely differences in
ideology or domestic vested interests. The second problem is a function of her level of analysis.
The first problem, ideology, seems less important than other factors. This could be a result of the
Cold War. That is, Doxey plays down the ideological problems because during the Cold War
there were principally two ideologies. These allowed for cooperation only within their spheres
of influence, East and West. Although this is a simplified statement about the ideological beliefs
at that time, a common enemy such as the Warsaw Pact improves c00peration for the opposing
side.

Doxey cites the next problem as the question of goals: the sender nations must decide
what their goals are. This does not create the same type of problem as the collective action above,
for according to Doxey the sole goal of any true sanction is that of peace keeping. Once again the
moral aspect of Idealism inhibits the researcher. Her definition of sanctions include only ones
enacted by international organizations against a wrong-doer nation. Thus, her question of goals
concerns not "The objective of peace, or of rectifying a situation which threatens peace..." which
is obvious to her, but "In the first place, the nature and circumstances of the crisis will determine
specific final goal, while the tactical decision to use a certain form of coercion will establish
intermediate goals."23 This strict definition causes a great deal of problems because it limits the
researcher to sanctions that are universal, legal and just. All of the times coercive economic power
is used by only one nation the researcher may not consider them sanctions and is left blind. In

fact, according to Hufbauer et al’s qualification, most of the sanctions are put in place by one

 

’2 Doxey, 1971, p. 90.

’3 Doxey, 1971, p. 92.

16
nation.“ As seen throughout this thesis the limitations set in place by the conventional tradition
are so stringent that they act as blinders to the scholar.

The next problem in sanctioning a country is the selection of measures. Doxey again
concentrates on the legally allowable sanctions according to the League and UN in their respective
periods. But here she takes an interesting turn, instead of fitting the punishment with the crime
so to speak, she posits the idea that:

Crucial factors influencing the choice of particular measures adopted singly or in

combination would be the objectives sought, the vulnerability and sensitivity of
the delinquent state, and the minimization of cost and damage to the sanctioning

group.25
This is odd because Doxey uses a more utilitarian approach instead of a moralistic view. This can
be explained by the fact that to the Idealist, policy change to preserve peace is the only goal of
sanctions. They are not seen as punishment or methods in which to bolster domestic support.
Thus, the Idealist uses a cost benefit analysis to determine what is the most efficient manner to
preserve peace, but not to determine whether to sanction.

The next consideration, costs tot he sender country, follows directly from the last. These
can be realized in both economic and political terms. One does not even have to regard counter
measures by the target nation to understand this:

American farmers were outraged by the partial grain embargo imposed by the

Carter Administration after the Soviet intervention in Afghanistan. Mr. Reagan’s

campaign promise to lift the embargo was carried out once he was elected

President, and it was not reinstated, in spite of his Administration’s imposition

of other penalties on the Soviet Union.“

This is an example of both economic and political costs, Reagan wanted the support of the farmers

who had a strong lobby, as well as lowering the costs inflicted on that group. Thus the sender

must take care in choosing which types of sanctions to impose, this is complicated by the problem

 

2‘ Data are from the Economic Sanctions Data Set.
’5 Doxey, 1987, p. 98.

2‘ Doxey, 1987, p. 101.

17

of universality. That is, many nations imposing the sanctions will be affected differently by
sanctions, and will therefore have a difficult time in choosing how to sanction, they will all have
different cost benefit analyses.

The last consideration is that of universality. This is the key part of any sanction in the
Idealist tradition. Doxey points out that universal coverage of sanctions is more of an ideal than
a true goal. She then moves to analyze what the UN can do legally to force others to comply to
the sanctions. What is not questioned is the importance of international c00peration in the success
of sanctioning, which should be a primary focus. Here again the Idealist is blinded by only
looking at how to get cooperation, without bothering to answer whether it is needed or not. This
is due to the normative approach. Thus, if the focus was changed from "why sanction" to "when
they wor " much more could be learned.
Robin Renwick

Robin Renwick strays from the Idealists as he attempts to answer the question of when
sanctions will work, and why. He sees sanctions in a different light than Doxey. To Renwick,
universality, morality and legality are not important. Also his theory is strongly influenced by
the Cold War, as can be seen below where sanctions are perceived as part of a package of options:

In deciding whether to impose economic sanctions, governments frequently find

themselves responding to an international crisis on the basis of three broad

options: (a) to do nothing; (b) to consider taking some form of military action; (c)

to seek to impose economic penalties. A decision to impose sanctions may be

taken less on its intrinsic merits than because of its attractions in relation to the

alternatives.”
It should be noted that by comparing the three options above, Renwick implies they are
interchangeable options, one of the basic tenets of the Cold War policies.

Renwick does rely on some of the Idealism in his work where he makes a distinction

between economic warfare and economic sanctions. Economic sanctions are defined as: "...the

imposition of economic penalties to bring about a change in the political behavior of the country

 

27 Renwick, 1981, p. 1.

18
against which they are directed"28 This is more utilitarian, or Realist, than Doxey due to the
absence of the moral, legal and universal aspects on which she concentrates. He does however
mention these as possible definitions that he chooses not to use. Economic warfare is not so well
defined:

In considering attempts to use the "economic weapon" it is necessary to

distinguish between the application of economic sanctions and the more limited

objective of halting trade in arms or other strategic materials.29

It is not clear what the difference here is; restricting military goods to a nation for strategic
reasons also implies that the sender does not trust the target and a change in policy would be
favored. This is no different than any other type of sanction which attempts a direct policy
change. A somewhat better definition is one Renwick gives by example. He considers the
"phoney war" that preceded WWII to be economic warfare, not economic sanctions. The
difference here seems to be the hostile outcome that resulted as well as the goal of the economic
measures: to weaken Germany enough to avoid war. This is still no different than any other
sanction though, as it is an economic measure attempting to change the policy of another nation.
The difference is the actual policy goal. Renwick’s distinctions between the two, sanctions and
warfare, are unneeded and as such only confuse the material while adding no additional
information.

The unproductive distinction above does not follow through in the analysis, nor does it
damage Renwick’s work. He theorizes that sanctions can have the reverse effect than what is
expected. That is, the target nation may be able to pull together and fight back against the
sanctions as the sender countries offer a common enemy. This almost punitive view of sanctions
does not go as far as saying they do not work. In fact, Renwick believes sanctions work to the
point of effectively damaging the target nation, but this does not equate to the desired policy

change. As a result economic sanctions are dependant on how they are imposed. According to

 

2' Renwick, 1981, p. 2.

2’ Renwick, 1981, p. 59.

19

Renwick, if the sanctions are imposed with strength and purpose their chances for success are
greatly enhanced, if not:

Minor or petty sanctions are virtually certain to produce a reverse political effect

without exerting any real pressure. A sanctions policy should never be based

on the initial fallacy, which was the supposition that it would prove possible to

deter aggression or otherwise to change fundamentally the political conduct of

states by the threat or economic penalties alone. The implementation of sanctions

is essentially punitive, the effect— if there is sufficient international support-

being to weaken the country to which they are applied.30

This is the crux of his theory. There is little here that is surprising; one would expect
sanctions that are applied willy nilly would fail. Renwick’s work does point out two main ideas
that are not so tautological. First, not only can the sanctions fail to be effective, but they can
actually reverse their effectiveness and blow up in the sender’ 3 face. The second main point is
what Renwick is fighting against. As mentioned in the historical overview, beginning in the
1960’s a growing disbelief in the effectiveness of sanctions began. This was much more apparent
in the 1980’s when Renwick was writing. He is writing counter to this belief, and although the
argument is not as strong as David Baldwin’s, it does not buy into the belief that sanctions are
impotent.
Johan Galtung

As a scholar, Johan Galtung does not fall into any of the theoretical traditions, rather he
attempts to explain the reaction of the target country (receiver in his terminology) to sanctioning.
His definition of sanctions, however, is Realist:

...actions initiated by one or more international actor (the "senders") against one

or more others (the"receivers") with either or both of two purposes: to punish the

receivers by depriving them of some value and / or to make the receivers comply
with certain norms the senders deem important.31

 

3° Renwick, 1981, p. 92.

3‘ Galtung, Johan On the Effects of International Economic Sanctions found in N incic and Wallensteen, 1983,
p. 19. .

20

He begins by looking only at "...negative, collective, and external sanctions, and like most other

analysts we shall concentrate on the theory of economic sanctions."32

His theory for how
sanctions work is that of a simple economizing cost/ benefit analysis: that is, the sender wants to
spend as little as possible for the most effect. He does not go much further with this general part
of his theory; in fact, this functions as his base assumption.

Galtung explains his

Political Intoc'atlon

 

 

 

 

 

 

l
theory by comparing it to
what he calls the naive theory

Gal '
of sanctions. As he defines :3: w Tot-I
Sltl. 0 lelt Lll'l'llt Destruction > 3:337. tl
nai‘oe theory it explains Am \‘\\\ N
‘7

political disintegration of the Low um : ‘
target as a function of value

l
deprivation, brought on by poumcu OI-Intocr-tlon ———r.rq.u- Reaction to Sanction.

economic sanctions.” FIGURE 2'1 Galtung’s The")?

According to mice theory one would expect a direct linear relationship between the pressure put
on a country and the time it takes to fall. Galtung creates a revised theory that predicts the first
reaction to sanctioning will be a more determined resistance in the form of political integration,
followed eventually by disintegration if deprivation is increased. This is shown in Figure 2.1.“
Both the naive view, the dashed line, and the revised theory, the solid arc are shown here. The
revised theory expects the initial resistance to occur because of internal adaptation. There are
three conditions for this to occur: first economic sanctions must affect the entire nation, so those

who are not in favor of the targeted policy will feel unjustly punished, pushing them to support

 

32 Galtung, found in Nincic and Wallensteen, 1983, p. 21.

33 Galtung defines economic sanctions as a type of warfare, this has been ignored so to avoid using too
many different terms. As such, value deprivation is caused by all types of warfare, economic, military etc
which leads to political disintegration.

3‘ Redrawn from the original: Galtung, in Nincic and Wallensteen, 1983, p. 28.

21
the policy. Second, the pe0ple have no attraction to the sender, or they dislike the sender,
therefore they will be even more likely to side with their own government. Lastly there must be
a strong belief among the people that they are in the right. Galtung theorizes that if these three
conditions are met there will be a strong resistance to the sanctions.

Much of this has psychological underpinnings. Galtung argues that sanctions are much
like a prison sentence, the criminal rarely feels as though all of him is to blame for the crime and
yet all of him is being punished. The same applies for the target country, not all the people are
to blame yet all are being punished. The prison metaphor is expanded by viewing the target as
"...an organism with a certain self-maintaining potential."35 Although this is interesting and surely
somewhat true, Galtung does not seem to feel a need for a leakage into the target country. His
organism theory predicts that a nation could go from relatively free trade to autarky and survive
just by its society adapting to the change. This is overstated. Galtung does realize an upper limit
where the nation would fall before total destruction.

Galtung’s main conclusion is not that sanctions will not work, rather, that it is
understandable that sanctions do not often work given the adaptable nature of nations. Each
nation is different in this ability; the more diversified a nation, the more readily it can adapt.
Also, the more universal and severe the sanctions, the more a nation will have to adapt, and the
less likely it will be able to, as it reaches the upper limit quicker. But this does not mean a nation
is sure to fall just because near universal sanctioning exists; it can still adapt. Thus, the strength
of the organism is the true determinant of the success of nations, according to Galtung.

Jerrold Green

These sanctions originate within the international economic system, and while not

denying that "organisms have a certain self—maintaining potential", we must also

recognize that no degree of self-reinforcement will produce needed commodities
which might be denied a target state. In the throes of economic sanctions, it is

 

35 Galtung, found in-Nincic and Wallensteen, 1983, p. 46.

22

unlikely that joking, mysticism, or conspicuous sacrifice can produce desperately
needed fuel or food without which a society cannot exist.36

Jerrold Green's attack on Galtung is along the same lines as the critical comments made above.
Green’s primary critique of Galtung centers on his stress on the organism adaptation idea. He
counters with a theory explaining the failure of sanctions from a much more economic view. He
states two main ideas in reference to the target’s reaction. First, the demands of the sender
nations are often so austere that it is cheaper for the target to resist the sanctions and attempt to
survive them than actually give in and lose power. This is true in the cases where the demands
essentially equate the dismantlement of the target’ 9 power structure.

Second, regardless of the severity of the sanctions, and how many nations are engaged
as senders, the target can still circumvent the effect by finding one nation from which to buy and
sell.37 This is what Hufbauer et al call the black knight effect, one black knight nation can effectively
render a multi-lateral sanctioning effort useless. This is a type of economic adaptation, the target
seeks out and finds new markets, for without them it would be unable to exist. There have been
cases where it was actually profitable for the target to have sanctions imposed on it, the most
prominent case being that of the 1978 Arab League sanctions on Egypt for signing the peace treaty
with Israel. Egypt gained an average of $77 million annually for four years because they were
able to sell oil to the Israelis, a new market. This is something Galtung misses because of his
concentration on political integration. The economic adaptation by a country can add new
markets which in turn can bring in more businesses fostering support for the dominant regime.

A criticism of Galtung that Green misses is the question of actually how important
political integration is within a country. If the target country is a democracy where the people’s

voices make a difference and their opinions are formed independently from what the government

 

3‘ Green, jerrold D. Strategies for Evading Economic Sanctions, found in Nincic and Wallensteen, 1983, p.
64.

37 This does not include universal sanctions which have only been enforced once, against Iraq in 1990-
1991.

23
wants, then integration would be very important. But in a nation where the only news people
get originates from the dictatorship, sanctions will have much less effect because the government
has control of their beliefs. Thus, who cares if sanctions take effect against a nation full of
peasants who don’t know the difference between government repression and economic sanctions
when it comes to answering why they are starving.

Green’s analysis is far more persuasive than Galtung’s because he uses both the ideas
from the organism theory and the economic analysis of how a nation can survive and even
prosper under sanctions. Although it is a simple test, one can compare the organism theory to
Green’s by applying them both to the Arab League v. Egypt. Galtung would be unable to answer
the question of why they prospered, only why the sanctions did not work. But Green could
answer both questions, and more accurately. That is, Galtung would say that the nation
experienced increased political integration, but this is doubtful as most Muslims hate the Jews,
and as such would not be very likely to support their government. Galtung’s three criteria are
violated, regardless of who was affected by the sanctions; the Egyptians certainly had more
attraction to the brother Arabs than to the Israelis, and they were not wholly convinced that their
cause was right, as Sadat’ s assassination testifies. Thus, Galtung would be at a loss. Green on
the other hand would note the new oil market and could bypass any need to speak of political
integration. This is a much more plausible answer.

David Deese

At Least since the dawn of the modern nation—state in the 1500’s, wealth and

power have been the most prominent and interactive of national objectives. ...In

the twentieth century economic sanctions have become a tool of diplomacy and

warfare. Yet, carefully conceived and well executed economic diplomacy, or its

avoidance when appropriate, remains the exception rather than the rule in
international relations.”

As one can see above, David Deese comes from the Realist Tradition and discredits the

current effectiveness of sanctions. Deese does not discount their use, saying they are ineffectual

 

38 Deese, David A., found in Nincic and Wallensteen, 1983, p. 155.

24

tools of foreign policy; rather it is in the manner they have been used that renders them impotent.
He then engages in a brief analysis questioning what will make them effective.

Deese uses two helpful concepts to understand the effectiveness of sanctions. First is what
he calls Comparative Economic leverage, which is the same as a comparative advantage in trade only
it considers what countries have and need. For example, Japan is disadvantaged in food
production compared to the US, thus we could seriously damage the Japanese by halting all food
stuffs exported to them, just as the OPEC could damage just about any country in the world by
halting oil exports. This type of economic sanction is meant to be the most damaging and
destructive to a country and thus most likely to succeed.

There is a draw back to this type of sanctioning: it has escalatory properties. The truly
useful aspect of economic sanctions is their variability; that is, they can be instituted on many
different levels, both extreme and temperate. Only using those sanctions which are the most
effective at damaging a country one could actually cause that nation to attack militarily. Thus,
although the concept is useful there are many times when sanctions are not meant to be full scale
attacks on the target’ 3 economy. Since 1914, 33% of the sanctions levied have been against nations
which were on friendly terms, thus it was allies sanctioning each other. Deese misses this point
due to his concentration on sanctions directed at enemies only.

The second point that Deese makes is that sanctions are not always aimed directly at the
target’s policies but aimed at showing resolve to the sender’ s allies. Thus, the possibility of third
party consideration enters into the picture. There are two possible considerations that could
weigh on the sender country: the first is mentioned above, the idea of reenforcing allegiances; the
second is the possibility of creating a reputation among other nations so that in the future only
the threat of economic sanctions is necessary for changing the policies of the target. This point
Deese does not touch on directly, but one can see how it relates to the third party influence.

Keying into this last point is the idea that sanctions are not always meant to change a

policy, they could be targeted for one of the options mentioned above. Sanctions can be meant

25

to not only coerce the target to shift a policy but "...to punish, discredit, or embarrass."39 Here it
is understood that sanctions can take on a myriad of different functions, which runs somewhat
contradictory to the idea of effectively damaging an economy. This is an interesting contrast, one
that does not have an answer. Deese seems to take both sides of the argument implying that
sanctions are both an all or nothing ordeal, and they are also meant to punish, discredit, etc. One
would not think that sanctions only meant to embarrass would be implemented with the same
level of severity and determination as ones aimed at changing an hostile policy.

Taking the best from Deese, the idea of Comparative Economic Leverage, the range of
sanction severity and using the third party as a possible influence for the decision to employ
sanctions. By doing this the researcher is allowed to visualize the entire range of use for sanctions
as well as why they are used and when they may be the most effective. These ideas feed right
into the last author, David Baldwin.

