Content uploaded by Dawood Mamoon
Author content
All content in this area was uploaded by Dawood Mamoon on May 16, 2014
Content may be subject to copyright.
Working Paper
No. 446
S Mansoob Murshed
and
Dawood Mamoon
July 2007
ON THE COSTS OF NOT LOVING THY NEIGHBOUR
AS THYSELF:
The trade, democracy and military expenditure
explanations behind India–Pakistan rivalry
2
ISSN 0921-0210
The Institute of Social Studies is Europe’s longest-established centre of higher education and
research in development studies. Post-graduate teaching programmes range from six-week
diploma courses to the PhD programme. Research at ISS is fundamental in the sense of laying a
scientific basis for the formulation of appropriate development policies. The academic work of ISS
is disseminated in the form of books, journal articles, teaching texts, monographs and working
papers. The Working Paper series provides a forum for work in progress which seeks to elicit
comments and generate discussion. The series includes academic research by staff, PhD
participants and visiting fellows, and award-winning research papers by graduate students.
Working Papers are available in electronic format at www.iss.nl/workingpapers.
Please address comments and/or queries for information to:
Institute of Social Studies
P.O. Box 29776
2502 LT The Hague
The Netherlands
or
E-mail: workingpapers@iss.nl
3
TABLE OF CONTENTS
ABSTRACT 4
1 INTRODUCTION 5
2 THEORETICAL MODEL 10
2.1 Costs of War 10
2.2 Costs of Peace 12
3 EMPIRICAL ANALYSIS 16
3.1 Hypotheses 16
3.2 Data and Methodology 16
3.2.1 Data 16
3.2.2 Methodology 20
3.2.3 Results with VAR models 21
3.2.4 Results with VECM models 32
4 CONCLUSIONS 39
REFERENCES 40
DATA AND SOURCES 41
Single Country Variables 41
Dyadic Variables 42
4
ABSTRACT
We examine whether greater inter-state trade, democracy and reduced military
spending lower belligerence between India and Pakistan. We begin with
theoretical models covering the opportunity costs of conflict in terms of trade
losses and security spending, as well as the costs of making concessions to
rivals. Conflict between the two nations can be best understood in a
multivariate framework where variables such as economic performance,
integration with rest of the world, bilateral trade, military expenditure,
population are simultaneously taken into account. Our empirical investigation
based on time series econometrics for the period 1950-2005 with causality tests
suggests that reduced trade, greater military expenditure, less development
expenditure, lower levels of democracy, lower growth rates and less general
trade openness are all conflict enhancing. Moreover, there is reverse causality
between bilateral trade, militarization and conflict; low levels of bilateral trade
and high militarization are conflict enhancing, equally conflict also reduces
bilateral trade and raises militarization. We also run forecasting simulations on
6 different VECM models. Globalization or a greater openness to international
trade in general are more significant drivers of a liberal peace, rather than a
common democratic political orientation suggested by the pure form of the
democratic peace.
Keywords
Inter-state conflict and trade, democracy and conflict, conflict and economic
development
J.E.L codes: D74, F13, F15, F51.
5
ON THE COSTS OF NOT LOVING THY NEIGHBOUR
AS THYSELF: The Trade, Democracy and Military
Expenditure Explanations Behind India–Pakistan Rivalry1
1 INTRODUCTION
Conflict may be motivated by factors such as historical grievances, the clash of
civilizations (Huntingdon, 1996), or pure avarice. Outright hostility between
states implies the absence of peaceful cooperation, manifesting itself in
diminished inter-state commerce, which in turn could further exacerbate the
rivalry between the same countries. In this paper we are concerned with inter-
state rivalry between India and Pakistan. Civil war, however, is the most
dominant form of war at present; see Harbom, Högbladh and Wallensteen
(2006) for data, and Murshed (2002) for a theoretical overview. Despite the
preponderance of civil war, inter-state rivalry has not withered away, and these
too can also divert substantial amounts of resources away from poverty
reduction in developing countries.
International trade allows one country to peacefully benefit from the
endowments of another nation through voluntary exchange. Furthermore, free
trade integrates the world economy. War is another way of expropriating the
endowments of another country, but it is costly as it destroys part of both
countries pre-existing wealth. Predation is an alternative to production, but it is
not usually the most efficient, as predation (war or other forms of larceny)
unnecessarily wastes resources. Such, unenlightened behaviour may be rational
or optimal from the standpoint of the individual person or a nation, but is
inefficient in the global sense. The work of Francis Edgeworth, writing in the
late 19th century, provides a useful starting point in understanding the
economic rationale for violence. Edgeworth distinguished between consent—
and its absence—in human economic interaction:
The first principle of Economics is that every agent is actuated only by self-
interest. The workings of this principle may be viewed under two aspects,
according as the agent acts without, or with, the consent of others affected by his
actions. In wide senses, the first species of action may be called war; the second,
contract. [Edgeworth, 1881, pp 16-17].
International economic interactions between nations may involve peaceful
trade, or it could be belligerent with reduced economic interaction. Outright
war is just one manifestation of the rivalry between nations; the armed peace is
equally consistent with aggressiveness. India and Pakistan are a case in point.
They have had at least four large scale military confrontations (1948, 1965,
1971 and 1999), but otherwise spend a great deal of time in uncompromising
posturing vis-à-vis each other. India, in particular, frequently accuses Pakistan
of sponsoring terrorism in her territory. But occasionally they make goodwill
1 We wish to thank Admasu Shiferaw and Arjun Bedi for valuable comments on
previous drafts of this paper.
6
1: Pakistan and India Hostility Levels
0
1
2
3
4
5
6
1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1988
1992
1995
1998
2001
2004
Years
Hstlev
gestures, such as sending out peace buses between cities like Delhi and Lahore,
and agree to cricket tours. Less frequently, major concessions are made mainly
by Pakistan, such as President Musharraf’s willingness to put aside the long
standing Pakistani demand and United Nations resolution for a plebiscite to
settle the future of Kashmir.2 Figure 1 charts the hostility levels of the two
states on a scale of 0-6. It has never been below 2, but is usually at a high level
of 4, which measures belligerency short of outright war.
FIGURE 1
Hostility between Pakistan and India
TABLE 1
The Military Burden in Selected Countries
Countries with Conflict
Defence
Expenditure
(% of GDP)
Countries without Conflict
Defence
Expenditure
(% of GDP)
India (2004) 2.34 Canada (2004) 1.19
Pakistan (2004) 4.14 Germany (2004) 1.38
Egypt (2004) 2.76 Holland (2004) 1.73
Syria (2003) 6.97 Sweden (2004) 1.73
Israel (2004) 9.30 Argentina (2004) 1.01
Lebanon (2003) 3.92 Mexico (2004) 0.51
Saudi Arabia (2004) 7.70 Nicaragua (2004) 0.69
Oman (2001) 12.16 Panama (2004) 0.97
Yemen (1999) 5.28 Paraguay (2004) 0.70
South Korea (2004) 2.45 Peru (2004) 1.20
USA (2004) 3.98 Guatemala (2004) 0.40
UK (2004) 2.57 El Salvador (2004) 0.66
The most recent year for which data is available is given in parenthesis.
Source: World Development Indicators (2006)
2 See http://news.bbc.co.uk/2/hi/south_asia/3330031.stm.
7
1.Pakistan- India Trade
0
2
4
6
8
10
12
14
16
18
20
1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
2004
Years
Tpitp
Tpitp
2. Pakistan's Exports and Imports to India
0
1
2
3
4
5
6
7
8
9
1958
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
Yea r s
Px-i
Pm-i
3.Total Trade/GDP for India and Pak ist an
0
10
20
30
40
50
60
70
80
1950
1954
1958
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
Years
Iop en
Pop e n
4. Pakistan's trade with Developed and
Developing Countries
0
10
20
30
40
50
60
70
80
90
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
Yea r
Tpdvd
Tpdvg
Both countries spend a considerable amount in military expenditure,
measured as a proportion of GDP (see table 1). In fact, these two countries
have among the highest military burdens in the world outside the Middle East.
One can surmise, that such large scale military expenditure detracts from
development and poverty reduction in South Asia which has the largest
concentration of the world’s poor, defined by below purchasing power parity
$1 a day per-person.
FIGURE 2
Patterns in India-Pakistan Trade
Polachek (1997) and Polachek and Seiglie (2006) argue that wars and
disputes between geographically contiguous states involve greater losses, as
more efficient geographically proximate trade is displaced. 3 This effect,
however, depends on the absence of alternative trading partners, who despite
3 When we come to comparing trade and conflict with many nations, not just dyadic
(pair-wise) interactions, Dorussen (1999) argues that although trade reduces conflict,
in the presence of many countries, an increase in the number of countries or the
world’s endowment may induce more conflict, as there are more countries to grab
from. Formally, it lowers the minimum probability of military success needed to make
conflict worthwhile in the presence or absence of trade with the target country. Hegre
(2002) shows that by taking ratios of the probabilities (rather than differences as in
Dorussen, 1999) the benefits of trade rise as the number of countries increase. Thus,
Dorussen (1999) establishes gains from conflict after globalization, whereas Hegre
(2002) models benefits from cooperation (or trade) as globalization gathers pace.
8
greater distance may be equally or more efficient. Figure 2 shows that India-
Pakistan official trade (as a proportion of Pakistan’s total international trade)
steadily declined from nearly 20% in the early 1950s, plummeting to almost
zero after their war in 1965, and has shown some signs of recovery in the
1990s. But it is still below the levels of the 1950s, which was shortly after the
two nations were separated politically. This is despite the fact that India and
Pakistan have fairly open economies at the present. Pakistan has traditionally
been more open than India (Figure 2, panel 3). Pakistan also trades more with
developing countries compared to developed countries, as shown by graph 4 in
figure 2. Conflict and rivalry are symptomatic of the absence of cooperation
including lower trade volumes. Equally, conflict may be said to be a
consequence of the lack of trade.
A related issue concerns the so-called democratic peace,4 see Polachek
(1997) and Polachek and Seiglie (2006) for a review of this substantial
literature. The idea is that democracies will not fight each other because they
share cultural norms that militate against forceful dispute resolution, or
alternatively the checks and balances that characterise political processes in
advanced democracies restrain violence. Put simply, the idea is that established
democracies do not go to war with each other, but cooperate instead. The
intellectual basis for this argument has been traced back to Immanuel Kant’s
(1795) work on the Perpetual Peace, where a like mindedness referred to as
cosmopolitanism would prevent outright war between republics; a tendency that
could be reinforced by commercial interdependence. Mirroring Kant’s
thoughts, the contemporary philosopher, John Rawl’s (1999) notion of peace
between liberal societies or nations, arguing that liberal societies do not go to
war with each other because their needs are satisfied, they are non-acquisitive
in the sense of not wishing to grow beyond an achieved steady-state level of
(presumably high) income, and they are tolerant of difference. They will only
fight in self-defence, and invade to prevent gross human rights abuses such as
genocide in other countries. They can, however, occasionally be duped into
supporting foreign wars. Polachek (1997) makes a case for the alternative
notion of the liberal peace, presenting empirical evidence to suggest that
advanced democracies cooperate, not because of their similar political systems,
but due to their vast and multiply intersecting economic interdependence.
Barbieri (1996) demonstrates that the liberal peace based upon the pacific
effects of economic interdependence may be a chimera. Oneal and Russett
(1999) and Hegre (2000), however, argue that economic interdependence
reinforces peace, particularly between democracies. Perhaps, we need a theory
that embeds democracy with economic interdependence. Democracies may,
however, go to war with other democracies that are distantly located, culturally
disparate and considerably poorer, something that is also echoed in Kant
(1795). Indeed, Robst, Polachek and Chang (2006) present some evidence to
suggest that more democratic nations could exhibit some degree of
4 Sometimes the literature refers to this concept also as the liberal peace, which is a
source of some confusion as some authors refer to the peace emanating from
economic interdependence as the liberal peace.
9
3.Paki stan and India's E ducati on
Expendit ures
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
1950
1954
1958
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
Years
Ped ug
Iedug
1. fatality Score
0
1
2
3
4
5
6
7
1950
1954
1958
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
Year
fatal
5. India's Polit y 2 sc ore
0
1
2
3
4
5
6
7
8
9
10
1950
1954
1958
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
Year
Ip2
6. Pakistan's Polity 2 score
-8
-6
-4
-2
0
2
4
6
8
10
1950
1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
Years
Pp2
2.Pakistan and India's Defence
Expenditures
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
1950
1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
Year
Pdg
Idg
India and Pakist an real GDP percapita
growth rates
-8
-6
-4
-2
0
2
4
6
8
10
1950 1954 1958 1962 1966 1970 1974 1978 1982 1986 1990 1994 1998 2002
Igp c
Pgp c
belligerence to less democratic countries, such as in the case of India’s actions
vis-à-vis Pakistan. Nevertheless, increased democratic levels can mandate
concessions and re-negotiation with neighbours.
The Polity score of democracy (see, Polity 4 project) ranges from 0-10.
Similarly there is an autocracy score of between -10 to 0. Together, the
autocracy and democracy score gives us an average score, acting as an indicator
of the overall political system, which is graphed above. Graph 5 in figure 3
shows that India has one of the highest democracy scores in the developing
world for the entire 50 year period (7-9), whereas Pakistan’s experience with
democracy fluctuates, with high autocracy scores associated with military coups
in 1958, 1969, 1977 and 1999.
FIGURE 3
Conflict, Development and Democracy Trends in India-Pakistan
Comparisons of graphs 2 and 3 in figure 3 indicates that military
expenditures tend to move inversely with development (education)
expenditure, providing prima facie evidence that large military expenditure
crowds out development in the social sector. Pakistan’s military expenditure is
consistently above India’s except in the mid-1960s when India had wars with
both China and Pakistan. In Pakistan’s case, military expenditure as a
proportion of GDP has historically been at 5%, but rising during and after its
1965 and 1971 wars with India to as high as 8%. The average defence
expenditure of Pakistan is 5.5% of GDP in the 1950-2005 period, whereas for
India it is about half at 2.8% of GDP. Since the 1990s Pakistan’s military
expenditure has been falling, and is now at a little above 4% of GDP, which
represents a historical low. As Indian education expenditure rose to 4 % of
GDP in the1990s, its defence expenditure fell from nearly 4% of GDP in the
mid-1960s to less than 3% of GDP (it has rarely been below 2% of GDP).
Pakistan’s education public expenditure is stagnating at around 2% of GDP.
10
The opportunity costs of conflict could rise when countries move to
higher stages of economic development as they have more to lose from
conflict, and have more resources to negotiate peaceful settlements. The 1990s
is considered to be a golden decade for India as GDP growth rates on average
the Indian economy grew at 5-6% annually. Pakistan has been growing at an
average of 6% for the last 3 to 4 years. Traditionally, from the early 1960s up
to the early 1990s, Pakistan’s was the faster growing economy of the two. Both
countries are in the second most rapidly growing region (South Asia) in the
world (World Development Indicators, 2006).
There is more to India-Pakistan conflict than merely Pakistan’s political
orientation and a comparison of bilateral economic growth rates. This is
because of the fact that despite high growth rates and relatively high
democracy scores in Pakistan up to 1999, conflict between the countries
escalated in the 1990s. By contrast, the current regime in Pakistan with a strong
military orientation (the military is highly influential and the President
continues to be the army chief), and therefore less democratic, is making major
unilateral concessions to India vis-à-vis their long standing disputes over
Kashmir. Could that be related to the very recent impressive growth record in
Pakistan? If anything, conflict between the two nations can be best understood
in a multivariate framework where the relevant variables and processes
(economic performance, integration with rest of the world, trade between the
conflicting nations, military expenditure, democracy, and population) are
simultaneously taken into account. The purpose of this paper is to examine
whether greater inter-state trade, democracy and reduced military spending
lower belligerence between India and Pakistan. We also investigate the causal
links between bilateral conflict and these variables in a time series framework,
between 1950 and 2005 in most cases. The rest of the paper is organised as
follows: section 2 contains the theoretical model, the econometric analysis is
presented in section 3, and section 4 concludes.
2 THEORETICAL MODEL
This section consists of two parts: the first deals with the costs of belligerent
behaviour in a single country context where the losses are displaced trade and
the crowding out effects of defence expenditure; the second looks at the costs
of peaceful behaviour where the disutility of making concessions to an
adversary is modelled in a two country setting. The situations we model either
pertain to limited warfare, with negligible effects on national endowments, or
alternatively we could be said to model the costs of an armed peace associated
with large security and military establishments. In many ways, conflict has
similar effects as other forms of trade wars.
2.1 Costs of War
We begin with a single country’s decision making with regard to belligerence,
based on Polachek (1997). The welfare of either country (U) depends upon
consumption (E), and security (S), entering the utility function in a separable
fashion:
11
),( SEuU = (1)
where:
TMXcQE −+−= (2)
Q is the total endowment of the country where a proportion c is devoted
to private and public non-military consumption and investment; a fraction 1-c
to a public good covering security or military expenditure. X and M denote
exports and imports to the rival country, and T represents trade (exports minus
imports) with the rest of the world. θ is the price of the exportable and the
price of the importable is the numeraire good, normalised at unity. There is
also a balance of trade constraint, the value of exports must equal imports:
0,0)(
<
=+− s
XTMSX K
θ
(3)
Following Polachek (1997) let us postulate that conflict disrupts trade.
Specifically, it lowers exports, but unlike in Polachek’s model both countries
are hostile towards each other, and not just one country (described as the
actor) against a passive target. So, in our model, both countries exports to each
other will decline, along with ambiguous effects on the terms of trade. The
country whose goods are demanded more elastically will experience the
negative terms of trade effect. Nevertheless, exports displaced by conflict are a
loss, as they represent foregone trade, especially in the context of neighbours
who might be expected to trade substantially in peaceful circumstances.
Substituting (3) as a constraint and (2) into (1) allows us to write a Lagrangian
function (L), where λ indicates the Lagrange multiplier:
)(])([);( SCTMSXSTMXcQuL
−
+
−
+
−+
−
=
θ
λ
(4)
The function C represents the distortionary (taxation and crowding out)
costs of security expenditure, which rises with S, so that the partial derivative is
positive. This is an additional cost associated with security spending, absent in
Polachek’s (1997) model. The first order condition with respect to S is:
0,0,
<
>+−= ssssss XCuCXu L
λ
(5)
In equation (5) the marginal utility of security (us) is equated to its marginal
costs. The latter (on the right-hand side of (5)) is comprised of the trade
disruption due to conflict, and the cost of diverting resources to military and
security expenditure. This, last effect, is absent from the Polachek models. The
cost of conflict is not just confined to displaced trade, but it also has a
distortionary resource cost because of security expenditure, either because of
distortionary taxation or due to the crowding out effect on other forms of
investment, including government spending on health and education; see
Deger and Sen (1990). Note, that security expenditure and benefits derived
from confronting one’s enemy does yield positive utility, but comes at a price.