David Baldwin

David Baldwin’s theory, found in Economic Statecraft, depicts for the Realist what
Margaret Doxey does for the Idealist. He argues that international economics should be viewed
as part of national foreign policy:

...not a bizarre, abnormal, nonroutine, extraordinary, unusual occurrence, but

rather a normal routine, everyday, ordinary, commonplace activity. Power

relations infuse every aspect of social life; there is no reason to make an exception

for international economic relations.‘0
This point is crucial to the study of economic sanctions because Baldwin is arguing against those
theorists who believe that sanctions are impotent while at the same time making a stand for the
Realist perspective. That is, Baldwin is not interested in the moral, ethical or legal questions of
the Idealists. Instead he finds the utility of economic statecraft a method of‘exercising power in

order to preserve national security.

 

3’ Deese, in Nincic and Wallensteen, 1983, p. 156.

‘° Baldwin, David A. Economic Statecraft, 1985, Princeton: Princeton University Press, pp. 60-61.

 

26

Although Baldwin defines economic statecraft in general terms: "Economic statecraft refers
to influence attempts relying primarily on resources which have a reasonable semblance of a
market price in terms of money.“1 I shall use this synonymously with sanctions. Actually
sanctions are a subset of statecraft, which can include foreign aid, import and export controls etc.
Everything said about statecraft applies to economic sanctions; for this reason they will be treated
synonymously.

His position on diplomatic, economic and military statecraft is that they are part of a
state’s instruments used to maintain itself. "Neither war nor economics can be divorced from
politics; each must be judged as an instrument serving the higher goals of the polity."42 There is
no reason to separate any of these tools from politics as they are inexorably intertwined. Baldwin
finds this to be the first reason for scholars nay-saying sanctions: they do not see economic
statecraft as foreign policy measures, instead they focus on economic statecraft as international
economics, i.e. trade policy etc. Once sanctions are seen as international economic relations and
not deliberate foreign policy goals they can fall into what Baldwin calls low politics, "...dealing with
such mundane and unimportant matters as trade and investment.“3 In so doing the researcher
cannot fully understand the proper setting in which to study sanctions. Along the same lines
Baldwin argues that statecraft is power, that is, it characterizes a nation’s power. This includes
all types of statecraft: "Economics as a separate science is unrealistic, and misleading if taken as
a guide in practice. It is one element—a very important element, it is true—in a wider study, the
science of power!“ Once again the study of Realist type issues are favored over the legal and
moral dilemmas of the Idealists. It is not the moral/ legal issues that drive nations, it is the quest

for power, the power that will insure their survival and success.

 

“ Baldwin, 1985, pp. 13-14.
‘2 Baldwin, 1985, p. 65.
‘3 Baldwin, 1985, p. 61.

“ Russell, Bertrand Power: A New Social Analysis, 1938, New York: W. W. Norton, p. 135.

27

Baldwin’s second critique of the nay-sayers, or reason they come to erroneous conclusions,
is that statecraft has a multitude of ends and means, not just one. "Means-ends analysis is
complicated by the fact that very few ends are ultimate or final values; most are intermediate or
instrumental in some sense.“5 Those analysts who attempt to simply use an ends-means analysis
to determine success of an economic venture will often find it to be a failure because of the many
different ends that can be studied. For example, if Nation A sanctions Nation B, is it a success
only if B changes its policies; or perhaps if C and D, A’s allies, are reassured that A will maintain
its alliance to them; or if E and F, two nations planning policies similar to 8’8, are deterred from
doing so.“ All of these are possible and as such the desired result of a sanction is not easily
detectable.

This keys in on another point: reputation in the international arena. "Images matter.
Statesmen care about how policy makers in other countries perceive their capabilities and
intentions. Economic instruments can be—and have been—used to affect images!” As Deese
alluded to in his work, the concept of reputation is vital to economic sanctions, an end that is not
easily determined. As mentioned above this is one of the main causes of sanction nay-saying.
Thus, Baldwin’s entire argument is focused on the study of economic statecraft as a part of foreign
policy, in arguing for this end he concludes with nine interesting ideas.

First, as discussed above, there are usually many targets and ends that a policy will aim
for. Second, "Success is a matter of degree." It is not often the case that sanctions, or any type of
statecraft, are dichotomously a victory or defeat. The Gulf War is a good example: immediately
following the collapse of the Iraqis, it was believed that total victory was ours, but now it is
apparent that although we won, there are still many of the problems that were there when the

war began. Third, alternative policy choices to economic statecraft are important. It is easy to

 

‘5 Baldwin, 1985, p. 16.

“3 These confusing alphabet nations point out that real world sanctions are complex affairs as well.

‘7 Baldwin, 1985, p. 101.

28

say that one type of policy will be ineffectual, but a much different thing to find a policy that will
work. Fourth, "A moderate degree of success in accomplishing a difficult task may seem more
impressive than an high degree of success in accomplishing an easy task." Fifth, as mentioned
above "Images matter." Sixth, "The basa of power are many and varied." This point addresses the
concept of power fungibility. Just as military power is not fungible across many areas, such as
protection of endangered species, neither is economic power.

Seventh, it is not always helpful to compare costs of the sender and the target. Even
though the sender may loose twice as much as the target, there still could be total success, and
even a net gain. This could occur because the target’ 5 economy may not be in a position to loose
much while the sender’ 5 economy is in such a position. The US could loose a great deal if it
imposed sanctions on the member countries of OPEC, but if the sanctions were successful, then
the long run could benefit the US This relates directly to the idea of multiple goals, the cost to
a sender could be tremendous, but not compared to the reputation it can develop. Eighth,
"Imposing costs for noncompliance is a measure of success." That is, just being able to punish a country
for noncompliance by causing it economic damage is successful as the sender has shown resolve,
power to influence and dissatisfaction. Lastly, and related to eight, "Costs have their usage." Thus
it is useful for the sake of reputation to show that you are willing to pay the price to impose a
sanction.“

Baldwin’s theory of economic statecraft is far more systematic than the Idealist tradition
because it ignores the moral, ethical and legal grounds of sanctions and concentrates on the
economic and political aspects. This makes much more sense because nations do not enact
sanctions for ethical reasons, they do it for reasons of power and national security. This may
sound cynical, but in fact it is irrational to enact a sanction that will impose heavy costs on the

sender without having any net benefit. In cases where morality seems to be the cause of the

 

‘3 All the above quotes may be found in Baldwin, 1985, pp. 371-372.

 

29
sanctions I suggest looking for other reasons, as Baldwin points out, there are many possible
targets.

This is not to say that the Realism school is the best option, or that Baldwin subscribes
to it wholly. International organizations play a key role in facilitating the relations between
nations, and should not be discounted as much as the Realists do. Instead a union of the two
should be adopted, much like Baldwin has done. These ideas are more fully developed in the last
section where possible theories are considered.

Problems defining economic sanctions

A major question which has remained unanswered through the literary review is: what
is really the definition of economic sanctions? This is not easily answered, but an attempt is made
to do so here. Kim Richard Nossal addresses this point, and lays out an argument contending
that sanctions are international punishments for some sort of wrong doing. This brings us
curiously back to Doxey and the Idealists for whom moral grounds were needed to make
economic sanctions anything more than coercion. As I understand Nossal, this is not the point
he is attempting to make. Instead he is attempting to distinguish between sanctions and statecraft:

Thus, all coercive economic policy instruments—embargoes, boycotts and indeed

the disruption in the "normal" or "customary" levels of economic intercourse

between states—are too often simply described as economic "sanctions"."

Thus, to N ossal, sanctions are punishments for wrong doing. This raises a touchy subject
though: what is wrong?

...the perception that the target state has violated norms of moral behavior valued

by the sender and thus deserves not only concrete penalties but also a public

proclamation of the target's impiety.so
This answer asks yet another question: What is moral behavior? Before delving into a deep

argument on international morality I shall side with the Realist perspective and assume that there

 

‘9 Nossal, Kim Richard International Sanctions as International Punishment International Organization, v.
43. no. 2 Spring 1989, p. 305.

5° Nossal, 1989, p. 306.

 

30

is no such thing. This may sound as though I am refuting Nossal’s definition of economic
sanctions, but I am not. I agree that sanctions are actions taken against a target because the
sender feels wronged in some way. But wronged can mean threatened, disapproving, angry, hurt
etc. Nations will always find reasons for their actions. For example, the US sanctions against the
Dominican Republic in 1960-1962 were said to have been enacted because of the Dominican
Republic’s hostilities toward Venezuela. But another possible reason is the Trujillo resistance to
US demands for change. Nations act out of a desire to survive and will create reasons for their
actions.

Nossal’s next point concentrates on the punishment aspect of sanctions. The Latin root
of the word sanction is Sanctio, meaning punishment for some kind of violations"1 But this
punishment is not to be equated as sadism; punishment has purpose. Nossal finds three purposes
of sanctions: first, as deterrence or prevention; second, as compulsion to end a behavior; and last,
as retribution. The first two are encompassed by Baldwin, and as I shall show below, Hufbauer
et al. These types are intuitive and are directly observable; the last type, retribution, is not.

Retribution plays an important role because it affects domestic politics, and the reputation
of allies. First of all, forceful expression of national resolve and power always seems to have a
rally around the flag effect in domestic politics. This has often been the purpose of many sanctions;
e. g., COMECON.52 Second, retributive sanctions relay a message to the allies of the sender nation
that resolve is still present.53 That is, the allied nations will perceive the sanctions as a show of

strength towards offending nations. This in turn is interpreted as the nation having strength

towards the alliance.

 

5‘ Galtung also refers to sanctions as a punishment when depicting the resistance of a nation under
sanctions as a prisoner.

5’ Not only did the US want to hurt the USSR, but show the public that their government was fighting
the Communists actively.

9 COMECON is also an example of this.

31

This type of sanction is not covered by Baldwin or Hufbauer et al, or for that matter any
of the scholars reviewed. Part of this is due to the need to shift one’s focus from sanctions
directed at policy change, the hub of Baldwin’s theory, to one that can encompass domestic
politics and international reputation. The international reputation easily falls within the policy
change rubric, but domestic politics does not. Although this point is extracted from Nossal’s
concept of retribution, it is unique and one of the more useful points brought out in his work.
Definition of economic sanctions

Using the ideas above, I shall attempt to formulate a definition now. This should act as
an heuristic during the critique of Hufbauer et al, as well as the model testing. Economic
sanctions are best thought of as a step on a ladder of sanctions, or what Baldwin calls statecraft. On
the bottom of this ladder are diplomatic sanctions, the least destructive, and least austere. On the
top are military sanctions, which include full scale war, the most severe and destructive. Between
these two extremes lie economic sanctions. As a nation becomes involved in a conflict with
another nation, they begin to climb this ladder, first diplomatic sanctions are placed into effect,
next economic and finally military sanctions. This is intended to show the reader how sanctions
will build on themselves. For example, nations do not usually engage in war without first
imposing economic sanctions on their enemy if possible.

Now that economic sanctions have been placed into a context, let me define what I will
consider to be a sanction. Any economic measure imposed by a nation(s) against another
nation(s) aimed at harming them in some way, and provoked by some change in the target’ 5
behavior. This includes any change that may not truly provoke the sender, only give the nation
a reason to impose sanctions. This definition allows for all sorts of conditions: threats, breech of
international norms, simple power plays by the sender nation, etc. Nossal would not agree totally
with this definition because it includes any change in the sender’ s policies that will harm the
target, including the cessation of trade for reasons of economic compcli“ Although the

purpose of this thesis is not to create a new theory, but test an existing one, one restriction should

32

be put on this definition. I do not propose that normal trade relations, including the trivial tit-for-
tat game that occurs during two nations’ trading, to be considered sanctions. This is not meant
to define sanctions by their severity, but to exclude normal adjustments that nations undergo to
stabilize their trade relations. Thus, the definition centers around the motive of the sender nation,
not acuteness of the sanctions.

Using this definition would expand what are considered sanctions, and effectually give
us a better understanding of the dynamics involved. As mentioned above, this is not followed
up in this study because of the thrust to test Hufbauer et al’s work. It does hold promise for
future research though, which shall be discussed. I now examine, analyze and critique Hufbauer

et al’s work, after which I begin the process of modeling and testing the hypothesis.

 

 

 

33
Chapter III

Hufbauer, Schott and Elliott

As mentioned above in the introduction, the question Hufbauer, Schott and Elliott address
in Economic Sanctions Reconsidered is not the question if economic sanctions work, but when they
work. They attempt to determine what is it that make sanctions most likely to succeed, most like
a policy cookbook, or a recipe for a succasful sanction. As one can imagine, their work is very closely
based on policy recommendations, as their nine commandments attest. Although they focus
primarily on the U.S., their model, data and testing encompass the whole world. This focus is due
partly to the overwhelming majority of sanctions in which the US was involved.“

The cookbook-like focus of Hufbauer et al makes their work not so much a theory that is
being tested, but a moderately systematic exploration of sanctions. Strictly speaking this is not
a theory, nor do I consider it one. But for reasons of testing their work, their recipe for a successful
sanction I shall act as though it is one. This entails testing Hufbauer et al’s results and deriving
conclusions from their work. It, however, does not include referring to it as a theory.

Hufbauer et al’s work is based on the following two basic questions which create their
recipe for successful sanctions: what contributes to success and what costs are incurred. Were it
possible to put their results into a sentence it would read: nations that are larger, patient, act
swiftly and stemly and do not attempt too much, succeed the most. The assumptions that
underlie these results are in the form of definitions. Hufbauer et al’s first assumption regards the
motives behind the study, efficiency. That is, they only want to find the most effective manner
in which to impose sanctions successfully. This is a clear break from the Idealist who concerns
herself with questions of morality and ethics. Hufbauer et al have no such agenda. The second
assumption/ definition is Hufbauer et al’s definition of sanctions. They cite three possible grounds

for imposing a sanction: "Demonstration of resolve...", "Deterrence..." and as a form of punishment

 

5‘ Almost two thirds of all sanctions involved the US. as a sender.

 

 

 

 

34

that "...add[s] teeth to international diplomacy“. Later they define sanctions as having five

possible goals:

Change target-country policies in a relatively modest way... Destabilize the target

government... Disrupt a minor military adventure... Impair the military potential

of the target country... Change target-country policies in a major way... 5"
There is a problem here: the goals and grounds for imposition are not fully connected. None of
these policy goals fall specifically into the original three reasons for the use of sanctions, and the
reader is left to interpret how this happens. This discrepancy is troublesome when attempting
to pin down Hufbauer et al’s definition of sanctions. The solution here is not to concentrate on
the goals or grounds for sanctions, but to derive just what qualifies as sanctions for them. For
Hufbauer et a1 sanctions would include any harmful economic measure implemented on the
instigation of the target’s policies and/ or actions. This does include sanctions aimed primarily
at alliance reinforcement.

Indeed, the international community often expects such action from the United

States, to demonstrate moral outrage and to reassure the alliance that America

will stand by its international commitments.”
The definition does not include sanctions being used to bolster public Opinion, something that
makes Hufbauer et al’s results suffer from its absence. Once again this thesis is unable to deal
with this problem as a result of its limited scope, nonetheless it will be followed up in the future.

Hufbauer et al are very explicit in forming their model that tests the questions previously
laid out. The level of analysis used constantly throughout their work is the nation state.
Although there is some brief mention of domestic interaction with economic sanctions, it is not
followed up, nor is it tested. As mentioned above this is a failure of their theory. This level of

analysis is common to just about all the authors reviewed, including Galtung, who views the

nation state as an organism, and to Doxey, who concentrates on international organizations but

 

s Hufbauer, Schott and Elliott, 1990, p. 11.
5‘ Hufbauer, Schott and Elliott, 1990, p. 38.

57 Hufbauer, Schott and Elliott, 1990, p. 11.

 

 

 

 

35

still sees the world as made up of national actors. I find only slight fault with this, and that is
the ignorance of how domestic politics and economic sanctions interact. A two level game should
be set up to more fully encompass this phenomenon, and enable the researcher to test the
relationships.58

The model is broken into three categories: foreign policy goals, political variables and
economic variables. These account for the specific types of policies initiated, the past relationships
of the countries and the international setting, as well as the economic costs and relations that exist.
The types of foreign policy goals are set out above, while the political and economic variables will
be specified below in the description and definition of the quantitative model. Hufbauer et al’s
use of case studies as a means to derive their results is done in a systematic manner, and there
are enough cases where the conclusions are generalizable.“

The testing done by Hufbauer et a1 has many faults, but before they are analyzed a short
diatribe. Methods are important tools that we as social scientists use, but they are only a tool and
not a substitute for substantive theory, which should always be our primary focus. This is not
meant to discount the importance of using the correct method for the job, one should attempt to
be as accurate as possible. However, if some reasonable amount of accuracy must be initially
sacrificed for interesting theory, one that is difficult to measure, then it is better to go ahead and
improve on it later. There are many faults with Hufbauer et al’s methods but one must not loose
sight of the substance they created, for that is the important part. Having said this I will now
attack their methods, and later I will make an attempt to improve them through scientific testing.

There are three main points of contention I have with Hufbauer et al’s methods. First is

their measure of duration. They want to use whole years as a unit, so they cut off any months

 

5"For a discussion of two-level games see: Putnam, Robert D. Diplomacy and Domestic Politics: The Logic
of Two-Level Games Integational Organization, 1988, v. 42 no. 3, 427-460.

5’ Hufbauer et a1 imply that their compendium of cases includes all sanctions from 1914 to 1991.
Following their definition this is probably the case, however, as mentioned above, definitions are difficult

to agree on. My feeling is there are more than 115 cases, but only if one changes the definition to the one
I propose.

 

 

 

 

36

left over or round up. Yet when they begin their analysis they talk about sanctions lasting 2.1
years. Does this mean they lasted 2 years, 36 days and 12 hours? Instead of complicating this
problem they should have used months as an even unit. The reason they give is to avoid the
problems with dating the beginning and termination of sanctions, but it is far more inaccurate to
round off to whole years than to attempt using months. Although this may not effect the
calculated results much, there is no way to tell. Their reason is poor and the problem should be
corrected.