12
There is, therefore, an additional cost of belligerent behaviour over and above
losses from trade displacement, and is likely to be substantial because it
detracts from poverty reduction directly. It is worth noting that trade costs and
losses from resource misallocation are a priori likely to be greater for the smaller
economy, Pakistan. The same will be true of the terms of trade which are likely
to deteriorate for Pakistan. This is because a smaller economy’s exports to its
larger neighbour are usually a greater proportion of its total exports, its goods
may be demanded more elastically, and the costs of an arms race are larger for
the smaller nation.
2.2 Costs of Peace
If peace is Pareto optimal, why don’t countries engage in it?5 In this section we
model the costs of peace, which include psychic non-pecuniary costs of
making concessions to one’s adversaries. Additionally, we try to demonstrate
how increased globalization and democratisation can help to reduce conflict by
lowering the cost of making concessions to one’s neighbours. To analyse these
factors we require a two country expected utility model of non-cooperative
strategic interaction.
The two countries: India and Pakistan are indexed by subscripts I for India
and P for Pakistan. There are two states of nature, denoted by superscripts:
one more peaceful or dovish (D), and the other associated with greater
hawkishness (H). Their probabilities are defined as
π
and 1 -
π
, respectively.
An important feature of our model is that states of conflict, or peace, are
relative. The probability of either state is in turn affected by an action (a) by
India and effort (e) by Pakistan. These are also the strategic variables employed
by the two sides to the conflict. We postulate that the probability of the
peaceful state
π
rises with the input of action and effort by the two sides, but at
diminishing rates. One can imagine a range of activities by one or both sides if
they wish to promote peace, including a greater willingness to compromise,
reduce military expenditure, devoting more resources to peaceful economic
development, or a greater willingness to respond to calls for peace by third-
parties such as the UN or under the influence of pressure from the United
States.
Actions and efforts to seek peace entail costs for each party. The costs of
actions to promote peace could take a variety of forms, but, above all, there is
the loss of face to either party’s hawkish domestic political constituencies,
including the military establishment. Increased globalization may, however,
augment the stock of rhetoric available to politicians who wish to push their
‘peace’ agenda through the political process. Secondly, and in a more palpable
sense, increased international trade and the growth it brings may provide the
5 Sir Normal Angell, winner of the 1933 Nobel peace price and former editor of
Foreign Affairs, in his great book The Great Illusion, asserted that nations could never
enrich themselves through war, and even a victorious nation would come off
economically worse from a war; see Angell-Lane (1910).
13
additional resources to buy off domestic ‘war’ lobbies. A more democratic
government, following military rule, may similarly use its mandate from the
people to justify greater peace and reduced military expenditure.
The expected utility of India is given by
++= )(),( D
I
D
I
D
II SEUeaU
π
))(()(),)(1( TaZSEUea H
I
H
I
H
I−+−
π
(6)
where I
D
Uand I
H
Udenote utilities or pay-offs in dovish and hawkish states
respectively, weighted by the probabilities of the two states. D
I
D
I
SE +,
H
I
H
I
SE +indicate the exogenous pair of payoffs from consumption and
security expenditure respectively in the less belligerent and more belligerent
states respectively. The difference is that in dovish state security spending is
lower and private consumption higher than in the hawkish state. There will
also be more trade between the two countries. Most importantly, the dovish
state of nature will imply greater poverty reduction. Z is the cost function of
undertaking the action, a. Action, a, increases the probability of peace,
π
,
however, undertaking it entails a cost, as described above. T indicates greater
globalization (more trade with the rest of the world), and this is postulated to
reduce the cost of making peace via the cost function (Z) as discussed above,
Za1 < 0.6 Also,
π
a > 0, but
π
aa < 0; there are diminishing returns to these
actions. Note, however, both Za > 0 and Zaa > 0.
Turning to Pakistan, we symmetrically have
++= )(),( D
P
D
P
D
PP SEUeaU
π
)),(()(),)(1( PTeLSEUea H
P
H
P
H
I−+−
π
(7)
L is the cost of effort, e, which increases the probability of peace,
π
. As
with India, greater globalization lowers the marginal cost of making peaceful
concessions, but so does a hybrid concept called increased democratisation (P)
for Pakistan only given the nature of swings there between democratically
elected governments and military rule; Le1 and Le2 < 0. Also,
π
e > 0, but
π
ee < 0,
Le > 0, and Lee > 0.
In the non-cooperative or Cournot-Nash game played by the two sides
both sides move simultaneously. Each side, therefore, maximises its own utility
function with respect to its own choice variable. For India, it implies
maximising utility, Equation (6), with respect to a as shown by
6 Increased globalization is unlikely to directly affect the marginal productivity of
actions or efforts (a, e) that raise the probability of peace (π).
14
[
]
a
H
I
D
Ia ZUU =⋅−⋅ )()(
π
(8)
Pakistan maximises Equation (7) with respect to e
[
]
e
H
P
D
Pe LUU =⋅−⋅ )()(
π
(9)
Note that in Equations (8) and (9) each side will equate its marginal
benefit from exercising their own strategic choice to the corresponding
marginal cost. Each side's strategic choices will depend on the first order
conditions given in Equations (8) and (9), along with a fixed conjecture about
the opposition’s strategic choice. These lead to the (linear) reaction functions
for both sides, obtained by totally differentiating Equations (8) and (9) with
respect to a and e. For India this is indicated by
[
]
[]
00
)()(
)()(
/≤
≥
≤
≥
⋅−⋅
⋅−⋅+
=ae
H
I
D
Iae
D
I
H
Iaaaa
Iif
UU
UUZ
Rdade
π
π
π
LKK
(10)
and for Pakistan by
[
]
[]
00
)()(
)()(
/≤
≥
≤
≥
⋅−⋅+
⋅−⋅
=ae
D
P
H
Peeee
H
P
D
P
ae
Pif
UUL
UU
Rdade
π
π
π
KKK
(11)
Note that
π
ae =
π
ea by symmetry.
The reaction functions are positively sloped if
π
ae > 0, implying that the
two strategies are complements. This is the standard assumption in the
literature on conflict. In our model, however, we postulate that
π
ae < 0, the
choice variables are strategic substitutes, and the reaction functions slope
downwards (Figure 4). This can occur because the strategy space is defined in
terms of peace. Thus, if one side behaves more peacefully it increases the utility
of both parties and the other side may free ride on this action by not bringing
about a corresponding increase in their action.
In Figure 4, two non-cooperative equilibria are illustrated by points N and
C respectively. Point C is more cooperative and peaceful with greater inter-
country trade and poverty reduction. A shift from N to C can occur because of
greater globalisation (rise in T) because of, say, the establishment of a free trade
area, and increased international (not necessarily just bilateral) trade lowers the
marginal cost of peaceful behaviour (Za1, Le1< 0). Analytically this means a
change in the first-order condition for India:
[
]
dTZUU a
H
I
D
Ia 1
)()( =⋅−⋅
π
(8')
and for Pakistan
15
[
]
dTLUU e
H
P
D
Pe 1
)()( =⋅−⋅
π
(9')
This pertains to the liberal peace. Alternatively, there could be a rise in the
exogenous pay-offs in terms of consumption expenditure (E) in (8) and (9)
above, leading to the same outcome in figure 4.
The costs of peaceful actions may be easier to bear when countries (in this
case only Pakistan) are more democratic, as there may be a mandate from the
people to engage in more poverty reduction, greater social sector spending and
lower military expenditure. This corresponds to the democratic peace and will
cause the first order condition for Pakistan to become:
[
]
dPLUU e
H
P
D
Pe 2
)()( =⋅−⋅
π
(9'')
FIGURE 4
Reaction Functions of India and Pakistan
S
e
a
RI2
RI1
RP1 RP2
C
N
16
This causes Pakistan’s reaction function to shift outwards along India’s,
with a new equilibrium at point S. Note, however, in the new equilibrium
(point S) India has effectively passed on some of the burden of adjustment to
Pakistan. In fact, the level of effort exercised by Pakistan is greater than even
in the more cooperative solution (C), but not India’s. This could be argued to
be the case at present. As India moves closer to the United States, and with the
latter’s global war on terror more pressure is exerted on Pakistan to make
unilateral concessions towards India since 2001. We could even argue that
India is free riding on Pakistan.
3 EMPIRICAL ANALYSIS
3.1 Hypotheses
H1: Greater bilateral inter-state commerce, as well as greater multilateral trade
with third countries lowers various forms of bilateral inter-state conflict. This
corresponds to the liberal peace. This hypothesis follows from our theoretical
discussion, specifically the first order conditions in (8’) and (9’), and in
inversely from the right-hand side of (5).
H2: More military spending as a result of increased insecurity raises conflict.
The hegemonic power, however, may have internal conflict (India has many
civil wars) and other neighbours to militarily confront. The marginal utility of
security spending rises in (5), as well as in (8) and (9).
H3: Development expenditure (such as public spending on education) should
lower conflict, because of economic growth which enables more consumption
in equations (4) to (9). This is also related to the increased democratisation
hypothesis, below.
H4: GDP growth will decrease inter-state conflict; there is more to lose from
war. This raises the utility from consumption in (4), (6) through to (9).
H5: Increases in dyadic democracy scores will lead to less conflict, related to
the notion of the democratic peace. Increased democracy may lower the cost
of concessions and compromise with former enemies, as in (9’’) above.
3.2 Data and Methodology
3.2.1 Data
Since inter-state conflict involves at least two parties, it is a dyadic concept. We
construct dyadic proxies for India-Pakistan inter-state trade, military burden,
development expenditure, economic development and democracy to test the
five hypotheses we have presented above. Data definitions are given in the
appendix.
3.2.1.1 Measuring conflict
The literature on inter-state conflict classifies conflict data sets into two
categories: 1) war data and 2) events data (Polachek and Seiglie, 2006). War
data sets focus on more hostile aspects of inter-state interactions such as crises,
wars or militarized inter-state disputes (Jones, Bremer and Singer, 1996). The
17
most comprehensive wars data set is available under the Correlates of War
Project (COW) which has updated war data sets employed by Wright (1942),
Richardson (1960), and Singer and Small (1972). The other major data set on
inter-state armed conflict is hosted by the Uppsala Conflict Data Project
(UCDP) with the collaboration of the International Peace Research Institute,
Oslo (PRIO) and is collected on an annual basis and covers the full post-World
War II period, 1946–2003. Events data focuses on all inter-state events and
bilateral interactions reported in newspapers. McClelland’s (1978) World
Events Interaction Survey (WIES) is probably the first of its kind based on
bilateral interactions, occurring between 1966-1992, reported in New York
Times. Azar’s (1980) Conflict and Peace Data Bank (COPDAB) is an extensive
longitudinal collection of about one million daily events reported from forty
seven newspaper sources between 1948 and1978. Since we are interested in the
evolution of India-Pakistan conflict over a period of the last 55 years, we will
use Uppsala/PRIO and COW inter-state war data set instead of events based
data sets because the former data sets provide conflict data which covers most
of the period of 55 years (1950-2005) which we have selected for our analysis.
Events data set is not available for the entire period. Though the events data
set captures daily observations, our macroeconomic and democracy data varies
annually which limits the use of daily information on conflict. Secondly, as
shown in figure 1, hostility between India and Pakistan has usually been high in
most of last 55 years, enabling the COW data set to capture the severity of
conflict in most years of the dispute. Greater coverage by the COW and
Uppsala data sets, and the availability of macroeconomic and democracy data
on an annual basis also limits the scope of using the events data sets.
3.2.1.2 Measuring international trade
Generally dyadic trade is captured by sum of imports and exports between
actor and target countries (Polachek and Seglie, 2006). Figure 2 shows that in
the last 55 years the patterns of inter-state trade between Pakistan and India
have changed. Before trade between both countries collapsed to near zero in
early 1970s, Pakistan was exporting more to India. Since the 1970s, Pakistan is
importing more from its neighbour. In the 1950s, Pakistan and India’s trade
with each other constituted a significant amount of their respective total trade.
However, after the 1965 war, India-Pakistan trade never reached more than 2
percent of their respective total trade levels. Till the late 1980s, India had been
a relatively closed economy, whereas Pakistan has traditionally been more
open. We construct two composite measures of India-Pakistan trade. They are
Pakistan’s total trade with India as percentage of Pakistan’s total trade (Tpitp),
and also India’s trade with Pakistan as a percentage of India’s total trade
(Tpiti). We expect both trade proxies to be negatively related with conflict. It
would be interesting to investigate whether trade between both countries as
share of each countries total trade also affects the responsiveness of bilateral
trade in conflict mitigation. If trade reduces conflict, trade with more countries
should reduce conflict even more (Dorussen, 1999). Thus, it is important to
investigate how more trade with the rest of the world affects India-Pakistan
hostilities. We construct a total of 8 dyadic proxies to capture the combined
international integration levels for both countries. Pakistan’s total trade as a
ratio of India’s total trade (Xmpi), and its inverse, India’s total trade as a ratio
18
of Pakistan’s total trade (Xmip) are the first two indicators. If both of these
trade proxies are negatively related with hostilities, we can conclude that any
external trade competition does not increase bilateral rivalry between India and
Pakistan, but instead both countries have similar trade policies or could
integrate within regional bodies like SAARC (the South Asian Association for
Regional Cooperation). However, any evidence of a positive relationship
between conflict and these two trade proxies would suggest that the
competition in international markets has significant implications in sustaining
their rivalry.
3.2.1.3 Measuring military expenditure
Military expenditures can reflect hostility, as well as deterrence (Polachek and
Seglie, 2006). In the India-Pakistan case, we would like to examine how each
county’s military expenditure/ military burden affects the dispute. Pakistan’s
spending on military expenditure as a proportion of GDP is higher than
India’s. Additionally, since military expenditures may also capture the capability
of a country to deal with civil unrest or intra-state conflict, Indian military
expenditure can also be explained in terms of the high prevalence of
continuing intra-state conflicts in various regions of India. Pakistan has had
fewer civil wars. This may mean that Pakistan’s military burden captures its
security concerns vis-à-vis India solely. If so, dyadic variables which take the
military burden of Pakistan as a ratio of the Indian military burden, should
affect conflict positively and vice versa. We construct 8 different dyadic proxies
of military burden utilizing data on military expenditures as well as military
personnel from Correlates of Wars: 1. The log of Pakistan’s defence
expenditure over GDP as a ratio of India’s defence expenditure over GDP
(Lmilbrd 1) 2, Log of India’s defence expenditure over GDP as a ratio of
Pakistan’s defence expenditure over GDP (Lmilbrd 2), 3. Log of Pakistan’s
defence expenditure over GDP as a ratio of Pakistan’s defence expenditure
over GDP plus India’s defence expenditure over GDP (Lmilbrd 3), 4. Log of
India’s defence expenditure over GDP as a ratio of Pakistan’s defence
expenditure over GDP plus India’s defence expenditure over GDP (Lmilbrd4).
Note that first two proxies are the inverse of each other and are expected to
reveal the relative sensitivity of each countries’ military expenditure to conflict.
The last two proxies are a robustness check where military expenditure of each
country is divided by the combined military expenditure score of both
countries. If Lmilbrd3 is positively associated with conflict, we can substantiate
our hypothesis for Lmilbrd1. If Pakistan’s military expenditure is more closely
associated with their bilateral conflict, and if Indian military expenditure
captures element of deterrence, as well as belligerence with other national and
international rivals, then the combined military expenditures should have lower
explanatory value than Pakistan’s military expenditure alone but the sign for
combined military score should remain positive. We investigate the average
effects of military expenditures by both countries on India-Pakistan rivalry by
taking two more proxies of military burden. This is to investigate whether
military burden has on average a conflict enhancing effect, irrespective of
country of origin, after analyzing its country specific application for deterrence
or belligerence. Thus we propose two further proxies: 5. the log of average of
India’s defense expenditure over GDP and Pakistan’s defense expenditure over
19
Paki stan and Indi a Combined Democrac y Score
0
50
100
150
200
250
300
350
400
1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
2001
Years
Demopi
GDP (Lmilbrd5), 6. Log of Pakistan and India’s GDP weighted average of
defense expenditures (Lmilbrd6). The proportion of military personnel to the
total population represents the extent of militarization in a society. Thus we
choose two further variable specifications for military burden: 7. Log of
Pakistan military personnel over Pakistan’s total population as a ratio of India’s
military personnel over India’s total population (LMilppi) and 8 Log of India’s
military personnel over India’s total population as a ratio of Pakistan’s military
personnel over Pakistan’s total population (LMilppi).
3.2.1.4 Measuring democracy, growth and other variables
To capture democracy levels for India and Pakistan, we turn to the Polity IV
project hosted by Center of International Development and Conflict
Management (CIDCM). Polity IV computes a combined polity score by
subtracting autocracy scores from the democracy scores for the corresponding
year. The value of this Polity score ranges from -10 to 10, where -10 denotes
the highest autocracy level, and 10 the maximum democracy score. Although
India always takes a high positive value of 7 or above, Pakistan frequently takes
on negative values. We construct a dyadic variable of democracy for both
countries by combining multiplying their Polity scores, following Polachek and
Seiglie (1969). We add 10 to each countries polity series to make the negative
Polity values positive so that our combined democracy score may capture the
variations in the democratization process only on a positive scale. The dyadic
democracy variable shows values as low as 50 on the scale of 0 to 400 when
there are high levels of political dissimilarities between Pakistan (dictatorship)
and India (democracy), and as high as 350 when both countries are governed
by democracies (see figure 5).
FIGURE 5
Dyadic democracy scores for Pakistan and India
20
We take the weighted average of India and Pakistan’s real GDP per capita
growth rates (Gpi) as the dyadic proxy of economic progress for both
countries. We constructed the series for both countries by dividing GDP at
constant prices taken from economic surveys, and dividing it by population
levels. The data was later tallied with GDP per capita series available at the
World Development Indicators (2006) data set. We also constructed 4 different
proxies of social development based on India and Pakistan’s education data7: 1.
GDP weighted average of per capita education expenditure, 2. Mean average
of per capita education expenditure, 3. Pakistan and India’s education
expenditures as a ratio of Pakistan and Indian’s GDP and, 4. The average of
Pakistan’s education expenditure as a percentage of its GDP and India’s
education expenditure as a percentage of its GDP. Note that the first two
proxies employ per capita education expenditure and the last two proxies
employ total education expenditure. The four education proxies are
constructed to carry out a robustness check on the role of education in conflict
mitigation. India and Pakistan are one of the most densely populated countries
in the world. Pakistan has 160 million inhabitants, and India has over a billion
citizens. In line with the earlier literature, we also take the mean average of
both countries populations as a standardising variable in our analysis (see
Polachek, 1997).