The second problem Hufbauer et al have is the success rating used is crude and leaves
little room for variation. They use a four point scale to determine whether a policy was successful
or not, and one to determine how much the sanctions contributed. These scales are created by
examining expert opinions on each case. These two scales are then multiplied resulting in the
measure for success. Hufbauer et al then set the minimum value for success at 9, moderate
success and moderate contribution by the sanctions. By setting the cut off level at 9 however, they
leave only four possible scenarios that equate to success: the sanctions must have had a modest
or significant contribution completing policy goals either somewhat or totally. This does not allow
for enough variation in the testing, making the results difficult to interpret. Also it seems
atheoretical to simply decide to multiply two scores together without a reason. Because of this
there is no way to evaluate whether or not the data is biased. This possible bias could be
mitigated somewhat by adding variability in the success variable. That is, as more room was
created on the scale of success different cases could be ordered in levels of their respective
successes. Thus, 115 cases would not be rammed into only 9 possible outcomes.

The last critique is the most problematic, Hufbauer et a1 use their data as if they were
interval, while much of the data are really ordinal. There is, however, a reasonably strong defense

for this, though. The data that are specifically the problem are the measure of success. It falls

 

 

 

 

37

into integers ranging from 1 to 16."0 These are continuous, and ordered only in the sense that
higher values indicate higher levels of success. Thus, one could interpret a mean of 6.83 as a
sanction that has had minor to moderate contribution to a result that is positive or almost so; that
is, the mean sanction is only moderately successful. Although mathematically this is inaccurate
because the measure cannot take on such a value, still intuitively it can be so interpreted. I am
not advocating this kind of use in statistics on a regular basis, I am however offering it as an
initial point at which to start. The problem of measures of success is their inherently inescapable
subjective bias; that is, there is no possible way to measure the success of a sanctioning episode
objectively. The only solution to this is to systematically define how sanctions will be measured.
That way all of the sanctions rated will be done so equally, and treated not in an ad hoc matter,
but in a methodical one. This is what Hufbauer et al have done, and although the measure needs
improvement, it will have to wait.

Hufbauer et al’s recipe, and the testing of it, are done in a systematic manner which allows
for falsification and later testing, of which this thesis is proof. Given their faults, the study is a
sound one and I would not expect to see much failure in it. However, as I show below there are
some .significant differences in the results I find and the ones they report. The rest of the thesis
will be devoted to this cause: defining the model and variables, discussing the expected results,

and then testing the model. So without further delay, let’s jump right into the fun stuff.

 

6° There are 9 different values that the measure can take on, this is caused by the multiplication used to
create the measure. This will be discussed formally below.

 

38

Chapter IV
Quantitative Model

To test Hufbauer, Schott and Elliott’s results, a specific model will be developed, following
what were referred to as the nine commandments in the introduction. First, the variables are
defined and described, second the model is explicated and last I write out the expected results.
List and definition of variables

There are 13 variables, not including lags or transformations, that will used to test
Hufbauer et al’s work. The dependant variable, SUCCESS, is one of the most problematic not
only for the lack of variation as discussed above but for the manner in which it was created. A
quick recap may help. Hufbauer et al used two four point scales, one for achievement of the
objective and one for the contribution the sanctions made. These were determined by the expert
opinions of several scholars in each case.61 These two values were then multiplied, implying an
interactive relationship. But in the process of doing this the number of possible values for a case
to occupy is only 9. To rectify this problem I have recoded the variable, now RSUCCESS, to make

it contiguous, the values are shown in Table l:

 

II success 1 2 3 4 6 8 9 12 16 H
H Rsuccrsss 1 2 3 4 5 6 7 s 9 II

 

 

 

 

 

 

 

 

 

 

 

The variable is now contiguous. The only possible qualitative effect on the variable from this
transformation is if the non-recoded values had some intuitive meaning. For example, if a 12 is
twice as successful as 6 instead of the difference between 5 and 8.62 There is no evidence to
support this ether way, and Hufbauer et al make no mention of its significance. For these reasons

I shall use the recoded variable.

 

‘1 Hufbauer, Schott and Elliott, 1990, p. 41-42.

‘2 I thank Scott Gates for pointing this out to me.

 

 

 

39
Of the 13 variables there are 5 dummy variables. The first is BLKNIGHT. This refers to

the presence of a nation or nations who openly aid the target nation, in effect becoming the black
knight to the sender country. This does not include covert aid to the target, instead "...we are
concerned with overt economic or military aid to the target country in response to the imposition
of sanctions."63 The second dummy is COOP, international cooperation on the side of the sender
nation. This is just the opposite as the black knight effect, although both can occur at the same
time. Cooperation is considered to be either zero or modest: "...meaningful restraints... from some
but not all the important trading partners of the target country" to significant: "the major trading
partners make a major effort to limit trade, although leakages may still exist through neutral
countries“

The economic vitality and political stability of the target are accounted for by the DSTRES
variable. A country is considered distressed when it has "..acute economic problems, exemplified
by high unemployment and rampant inflation, coupled with political turmoil bordering on
chaos"65 The RELATE dummy refers to the prior relations between the sender and the target,
specifically if the relations were friendly. That is, considered "...close friends and allies"“ The
last dummy variable, POLICIES, determines if the sender nation used any other policies in
conjunction with the economic sanctions. This includes covert, quasi-military and full scale
military sanctions.

The variables that remain fall into two groups, those that are characteristic of the nations
involved and those that concern the consequences of each sanction. In the first group PERCENT'M

and PERCENTX are measures of the percent of pre-sanction imports coming from the sender

 

‘3 Hufbauer, Schott and Elliott, 1990, p. 45.
6‘ Hufbauer, Schott and Elliott, 1990, p. 44.
‘5 Hufbauer, Schott and Elliott, 1990, p. 46.

6‘ Hufbauer, Schott and Elliott, 1990, p. 47.

 

 

 

 

40

nation(s) and the exports being received by the senders, respectively. GNPRATIO is a measure
of sender to target GNP, a relative measure of the nations’ strength."7

The second group begins with TARCST, which is the actual cost to the target in current
dollars (measured in millions). This does not include any political costs to the target, only
economic ones. The methods used in determining this is simply determining the pre-sanction
supply and the post-sanction implementation supply, and then determining the costs from the lost
supply.68 Similarly, GNPCST is the same figure only controlling for the target’s GNP, thus cost
as a percent of GNP.

The last two variables are LENGTH and COST, where LENGTH is simply the duration
of the sanctions in years. COST is an analysis of gains or losses to the sender, as rated on a four
point scale, four being a major loss and one being a net gain.

Description of model

Creating the model is a rather simple task because the variables and theory come from

Hufbauer et al’s work. In addition to Hufbauer et al’s model, past successes (RSUCCESS) will
be added along with lagged values of COST, GNPCST, TARCST and LENGTH. This will
determine whether nations have a tendency to look back on past sanctions and react to them.
That is, do past sanctions effect Nation B’s decision to resist Nation A, or not. The initial
unreduced model will be the following:
RSUCCESS = BLKNIGHT + COOP + DSTRES + POLICIES + RELATE + GNPRATTO +
PERCENTM + PERCENTX + COST(AND LAGGED COST) + GNPCST(AND LAGGED GNPCST)
+ LENGTH(AND LAGGED LENGTH) + TARCST(AND LAGGED TARCST) + LAGGED
RSUCCESS

There are eight variables that theoretically cannot be lagged: the dummies and

GNPRATIO, PERCENTM and PERCENTX. These are situational factors, that is, they comprise

 

‘7 All three of these measures are discussed briefly in Hufbauer, Schott and Elliott, 1990, p. 48.

6‘ For a full discussion of this see Appendix A, Hufbauer, Schott and Elliott, 1990, pp. 120-122.

 

 

 

f. ,

 

41
the environment in which the sanctions take place, past values should not effect them. For
example, it is expected that Nation B will be concerned with the successes, costs and lengths of
past sanctions, but not with the characteristics of the target and sender. Thus, these variables will
not be lagged.

One important note must be made about the lagged variables. The cases are ordered by
the time in which they occurred, not by each year. That is, some years no sanctions occurred and
in others many did. The most sanctions imposed in a single year was in 1978 when 8 sanctions
were imposed. The mean sanctions applied in a given year from 1914 to 1990 is slightly over 1.5,
but as one can see this is meaningless, much as the average family with 2.4 children. Thus, the
lags are event lags, past sanctions, whether that means in the same year, or several years in the
past. The concept behind the lags is that the target country will look back to the last few
sanctions, not the last few time periods, to estimate what consequences of the sanctions being
presently imposed may be.

Expected results

I now lay down the results I expected to find when testing the model, this is based on the
Nine Commandments from Hufbauer et al, and additional expectations on how the lags work. In
order to do this I discuss what each variable was expected to do and then summarize these
expectations.

Of the dummy variables Hufbauer et a1 expect that BLKNIGHT, DSTRES, POLICIES and
RELATE all to be significant in estimating success. They find COOP to be insignificant. This
insignificant result could be due to the black knight effect, which is expected to be strongly
negative. The economic and political stability of the target should have a weakly positive effect
on sanctions, that is, a distressed nation will be more likely to fold under sanctions. Non-
economic policies should have a reasonably positive effect as should the relations between sender

and target. These relate to the ideas of choosing the right tools for the job and picking on your friends.

 

 

 

 

42

In considering the rest of the variables all were expected to be significant. The greater the
size difference between the two nations and their previous attachments, measured by GNPRAT‘IO,
PERCENTM and PERCENTX, respectively, should all have positive effects on success. The two
variables, ATARCST and AGNPCST, essentially measure the same thing, cost to the target. These
should positively effect success as well. COST is expected to be significant but on closer
examination of their tables it seems that it is borderline.69 The cost should have a negative effect
on success, but I am dubious of its significance. The last variable Hufbauer et al use is
LENGTH.7o Unlike the other variables, the length of the sanctions has a negative effect on
success, thus the quote: "If It Were Done, When 'Tis Done, Then ’Twere Well It Were Done Quickly "71

Thus, sanctions which hit hard and quick, and are directed against a nation which has
close, friendly ties with the sender, and is isolated, are the most likely to succeed.

The additional variables that are added are the lagged values of COST, GNPCST,
LENGTH, TARCST and RSUCCESS. I expect that nations look back on the results of past
sanctions, and their factors. That is, Nation B, when deciding to resist sanctions from Nation A,
will look back and see if past sanctions were successful, long lived and / or costly to either target
or sender. GNPCST, TARCST and RSUCCESS should all positively effect success, with a
diminishing effect. Lagged COST and LENGTH, unlike their present values, should also have a
positive effect on success. This is due to the target nation seeing the resolution of past senders,
that is they were willing to bear heavy costs and wait out the sanctions for a long time.

Now with the model specified, and the expected results laid out we can move on to the

statistical testing.

 

‘9 Hufbauer, Schott and Elliott, 1990, pp. 92-106.

7° Due to restrictions in methodology and data the type of sanction and goal cannot be tested.

7‘ Hufbauer, Schott and Elliott, 1990, p. 100.

 

 

 

43

Chapter V
Statistical results and problems encountered

I have taken an econometric approach for the data analysis as developed by Robert Engle,
David Hendry and Jean-Francois Richard.72 I have had to modify this time series approach
somewhat. As mentioned above the time is not a function of years, months or for that matter any
time, it is a function of sanctions. One could argue that this is an event count model, which it is
in many ways.’3 But on the same grounds all data are event count, for anything that happens
is an event. This point is more nit picking than helpful though, so let us move on.

The original model I began with contains 30 variables, including lags and a constant.
Listed below are the variables and their respective signs, I shall refrain from listing their
coefficients and t-scores until I begin the reductions. Each equation, that is all the print out, can
be found in the Appendix A.
RSUCCESS = CONSTANT - BLKNIGHT' - COOP + DSTRES - POLICIES + RELATE - COST -
COSTl + COST2 + COSTS + GNPCST - GNPCST] + GNPCST2 - GNPCST3 + GNPRATIO -
LENGTH + LENGTH] - LENGTI-IZ -LENGTH3 + LENGTH4 + PERCENTM + PERCENTX +
TARCST - TARCSTI + TARCST2 - TARCST3 + RSUCCESSl - RSUCCESSZ + RSUCCESS3 +
RSUCCESS4

The reader will note that the dynamic specification is AD(1,1). I found at this point in
time this specification was more appropriate than the Error Correction Model(ECM) or Partial
Adjustment(PA) models due to the requirements for common factors." I do intend to use the

ECM in the future because of its theoretical attractions.

 

7’ Engle, Robert and David Hendry Testing Super Exogeneity and Invariance in Regression Models Applied
Economics Discussion Paper no. 100, University of Oxford, 1990; and Engle, Robert, David Hendry and Jean-
Francois Richard Exogeneity Econometricg, v. 51, 1983, pp. 277-304.

73 The data will be modeled as an event count following the econometric model.

7‘ For a full discussion of this see: Beck, Nathanial Comparing Dynamic Specifications: The Case of Presidential
Approval Forth Corning: Political Analysis, March 1991.

 

 

44

In reducing the model I used 6 different equations.’5 I shall now list the reductions
along with their coefficients, t-statistics and the new, that is post-reduced, residual sum of
squares, model variance and Schwarz Criteria. The first reductions, shown in Table 2 were to

remove GNPCST, and its three lags, GNPRATIO, LENGTH and RSUCCESSI.

 

 

 

 

 

 

 

 

 

Variable fl t-stat
GNPCST .0558771 .92841
GNPCST] -.0358706 -.35541
GNPCST2 .0126209 .12142
GNPCST3 -.1150506 -1.20178

GNPRATIO .0000189 .22310
LENGTH3 -.0199133 -.55236

RSUCCESSI .0315563 .28559

RSS = 502.84 0’ = 2.3904 SC = 2.49

 

 

 

 

 

The initial values, before the reduction were: RSS = 484.32, 0 = 2.4452, sc = 2.75, R2 = .42056, F(29,
81) = 2.03, DW = 2.121. This first round of reductions is very surprising because both Hufbauer,
Schott and Elliott’ 8 ideas of relative size and cost to the target show themselves as unimportant,
as well as my idea of past values of success holding some importance. The reason that
GNPRATTO was found to be insignificant is probably due to its large range and not the fact that
relative size matters when applying sanctions. For all observations the minimum value for
GNPRATIO is .1, while the maximum is 32900, with a standard deviation of 3103. If you restrict
the values to be under 2500, the standard deviation falls to 384 while only loosing five
observations. After restricting the variable to under 1000 the standard deviation becomes 224,
while loosing only nine observations. When the regression analysis was re-run with a limit of

1006 (there was only one observation that close to the cut off) GNPRATIO was still insignificant.

 

75 In the Appendix there are more than 6 reductions, for purposes of reporting the results however I only
listed 6 to avoid being tedious.

 

 

 

— T—_________———

45

Thus, I must conclude that the variable is not important in its present state. However statistics
should not be taken as the word of God, so considering this and the fact that relative size seems
intuitively important, I suggest that in the future it should be remeasured to see if it is important
after all.

The problems with the ratio of GNP’s is a difficult quandary. The use of the measure is
not unusual. Organski and Kugler used total national output as a measure for national power
in The War Ledger, actually finding little difference between the Singer-Bremer-Stuckey model
which accounts for many aspects of a nation’s power and GNP.’6 Although this was used as a
measure for military power, it should hold for economic power as well. Thus, a ratio between
the sender and target should be a good measure. The point here is that it is not the measure, but
the great variation in national sizes. Two possible solutions come to mind. First, instead of using
a ratio, the straight GNP figures could be used as variables. Unfortunately this gives no idea of
relative size since the variables are separated. The second possibility is to reverse the direction
the ratio is calculated in, that is instead of it being sender/ target, it could be calculated as a
relative weakness variable: target/ sender. This would crush the variation down, such that. 10 2
8 2 .000000303951, where 8 is the new ratio. Although this is a little hocus pocus, GNP is a good
measure of power, and reversing the ratio seems like it solve the problem. This shall not be done
here but is reserved for, and is currently being calculated, other work.

GNPCST was also reduced, but it does not suffer from the same problem as GNPRATIO.
This may lend us the possibility that the decision maker is not told the costs of the sanctions
relative to the nation’s GNP but as a total cost, TARCST. This should not be the case however,
as one would expect to find a decision maker to generally know what her nation’s GNP is, and

as such would take it in to account. I can only hypothesize that once again the data is not

 

7‘ Organski, A.F.K. and Jacek Kugler The War Ledger, 1980, Chicago: University of Chicago Press, pp.
33-38.

A

 

 

 

 

46
accurate in measuring the concept. Cost to the target, TARCST remains significant throughout
the reductions.

LENGTH 3 also was dropped, this is not counter to the expectations, as it seems the
decision makers do not look back to the third sanction. As mentioned above they are expected
to have some effect, but at which point is unknown. This is also true with RSUCCESS 1. Until
all the lags have been deleted I shall not make any conclusions about their effectiveness, or lack
thereof.

The tests for proper reductions were satisfied, the Schwarz Criteria went down, as did the
0, while the residual sum of squares went up. Thus the first round of reductions was successful.

The second round of reductions eliminated the following variables in Table 3:

 

 

 

 

 

 

 

 

Variable [3 t-stat
POLICIES -.5641345 -.90012
RELATE .1554511 .27701
COST -.2886574 -.77002
RSUCCESS 2 -.1237286 4.30042
RSUCCESS 3 .0101487 .10941
RSUCCESS 4 -.0010201 -.01123
RSS = 525684 0' = 2.3648 SC = 2.28

 

 

 

 

 

This is one of the most surprising set of reductions as four main expectations fall out. First, it
seems that added policies are do not significantly change the chances for success of sanctions.
I can think of no counter argument to this result so I must conclude that policies do not effect
the success of sanctions. This is also true of the prior relations between the sender and target.
Both of these results are a complete surprise and bear witness for the need for further research
and especially theoretical work on sanctions.

The current cost to the sender country was dropped, but this is less of a surprise. It

seems that countries often will maintain a course of action once they commit even if it is costly

 

 

 

47

to them; Also, the political costs of terminating the sanctions may outweigh the fiscal costs of
maintaining them. Thus, the cost they are bearing has no influence on their decision to cease the
sanctions.