3.2.2 Methodology
Any simple least square regression analysis may lead to spurious results due to
the endogeniety problems among our variables (from trade, military spending,
social sector expenditure and growth to conflict and vice-versa). We need to
utilize a simultaneous equation model where potential endogenieties between
various variables are addressed. Since our data is time series, we will use Vector
Autoregressive model (VAR) which is an extension of univariate autoregressive
(AR) models to capture the evolution and the interdependencies between our
multiple time series (Sims, 1980). All variables in a VAR are treated
symmetrically by including for each variable an equation explaining its
evolution based on its own lags and the lags of other variables in the model.
The number of equations in a VAR model depends upon the number of
endogenous variables; each endogenous variable is regressed on its lagged
value, and the lagged values of all other endogenous variables as well as any
number of exogenous variables. This solves the problem of endogeniety
among variables. In this sense VAR model is just a seemingly unrelated
regression (SUR) model with lagged variables and/or deterministic terms as
common regressors so that the regression results for each equation can be
interpreted in the same manner as we do for ordinary least square estimators.
The basic −
p
lag vector autoregressive (VAR )( p) model has the form
tptpttt yyycY
ε
+
Π
+
+
Π+
Π
+= −−− ......
2211 (12)
7 There is an insufficiently long time series for public health spending data for India.
21
where cis a )1( ×nvector of constants (intercept), i
Π
is a )( nn
×
matrix (for
every pi ,....,1=) and t
ε
is a )1(
×
nvector of error terms.
A bivariate VAR(2) can be written as the following system of equations:
tttttt yyyycy 12,2
22,12,1
21,11,2
12,11,1
11,111
ε
+Π+Π+Π+Π+= −−−− (13)
tttttt yyyycy 22,2
22,22,1
21,21,2
12,21,1
11,222
ε
+Π+Π+Π+Π+= −−−− (14)
The lag length
p
has to be determined by model selection criterion (MSC)
because too many lagged terms will consume more degrees of freedom and
may introduce the problem of multicollinearity. Introducing too few lags will
lead to specification errors. One way of deciding this question is to use Akaike
(AIC), Schwarz-Bayesian (BIC) or Hannan Quinn (HQ) criteria and choose
that model which gives the lowest values of these criteria. AIC criterion
asymptotically overestimates the order with positive probability, whereas BIC
and HQ criterion estimate the order consistently under general conditions if
the true order
p
is less than or equal to max
p
.
After fitting a VAR we may want to know which way causalities run. One
way to do that is by running Granger causality tests after the VAR analysis. In a
bivariate VAR model, a variable 2
y is said to Granger-cause a variable 1
y if,
given the past values of 1
y, past values of 2
yare useful for predicting 1
y
(Granger, 1969). Similarly, we can extend our analysis to test Granger-causality
for multivariate VAR (
p
), where ),......,,( 21
′
=
ntttt yyyY .
3.2.3 Results with VAR models
This section reports the results of the multivariate VAR regression analysis.
Proxies for conflict, bilateral and multilateral trade, economic progress, military
burden and social development will be treated as endogenous variables,
whereas dyadic democracy and population will be treated as purely exogenous
concepts. Before we carry out the regression analysis, a test for stationarity is in
order for all dyadic variables employed in our analysis. If any of the time series
variables are non-stationary, appropriate lags are taken to solve for
autocorrelation. Stationarity tests are carried out by running the modified
Dicky-Fuller t-test also known as the DF-GLS test proposed by Elliot,
Rothenberg and Stock (1996). Table 2 provides unit root test results based on
these criteria.
Table 2 shows that nearly all variables have unit roots. Since our time
series variables are stationary at levels, though with some time lags, we can use
unrestricted VAR analysis instead of restricted VECM methodology. We can
now proceed to VAR analysis. Our reduced form VAR model for conflict is as
follows
ititititititt MilTrConfConf −−−−−−
+
+
+
=,4,3,21
α
α
α
α
tttitititit PDemoGE
Ε
+
+
+
+
+
−−−− 87,6,5
α
α
α
α
(15)
22
where t
Conf , it
Tr−, it
Mil −,it
E−,it
G−,t
Demo and t
Pdepict inter-state conflict,
bilateral or multilateral trade, military burden, education expenditure, real
growth rate of GDP per capita, dyadic democracy score and population
respectively; t ranges from 1950-2005 and pi ,....,1
=
. Here
p
is the optimal lag
structure for the VAR model. it−,2
α
it−,3
α
it−,4
α
it−,5
α
and it−,6
α
are
)66( ×metrics (for every pi ,....,1=).
The model above is run for the number of fatalities, Fatal because it best
captures the severity of the militarized conflict between the two nations. Later,
we also employ other conflict proxies in our analysis.
TABLE 2
DF-GLS Unit Root Tests
Variables Lag length With intercept With intercept and trend
Fatal 1 -3.528* (Ng-Perron) -3.774* (Ng-Perron)
Volfatal 1 -4.789* (Ng-Perron)
-4.844* (Ng-Perron)
Dur 1 -4.058* (Ng-Perron)
-4.233* (Ng-Perron)
Hiact 1 -2.382** (Ng-Perron)
-2.590 (Ng-Perron)
Hstlev 1 -2.371** (Ng-Perron)
-2.512 (Ng-Perron)
Cnf 1 -3.025* (Ng-Perron)
-4.082* (Ng-Perron)
Tpitp 15 -1.112*** (Ng-Perron)
-1.861 (Ng-Perron)
Tpiti 15 -3.856* (MAIC) -3.319** (Ng-Perron)
Xmpi 2 -2.710* (Ng-Perron)
-2.860*** (Ng-Perron)
Xmip 8 -4.951* (MAIC) -4.923* (MAIC)
Lxpi1 0 2.951** (D-Fuller) 2.951** (D-Fuller)
Lxpi2 0 -4.769* (SIC) -4.929* (SIC)
Lmpi1 1 -4.049* (SIC) -3.961* (SIC)
Lmpi2 1 -4.511* (SIC) -4.382* (SIC)
Lmilbrd1 5 -2.209** (Ng-Perron)
-2.795*** (Ng-Perron)
Lmilbrd2 5 -2.209**(Ng-Perron)
-2.795***(Ng-Perron)
Lmilbrd3 5 -1.911***(Ng-Perron)
-2.686***(Ng-Perron)
Lmilbrd4 5 -2.128***(Ng-Perron)
-2.831***(Ng-Perron)
Lmilbrd5 1 -4.735* (SIC) -4.748* (SIC)
Lmilbrd6 0 - -4.308* (SIC)
Lmilppi 1 -4.082* (SIC) -4.098* (SIC)
Lmilpip 1 -4.082* (SIC) -4.098* (SIC)
Ledupi1 1 - -5.374* (SIC)
Ledupi2 1 - -5.478* (SIC)
Ledupi3 1 -5.918* (SIC) -5.907* (SIC)
Ledupi4 1 - -5.642* (SIC)
Gpi 0 -4.256* (Ng-Perron)
-4.276* (Ng-Perron)
Demopi 7 -2.790* (Ng-Perron)
-2.997* (Ng-Perron)
Poppi 10 - -7.392* (MAIC)
-*, ** and *** shows significance at 1%, 5%and 10% level
- The Lag structure is selected through (1) Ng - Perron sequential t (Ng-Perron), (2) the minimum
Schwarz information criterion (SIC), (3) the Ng-Perron modified information criterion (MAIC) and (4)
Dickey-Fuller test (D-Fuller).
23
TABLE 3a
VAR Regression Equations for Fatal under multiple specifications of BiLateral Trade and Military Burden
Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Bilateral Trade
Tpitp (16) -0.30* -0.30* -0.32* -0..28* -0.24** -0.23** -0.22**
Tpiti (16) -0.76*** -0.76*** -0.83** -0.70*** -0.61*** -0.64*** 0.55***
Military Burdeñ̃̃̃̃̃
lmilbrd1 (6) 2.33*** 2.02
lmilbrd2 (6) -2.33*** -2.02
lmilbrd3 (6) 6.53*** 6.03
lmilbrd4 (6) -3.45 -2.84
lmilbrd5 (2) 6.84** 6.54**
lmilbrd6 (1) 3.26*** 3.52***
Lmilppi(2) -1.80
Lmilpip(2) 1.79
Social Development
Ledupi1(2) -4.98 -4.98 -4.83 -5.09*** -6.35** -8.34* -6.08** -6.07*** -6.07*** -6.19*** -6.02*** -5.97** -8.35* -6.10**
Economic Growth
Gpi (1) -0.40* -0.40* -0.41* -0.40* -0.28* -0.35* -0.34* -0.39* -0.39* -0.39* -0.39* -0.31* -0.38* -0.37*
Exogenous Variables
Demopi (7) -0.003 -0.003 -0.003 -0.003 -0.003 -0.004*** -0.004*** -0.003 -0.003 -0.003 -0.004 -0.003 -0.003*** -0.004***
Poppi (10) 0.064* 0.064* 0.063* 0.066* 0.112* 0.094* 0.076* 0.063* 0.063* 0.062* 0.064* 0.101* 0.088* 0.072*
N 38 38 38 38 38 38 38 38 38 38 38 38 38 38
R2 0.61 0.61 0.62 0.61 0.63 0.61 0.59 0.57 0.57 0.58 0.57 0.61 0.59 0.57
VAR(p) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2) VAR(2)
-*, **, *** shows significance at 1%, 5% and 10% level
- VAR(p) reports lag-order for each VAR model based on final prediction error (FPE), Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion (SBIC) and the Hannan
and Quinn information criterion (HQIC),
24
TABLE 3b
VAR Regression Equations for Fatal under multiple specifications of Multilateral Trade and Military Burden
Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Mutilateral Trade
Xmpi(3) -0.71 -0.71 -0.75 -0.74 -0.62 -0.77*** -0.75***
Xmip(9) -3.74* -3.74* -3.77* -3.74* -3.89* -2.68* -3.83*
Military Burdeñ̃̃̃̃̃
lmilbrd1 (6) 0.08 -0.18
lmilbrd2 (6) -0.08 0.18
lmilbrd3 (6) 0.91 0.27
lmilbrd4 (6) -0.58 0.50
lmilbrd5 (2) 0.04 -0.49
lmilbrd6 (1) 3.38** 2.26***
Lmilppi(2) -1.02
Lmilpip(2) 0.92
Social Development
Ledupi1(2) -3.64* -3.64* -3.59* -3.69* -3.60* -8.07* -2.85* -4.73* -4.73* -4.67* -4.79* -4.44** -7.70* -4.22*
Economic Growth
Gpi (1) -0.37* -0.37* -0.37* -0.38* -0.37* -0.34* -0.37* -0.40* -0.40* -0.39* -0.40* -0.40* -0.36* -0.39*
Exogenous Variables
Demopi (7) -0.006* -0.006* -0.006* -0.006* -0.006* -0.006* -0.005* -0.006* -0.006* -0.006* -0.005* -0.006* -0.006* -0.005*
Poppi (10) 0.067* 0.067* 0.066* 0.067* 0.066* 0.094* 0.062* 0.083* 0.083* 0.082* 0.084* 0.078* 0.101* 0.075*
N 45 45 45 45 45 45 45 45 45 45 45 45 45 45
R2 0.42 0.42 0.42 0.42 0.42 0.46 0.42 0.45 0.45 0.45 0.45 0.45 0.47 0.46
VAR(p) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1)
-*, **, *** shows significance at 1%, 5% and 10% level
- VAR(p) reports lag-order for each VAR model based on final prediction error (FPE), Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion (SBIC) and the Hannan
and Quinn information criterion (HQIC),
- VAR(p) reports lag-order for each VAR model based on final prediction error (FPE), Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion (SBIC) and the Hannan
and Quinn information criterion (HQIC),
25
TABLE 3c
VAR Regression Equations for Fatal under multiple specifications of Exports, Imports and Military Burden
Variables 1 2 3 4 5 6 7 8 9 10 11 12
Mutilateral Trade
Lxpi1(1) -4.24* -4.03* -3.96*
Lxpi2(1) -7.89* -7.15* -4.78*
Lmpi1(2) -0.36 -0.17 0.03
Lmpi2(2) -0.71 -0.59 -0.33
Military Burdeñ̃̃̃̃̃ª
Lmilbrd3 (6) 2.19 5.84** 0.39 0.30
Lmilbrd4(6) -0.66 -2.34*** 0.44 0.37
lmilbrd6 (1) 3.51* 2.42*** 3.19** 3.09**
Social Development
Ledupi1(2) -1.96 -2.08*** -7.13* -2.87* -2.89* -7.02* -3.97* -4.19* -8.66* -4.01* -4.13* -8.43*
Economic Growth
Gpi (1) -0.36* -0.36* -0.35* -0.39* -0.39* -0.39* -0.34* -0.36* -0.33* -0.34* -0.35* -0.33*
Exogenous Variables
Demopi (7) -0.004*** -0.004*** -0.003*** -0.002 -0.002 -0.002 -0.006* -0.006* -0.006* -0.006* -0.005* -0.005*
Poppi (10) 0.122* 0.120* 0.154* 0.077* 0.075* 0.103* 0.077* 0.078* 0.104* 0.074* 0.075* 0.103*
N 45 45 45 45 45 45 45 45 45 45 45 45
R2 0.50 0.49 0.55 0.58 0.55 0.55 0.40 0.40 0.45 0.40 0.40 0.44
VAR(p) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1)
-*, **, *** shows significance at 1%, 5% and 10% level
- VAR(p) reports lag-order for each VAR model based on final prediction error (FPE), Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion (SBIC) and the Hannan
and Quinn information criterion (HQIC),
- ª Results for Lmilbrd1, Lmilbrd2, Lmilbrd5, Lmilppi and Lmilpip are also utilised and the results do not change. (See tables 3a and 3b for details)
26
TABLE 3d
VAR Regression Equations for Fatal under multiple specifications of Education and Multilateral Trade
Variables 1 2 3 4 5 6 7 8 9 10 11 12
Social Development
Ledupi1(2) -7.70* -7.13* -7.01*
Ledupi2(2) -8.17* -7.52* -7.44*
Ledupi3(2) -4.06*** -6.29* -5.79*
Ledupi4(2) -7.96* -8.93* -8.91*
Multilateral Trade
Xmip(8) -2.68 -2.68 -3.92*** -3.14
Lxpi1(1) -3.96* -3.92* -5.46* -4.74*
Lxpi2(1) -4.78* -4.75* -6.35* -5.76*
Military Burdeñ̃̃̃̃̃ª
Lmilbrd6 (1) 2.26 3.50** 2.42*** 2.44 3.62* 2.58*** -0.96 2.02*** 0.45 0.51 2.73** 1.52
Economic Growth
Gpi (1) -0.36* -0.35* -0.39* -0.37* -0.36* -0.39* -0.42* -0.41* -0.45* -0.39* -0.38* -0.42*
Exogenous Variables
Demopi (7) -0.006* -0.003*** -0.003 -0.006* -0.004* -0.003 -0.005* -0.001 -0.001 -0.006* -0.003*** -0.002
Poppi (10) 0.101* 0.154* 0.103 0.107* 0.158* 0.109* 0.031* 0.107* 0.038* 0.021 0.087* 0.028**
N 45 45 45 45 45 45 45 45 45 45 45 45
R2 0.47 0.54 0.55 0.47 0.55 0.55 0.39 0.53 0.53 0.44 0.55 0.56
VAR(p) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1)
-*, **, *** shows significance at 1%, 5% and 10% level
- VAR(p) reports lag-order for each VAR model based on final prediction error (FPE), Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion (SBIC) and the Hannan
and Quinn information criterion (HQIC),
- ª Results for Lmilbrd1, Lmilbrd2, Lmilbrd3, Lmilbrd4, Lmilbrd5, Lmilppi and Lmilpip are also utilised and the results do not change. (See tables 3a and 3b for details)
27
TABLE 4
VAR Regression results for Various Measures of Conflict
VAR Regression Equations under multiple Specifications for Conflict and Military Burden
Volfatal Cnfpi Dur Hstlvl Hiact
Variables
(1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3)
Mutilateral Trade
Lxpi2 (1) -10996* -9971* -6662* -2.60* -2.48* -1.22*** -451.46** -413.04** -182.81 -6.81* -6.60* -4.07** -25.75** -25.32* -16.08***
Military Burdeñ̃̃̃̃̃ª
Lmilbrd3 (6) 8276* 2.91* 604.72* 5.33*** 19.09
Lmilbrd4(6) -3352** -1.46* -
283.85** -2.68*** -9.98
Lmilbrd6 (10) 3255*** 0.31 -55.94 0.97 4.47
Social Development
Ledupi (2) -397.02 -435.58 -
6011.6** -0.74*** -0.69 -1.48 -146.53 -130.7 -180.69 -1.56 -1.47 -3.34 -9.09*** -8.75*** -17.08
Gpi (1) -517.07* -524.78* -554.46* -0.86** -
0.084*** -0.09** 4.89 4.97 3.63 -0.25*** -0.25*** -0.26*** -1.28** -1.26** -1.38***
Exogenous Variables
Demopi (8) 1.36 1.06 0.06 -0.001*** -0.001*** -0.002*** -0.336*** -0.342*** -0.372*** -0.001 -0.001 -0.001 -0.011 -0.012 -0.012
Poppi (11) 36.38*** 34.66*** 71.54* 0.023* 0.021* 0.027* 3.531*** 3.209*** 4.248*** 0.051** 0.048** 0.058** 0.253* 0.247* 0.295*
N 45 45 45 45 45 45 45 45 45 45 45 45 45 45 45
R2 0.45 0.44 0.42 0.53 0.51 0.42 0.40 0.37 0.31 0.42 0.42 0.38 0.39 0.40 0.37
VAR(p) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1) VAR(1)
-*, **, *** shows significance at 1%, 5% and 10% level
- VAR(p) reports lag-order for each VAR model based on final prediction error (FPE), Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion (SBIC) and the Hannan
and Quinn information criterion (HQIC),
- ª Results for Lmilbrd1, Lmilbrd2, Lmilbrd5, Lmilppi and Lmilpip are also utilised and the results do not change. (See tables 3a and 3b for details)
28
Table 3a shows the results for bilateral trade with 8 proxies of the military
burden we propose in section B.1. The evidence suggests that trade between
Pakistan and India significantly decrease hostilities between both nations.