Lastly, all lagged values of RSUCCESS proved to be unimportant to the decision maker.
This seems mainly the result of the decision maker’ s lack of interest in the outcomes of a previous
sanctions because the past target could be totally different from the present decision maker. What
is of interest is the past lengths the sender went to punish the target. This shall be shown in the
final model.

- The third group of reductions are shown below in Table 4:

 

 

 

 

 

 

 

Variable f3 t-stat
COST 3 .0939018 .29828
LENGTH -.0318588 -1.05230
LENGTH 1 .0500019 1.72475
PERCENTX .0123832 .63382
TARCST 2 .0002646 .71805
RSS = 548.34 0' = 2.3534 SC = 2.11

 

 

 

 

 

This round proved to be less surprising. As mentioned above I will ignore the reduction of any
lagged variables until there are none left. LENGTH has been deleted showing that there is little
effect on the amount of time that sanctions are imposed. This could be caused by the variation
in the length, where the maximum is 44 years, and the mean is only 6.052. When the value was
restricted to less than or equal to 10 years, the mean became 3.152 and the skewness dropped
from 2.845 to 1.089, while only loosing 16 observations. Although this may be the cause, one
cannot delete every outlier to make the goodness of fit better, all the while it is worth noting.
The other non-lagged variable that was removed was PERCENTX. Although Hufbauer

et a1 predicted this to remain important they also combined both the percentages of imports and

 

 

48

exports in their analysis?7 This result is not that surprising so long as the percent of imports
remains significant, which it does. The conclusion to be drawn here is that the percent of
exported goods to the sender is not as important a tie as imported goods are. This is possibly
because export markets are easier to replace than finding importers.

The fourth round of reductions leaves us with only eight variables in Table 4:

 

 

 

 

 

 

Variable f3 t-stat
COST 1 ' -.3802970 -1 27243
LENGTH 2 -.0390831 -1.41936
TARCST 1 -.005914 -1.65252
TARCST 3 -.0005596 -1.534485
RSS = 595.45 0 = 2.4044 sc = 2.02

 

 

 

 

 

All the lags of TARCST are now gone, obviously the decision maker does not look back on what
past target’ s lost. This could be a result of the decision maker seeing his costs as being unique,
unlike the extent to which the sender will go, as that seems to weigh on his decision.

For the first time there is a sign that we may have reduced to far, the variance increased.
To check this an F—test was run comparing all past models with this one, at no point did the P-
value reach .05, the highest it attained was .0831, showing no threat to the reductions. Backing
this up is the lower value of the Schwarz Criteria and the increase in the residual sum of squares.

The last reduction made was to remove COOP, whose 13 = -.7545279, and the t-stat = -
1.33018. As mentioned when discussing the expected results, international cooperation was
supposed to drop out. Although it persisted to the last reduction, it did follow the expectations.

Still, this deserves further research, perhaps, if it did prove to be a significant factor. The saying:

 

” This was called "Average trade linkage (percentage of total trade)", Hufbauer, Schott and Elliott, p.
99. I find it more accurate and informative to separate imports and exports.

 

 

 

49
”More Is Not Necessarily Merrier"78 should be amended to say Too Many Cooks Spoil the Soup.

Needless to say this possibility should be researched further.

The final model that is left after the 30 variables is reported below in Table 5:

 

 

 

 

 

 

 

 

 

Variable f? t-stat
Constant 2.2579175 3.24880
BLKNIGHT -1.6686837 -2.99202
DSTRES 1.4482112 2.53425
COST 2 .7906051 2.64900
LENGTH 4 .0650850 2.43503
PERCENTM .0331413 3.18519
TARCST .0008640 2.32053
RSS = 605.68 a = 2.4132 SC = 1.99

 

 

 

 

 

As one can see from the table below reporting the results of the diagnostic tests, the equation is
significant, but the goodness of fit is not that impressive, according to the R2. There is certainly
some room for improvement. For tests of first order autocorrelation I used the Durbin-Watson
statistic, and then preceded to test up to the seventh order, AR 1 - 7, the results of which are
distributed as an F-test. I confidently reject the possibility that there is any error autocorrelation
present in the data. Next was a test for normality, checking for small sample inference, which
was not significant. I consider my sample size large, but according to Peter Schmidt, all
researchers do, thus I tested for normality.79 There was a slight problem with excess kurtosis,
which is -.796385, but the curve is normal with almost no skew to it whatsoever. At this point
I tested for auto regressive conditional heteroskedasticity, or ARCH. This is best shown in Figure

5.1,30 notice there is no regular grouping of residuals, that is, the large residuals do not cluster

 

7’ Hufbauer, Schott and Elliott, 1990, p. 95.
7’ Stated in an Econometrics class, Michigan State University, Spring, 1992.

3° FIGURES 5.1 through 5.5 are found in Appendix A.

50

together nor do the small residuals. The test was negative as well. Lastly I checked for mis-
specification using the RESET tests, the general test for functional form and of course the Schwarz
Criterion throughout the reductions. These tests are really aimed at discovering an omitted

variable problem, which was not found. The results are shown in Table 6:

 

 

 

 

 

 

 

 

 

 

 

 

 

R2 .2753667

F (6, 104) 6.59
Durbin-Watson 2.164
AR 1-7(7, 97) .44
Normality-C1190) 2.790

Func. Form(25, 77) 1.3536
ARCH(7, 90) 1.35
RESET(1, 103) .804
RESET(Z, 102) .464
RESET(3, 101) .34

 

The next step is to use Recursive Least Squares(RLS) to estimate the model and
subsequently test for parameter consistency. RIS estimates its parameters incrementally, and as
such tests their significance one by one. This allows the parameters to be tested for constancy.
These tests are checking to see if the model is weakly exogenous, which allows us to make valid
inferences about the 8’s. This is best defined as: "A parameter, [3 is constant if it holds the same
value over the sample period."81 These tests are checked visually, that is they are graphs of
residual variation. The first, Figure 5.5 shows the residuals plotted over time, between 20.
Note the residuals do not break this line and as such remain constant over time. Also the
residuals show a stable pattern, just as they do in Figure 5.1, lending more evidence to their

consistency. Figure 5.2, depicts a One-Step Chow test for the residuals. Although it breaks the

 

'1 Granato, Jim An Agenda for Econometric Model Building Political Analysis, 1992, vol.3, Ann Arbor: The
University of Michigan Press, p. 129.

51

critical value of .05 three times, this is not enough to scrap the model by any means, but
something to watch for in the next two tests. Figure 5.3 represents a decreasing forecast horizon,
which tests the whole model against a restricted model, which is increased step by step. Each
step is then tested with a Chow test. As you can see there is no point at which the residuals
break the .05 critical value line. The last graph, Figure 5.4, is an increasing forecast horizon,
which does just the opposite of the decreasing horizon. And, just like Figure 5.3 the critical value
is not broken. This lends support to the conclusion that the model does indeed have stable

parameters, and thus is weakly exogenous.

52
Chapter VI

Policy implications and conclusions

The first thing to note is: This study is far from over. Not only do the results and problems
with the analysis call for more testing, but this subject is in need of more study. Having said this
I will now proceed to discuss the main findings and recap the conclusions that should be drawn
from them.

The first thing to note is the effect, or more accurately the problems caused by the ordinal
data. Although the results were not made invalid because of it, but I would bet there would be
a better fit if the data were interval. The last point I will make concerning methodology concerns
Hufbauer et al’s predictions. The testing they did was not very rigorous, and their statistics are
seriously flawed. As such the separate analysis of each variable, and atheoretical combination of
some”, did not properly test the interaction of the exogenous variables. This sounds like a plug
for the use of more rigorous statistics, which it is. By using regression one can better understand
the variables together, something that Hufbauer et a1 failed to do and as such their analysis was
faulty. The real proof of this is the dummy variables that fell out as well as the percentage of
exports going to the sender nation. Not only do faulty conclusions come from their analysis but
just as importantly there is no reason to do further analysis if those conclusions are accepted. It
seems better to continue questioning, than to stop at what could be invalid answers.83

The last part of this discussion is an attempt to answer the question: So What! .7 If you
observe all the sanctions imposed in the data set, it seems there is no consideration on the part
of the sender for their possible success, and as Mr. Spock would say "That is illogical Doctor". Or
is it. This is not a study of why sanctions are imposed, but curiously nations do not seem to

concern themselves with the possible outcome of their actions, or they may not know what the

 

‘2 This refers to the combination of imports and exports into total trade.

'3 Hufbauer, Schott and Elliott did use regression analysis in the First Edition of Economic Sanctions
Reconsidered, however it was poorly constructed and estimated. They dropped is from the Second Edition
saying it did not add any information to their results.

 

 

53

best manner to impose sanctions. I believe that the reasons for imposing sanctions are not
connected to success the way Hufbauer et al define it. This should change however when the
specific goals are redefined and then tested.

The worth of this type of study, and this one in particular, is still in question. The answer
is rather simple: by better understanding what makes sanctions work, more efficient policies can
be followed in order to attain success. This is the first step towards a general understanding of
sanctions and their effects, which should lower the wasted attempts and great losses to both
sender and target.

- As a final conclusion I will lay down my version of Hufbauer et al’s "Commandments":
First, isolate the target as best possible, and give no other nation a reason to be a black knight. If
there is such a nation, the sender should reconsider. Second, pick on the weak and helpless; not
only are they the most vulnerable, but in all the cases where the target was distressed, only 16.7%
of the time did a black knight appear. Third, the sender should sanction those whom she imports
a good deal from. Fourth, impose as much cost on the target as possible. Last, always show
resolve, and hope that others did in the past as well. This relates to the significance of LENGTH
4 and COST 2. It follows that if the target decision maker looks back on past values of those
variables, that the sender’s past record would also be significant.“

Economic sanctions are a useful tool to nation-states, and to international organizations.
They reduce the cost of human lives in many cases by weakening an adversary, or avoided
military confrontation all together. Their uses are far more fungible than military sanctions, and
have more bite than diplomatic sanctions. The study of them however is sparse to date, a gap

that should be filled through rigorous theories and testing.

 

“This is untested, and as such should be taken with a grain of salt.

 

APPENDD( A

APPENDIX A

Output for c:\thesis\append.OUT
Begun at
11: 2:50 on 313t July 1992

Data came from the Input Files: c:\thesis\ecnscn.INF and c:\thesis\ecnscn.BIN

Model OUTPUT from P C - G I V E

CORRELATION MATRIX
RSUCCESS RSUCCESI RSUCCESZ RSUCCES3 RSUCCES4 CONSTANT BLKNIGHT COOP

The Sample is 5 to 115 less 0 Forecasts
M E A N S of V A R I A B L E S
RSUCCESS RSUCCESl RSUCCESZ RSUCCES3 RSUCCES4 CONSTANT BLKNIGHT
' 4.9640 4.9820 4.9820 5.0090 5.0180 1.0000 .2252
COOP DSTRES POLICIES RELATE COST COST 1 COST 2
.2793 .2072 .3333 .3423 1.8559 1.8378 1.8559
COST 3 GNPCST GNPCST 1 GNPCST 2 GNPCST 3 GNPRATIO LENGTH
1.8559 1.7982 1.3658 1.4018 1.4081 593.6694 6.1982
LENGTH 1 LENGTH 2 LENGTH 3 LENGTH 4 PERCENTM PERCENTX TARCST
6.1982 6.2072 6.2072 6.2342 22.7613 ‘22.8784 178.7622
TARCST 1 TARCST 2 TARCST 3
178.5676 181.7658 179.0721
STANDARD DEVIATIONS OF VARIABLES
RSUCCESS RSUCCESI RSUCCESZ RSUCCES3 RSUCCES4 CONSTANT BLKNIGHT
2.7566 2.7765 2.7765 2.7518 2.7600 .0000 .4196
COOP DSTRES POLICIES RELATE COST COST 1 COST 2
.4507 .4071 .4735 .4766 .8073 .7809 .7845
COST 3 GNPCST GNPCST 1 GNPCST 2 GNPCST 3 GNPRATIO LENGTH
.7845 5.2085 2.7502 2.7596 2.7574 3157.3927 8.8254
LENGTH 1 LENGTH 2 LENGTH 3 LENGTH 4 PERCENTM PERCENTX TARCST
8.8254 8.8206 8.8206 8.8090 22.6959 24.8136 634.4464
TARCST 1 TARCST 2 TARCST 3
634.4983 634.9474 634.9813
D U R B I N - W A T S O N TESTS
RSUCCESS RSUCCESl RSUCCESZ RSUCCES3 RSUCCES4 CONSTANT BLKNIGHT
2.0913 2.0190 2.0944 2.0997 2.0956 .0000 1.7553
COOP DSTRES POLICIES RELATE COST COST 1 COST 2
1.9694 1.9743 1.6622 1.9207 1.9528 1.9529 1.9352
COST 3 GNPCST GNPCST 1 GNPCST 2 GNPCST 3 GNPRATIO LENGTH
1.9500 1.3582 2.1139 2.1195 2.1117 2.0216 1.5360
LENGTH 1 LENGTH 2 LENGTH 3 LENGTH 4 PERCENTM PERCENTX TARCST
1.5360 1.5378 1.5379 1.5429 1.7530 1.7852 1.9728
TARCST 1 TARCST 2 TARCST 3
1.9724 1.9729 1.9759

 

RSUCCESS 1.0000

RSUCCESl -.0381 1.0000

RSUCCESZ -.1391 —.0472 1.0000

RSUCCES3 .0564 -.1344 -.O619 1.0000

RSUCCES4 -.0847 .0653 -.1388 -.0563 1.0000

CONSTANT .0000 .0000 .0000 .0000 .0000 1.0000

BLKNIGHT -.2287 .0347 —.1369 .0770 .1534 .0000 1.0000

COOP —.1016 .0767 -.1340 -.1340 .1640 .0000 .3374 1.0000
DSTRES .2497 .1240 -.1253 -.0098 -.1085 .0000 -.0628 -.0210
POLICIES -.0186 .0530 -.1268 —.1419 .0023 .0000 .3050 .2414
RELATE .0717 .0116 .0459 -.0370 -.O462 .0000 .0201 .0164
COST -.1372 .1083 —.1066 -.0608 .0542 .0000 .0967 .3615
COST 1 -.1294 -.1565 .1454 -.0755 -.0788 .0000 .1680 .0007
COST 2 .1826 -.1014 —.1890 .1269 —.0534 .0000 .0719 .0635
COST 3 -.0781 .1824 —.1014 -.1889 .1272 .0000 .2652 .2692

55

GNPCST .1595 -.1050 -.0861 .0754 -.0233
GNPCST 1 -.1016 .1803 .0286 .0601 .0263
GNPCST 2 .1128 -.0845 .1608 .0194 .0724
GNPCST 3 —.0714 .1124 —.0879 .1568 .0203
GNPRATIO .0519 -.0115 -.1305 -.1582 .1332
LENGTH -.1428 .0124 -.1215 .0246 .1189
LENGTH 1 -.0064 -.1457 .0124 -.1285 .0226
LENGTH 2 -.0532 -.0088 -.1472 .0063 -.1290
LENGTH 3 -.1216 -.0566 -.0088 -.1544 .0043
LENGTH 4 .1505 -.1203 -.0612 -.0158 -.1527
PERCENTM .2531 .2023 -.0023 -.0143 .0069
PERCENTX .2281 .0964 -.0082 .0041 -.0066
TARCST .1127 .0746 -.1599 .0154 .0360
TARCST 1 -.1351 .1098 .0751 -.1637 .0142
TARCST 2 .1205 -.1276 .1025 .0724 -.1580
TARCST 3 -.1669 .1149 -.1214 .1071 .0683
DSTRES POLICIES RELATE COST COST 1
DSTRES 1.0000
POLICIES .0629 1.0000
RELATE .0528 .1343 1.0000
COST -.2955 .3646 -.0123 1.0000
COST 1 -.0935 -.0738 .0528 -.0086 1.0000
COST 2 -.0764 .0571 -.0613 .1535 .0209
COST 3 .0374 .2039 -.1586 .1679 .1544
GNPCST .0405 .3128 .1068 .3002 -.0933
GNPCST 1 -.0577 —.1070 —.0028 -.0477 .1637
GNPCST 2 .0603 .0719 .1695 .0156 -.0357
GNPCST 3 .0932 -.0390 -.1633 -.0783 .0276
GNPRATIO .2795 .1559 -.0647 -.0299 -.1065
LENGTH -.1254 .1385 -.1913 .1457 -.0019
LENGTH 1 -.1279 .0645 .1501 -.0419 .1643.
LENGTH 2 -.1133 .0617 .1603 -.0966 -.0294
LENGTH 3 -.1057 .1117 -.O905 .0974 -.0861
A LENGTH 4 .0776 .1467 -.0149 -.1665 .1153
PERCENTM .0369 .2204 .3316 .1157 -.0937
PERCENTX .0079 .2374 .3068 .1471 -.0383
TARCST .0546 .0635 -.0057 .0424 .0650
TARCST 1 -.0259 -.0779 .1040 -.1418 .0495
TARCST 2 .1329 -.0372 -.O616 -.0876 -.1422
TARCST 3 .0542 -.0620 .0807 -.0989 -.0966
GNPCST 1 GNPCST 2 GNPCST 3 GNPRATIO LENGTH
GNPCST 1 1.0000
GNPCST 2 -.0633 1.0000
GNPCST 3 .1327 -.0671 1.0000
GNPRATIO -.O454 -.0581 -.0593 1.0000
LENGTH -.0788 -.1598 .0683 —.0582 1.0000
LENGTH 1 .0688 -.0863 -.1613 -.0314 .2320
LENGTH 2 .2417 .0619 -.0880 —.0671 -.1625
LENGTH 3 .2198 .2331 .0606 -.0470 -.1632
LENGTH 4 -.0737 .2145 .2316 -.0170 .0491
PERCENTM -.0248 -.0026 -.0614 -.0754 -.0409
PERCENTX -.0386 -.0129 -.0596 -.0567 -.0227
TARCST .0064 -.0147 —.0346 -.0456 -.0284
TARCST 1 .0495 .0029 -.0152 -.0204 .0228
TARCST 2 .0028 .0525 .0014 -.0299 .0403
TARCST 3 .0256 .0015 .0540 -.0284 -.0141
LENGTH 4 PERCENTM PERCENTX TARCST TARCST 1
LENGTH 4 1.0000
PERCENTM -.1607 1.0000
PERCENTX -.1266 .8705 1.0000
TARCST -.0536 -.1127 -.1068 1.0000
TARCST 1 -.0412 -.0138 .0033 .0137 1.0000
TARCST 2 -.0569 .0346 —.0251 -.0169 .0123
TARCST 3 .0296 -.0578 —.0369 -.0225 -.0158
E0( 1) Modelling RSUCCESS by OLS
The Sample 15 5 to 115 less 0 Forecasts
VARIABLE COEFFICIENT STD ERROR H.C.S.E.
RSUCCESI .0315563 .11049 .11180
RSUCCESZ -.1207291 .10306 .09496
RSUCCESB .0143481 .10291 .10395
RSUCCES4 .0107587 .09870 .10441
CONSTANT 4.5271282 1.90041 1.70365
BLKNIGHT —.9C30361 .76203 .75979
COOP —1.0657853 .68076 .68213