However, the low values of it −,3
α
coefficients suggest that bilateral trade has a
limited role to play in conflict mitigation. This is not surprising because we
know from figure 2, that trade between Pakistan and India has remained very
low, and comprises only a very small fraction of each country’s total
international trade. Though low trade levels between both countries may very
well be the cause of the ongoing conflict, here we do not need to worry about
reverse causality because our VAR model takes care of potential endogeneity
problems between Fatal and Tpitp or Tpiti. On the other hand, Lmilbrd1,
Lmilbrd2, Lmilbrd3, Lmilbrd4, Lmilbrd5 and Lmilbrd6 all are significantly related
with conflict especially in case of Tpitp. Lmilbrd1 and Lmilbrd3 are negatively
related with conflict, and Lmilbrd2 and Lmilbrd4 are positively related with
conflict. This confirms our hypothesis that Pakistan’s high military expenditure
is a close determinant of the India-Pakistan conflict. The high values of the
it−,4
α
coefficients in this case indicate that any increase in military expenditure
by Pakistan when compared to India will be correlated with higher conflict.
However negative signs of Lmilbrd2 and Lmilbrd4 also suggest that India’s
military expenditure is weakly related with conflict whereas as argued Indian
military expenditure is also directed to its domestic civil wars and security
concerns with other states and thus in case of Lmilbrd1, Lmilbrd2, Lmilbrd3 and
Lmilbrd4 the explanatory power comes from Pakistan’s military expenditure.
Furthermore, combined military scores in Lmilbrd5 and Lmilbrd5 are
positively related with conflict and the relationship is significant for both
proxies of bilateral trade. This result suggests that irrespective of Indian
security concerns national or international, or Pakistan’s anxieties about Indian
hegemony, military expenditures on average do not have deterrent effect (in
terms of fewer fatalities), but high military expenditures by both sides show
some evidence of an arms race. The insignificance of Lmilppi and Lmilpip may
also indicate the transformation of contemporary conventional war tactics, in
which military size per se has a limited role in providing strategic depth.
However, the negative sign of Lmilppi and the positive sign of Lmilpip hints
that higher militarization in Pakistan may very well be an outcome of the
ongoing hostilities between two nations, as higher Pakistani military personnel
has a deterrent effect, and the converse is true for India. Education
expenditures Ledupi1 and growth rates Gpi are significantly related to conflict
mitigation, and the size of coefficients suggests that the potential of spending
on education in decreasing hostilities is quite substantial. Democracy also
decreases the severity of conflict, but the low values of coefficients show the
relationship is quite weak.
Table 3b and 3c show the results for multilateral trade with various
proxies of military burden. In combination with various proxies of multilateral
trade the explanatory power of Lmilbrd1, Lmilbrd2, Lmilbrd3 and Lmilbrd4 have
reduced as they are generally insignificant but the coefficients have also been
reduced especially for Xmpi and Xmip. The only military burden proxy which is
consistently significant and also comes out with the right sign is Lmilbrd6. This
means that our conclusion about the average conflict enhancing role of military
29
expenditures has not been altered. Results in table 3b show that Xmpi is
generally insignificant, whereas Xmip is significant in nearly all specifications.
This is an interesting result, which suggests that higher Indian levels of trade
integration mitigate conflict more than when Pakistani openness rises.
However, the negative signs for both proxies confirm that greater openness in
either country would significantly decrease conflict. Furthermore, we can also
conclude that there is no rivalry between India and Pakistan in terms of their
trade with rest of the world, and any competition to capture international
market share is healthy. Table 3c shows results for average trade scores for
both countries differentiated by exports and imports. Exports by both
countries to rest of the world are negatively related with conflict and the
relationship is significant at the 1% level. Also, note that the values
of it −,3
α
have increased further for combined exports when compared with the
results in table 3b, indicating that the more these two countries are able to
export to rest of the world, the lower are the levels of bilateral conflict. The
high coefficients of Xmpi can lead us to infer that the explanatory power for
Xpi comes more from the Indian side. Both countries are at similar rungs on
the technological ladder, and share the potential to export to the rest of the
world, along with the likes of China. In contrast to exports, results on Lmpi1
and Lmpi2 show that rising imports although do not increase hostilities as the
signs are always negative but the overall insignificance of combined import
scores mean imports may not exert any negative pressure on hostilities either.
The results for education expenditure, economic performance and democracy
remain unchanged.
We investigate the potential of development expenditure in conflict
mitigation in detail by employing 4 proxies of education expenditure, with 3
different combinations of multilateral trade, while using Lmilbrd6 as a common
proxy for military burden. The results are presented in table 3d. In contrast to
defence expenditure, which is positively related to conflict, efforts to improve
human capital by allocating more funds to education are a strong determinant
of conflict mitigation as the results in table 3d demonstrate. All four proxies of
education expenditure always enter the conflict regression equation with a
negative sign, and are significant in all specifications. The high values of
it −,5
α
indicate that channeling resources to development sector in general, and
investment in education in particular, may go a long way in building peace. The
weighted average of Pakistan and Indian per-capita growth rates (Gpi) are
negatively and significantly related with Fatal in all specifications confirming
the hypothesis that countries are more peaceful when they are moving forward
economically. The combined democracy score (Demopi) is always negatively
related to conflict, and also significant. However, the low values of democracy
coefficients suggest that political orientation has played a more limited role in
the India-Pakistan conflict. Our results also show that the high levels of
population in both countries, where a significant proportion are uneducated
and poor on both sides, contribute positively to the conflict, although the
effect is small. The results on Xmip, Lxpi1 and Lxpi2 confirm yet again that
India and Pakistan should open up further, as conflict mitigation is highly
responsive to multilateral trade. In other words, we can conclude that a lower
military burden would mean both countries can invest more on education,
30
higher multilateral trade combined with increased education level will seriously
contribute to peace between Pakistan and India on a sustainable basis. Though
democracy is also good for peace, economics clearly trumps democracy as a
conflict mitigating factor.
Further robustness checks, under additional specifications of the conflict
dependent variable, are carried out on (15) with different proxies of conflict
(i.e, Volfatal, Cnfpi, Dur, Hstlvl and Hiact). Each definition of conflict is
regressed on Lmilbrd3, Lmilbrd4 and Lmilbrd6, whereas Lxpi2, Ledupi1, Gpi,
Demopi and Poppi make up the common set of regressors in a total of 15
specifications. The results are given in Table 4. They confirm the validity of all
the 5 hypotheses proposed at start of our empirical section, and our theoretical
model. More trade, increased education expenditure, higher GDP per capita
growth rates, a greater democratic orientation, all exert downward pressure on
conflict, as all of these variables are significant in most cases, and always carry
the right signs. A comparison of coefficients suggests that integration with the
world has by far the most dominant effect on conflict mitigation than any
other variable. Note that in table 4 we only consider multilateral trade, and not
bilateral India-Pakistan trade. Education spending comes second in its
effectiveness in enhancing peace. The results in Table 4 also show that annual
battle deaths, severity of conflict, duration of escalation, hostility levels and
highest hostility level decrease when both countries score high on democracy.
Again, persistently low values taken by democracy t,7
α
means that political
orientation plays a less prominent role in explaining the severity of dispute or
levels of escalation. There is some evidence that these countries have entered
into outright war even when both were democracies. The ‘Kargil’ war of 1999
is a case in point. By contrast, the long military stand off between India and
Pakistan in 2002, occurred at a time when Pakistan was highly autocratic. The
results in the table 4 also indicate that Lmilbrd3 always enters (15) with a
positive sign and is significant in 4 out of 5 cases. The coefficients of LMilbrd3,
Lmilbrd4 and Lmilbrd6 are highest for Volfatal and Dur suggesting that military
expenditures significantly increase the severity of conflict as more days of the
year are spent in hostility and higher fatalities are borne by both sides.
It would be interesting to run multivariate Granger causality tests to see if
causality runs from the determinants of conflict to conflict, and whether there
are also cases of reverse causality. We ran Granger causality test for each VAR
specification for which we present results in tables 3a, 3b, 3c, 3d and 4. A
summary of Granger causality tests are provided in table 5 for all endogenous
regressors of conflict, and where there is an instance of reverse causality it is
noted. The results in table 5 show that all regressors except Lmilppi, Lmilpip,
Lmpi1 and Lmpi2 Granger cause conflict. We also witness some instances of
reverse causality especially for Tpitp, Tpiti, Lmilbrd5, Lmilbrd6, Ledpi1, Ledupi2
and Ledupi4 in case of Fatal, Lmilbrd6 in case of Volfatal, lmilbrd6 and Ledupi1 in
case of Cnfpi, Lmilbrd6 and Ledupi1 in case of Dur, Lxpi2, Lmilbrd6 and Ledupi1
in case of Hstlvl and Lxpi2, Lmilbrd6 and Ledupi1 in case of Hiact.