.0000 .0093 .1431
.0000 -.1232 -.1477
.0000 -.1636 -.0362
.0000 .0102 .0026
.0000 -.0319 .1233
.0000 .3880 .2077
.0000 .0787 -.0140
.0000 -.0569 -.2114
.0000 -.0373 -.1153
.0000 -.0070 -.0670
.0000 -.0194 .1110
.0000 .0584 .0883
.0000 .1475 .2874
.0000 .0508 -.0862
.0000 -.0894 .0091
.0000 .0988 -.0506
COST 2 COST 3 GNPCST
1.0000
.0250 1.0000
.0222 .0217 1.0000
-.0179 .0129 -.0686
.1710 -.0154 .0311
-.0432 .1716 .0135
.0084 .0228 .0019
.0593 .1722 -.0130
-.0155 .0593 .0785
.1515 -.0154 .0666
-.0429 .1515 -.0875
-.0951 -.0424 -.0895'
.1035 .0160 .4728
.1242 .0096 .4139
-.0820 -.0078 .0053
.0585 -.0820 -.0088
.0535 .0594 .0309
-.1427 .0527 -.0144
LENGTH 1 LENGTH 2 LENGTH 3
1.0000
.2315 1.0000
-.1625 .2310 1.0000
-.1652 -.1649 .2295
-.0170 .1889 -.2156
-.0936 .0850 -.2208
.0309 -.0554 -.0405
-.0282 .0308 -.0552
.0198 -.0308 .0278
.0428 .0221 —.0283
TARCST 2 TARCST 3
1.0000
.0116 1.0000
t-VALUE PARTIAL I2
.28559 .0010
—1.17147 .0167
.13942 .0002
.10900 .0001
2.38219 .0655
-1.18504 .017C
-1.56558 .0294

56

DSTRES .8788174 .71292 .78315 1.23270 .0184
POLICIES -.7117061 .66124 .64922 -1.07632 .0141
RELATE .1542180 .59711 .58756 .25827 .0008
COST -.3654394 .41687 .41584 -.87662 .0094
COST 1 —.4721548 .36948 .34874 -1.27789 .0198
COST 2 .7942759 .35019 .32225 2.26812 .0597
COST 3 .2915646 .36688 .32748 .79472 .0077
GNPCST .0558771 .06019 .05540 .92841 .0105
GNPCST 1 -.0358706 .10093 .09065 -.35541 .0016
GNPCST 2 .0126209 .10394 .08800 .12142 .0002
GNPCST 3 -.1150506 .09573 .07429 -1.20178 .0175
GNPRATIO .0000189 .00008 .00004 .22310 .0006
LENGTH -.0311800 .03347 .02946 -.93158 .0106
LENGTH 1 .0396715 .03355 .03704 1.18237 .0170
LENGTH 2 -.0479502 .03521 .03844 -1.36173 .0224
LENGTH 3 -.0199133 .03605 .03212 -.55236 .0038
LENGTH 4 .0841870 .03491 .03550 2.41184 .0670
PERCENTM .0174733 .02577 .02427 .67813 .0056
PERCENTX .0149661 .02121 .02065 .70549 .0061
TARCST .0009433 .00042 .00032 2.25577 .0591
TARCST 1 -.0006301 .00041 .00027 -1.55032 .0288
TARCST 2 .0003399 .00041 .00037 .83294 .0085
TARCST 3 -.0007998 .00040 .00024 -2.00842 .0474
R2 - .4205625 0 = 2.4452681 F(29, 81) - [ .0070] DW = 2.121
RSS = 484.3262209063 for 30 Variables and 111 Observations
Information Criteria: SC = 2.75; HQ = 2.31; FPE = 7.60
R2 Relative to DIPFERENCE+SEASONALS = .72292
SEASONAL MEANS of DIFFERENCES are
-.01818
Solved STATIC LONG RUN Equation
RSUCCESS == 4 . 255 -. 849 BLKNIGHT -1 . 002 COOP

S.E. 1.27316) ( .77727) ( .65246)

+ .826 DSTRES -.669 POLICIES+ .145 RELATE .233 COST

( .71233) ( .63395) ( .56406) ( .61135)

—.077 GNPCST + .000018 GNPRATIO+ .023 LENGTH .016 PERCENTM

( .18047) ( .00008) ( .07056) ( .02351)

+ .014 PERCENTX -.000138 TARCST

( .02076) ( .00075)

WALD Test Chizll3) = 520.186
ANALYSIS of LAG STRUCTURE
Var \ 0 1 2 3 4 5 6 7 8 2
RSUCCES -1.000 .032 -.121 .014 .011 0. 0. 0. 0. -1.064
S.E. 0. .110 .103 .103 .099 0. 0. 0. 0. .231
CONSTAN 4.527 O. O. 0. 0. 0. O. 0. 0. 4.527
S.E. 1.900 0. 0. 0. 0. 0. 0. 0. 0. 1.900
BLKNIGH -.903 O. 0. 0. 0. 0. 0. 0. 0. -.903
S.E. .762 0. 0. 0. 0. 0. 0. O. 0. .762
COOP —1.066 0. 0. 0. 0. 0. 0. 0. 0. —1.066
S.E. .681 0. 0. 0. 0. 0. 0. O. 0. .681
DSTRES .879 0. 0. 0. O. 0. 0. O. O. .879
S.E. .713 0. 0. O. 0. O. O. O. 0. .713
POLICIE -.712 O. O. 0. 0. O. 0. 0. 0. -.712
S.E. .661 0. O. 0. 0. 0. 0. 0. 0. .661
RELATE .154 0. 0. 0. O. 0. 0. O. 0. .154
S.E. .597 0. 0. 0. 0. O. 0. O. 0. .597
COST -.365 —.472 .794 .292 0. 0. 0. 0. 0. .248
S.E. .417 .369 .350 .367 O. O. O. O. 0. .640
GNPCST .056 —.O36 .013 -.115 O. 0. O. 0. O. —.082
S.E. .060 .101 .104 .096 O. O. O. 0. O. .185
GNPRATI .00002 0. 0. 0. 0. 0. O. 0. 0. .00002
S.E. .00008 0. 0. 0. O. 0. 0. O. 0. .00008
LENGTF -.031 .040 -.048 -.020 .084 O. 0. 0. 0. .025
S.E. .033 .034 .035 .036 .035 0. 0. 0. C. .073
PERCENT .017 0. 0. 0. O. O. 0. 0. 0. .017
S.E. .026 0. 0. 0. O. 0. 0. O. 0. .026
PERCENT .015 O. 0. 0. O. O. O. 0. 0. .015
S.E. .021 0. 0. 0. 0. C. 0. C. C. .021
TARCST .00094-.0006 .00034—.00080 0. 0. O. 0. O.-.OOC:5
S.E. .00042 .00041 .00041 .00040 0. 0. 0. 0. C. .00079
Tests on the Significance of each Variable
Variable F(NUM,DENOM] = Value Probabili: Uni: Roo: t-tes:

57

RSUCCESS F[ 4, 81] .397 .810 -4.602
CONSTANT F[ 1, 81] 5.675 .020 2.382
BLKNIGHT F[ 1, 81] 1.404 .239 ~1.185
COOP F[ 1, 81] 2.451 .121 -1.566
DSTRES F[ 1, 81] 1.520 .221 1.233
POLICIES F[ 1, 81] 1.158 .285 -1.076
RELATE F[ 1, 81] .067 .797 .258
COST F[ 4, 81] 1.858 .126 .388
GNPCST F[ 4, 81] .609 .657 -.445
GNPRATIO F[ 1, 81] .050 .824 .223
LENGTH F[ 5, 81] 2.097 .074 .341
PERCENTM F[ 1, 81] .460 .500 .678
PERCENTX F[ 1, 81] .498 .483 .705
TARCST F[ 4, 81] 3.080 .021 —.185
Tests on the Significance of each LAG
LAG F[NUM,DENOM] = Value Probability

4 F[ 2, 81] 3.035 .054

3 F[ 4, 81] 1.692 .160

2 P[ 4, 81] 1.411 .238

1 F[ 4, 81] .832 .509

Testing for Serial Correlation from Lags 1 to 2
CH12( 2) = 4.608 and F-Form( 2, 79) = 1.71 [ .1873]
Error Autocorrelation Coefficients:

-.2185 .2935

ARCH TEST
Residuals Scaled by .244SD+01

CNST 1 LAG

COEFP. .6595 .0924
S.E.’s .1275 .1121
RSS = .82699D+02 O = 1.02314
CHIZ( 1)= .938 with F( l, 79) = .68 [ .4123]
CHI-SQUARED Test for NORMALITY :CHI2(2) = 1.799
TEST for HETEROSCEDASTIC ERRORS 111*R2 = 26.3303 with 34 Variables

F(33, 57) = .5371 [ .9715]

RESET F-TEST for adding Yhat2
F( 1, 80) = 1.616 [ .2074]

Testing for Serial Correlation from Lags l to 7
CHI2( 7) = 7.414 and F-Form( 7, 74) = .76 [ .5252]
Error Autocorrelation Coefficients:

-.2071 .2359 -.1793 .0815 .0712 .1134 .0145

ARCH TEST
Residuals Scaled by .2445D+01

CNST 1 LAG 2 LAG 3 LAG 4 LAG 5 LAG 6 LAG 7 LAG
COEFF. .6413 .1125 -.1649 -.1795 .1582 -.0144 .1405 .0307
S.E.’s .2535 .1224 .1253 .1270 .1274 .1258 .1241 .1247

RSS = .70306D+02 O = 1.02437
CH12( 7)= 11.320 with P( 7, 67) = 1.17 [ .3323]
ANALYSIS of SCALED RESIDUALS
Sample Size 111
Mean .000000
Std.Devn. .858116
Skewness -.201071
Excess Kurtosis -.609255
Minimum —1.837720
Maximum 1.797985

CHI-SQUARED Test for NORMALIT :CHIZ(2) = 1.799

 

 

TEST for HETEROSCEDASTIC ERRORS

F(33,

Regressors used for forming the Quadratic are:

57) -

.5371 [

58

111*R2 =
.9715]

26.3303 with 34 Variables

RSUCCESl RSUCCESZ RSUCCES3 RSUCCES4 CONSTANT BLKNIGHT COOP DSTRES

POLICIES RELATE COST COST 1 COST 2 COST 3 GNPCST GNPCST 1

GNPCST 2 GNPCST 3 GNPRATIO LENGTH
HETEROSCEDASTICITY Coefficients and t-Values are:

Vars: V 1 V 2 V 3 V 4 V 5 V 6
Coeff: —.33724 .68535 .17350 —1.15780 -7.10475 .83304
t-Value -.31 .61 .16 -1.04 -.85 .39

Vars: V 7 V 8 V 9 V10 V11 V12
Coeff: -1.07024 1.43615 .10691 1.66641 .13756 3.14939
t-Value -.58 .68 .06 1.05 .04 .85

Vars: V13 V14 V15 V16 V17 V18
Coeff: 5.27159 1.00520 -.36585 .13827 .48194 -.49678
t—Value 1.40 .27 -.97 .28 .91 -.94

Vars: V19 V20 V 12 V 22 V 32 V 42
Coeff: .00036 -.09344 .03363 -.05506 .02217 .14478
t-Value .23 -.36 . -.48 .21 1.27

Vars: V112 V122 V132 V142 V152 V162
Coeff: .05562 -.71174 —1.13401 .00470 .00772 -.02357
t-Value .06 -.83 -1.30 .01 .90 -.57

Vars: V172 V182 V192 V202
Coeff: -.05509 .01242 -.000000016 .00023
t-Value -1.31 .29 -.34 .03
Model OUTPUT from P C - G I V E

E0( 2) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts

VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
RSUCCESl .0159221 .10943 .10785 .14550 .0003
RSUCCESZ -.1197499 .10259 .09385 —1.16724 .0161
RSUCCESB -.0110007 .10027 .09876 -.10971 .0001
RSUCCES4 .0082826 .09706 .10053 .08534 .0001
CONSTANT 4.7312697 1.88045 1.66507 2.51603 .0709
BLKNIGHT -.8555781 .75602 .74695 -1.13169 .0152
COOP -1.0777398 .67131 .66615 -1.60543 .0301
DSTRES .8863726 .68649 .73321 1.29117 .0197
POLICIES -.6514188 .64807 .63034 —1.00517 .0120
RELATE .1877313 .59322 .57210 .31646 .0012
COST -.3338227 .41424 .41483 -.80586 .0078
COST 1 —.4887210 .36769 .34256 —l.32918 .0208
COST 2 .7976999 .34892 .31711 2.28617 .0592
COST 3 .2086473 .35977 .31713 .57994 .0040
GNPCST .0480145 .05967 .05590 .80460 .0077
GNPCST 1 -.0477486 .10004 .09531 -.47731 .0027
GNPCST 2 .0235643 .10268 .08606 .22950 .0006
LENGTH —.0338515 .03320 .03035 -1.01953 .0124
LENGTH 1 .0433168 .03331 .03686 1.30024 .0200
LENGTH 2 -.0473561 .03511 .03783 -1.34874 .0214
LENGTH 3 -.0212902 .03585 .03175 -.59387 .0042
LENGTH 4 .0746940 .03396 .03314 2.19964 .0551
PERCENTM .0179284 .02555 .02358 .70172 .0059
PERCENTX .0149657 .02115 .02033 .70770 .0060
TARCST .0009515 .00041 .00029 2.30615 .0602
TARCST 1 -.0006408 .00041 .00027 -1.58145 .0293
TARCST 2 .0003386 .00041 .00037 .83257 .0083
TARCST 3 —.0008018 .00040 .00023 ~2.01963 .0468
R2 = .4095044 0 = 2.4385686 F(27, 83) 2.13 [ .0047] UN = 2.065

RSS = 493.5692124498 for 28 Variables and 111 Observations
Information Criteria: SC = 2.68; HQ 2.27; PPE = 7.45

R2 Relative to DIFFERENCE+SEASONALS = .71763

SEASONAL MEANS of DIFFERENCES are
-.01818

Testing for Serial Correlation from Lags 1 2

CHI’( 2) 3.174 and F-Form( 81) = 1.19 [ .3089}

59

Error Autocorrelation Coefficients:

-.1142 .2942
ARCH TEST
Residuals Scaled by .2439D+01
CNST l LAG
COEFF. .6760 .0916
S.E.'s .1267 .1106
RSS = .81694D+02 O = 1.00428
CHI3( 1): .923 with F( 1, 81) = .69 [ .4101]

CHI-SQUARED Test for NORMALITY :CHIZ(2) 1.963

111*R2 25.8603 with 34 Variables

TEST for HETEROSCEDASTIC ERRORS

F(33, 57) = .5246 [ .9758]

RESET F-TEST for adding Yhat2
F( 1, 82) = 1.532 [ .2194]
Testing for Serial Correlation from Lags 1 to 7
F-Form( 7, 76) = .58 [ .7693]

CHIz( 7) = 5.638 and

Error Autocorrelation Coefficients:

—.0573 .2600 -.1663 -.0380 .0336 .1443 .0228
ARCH TEST
Residuals Scaled by .2439D+01
CNST 1 LAG 2 LAG 3 LAG 4 LAG 5 LAG 6 LAG 7 LAG
COEFP. .6638 .1369 -.1989 -.1602 .1668 -.0702 .1776 .0215
S.E.‘s .2538 .1207 .1254 .1277 .1275 .1255 .1232 .1238
RSS = .67697D+02 O = .99051
CHI2( 7)= 12.266 with F( 7, 69) = 1.32 [ .2550]
ANALYSIS of SCALED RESIDUALS
Sample Size 111
Mean .000000
Std.Devn. .868646
Skewness -.149326
Excess Kurtosis -.691700
Minimum -1.959489
Maximum 1.796912

CHI-SQUARED Test for NORMALITY :CHI2(2) = 1.963

TEST for HETEROSCEDASTIC ERRORS 111*R2 = 25.8603 with 34 Variables

F(33, 57) = .5246 [ .9758]

Regressors used for forming the Quadratic are:

RSUCCESl RSUCCESZ RSUCCES3 RSUCCES4 CONSTANT BLKNIGHT COOP DSTRES

POLICIES RELATE COST COST 1 COST 2 COST 3 GNPCST GNPCST 1

GNPCST 2 LENGTH LENGTH 1 LENGTH 2
HETEROSCEDASTICITY Coefficients and t—Values are:

Vars: V 1 V 2 V 3 V 4 V 5 V 6
Coeff: .24706 .46895 .48137 -.61463 -12.16784 .48501
t—Value .22 .41 .46 —.55 -1.43 .24

Vars: V 7 V 8 V 9 V10 V11 V12
Coeff: -.31881 1.46260 -.56452 1.23465 -.91516 4.16807
t-Value -.17 .79 -.32 .78 -.24 1.14

Vars: V13 V14 V15 V16 V17 V18
Coeff: 4.99185 1.72709 -.32373 .16543 61351 -.02079
t-Value 1.34 46 —.92 .33 1.14 - 08

Vars: V19 V20 V 12 V 22 V 32 V 42
Coeff: .10811 .17581 —.02138 -.01974 —.01060 .09309
t—Value .39 .64 —.18 —.17 —.10 .83

Vars: V112 V122 V132 V142 V152 V162
Coeff: . 5759 -.93826 -1.10285 -.21971 .00645 —.03335
t-Value .41 -1.11 -1.27 —.26 .80 -.77

Vars: V172 V182 V192 V202

Coeff:
t—Value

-.06312
-l.48

-.00129
-.19

Model OUTPUT from P C - G I V E

EQ( 3) Modelling RSUCCESS by OLS

-.00090
-.13

60

-.00142

The Sample is 5 to 115 less 0 Forecasts
VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
RSUCCESl .0063107 .10277 .10149 .06141 .0000
RSUCCESZ -.1185369 .09605 .08507 -l.23411 .0174
RSUCCES3 -.0045892 .09529 .09122 -.04816 .0000
RSUCCES4 .0020977 .09168 .09332 .02288 .0000
CONSTANT 4.6521544 1.73594 1.46265 2.67991 .0771
BLKNIGHT -.8699997 .71576 .66545 -1.21549 .0169
COOP -1.0014149 .65376 .65612 -1.53178 .0266
DSTRES .9477207 .67051 .70772 1.41344 .0227
POLICIES -.6506728 .63967 .61723 -1.01720 .0119
RELATE .2245900 .57202 .55377 .39262 .0018
COST -.3758973 .39890 .39652 -.94233 .0102
COST 1 -.4991474 .34198 .32238 -1.45959 .0242
COST 2 .8372078 .32805 .30358 2.55209 .0704
COST 3 .1723284 .34666 .31465 .49711 .0029
GNPCST .0482891 .05880 .05558 .82130 .0078
LENGTH -.0310007 .03232 .02872 -.95905 .0106
LENGTH 1 .0467143 .03198 .03590 1.46092 .0242
LENGTH 2 -.0580629 .03185 .03540 -1.82310 .0372
LENGTH 4 .0710955 .03090 .02765 2.30106 .0580
PERCENTM .0202918 .02498 .02337 .81220 .0076
PERCENTX .0150015 .02070 .02040 .72476 .0061
‘TARCST .0009537 .00041 .00030 2.35227 .0604
TARCST 1 -.0006305 .00040 .00029 -1.57879 .0282
TARCST 2 .0002985 .00040 .00036 .74926 .0065
TARCST 3 -.0007832 .00039 .00024 -2.00388 .0446
R2 = .4033957 0 = 2.4080177 F(24, 86) = [ .0015] = 2.082
RSS = 498.6752381588 for 25 Variables and 111 Observations
Information Criteria: SC = 2.56; HQ = 2.20; FPE = 7.10
R2 Relative to DIFFERENCE+SEASONALS = .71471
SEASONAL MEANS of DIFFERENCES are
-.01818
Testing for Serial Correlation from Lags 1 to 2
CHI’( 2) = 4.214 and F-Form( 2, 84) = 1.66 [ .1968]
Error Autocorrelation Coefficients:
-.1474 .3260
ARCH TEST
Residuals Scaled by .2408D+01
CNST 1 LAG
COEFF. .7031 .0860
S.E.'s .1272 .1086
RSS = .83723D+02 o = .99835
CH12( 1)= .816 with F( 1, 84) = .63 [ .4305]

CHI-SQUARES Test for NORMALITY :CHIZ(2)

2.165

TEST for HETEROSCEDASTIC ERRORS 111*R2 24.7806 with 34 Variables

F(33, 57) = .4964 [ .9837}

RESET F-TEST for adding Yhat2
F( 1, 85) = 2.133 [ .1478]

Model OUTPUT from P C - G I V E

EQ( 4) Modelling RSUCCESS by OLS

61

The Sample is 5 to 115 less 0 Forecasts
VARIABLE COEFFICIENT STD ERROR H.C.S.E.
RSUCCESZ -.1237286 .09514 .08216
RSUCCE53 .0101487 .09276 .08547
RSUCCES4 -.0010201 .09082 .09052
CONSTANT 4.5164064 1.66138 1.36079
BLKNIGHT -.8913514 .70964 .65970
COOP -.9822509 .64648 .64835
DSTRES 1.0066203 .65332 .66603
POLICIES -.5641345 .62674 .60781
RELATE .1554511 .56118 .54736
COST -.2886574 .37487 .37603
COST 1 -.5103976 .33441 .31858
COST 2 .8030402 .32080 .30498
COST 3 .1648831 .33747 .32413
LENGTH -.0340736 .03187 .02707
LENGTH 1 .0517298 .03093 .03503
LENGTH 2 —.0589184 .03158 .03511
LENGTH 4 .0715200 .03055 .02750
PERCENTM .0257079 .02231 .02039
PERCENTX .0148377 .01999 .01930
TARCST .0009551 .00040 .00028
TARCST 1 -.0005808 .00039 .00027
TARCST 2 .0002989 .00039 .00035
TARCST 3 -.0007738 .00038 .00023
R2 = .3984114 0 - 2.3904197 F(22, 88) = 2.6
RSS = 502.8413420234 for 23 Variables and 11
Information Criteria: SC = 2.49; HQ =

R2 Relative to DIFFERENCE+SEASONALS c .71233

SEASONAL MEANS of DIFFERENCES are
-.01818
Testing for Serial Correlation from Lags 1 to 2
CHI’( 2) = 4.479 and F—Form( 2. 86) =
Error Autocorrelation Coefficients:
-.0699 .3450
ARCH TEST
Residuals Scaled by .2390D+01
CNST 1 LAG

COEFF. .7238 .0809

S.E.‘s .1301 .1077
RSS = .898050+02 O = 1.02188
CH12( 1)= .717 with F( 1, 86) = .56 [
CHI-SQUARED Test for NORMALITY :CH12(2) = 2.146

TEST for HETEROSCEDASTIC ERRORS

111*R2

F(33, 57) = .6292 [ .9230]

RESET F-TEST for adding Yhat2

F( 1, 87) = 1.808 [ .1823]

Model OUTPUT from P C - G I V E

E0( 5) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts
VARIABLE COEFFICIENT STD ERROR H.C.S.E.
RSUCCE52 -.1229717 .09236 .08216
CONSTANT 4.6040584 1.36062 1.19435
BLKNIGHT —.8615256 .68211 .65292
COOP —.9811859 .62414 .62800
DSTRES 1.0203156 .63331 .64895
POLICIES -.5628794 .61013 .59332

5
1

2.

t-VALUE PARTIAL r2

-1.30042 .0189
.10941 .0001
-.01123 .0000
2.71847 .0775
-1.25607 .0176
-1.51938 .0256
1.54079 .0263
-.90012 .0091
.27701 .0009
-.77002 .0067
-1.52625 .0258
2.50324 .0665
.48859 .0027
-1.06916 .0128
1.67256 .0308
-1.86548 .0380
2.34104 .0586
1.15247 .0149
.74223 .0062
2.39298 .0611
—1.50702 .0252
.77420 .0068
-2.01872 .0443
[ .0007] DW = 2.156
Observations
15; FPE = 6.90
1.81 [ .1702]
.4547]

29.6395 with 34 Variables

t-VALUE PARTIAL :2

-;.33147 .0191
3.38379 .1118
~1.26302 .0172
-1.57207 .0264
1.61109 .0277
-.92256 .0093

62

 

COST ‘ -.2841768 .36689 .37542 -.77455 .0065
COST 1 -.5054776 .32690 .31005 -1.54629 .0256
COST 2 .8004167 .31005 .30124 2.58161 .0682
COST 3 .1388152 .31788 .31962 .43669 .0021
LENGTH -.0358685 .03048 .02657 -1.17671 .0150
LENGTH 1 .0530569 .02939 .03418 1.80522 .0346
LENGTH 2 -.0582409 .03045 .03437 -1.91273 .0386
LENGTH 4 .0720928 .02948 .02700 2.44552 .0617
PERCENTM .0264486 .02173 .02044 1.21689 .0160
PERCENTX .0151832 .01958 .01886 .77558 .0066
TARCST .0009551 .00039 .00028 2.44088 .0614
TARCST 1 -.0005771 .00036 .00027 -1.58367 .0268
TARCST 2 .0002999 .00037 .00034 .80545 .0071
TARCST 3 -.0007590 .00037 .00021 -2.03546 .0435
R2 = .3977976 6 = 2.3518858 F(19, 91) = 3.16 [ .0001] DW a 2.156
RSS = 503.3543726349 for 20 Variables and 111 Observations
Information Criteria: SC - 2.36; HQ = 2.07; FPE a 6.53
R2 Relative to DIFFERENCE+SEASONALS = .71203
SEASONAL MEANS of DIFFERENCES are
-.01818
Testing for Serial Correlation from Lags 1 to 2
CHI2( 2) = 3.967 and F-Form( 2, 89) = 1.65 [ .1980]
Error Autocorrelation Coefficients:
-.0735 .3048
ARCH TEST

Residuals Scaled by .2352D+01

CNST 1 LAG
COEFF. .7485 .0815
S.E.‘s .1312 .1059
RSS 8 .93442D+02 O = 1.02465
CH12( 1)= .727 with F( 1, 89) = .59 [ .4435]
CHI—SOUARED Test for NORMALITY :CHI’(2) = 2.359

TEST for HETEROSCEDASTIC ERRORS 111*R2 = 28.3175 with 35 Variables

F(34, 56) = .5641 [ .9619]

RESET F—TEST for adding Yhat2

F( 1, 90) = 1.905 [ .1710]

Model OUTPUT from P C — G I V E

E0( 6) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts
VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r3
CONSTANT 3.2821609 1.00916 1.00765 3.25238 .1011
BLKNIGHT -.9609569 .65130 .64601 -1.47544 .0226
COOP -1.1413293 .60315 .62852 —1.89229 .0367
DSTRES 1.2657470 .58036 .57883 2.18096 .0482
COST 1 —.5123696 .31633 .30747 -1.61974 .0272
COST 2 .8580056 .30253 .30439 2.83613 .0788
COST 3 .0939018 .31481 .33400 .29828 .0009
LENGTH -.0318588 .03028 .02580 -1.05230 .0116
LENGT' 1 .0500019 .02899 .03396 1.72475 .0307
LENGTH 2 -.0509018 .02949 .03292 —1.72616 .0307
LENGTH 4 .0723138 .02723 .02567 2.65524 .0698
PERCENTM .0253593 .02182 .02064 1.16203 .0142
PERCENTX .0123832 .01954 .01858 .63382 .0043
TARCST .0010334 .00039 .00021 2.67506 .0707
TARCST 1 -.0005484 .00036 .00023 -1.51565 .0239
TARCST 2 .0002646 .00037 .00032 .71805 .0055
TARCST 3 —.0006309 .00037 .00023 -1.71112 .0302
R2 = .3710834 0 = 2.3648211 F(l6, 94) = 3.47 [ .0001] ON = 2.161

 

63

RSS = 525.6836311271 for 17 Variables and 111 Observations
Information Criteria: SC - 2.28; HQ = 2.03; FPE = 6.45
R2 Relative to DIFFERENCE+SEASONALS = .69926

SEASONAL MEANS of DIFFERENCES are
-.01818

Testing for Serial Correlation from Lags 1 to 2
CHIZ( 2) = 1.011 and F-Form( 2, 92) = .42 [ .6566]
Error Autocorrelation Coefficients:

-.0987 .0027

ARCH TEST
Residuals Scaled by .2365D+01

CNST 1 LAG
COEFF. .7595 .0881
S.E.‘s .1314 .1033

RSS = .96062D+02 0 = 1.02184
CHI’( 1)= .863 with F( 1, 92) s .73 [ .3960]
CHI-SQUARED Test for NORMALITY :CHI’(2) 2.553

TEST for HETEROSCEDASTIC ERRORS 111*R2 28.5938 with 30 Variables

F(29, 64) = .7658 [ .7834]
RESET F-TEST for adding Yhat2
F( 1, 93) = 1.363 [ .2461]

Model OUTPUT from P C - G I V E

E0( 7) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts

VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
CONSTANT 3.4543143 .90387 .91392 3.82171 .1309
BLKNIGHT -.9166008 .62787 .62053 -1.45986 .0215
COOP —1.1343939 .57879 .60319 -1.95994 .0381
DSTRES 1.3115958 .56616 .56039 2.31667 .0524
COST 1 -.5056587 .30550 .29172 -1.65519 .0275
COST 2 .8874334 .29741 .30037 2.98386 .0841
LENGTH —.0299710 .02979 .02499 -1.00607 .0103
LENGTH 1 .0481723 .02836 .03288 1.69872 .0289
LENGTH 2 -.0541441 .02869 .03181 -1.88701 .0354
LENGTH 4 .0704873 .02686 .02573 2.62451 .0663
PERCENTM .0374672 .01046 .00909 3.58254 .1169
TARCST .0010184 .00038 .00021 2.68179 .0690
TARCST 1 -.0005504 .00036 .00024 -1.54360 .0240
TARCST 3 -.0006100 .00036 .00024 -1.67677 .0282
R2 = .3646628 0 = 2.3398174 F(13, 97) = 4.28 [ .0000] DW = 2.173
RSS = 531.0503054067 for 14 Variables and 111 Observations
Information Criteria: SC = 2.16; HQ = 1.96; FPE = 6.17
R2 Relative to DIFFERENCE+SEASONALS = .69619

SEASONAL MEANS of DIFFERENCES are
—.01818

Testing for Serial Correlation from Lags 1 to 2
CH12( 2) = 1.158 and F-Form( 2, 95) = .50 L .6076}
Error Autocorrelation Coefficients:

-.1052 -.0104

ARCH TEST
Residuals Scaled by .234OD+01

CNST l LAG

 

COEFF. .8112 .0555

S.E.’s .1336 .1017

RSS = .10250D+03 O = 1.03870

CHI3( 1)= .344 with F( 1, 95) = .30 [ .5865]

CHI-SQUARED Test for NORMALITY :CHI2(2) 2.737

TEST for HETEROSCEDASTIC ERRORS 111*R2

22.5885 with 24 Variables
F(23, 73) = .8109 [ .7074]

RESET F-TEST for adding Yhat2

F( 1, 96) = 1.474 [ .2277]

Model OUTPUT from P C - G I V E

E0( 8) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts
VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
CONSTANT 3.4433000 .88672 .92741 3.88318 .1322
BLKNIGHT -1.0997460 .58835 .58050 -1.86920 .0341
COOP —1.1735997 .58024 .58601 -2.02263 .0397
DSTRES 1.3063889 .56441 .55073 2.31459 .0513
COST 1 -.3802970 .29887 .33327 -1.27243 .0161
COST 2 .8423133 .29804 .31493 2.82619 .0747
LENGTH 2 -.0390831 .02754 .03190 -1.41936 .0199
LENGTH 4 .0613442 .02651 .02491 2.31371 .0513
PERCENTM .0367540 .01050 .00939 3.50022 .1101
TARCST .0010660 .00038 .00019 2.80942 .0738
TARCST 1 -.0005914 .00036 .00024 —1.65252 .0268
TARCST 3 -.0005596 .00036 .00020 -1.53485 .0232
R2 = .3439764 0 = 2.3534655 F(ll, 99) - 4.72 [ .0000] DW = 2.169
RSS - 548.3411641640 for 12 Variables and 111 Observations
Information Criteria: SC - 2.11; HQ = 1.93; FPE = 6.14
R2 Relative to DIFFERENCE+SEASONALS = .68630

SEASONAL MEANS of DIFFERENCES are
-.01818

Testing for Serial Correlation from Lags 1 to 2
CHI2( 2) = 1.417 and F-Form( 2, 97) = .63 [ .5363]

Error Autocorrelation Coefficients:

-.1076 -.0569
ARCH TEST

Residuals Scaled by .2353D+01

CNST 1 LAG
COEFF. .8930 -.0158
S.E.'s .1390 .1010
RSS = .118020+03 o = 1.10306
CHI2( 1)= .028 with F( 1, 97) = .02 [ .8758]
CHI—SQUARED Test for NORMALITY :CHIz(2) = 1.801
TEST for HETEROSCEDASTIC ERRORS 111*R2 = 14.2013 with 20 Variables

F(19, 79) = .6100 [ .8881]
RESET F-TEST for adding Yhat2
F( 1, 98) = 1.829 [ .1794]
Model OUTPUT from P C - G I V E

E0( 9) Modelling RSUCCESS by OLS

 

 

65

The Sample is 5 to 115 less 0 Forecasts

VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
CONSTANT 2.6005438 .70076 .66745 3.71103 .1200
BLKNIGHT -1.2375551 .58368 .57653 -2.12027 .0426
COOP -.9214438 .56495 .56405 -1.63101 .0257
DSTRES 1.4627989 .56270 .53175 2.59958 .0627
COST 2 .7778134 .29722 .30607 2.61695 .0635
LENGTH 4 .0623961 .02630 .02539 2.37236 .0528
PERCENTM .0345574 .01035 .00997 3.33891 .0994
TARCST .0010083 .00038 .00016 2.64515 .0648
TARCST 1 —.0006032 .00036 .00030 -1.67335 .0270
TARCST 3 -.0005317 .00036 .00017 -1.45899 .0206
R2 = .3201353 0 = 2.3720086 F( 9,101) = 5.28 [ .0000] DW = 2.130
RSS = 568.2689094206 for 10 Variables and 111 Observations
Information Criteria: SC = 2.06; HQ = 1.91; FPE = 6.13
R2 Relative to DIFFERENCE+SEASONALS = .67490

SEASONAL MEANS of DIFFERENCES are
-.01818
Testing for Serial Correlation from Lags 1 to 2
CHIZ( 2) = 1.229 and F-Form( 2, 99) = .55 [ .5764]
Error Autocorrelation Coefficients:
-.0904 —.0745
ARCH TEST
Residuals Scaled by .2372D+01
CNST 1 LAG

COEFF. .9106 -.0147
S.E.‘s .1372 .1000
RSS = .11334D+03 o = 1.06999
CHIZ( 1)= .024 with F( l, 99) = .02 [ .8833]
CHI-SQUARED Test for NORMALITY :CHI’(2) = 2.663

TEST for-HETEROSCEDASTIC ERRORS 111*R2

10.6738 with 16 Variables

F(15, 85) = .6029 [ .8649]

RESET F-TEST for adding Yhatz
F( 1,100) = 1.263 [ .2637]

Model OUTPUT from P C - G I V E

20(10) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts

VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
CONSTANT 2.3135143 .69370 .69684 3.33502 .0975
BLKNIGHT -1.4281698 .58434 .61182 -2.44408 .0548
COOP -.7545279 .56724 .58845 -1.33018 .0169
DSTRES 1.4344275 .56945 .53838 2.51898 .0580
COST 2 .8104968 .29773 .32450 2.72224 .0671
LENGTH 4 .0641996 .02664 .02915 2.41001 .0534
PERCENTM .0352171 .01048 .01039 3.35934 .0987
TARCST .0010048 .00039 .00017 2.60470 .0618
R2 = .2876045 U = 2.4044053 F( 7,103) = 5.94 [ .0000] DW =

RSS = 595.4599724654 for 8 Variables and 111 Observations
Information Criteria: SC = 2.02; HQ = 1.90; FPE = 6.
R2 Relative to DIFFERENCE+SEASONALS = .65934
SEASONAL MEANS of DIFFERENCES are
-.01818

Testing for Serial Correlation from Lags 1 to 2
CH12( 2) = 1.296 and F-Form( 2.10;) = .60 E .5525?

2.159

Error Autocorrelation Coefficients:

-.1039 -.0571
ARCH TEST

Residuals Scaled by .2404D+01

CNST 1 LAG
COEFF. .9311 -.0161
S.E.'s .1378 .0991
RSS = .11521D+03 O = 1.06805
CHI2( 1)= .029 with F( 1,101) = .03 [ .8715]

CHI-SQUARED Test for NORMALITY :CHI2(2) 2.780

TEST for HETEROSCEDASTIC ERRORS 111*R2 10.4211 with 12 Variables

F(11, 91) = .8571 [ .5844]
RESET F-TEST for adding Yhat2
F( 1,102) = 1.147 [ .2868]

Model OUTPUT from P C — G I V E

EQ(11) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts

VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
CONSTANT 2.2579175 .69500 .69735 3.24880 .0921
BLKNIGHT —1.6686837 .55771 .59821 -2.99202 .0793
DSTRES 1.4482112 .57145 .52523 2.53425 .0582
COST 2 .7906051 .29845 .31967 2.64900 .0632
LENGTH 4 .0650850 .02673 .03051 2.43503 .0539
PERCENTM .0331413 .01040 .01026 3.18519 .0889
TARCST .0008640 .00037 .00015 2.32053 .0492
R2 = .2753667 0 = 2.4132825 F( 6,104) = 6.59 [ .0000] DW = 2.164
RSS = 605.6889938162 for 7 Variables and 111 Observations
Information Criteria: SC = 1.99; HQ = 1.89; FPE = 6.19
R2 Relative to DIFFERENCE+SEASONALS = .65349

SEASONAL MEANS of DIFFERENCES are
-.01818

Testing for Serial Correlation from Lags 1 to 2
CHI2( 2) = 1.154 and F-Form( 2,102) = .54 [ .5869]

Error Autocorrelation Coefficients:

-.1014 -.0436
ARCH TEST

Residuals Scaled by .2413D+01

CNST 1 LAG
COEFF. .9268 .0047
S.E.’s .1392 .0992
RSS = .11884D+03 O = 1.07938
CHIZ( 1)= .002 with F( 1,102) = .00 [ .9625]

2.790

CHI-SOUARED Test for NORMALITY :CH12(2)

TEST for HETEROSCEDASTIC ERRORS 111*R2

11.3404 with 11 Variables
F(IO, 93) = 1.0583 [ .4023]
RESET F-TEST for adding Yhatz
F( 1,103) = .804 [ .3720]
RESIDUAL CORRELOGRAM

111*(Sum of 16 Squared Residual Autocorrelations) = 12.540
1 2 3 4 5 6 7

-.0902
8
-.0197

-.O308
9
.0478

—.0226
-.1420

67

.0954
11
-.1920

-.0336 .
12
.0258

10

15
.1065

RESIDUAL AUTOREGRESSION of Order

LAG

COEFF. —
S.E.'s

LAG

COEFF.
S.E.'s

RSS =
CHI’(12)=

16
.0258

l
.0745
.1118

9 10
.0162
.1085

-.0336
.1123

-.1910
.1068

-.0261
.1098
11

-.1879
.1088

12
4
.0917
.1109
12
-.0259
.1092

5
0468

.1113

.46489D+03 -
10.729 with F(12.

2.41062
80) =

0—

General TEST of FUNCTIONAL FORM 111*R2 = 3

F(25.

77) =

1.3536 [ .1578]

Regressors used for forming the Quadratic are:

CONSTANT BLKNIGHT DSTRES

COST

HETEROSCEDASTICITY Coefficients and t—Values are:

0927 -.0860
13 14
0055 -.0689
6 7 8
.1418 —.0717 -.0467
.1084 .1081 .1092
.81 [ .6390]
3.8897 with 26 Variables

2 LENGTH 4 PERCENTM TARCST

Vars: V1 * V1 V2 * V1 V3 * V1 V3 * V2 V4 * V1 V4 * V2
Coeff:. 2.37940 -14.20482 4.18022 2.70444 2.02983 5.07042
t-Value .49 -2.87 .90 .63 .52 2.54

Vars: V4 * V3 V4 * V4 V5 * V1 V5 * V2 V5 * V3 V5 * V4
Coeff: -1.07284 —.05898 -.12264 .24621 -.09730 -.06673
t-Value -.54 —.07 -.29 1.14 -.57 —.47

Vars: V5 * V5 V6 * V1 V6 * V2 V6 * V3 V6 * V4 V6 * V5
Coeff: .00382 .18139 .09912 -.10379 -.04103 .00416
t-Value .50 1.58 1.45 -1.19 -1.17 .64

Vars: V6 * V6 V7 * V1 V7 * V2 V7 * V3 V7 * V4 V7 * V5
Coeff: -.00115 .00824 -.00429 ~.00467 -.000030259 -.00093
t-Value -1.09 .56 -.49 -.71 -.01 -1.11

Vars: V7 * V6 V7 * V7
Coeff: —.00048 .000000641
t-Value -1.78 .27
Model OUTPUT from P C - G I V E
Outcomes for the first 28 OBSERVATIONS

M E A N S of V A R I A B L E S
RSUCCESS CONSTANT BLKNIGHT DSTRES COST 2 LENGTH 4 PERCENTM
5.1071 1.0000 .3214 .0714 2.0714 9.1786 25.4286
TARCST
105.7857
STANDARD DEVIATIONS OF VARIABLES
RSUCCESS CONSTANT BLKNIGHT DSTRES COST 2 LENGTH 4 PERCENTM
3.0712 .0000 .4756 .2623 .8133 14.2180 25.9833
TARCST
191.4199
D U R B I N - W A T S O N TESTS
RSUCCESS CONSTANT BLKNIGHT DSTRES COST 2 LENGTH 4 PERCENTM
2.0182 .0000 1.9649 1.6154 1.9040 1.4694 2.2352
TARCST
2.5290

CORRELATION MATRIX

RSUCCESS CONSTANT BLKNIGHT DSTRES

RSUCCESS 1.0000

CONSTANT .0000 1.0000

LKNIGHT -.5569 .0000 1.0000

DSTRES .2660 .0000 -.1909 1.0000
COST 2 .1896 .0000 -.0616 -.1985
LENGTH 4 .0284 .0000 —.0964 .2348
PERCENTM .2175 .0000 —.1998 .0660
TARCST .0645 .0000 .4695 -.1406
EQ(12) Modelling RSUCCESS by RLS

COST 2 LENGTH 4 PERCENTM TARCST
1.0000
-.1645 1.0000
.0809 -.4121 1.0000
.2632 -.2274 -.1336

54
O
O
('0
(7

 

68

The Sample is 5 to 115 less 0 Forecasts

VARIABLE COEFFICIENT STD ERROR H.C.S.E. t-VALUE PARTIAL r2
CONSTANT 2.2579175 .69500 .69735 3.24880 .0921
BLKNIGHT -1.6686837 .55771 .59821 -2.99202 .0793
DSTRES 1.4482112 .57145 .52523 2.53425 .0582
COST 2 .7906051 .29845 .31967 2.64900 .0632
LENGTH 4 .0650850 .02673 .03051 2.43503 .0539
PERCENTM .0331413 .01040 .01026 3.18519 .0889
TARCST .0008640 .00037 .00015 2.32053 .0492
R2 = .2753667 0 = 2.4132825 F( 6,104) = 6.59 [ .0000] DW = 2.164
RSS = 605.6889938162 for 7 Variables and 111 Observations
Information Criteria: SC = 1.99; HQ = 1.89; FPE = 6.19
t( 82) for a Zero Sample Innovation Mean - -1.09

R2 Relative to DIFFERENCE+SEASONALS = .65349

SEASONAL MEANS of DIFFERENCES are
-.01818

Testing for Serial Correlation from Lags 1 to 2

CH12( 2) = 1.154 and F-Form( 2,102) = .54 [ .5869]
Error Autocorrelation Coefficients:

-.1014 -.0436
ARCH TEST
Residuals Scaled by .2413D+01
CNST 1 LAG

COEFF. .9268 .0047

S.E.'s .1392 .0992

Rss = .11884D+03 O = 1.07938

CHIZ( 1)= .002 with F( 1,102) = .00 [ .9625]

CHI-SQUARED Test for NORMALITY :CHI’(2) = 2.790

11:39:

q
.1

TEST for HETEROSCEDASTIC ERRORS 111*R2 = 11.3404 with 11 Variables
F(lO, 93) = 1.0583 [ .4023]
' PROGRESS to DATE:
MODEL PARAMETERS RSS 0 SCHWARZ Criterion
1 30 484.32622091 2.44526810 2.74607449
2 28 493.56921245 2.43856864 2.68012250
3 25 498.67523816 2.40801771 2.56312986
4 23 502.84134202 2.39041966 2.48659310
5 20 503.35437263 2.35188578 2.36032824
6 17 525.68363113 2.36482115 2.27644877
7 14 531.05030541 2.33981739 2.15932135
8 12 548.34116416 2.35346545 2.10650585
9 10 568.26890942 2.37200861 2.05734654
10 8 595.45997247 2.40440529 2.01922957
11 7 605.68899382 2.41328254 1.99383384

Econometric Modelling Session Finished at

9 on 3lst July
Output for c:\thesis\a
Begun at

1: 2:52 on 3lst July

Data came from the Inp

1992
ppend.EQN
1992

ut Files:

c:\thesis\ecnscn.INF

and c:\thesis\ecnscn.BIN

Q( 1) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = .032 RSUCCESl -.121 RSUCCE52+ .014 RSUCCESS

SE ( .11049) ( .10306) ( .10291)

+ .011 RSUCCES4+ 4.527 -.903 BLKNIGHT -l.066 COOP

( .09870) ( 1.90041) ( .76203) ( .68076)

+ .879 DSTRES -.712 POLICIES~ .154 RELATE -.365 C05?

( .71292) ( .66124) ( .59711) ( .41687)
-.472 COST 1* .794 COST 2* .292 COST 3* 056 GNPCST

( .36948) ( .35019) ( .36688) ( 06019)
-.036 GNPCST 1+ .013 GNPCST 2 -.115 GNPCST 3+ .000019 GNPRATIO

69

 

( .10093) ( .10394) ( .09573) ( .00008)
-.031 LENGTH + .040 LENGTH 1 -.048 LENGTH 2 -.020 LENGTH 3
( .03347) ( .03355) ( .03521) ( .03605)
+ .084 LENGTH 4+ .017 PERCENTM+ .015 PERCENTX+ .001 TARCST
( .03491) ( .02577) ( .02121) ( .00042)
-.001 TARCST 1+ .000340 TARCST 2 -.001 TARCST 3
( .00041) ( .00041) ( .00040)
R2 - .4205625 0 - 2.4452681 F(29, 81) - 2.03 [ .0070] DW = 2.121
RSS = ' 484.3262209063 for 30 Variables and 111 Observations
Information Criteria: SC = 2.75; HQ = 2.31; FPE = 7.60
R2 Relative to DIFFERENCE+SEASONALS - .72292
STATIC LONG RUN
RSUCCESS = 4.255 -.849 BLKNIGHT -1.002 COOP
S.E. ( 1.27316) ( .77727) ( .65246)
+ .826 DSTRES -.669 POLICIES+ .145 RELATE .233 COST
( .71233) ( .63395) ( .56406) ( .61135)
. -.077 GNPCST + .000018 GNPRATIO+ .023 LENGTH .016 PERCENTM
( .18047) ( .00008) ( .07056) ( .02351)
+ .014 PERCENTX -.000138 TARCST
( .02076) ( .00075)
Mean = 4.963964 S.D. = 2.756572 0 = 2.4452681
Chow F[ 0., 0.] = .00 Normality Chi2(2) = 1.80
AR 1- 2F[ 2., 79.] = 1.71 ARCH 1 F[ 1., 79.] = .68
Xi2 F[33., 47.] = .54 RESET F[ 1., 80.] = 1.62
AR 1- 7F[ 7., 74.] = .76 ARCH 7 F[ 7., 67.] = 1.17
Xi2 F[33., 47.] = .54
EQ( 2) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = .016 RSUCCESl -.120 RSUCCES2 -.Oll RSUCCES3
SE ( .10943) ( .10259) ( .10027)
+ .008 RSUCCES4+ 4.731 -.856 BLKNIGHT -1.078 COOP
( .09706) ( 1.88045) ( .75602) ( .67131)
+ .886 DSTRES -.651 POLICIES+ .188 RELATE -.334 COST
( .68649) ( .64807) ( .59322) ( .41424)
-.489 COST 1+ .798 COST 2+ .209 COST 3+ .048 GNPCST
( .36769) ( .34892) ( .35977) ( .05967)
-.048 GNPCST 1+ .024 GNPCST 2 -.034 LENGTH + .043 LENGTH 1
( .10004) ( .10268) ( .03320) ( .03331)
-.047 LENGTH 2 -.021 LENGTH 3+ .075 LENGTH 4+ .018 PERCENTM
( .03511) ( .03585) ( .03396) ( .02555)
+ .015 PERCENTX+ .001 TARCST -.001 TARCST 1+ .000339 TARCST 2
( .02115) ( .00041) ( .00041) ( .00041)
-.001 TARCST 3
( .00040)
R2 = .4095044 0 = 2.4385686 F(27, 83) = 2.13 [ .0047] DW = 2.065
RSS = 493.5692124498 for 28 Variables and 111 Observations
Information Criteria: SC = 2.68; HQ = 2.27; FPE = 7.45
R2 Relative to DIFFERENCE+SEASONALS = .71763
Mean = 4.963964 S.D. = 2.756572 0 = 2.4385686
Chow F[ 0., 0.] = .00 Normality Chi2(2) = 1.96
AR 1— 2F[ 2., 81.] = 1.19 ARCH 1 F[ 1., 81.] = .69
Xi2 F[33., 49.] = .52 RESET F[ 1., 82.] = 1.53
AR 1- 7F[ 7., 76.] = .58 ARCH 7 F[ 7., 69.] = 1.32
Xi2 F[33., 49.] = .52
E0( 3) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = .006 RSUCCESl -.119 RSUCCES2 -.005 RSUCCE53
SE ( .10277) ( .09605) ( .09529)
+ .002 RSUCCES4+ 4.652 —.870 BLKNIGHT -1.001 COOP
( .09168) ( 1.73594) ( .71576) ( .65376)
+ .948 DSTRES -.651 POLICIES+ .225 RELATE -.376 COST
( .67051) ( .63967) ( .57202) ( .39890)
-.499 COST 1+ .837 COST 2+ .172 COST 3+ .048 GNPCST
( .34198) ( .32805) ( .34666) ( .05882)

 

70

-.031 LENGTH + .047 LENGTH 1 -.058 LENGTH 2+ .071 LENGTH 4
( .03232) ( .03198) ( .03185) ( .03090)
+ .020 PERCENTM+ .015 PERCENTX+ .001 TARCST -.001 TARCST 1
( .02498) ( .02070) ( .00041) ( .00040)
+ .000299 TARCST 2 -.001 TARCST 3
( .00040) ( .00039)
R2 = .4033957 0 = 2.4080177 F(24, 86) = 2.42 [ .0015] Dw = 2.082
RSS = 498.6752381588 for 25 Variables and 111 Observations
Information Criteria: SC = 2.56; HQ = 2.20; FPE = 7.10
R2 Relative to DIFFERENCE+SEASONALS = .71471
Mean = 4.963964 S.D. = 2.756572 0 - 2.4080177
Chow F[ 0., 0.] = .00 Normality Chi2(2) = 2.17
AR 1— 2F[ 2., 84.] = 1.66 ARCH 1 F[ 1., 84.] = .63
Xi2 F[33., 52.] = .50 RESET F[ 1., 85.] = 2.13

E0( 4) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = -.124 RSUCCE52+ .010 RSUCCES3 -.001 RSUCCES4
SE ( .09514) ( .09276) ( .09082)
+ 4.516 -.891 BLKNIGHT -.982 COOP + 1.007 DSTRES
( 1.66138) ( .70964) ( .64648) ( .65332)
-.564 POLICIES+ .155 RELATE -.289 COST -.510 COST 1
( .62674) ( .56118) ( .37487) ( .33441)
+ .803 COST 2+ .165 COST 3 -.034 LENGTH + .052 LENGTH 1
( .32080) ( .33747) ( .03187) ( .03093)
-.059 LENGTH 2+ .072 LENGTH 4+ .026 PERCENTM+ .015 PERCENTX
( .03158) ( .03055) ( .02231) ( .01999)
+ .001 TARCST -.001 TARCST 1+ .000299 TARCST 2 -.001 TARCST 3
( .00040) ( .00039) ( .00039) ( .00038)
R2 = .3984114 0 = 2.3904197 F(22, 88) = 2.65 [ .0007] DW = 2.156
RSS = 502.8413420234 for 23 Variables and 111 Observations
Information Criteria: SC = 2.49; HQ = 2.15; FPE = 6.90
R2 Relative to DIFFERENCE+SEASONALS = .71233
Mean = 4.963964 S.D. = 2.756572 0 = 2.3904197
Chow F[ 0., 0.] = .00 Normality Chi’(2) = 2.15
AR 1- 2F[ 2., 86.] = 1.81 ARCH 1 F[ 1., 86.] = .56
Xi2 F[33., 54.] = .63 RESET F[ 1., 87.] = 1.81
EQ( 5) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = -.123 RSUCCESZ+ 4.604 -.862 BLKNIGHT
SE ( .09236) ( 1.36062) ( .68211)
-.981 COOP + 1.020 DSTRES -.563 POLICIES -.284 COST
( .62414) ( .63331) ( .61013) ( .36689)
-.505 COST 1+ .800 COST 2+ .139 COST 3 -.036 LENGTH
( .32690) ( .31005) ( .31788) ( .03048)
+ .053 LENGTH 1 —.058 LENGTH 2+ .072 LENGTH 4+ .026 PERCENTM
( .02939) ( .03045) ( .02948) ( .02173)
+ .015 PERCENTX+ .001 TARCST -.001 TARCST 1+ .000300 TARCST 2
( .01958) ( .00039) ( .00036) ( .00037)
—.001 TARCST 3
( .00037)
R2 = .3977976 0 = 2.3518858 F(l9, 91) = 3.16 [ .0001] DW = 2.156
RSS = 503.3543726349 for 20 Variables and 111 Observations
Information Criteria: SC = 2.36; HQ = 2.07; FPE = 6.53
R2 Relative to DIFFERENCE+SEASONALS = .71203
Mean = 4.963964 S.D. = 2.756572 0 = 2.3518858
Chow F“ 0., 0.] = .00 Normality Chi2(2) = 2.36
AR 1— 2F[ 2., 89.] = 1.65 ARCH 1 F[ 1., 89.] = .59
Xi2 F[34., 56.] = .56 RESET F[ 1 , 90.1 = 1.90

E0( 6) Modelling RSUCCESS by OLS

The Sample is 5 to 115 less C Forecasts

71

RSUCCESS = 3.282 -.961 BLKNIGHT -1.141 COOP

SE ( 1.00916) ( .65130) ( .60315)
+ 1.266 DSTRES -.512 COST 1+ .858 COST 2+ .094 COST 3
( .58036) ( .31633) ( .30253) ( .31481)

—.032 LENGTH + .050 LENGTH 1 -.051 LENGTH 2+ .072 LENGTH 4
( .03028) ( .02899) ( .02949) ( .02723)
+ .025 PERCENTM+ .012 PERCENTX+ .001 TARCST —.001 TARCST 1
( .02182) ( .01954) ( .00039) ( .00036)
+ .000265 TARCST 2 -.001 TARCST 3
( .00037) ( .00037)
R2 n .3710834 0 3 2.3648211 F(l6, 94) a 3.47 [ .0001] DW - 2.161
RSS = 525.6836311271 for 17 Variables and 111 Observations
Information Criteria: SC = 2.28; HQ = 2.03; FPE = 6.45

R2 Relative to DIFFERENCE+SEASONALS = .69926

Mean = 4.963964 S.D. = 2.756572 0 = 2.3648211
Chow F] 0., 0.] = .00 Normality Chiz(2) = 2.55
AR 1- 2F[ 2., 92.] = .42 ARCH 1 F[ 1., 92.] = .73
X12 F[29., 64.] = .77 RESET F[ 1., 93.] = 1.36
EQ( 7) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = 3.454 -.917 BLKNIGHT -1.134 COOP
SE ( .90387) ( .62787) ( .57879)
+ 1.312 DSTRES -.506 COST 1+ .887 COST 2 -.030 LENGTH
( .56616) ( .30550) ( .29741) ( .02979)
+ .048 LENGTH 1 -.054 LENGTH 2+ .070 LENGTH 4+ .037 PERCENTM
( .02836) ( .02869) ( .02686) ( .01046)
+ .001 TARCST -.001 TARCST 1 -.001 TARCST 3
( .00038) ( .00036) ( .00036)
R2 = .3646628 0 = 2.3398174 F(13, 97) = 4.28 [ .0000] DW = 2.173
RSS = 531.0503054067 for 14 Variables and 111 Observations
Information Criteria: SC = 2.16; HQ = 1.96; FPE = 6.17
R2 Relative to DIFFERENCE+SEASONALS = .69619
Mean = 4.963964 S.D. = 2.756572 0 = 2.3398174
Chow F[ 0., 0.] = .00 Normality Chi2(2) = 2.74
AR 1- 2F[ 2., 95.] = .50 ARCH 1 F[ 1., 95.] = .30
Xi2 F[23., 73.] = .81 RESET F[ 1., 96.] = 1.47
E0( 8) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = 3.443 -l.100 BLKNIGHT -1.174 COOP
SE ( .88672) ( .58835) ( .58024)
+ 1.306 DSTRES —.380 COST 1+ .842 COST 2 —.039 LENGTH 2
( .56441) ( .29887) ( .29804) ( .02754)
+ .061 LENGTH 4+ .037 PERCENTM+ .001 TARCST -.001 TARCST 1
( .02651) ( .01050) ( .00038) ( .00036)
-.001 TARCST 3
( .00036)
R2 = .3439764 U = 2.3534655 F(ll, 99) = 4.72 [ .0000] DW = 2.169
RSS = 548.3411641640 for 12 Variables and 111 Observations
Information Criteria: SC = 2.11; HQ = 1.93; FPE = 6.14
R2 Relative to DIFFERENCE+SEASONALS = .68630
Mean = 4.963964 S.D. = 2.756572 0 = 2.3534655
Chou F’ 0., 0.] = .00 Normality Chi2(2) = 1.80
AR 1— 2F: 2., 97.‘ = .63 ARCH 1 F[ 1., 97.} = .02
Xi? F(19., 79. = .61 RESET F[ 1., 98.] = 1.83
E0( 9) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = 2.601 -1.238 BLKNIGHT -.921 C00?
HCSE I .66745] [ .57653] { .56405j
* 1.463 DSTRES + .778 COST 2* .062 LENGTH 4+ .035 PERCENTM

[ .53175: E .30607; E .02539} L .00997;

72

+ .001 TARCST -.001 TARCST 1 -.001 TARCST 3
[ .00016] [ .00030] [ .00017]
R2 = .3201353 6 = 2.3720086 F( 9,101) - 5.28 [ .0000] DW = 2.130
RSS = 568.2689094206 for 10 Variables and 111 Observations
Information Criteria: SC - 2.06; HQ = 1.91; FPE = 6.13
R2 Relative to DIFFERENCE+SEASONALS = .67490
Mean = ‘ 4.963964 S.D. = 2.756572 6 = 2.3720086
Chow F[ 0., 0.] = .00 Normality Chi’(2) = 2.66
AR 1- 2F[ 2., 99.] = .55 ARCH 1 F[ 1., 99.] = .02
Xi2 F[15., 85.] = .60 RESET Ff 1.,100.] = 1.26
EQ(10) Modelling RSUCCESS by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = 2.314 -1.428 BLKNIGHT -.755 COOP
SE ( .69370) ( .58434) ( .56724)
+ 1.434 DSTRES + .810 COST 2+ .064 LENGTH 4+ .035 PERCENTM
( .56945) ( .29773) ( .02664) ( .01048)
+ .001 TARCST
( .00039)
R2 = :2876045 0 = 2.4044053 F( 7,103) = 5.94 [ .0000] DW = 2.159
RSS = 595.4599724654 for 8 Variables and 111 Observations
Information Criteria: SC - 2.02; HQ = 1.90; FPE = 6.20
R2 Relative to DIFFERENCE+SEASONALS = .65934
Mean = 4.963964 S.D. = 2.756572 0 = 2.4044053
Chow F[ 0., . = .00 Normality Chi2(2) = 2.78
AR 1- 2F[ 2.,101.] = .60 ARCH 1 F[ 1.,101.] = .03
Xi? F[11., 91.] = .86 RESET F[ 1.,102.] = 1.15
30(11) Modelling Rsuccsss by OLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = 2.258 -1.669 BLKNIGHT+ 1.448 DSTRES
SE ( .69500) ( .55771) ( .57145)
+ .791 COST 2+ .065 LENGTH 4+ .033 PERCENTM+ .001 TARCST
( .29845) ( .02673) ( .01040) ( .00037)
R2 = .2753667 0 = 2.4132825 F( 6,104) = 6.59 I .0000] DW = 2.164
RSS = 605.6889938162 for 7 Variables and 111 Observations
Information Criteria: SC - 1.99; HQ = 1.89; FPE = 6.19
R2 Relative to DIFFERENCE+SEASONALS = .65349
Mean = 4.963964 S.D. = 2.756572 0 = 2.4132825
Chow F[ 0., 0.] = .00 Normality Chi2(2) = 2.79
AR 1- 2F[ 2.,102.] = .54 ARCH 1 F[ 1.,102.] = .00
Xi2 F[10., 93.] = 1.06 RESET F[ 1.,103.] = .80
Xi*Xj F[25., 78.] = 1.35
EO(12) Modelling RSUCCESS by RLS
The Sample is 5 to 115 less 0 Forecasts
RSUCCESS = 2.258 -1.669 BLKNIGHT+ 1.448 DSTRES
SE ( .69500) ( .55771) ( .57145)
+ .791 COST 2+ .065 LENGTH 4+ .033 PERCENTM+ .001 TARCST
( .29845) ( .02673) ( .01040) ( .00037)
R2 = .2753667 0 = 2.4132825 F( 6,104) = 6.59 [ .0000] DW = 2.164
RSS = 605.6889938162 for 7 Variables and 111 Observations
Information Criteria: SC = 1.99; HQ = 1.89; FPE = 6.19
t( 82) for a Zero Sample Innovation Mean = -1.09
R2 Relative to DIFFERENCE+SEASONALS = .65349
Mean = 4.963964 S.D. = 2.756572 0 = 2.4132825
Chow F[ 0., 0.3 = .00 Normality Ch12(2) = 2.79
AR 1- 2F[ 2.,102.] = .54 ARCH 1 F[ 1.,102.] = .00
Xi2 F[10., 93.] = 1.06

Econometric Modelling Session Finished at
11:39: 9 on 3lst July 1992

73

APPENDD( B

Figure 5.1: Scaled Residuals

.59

.50

.50

 

 

74

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

58

 

 

 

A 1
60

 

79

 

 

 

I l
89

 

 

A l
90

 

 

100

 

 

110

 

 

i 1
120

75

Figure 5.2: One Step Chow
1 . 40 -

)—

1.20 '-

 

.OO —

 

.6. )-

.40 *-

 

.2° '-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

oo so so ,

7o so A so 100

11°

12°

76

Figure 5.3: Decreasing Horizon

 

.00 >—

.60 )—

.20 '—

 

 

 

4o so so 7o oo ‘ so A 1.00 ‘ no 120

Figure 5.4: Increasing Horizon

1.20

.60

.40

 

 

 

A 1.00

A no

A 12o

 

 

 

 

 

 

 

 

 

 

 

_______

 

 

 

BIBLIOGRAPHY

79

Bibliography

Achen, Christopher H. Interpreting and Using Rgggession, 1989, Newbury Park: Sage Publications.

Amato, Timothy and Jim Granato Formulating and Tating Macro Political Models in the Lucas
Critique Era Midwest Chicago, April 1992.

 

Aron, R Peace and War: a Theog of International Relations, 1966, Garden City: Doubleday.

Baldwin, Robert E. and J. David Richardson International Trade and Finance: Readings, 1986,
Boston: Little, Brown and Company.

Baldwin, David A. Economic Statecraft 1985, Princeton: Princeton University Press.

 

Beck, Nathanial Comparing Dynamic Specifications: The Case of Praidential Approval Forthing
Coming: Political Analysis, March 1991.

Daoudi, M. 8., and M. S. Dajani Economic Sanctions: Ideals and Exgrience, 1983, London:
Routledge 8: Kegan Paul.

Dixit, Avinash and Albert 5. Kyle The Llse of Promotion and Subsidies for Entry Promotion and
Deterrence The American Economic Review, v. 75 no. 1, March 1985, pp. 139-152.

Doxey, Margaret P. International Sanctions in Contemmrg Persgtive, 1987, London: Macmillan
Press Ltd.

Doxey, Margaret R, Economic Sanctiomnd International Enforcement, 1971, London: Oxford
University Press.

 

Dufour, Jean-Marie Recursive Stability Analysis of Linear Regression Relationships [ournal of
Econometrics v. 19, 1982, pp. 31-76.

Engle, Robert and David Hendry Testing Super Exogeneity and Invariance in Regression Models
Applied Economics Discussion Paper no. 100, University of Oxford, 1990.

Engle, Robert F. and C. W. J. Granger Co-Intigration and Error Correction: Representation, Estimation
and Testing Econometrica, v. 55, no. 2, March 1987.

Engle, Richard, David Hendry and lean-Francois Richard Exogeneity Econometrica, v. 51, 1983, pp.
277-304.

Engle, Robert Autoregressive Conditional Heteroskedasticity with Estimates of the variance of United
Kingdom Inflation Econometrica, v. 50, no. 4, July 1982.

Freeman, John R. Granger Causality and the Times Series Analysis of Political Relationships pp. 328-358.

80
Friedman, Milton Newsweek, January 21 1980, p. 76.

Gilpin, Robert The Political Economy of International Relations, 1987, Princeton: Princeton
University Press.

Granato, James The Use of Encompassing Theory to Test Competing Models MPSA, April 1991.

Granato, James An Agenda for Econometric Model Building Political Analysis, 1992, vol. 3, Ann
Arbor: The University of Michigan Press.

Hall, 8. G. Maximum Likelihood Estimation of Cointigration Vectors: An Example of the Iohansen
Procedure Oxford Bulliten of Economics and Statistics, v. 51, no. 2, 1989.

Hanushek, Eric A. and John E. Jackson Statistical Methods for Social Scientists 1977, New York:
Academic Press inc.; Harcourt, Brace Jovanovich.

 

Hendry, David F. and Adrian J. Neale Interpreting Long-Run Equalibrium Solutions in Conventional
Macro Models: A Comment The Economic loumal, v. 98, September 1988.

Hendry, David F. and Jean-Francois Richard On the Formulation of Empirical Models in Dynamic
Econometrics lournal of Econometrics, 1982, pp. 3-33.

Hendry, David and Mary Morgan A Re-Analysis of Confluence Analysis Oxford Economic Papers,
v. 41, 1989.

Hufbauer, Gary Clyde Jeffrey J. Schott and Kimberly Ann Elliott Economic Sanctions
Reconsidered: History and Currant Polig, 1990, Washington, DC: Institute for International
Economics.

Hufbauer, Gary Clyde Jeffrey J. Schott and Kimberly Ann Elliott Economic Sanctions
Reconsidered: Supplemental Case Histories, 1990, Washington, DC: Institute for International
Economics.

Johansen, Soren and Katerina Juselius Maxim um Likelihood Estimation And Inference in Cointigration-
With Applications to the Demand For Money Oxford Bulliten of Economics and Statistics v. 52, no.
2, 1990.

 

Johansen, Soren Statistical Analysis of Cointigration Vectors Journal of Economic Dynamics and
Control, v. 12, 1988, pp. 231-254.

Kennedy, Peter A Guide to Econometrics, 1989, Second Edition, Cambridge: The MIT Press.
Keohane, Robert 0. ed. Neorealism and Its Critics, 1986, New York: Columbia University Press

Leyton-Brown, David, ed. The Utilig of International Economic Sanctions, 1987, London: Croom
Helm.

Lindsay, James M. Trade Sanctions as Policy Instruments: A Re-examination International Studies
Quarterly, June 1986, v. 30, pp. 153-173.

MacKinnon, James G. Critical Valua for Cointigration Tests 24 January 1990.

81

Nincic, Miroslav and Peter Wallensteen eds. Dilemmas of Economic Coercion: Sanctions in World
Politics, 1983, New York: Praeger.

 

Nossal, Kim Richard International Sanctions as International Punishment International Organisation,
v. 43. no. 2 Spring 1989, pp. 301-322.

Organski, A.F.K. and Jacek Kugler The War Ledger, 1980, Chicago: University of Chicago Press.

Ostrom, Charles and Renee Smith Routinizaiton and Reciprocity in 11.5. Defence Expenditure Policy
Making: An Analysis of Countervailing Form, APSA, Washington DC, 1991.

Putnam, Robert D. Diplomacy and Domestic Politics: The Logic of Two-Level Games International
Organization, 1988, v. 42, no. 3, pp. 427-460.

Renwick, Robin Economic Sanctions 1981, Rensselaer: Hamilton Printing Company.

 

Rogowski, Ronald Commerce and Coalitions: How Trade Affects Domestic Political Alignments,
1989, Princeton: Princeton University Press.

Russell, Bertrand Power: A New Social Analysis, 1938, New York: W. W. Norton.

Spanos, Aris Statistical Foundation of Econometric Modelling, 1990, Cambridge: Cambridge
University Press.

 

1L.

.

.4:

...t.

..»

r...

4..

_ ..

..s.

. c
I.»

..e

if

.i.x.~‘v

.11....

1..

e .

t...

 

.12 . .
Lila...
z .17... ..
;.........‘
.13....
.111 i :23...
.x. x111...
...zllx
3.7.;‘1111...
t: L 12;:
_ xixtlt.
11.14.:

i . . ts.
.C... ...ixzxi