31
TABLE 5
Granger Causality Wald Tests
Direction of Causality Causes RC Direction of Causality Causes RC
FatalTpitp → (√)* (√)*** VolfatalGpi → (√)* ×
FatalTpiti → (√)*** (√)** CnfpiLxpi →2 (√)* ×
FatalXmpi → (√)** × CnfpiLmilbrd →3 (√)* ×
FatalXmip → (√)* × CnfpiLmilbrd →4 (√)* ×
FatalLxpi →1 (√)* × CnfpiLmilbrd →6 × (√)***
FatalLxpi →2 (√)* × CnfpiLedupi →1 (√)*** (√)***
FatalLmpi →1 × × CnfpiGpi → (√)* ×
FatalLmpi →2 × × DurLxpi →2 (√)* ×
FatalLmilbrd →1 (√)** × DurLmilbrd →3 (√)* ×
FatalLmilbrd →2 (√)** × DurLmilbrd →4 (√)** ×
FatalLmilbrd →3 (√)* × DurLmilbrd →6 × (√)***
FatalLmilbrd →4 (√)* × DurLedupi →1 (√)* (√)*
FatalLmilbrd →5 (√)* (√)** DurGpi → (√)* ×
FatalLmilbrd →6 (√)* (√)* HstlvlLxpi →2 (√)* (√)*
FatalLmilpip → × × HstlvlLmilbrd →3 (√)*** ×
FatalLmilppi → × × HstlvlLmilbrd →4 (√)*** ×
FatalLedupi →1 (√)* (√)* HstlvlLmilbrd →6 × (√)*
FatalLedupi →2 (√)* (√)* HstlvlLedupi →1 × (√)*
FatalLedupi →3 (√)* × HstlvlGpi → (√)*** ×
FatalLedupi →4 (√)* (√)*** HiactLxpi →2 (√)** (√)***
FatalGpi → (√)* × HiactLmilbrd →3 × ×
VolfatalLxpi →2 (√)* × HiactLmilbrd →4 × ×
VolfatalLmilbrd →3 (√)* × HiactLmilbrd →6 × (√)*
VolfatalLmilbrd →4 (√)* × HiactLedupi →1 (√)*** (√)**
VolfatalLmilbrd →6 (√)* (√)*** HiactGpi → (√)*** ×
VolfatalLedupi →1 (√)* ×
*, **, *** shows significance at 1%, 5% and 10% level, RC stands for reverse causation, √ means
causes and × means not causes
The reverse causality in the India-Pakistan bilateral trade measures show
that low levels of trade are also an outcome of India-Pakistan conflict which
has spanned more than 50 years. Thus any decrease in hostility levels would
also exert a positive and favourable effect on bilateral trade which would create
fertile grounds for dispute resolution. Thus more bilateral trade through
reduction of tariffs is a noteworthy confidence building measure. The presence
of reverse causality in average military spending is also not a surprise. This
means that India-Pakistan conflict is a significant cause of historically high
military expenditures between both countries. Especially, if high levels of
conflict between India and Pakistan lower India’s military expenditure as a
proportion of Pakistan’s military expenditure, then Lmilbrd1 and Lmilbrd3
32
would be positively related with conflict, which is the case in table 3a, 3b, 3c
and 4 . In the light of the results one interpretation may be that a military build
up by Pakistan increases as a response to conflict. This may be true because of
the dominant role of the army and high military expenditures in Pakistan are
justified due to continuous high levels of hostility with its neighbour.
Otherwise, Pakistan doesn’t have any major dispute with any other nation, or
frequent instances of intra-state disputes to justify the high budget allocation
for defense. Reduction of hostilities would thus favourably affect the military
burden in both countries, and both India and Pakistan can have more
resources to channel towards its development and poverty reduction strategies.
The reverse causality from conflict to education expenditure could explain this
process. Reverse causality between conflict measures and proxies of education
expenditure highlight the resource constraints faced by both sides due to their
rivalry where funds allocated to defense seem to crowd out public investment
in development sector. We also find that there is reverse causality between
Lxpi2 and Hstslvl and Hiact. This result highlights the economic implication of
conflict. If hostility levels rise and conflict moves closer to outright war, it will
strangle export capability with rest of the world for both countries. This will
have negative effects on growth potentials also. For example one can observe
from figure 2, section 1, that right after 1971 and 1999 wars between Pakistan
and India, total trade shares for both countries witnessed a deep decline.
Economic growth Granger causes conflict and the relationship is negative. The
growth patterns of both countries are independent of conflict, as far as reverse
causality is concerned. The relationship is highly significant at a 1% level in all
the observed instances of table 5. These results substantiate our graphical
analysis, where hostilities between both countries seem to go down when both
countries are performing well on the macroeconomic front. Any slow down in
growth rates in any of the two nations seem to be positively correlated with the
conflict and this trend has been very much present since 1950.
3.2.4 Results with VECM models
Our analysis above establishes an average relationship between conflict and
some of its identified determinants in a pure dyadic setting. We now wish to
further analyze country specific effects in order to investigate in detail the
potential of each country’s trade levels, military burden, development
expenditure and economic performance in enhancing peace and mitigating
conflict. For Pakistan, we use Pakistan’s trade share with rest of the world
(Popen), Pakistan’s total exports to GDP ratio (Pexpg) and Pakistan’s imports to
GDP ratio (Pimpg) as proxies of Pakistan’s multilateral trade. Pakistan’s exports
to India (Pxi) are a proxy for bilateral trade. Pakistan’s defence expenditure as a
percentage of its GDP (Pdg) is a proxy for the military burden, and Pedug is
Pakistan’s education expenditure as a percentage of its GDP. Similarly for
India, we employ 3 proxies of multilateral trade namely Iopen, Iexpg and Iimpg, 1
proxy of bilateral trade (Ixp), 1 proxy of military burden (Idg) and 1 proxy for
education expenditure (Iedug). We will not use separate Polity scores for India
and Pakistan, as any changes in combined democracy scores are due to
Pakistan. Before we carry out our econometric analysis, we undertook the
stationary test. Here note that our new variables are not a complex
33
combination of weighted proxies of dyadic nature and thus may show higher
levels of autocorrelation because they are simple percentages of times series
variables which are mostly capturing single country time dynamics. Achieving
stationarity in such a series at their level may be difficult.
TABLE 6
Augmented Dickey Fuller Test
Variables Lag length With intercept With intercept and trend
∆Fatal 1 -0.875* -0.929*
∆Popen 1 -0.977* -0.984*
∆Iopen 1 -1.192* -1.495*
∆Pexpg 1 -0.937* -0.965*
∆ Iexpg 1 -0.940* -1.257*
∆Pimpg 1 -1.125* -1.121*
∆Iimpg 1 -1.321* -1.449
∆Pxi 1 -1.692* -1.702*
∆Ixp 1 -1.971* -2.328*
∆Pedu 1 -0.946 -1.025*
∆Iedu 1 -0.841* -0.879*
∆Pgpc 1 -1.992* -1.995*
∆Igpc 1 -2.292* -2.293*
∆Pdg 1 -1.421* -1.441*
∆Idg 1 -0.899* -0.877*
∆Pmilpop 1 -1.289* -1.292*
∆Imilpop 1 -0.756* -0.766*
∆Demopi 1 -0.982* -0.982*
-*, ** and *** shows significance at 1%, 5%and 10% level
For time series variables, it is quite possible for random walks to be related
to each other so that a regression of one random walk on the other has a
stationary error term. As a simple example, consider a two variable system:
ε
=∆ t
y,1 and uy t=∆ ,2 let tt yy ,2,1
+
be stationary. The simplest example
is that .
,1,2 vyy tt +
−
=
That is, let one random walk be the negative of the other and allowing for
some error. Then the sum is simply a random error with no unit root or
autocorrelation. If the combination of unit root variables is not a unit root
then there must be some relationship between them. If there is co-integration
then a relationship exists, if not it does not. Therefore establishing that a
relationship exists between unit root variables is equivalent to establishing co-
integration. That relationship is called the co-integrating vector, which for our
example is (1, 1) since the sum is stationary. There is a way to write a system
that captures all the relationships and avoids unit roots. Consider
34
ttttt vyyy
+
+
+=∆ −−
ε
β
β
α
)( 1,121,211,1 ,
ttttt vuyyy
+
+
+=∆ −− )( 1,121,212,2
β
β
α
This is called a vector error correction model. The error correction comes
from the co-integrating relationship. The betas contain the co-integrating
equation and the alphas the speeds of adjustment. If t
y,1 and t
y,2 are far from
their equilibrium relationship, either t
y,1 or t
y,2 or both must change, the
alphas let the data choose. The vector part of the name does not apply to the
model above, but it will if the error terms are autocorrelated.
We ran unit root tests on the above variables and find that the unit root is
only solved at first differences, as shown by table 6. Since at levels, nearly all
variables have unit roots, there should be at least one co-integrating
relationship for our analysis to move forward. In other words, we can no more
use unrestricted VAR analysis but need to under take Vector Error Correction
Methodology (VECM) which is only a restricted VAR, where we first find the
presence of the number of co-integration equations in each VECM
specification and then run the regression analysis. As mentioned above, VECM
also allows us to have a rich set of information among variables including their
short and long-term adjustment dynamics and thus provides more
comprehensive insights into the relationship among variables than an
unrestricted VAR would do.
The three reduced form VECM equations for Conflict would be as
follows then:
+
+
+
+= −−−−−−−− ititititititititt PdgItrPtrConfConf ,4,3,2,111 (
β
β
β
β
α
t
yityatititit CDemoIdg 1
6
1,,6,5 )Ε+++ ∑
=
−−−−
ββ
(16)
+
+
+
+= −−−−−−−− ititititititititt PdgIeduPeduConfConf ,10,9,8,722 (
β
β
β
β
α
t
yitybtititit CDemoIdg 2
6
1,,12,11 )Ε+++ ∑
=
−−−−
ββ
(17)
+
+
+
+= −−−−−−−− ititititititititt PdgIgpcPgpcConfConf ,16,15,14,1333 (
β
β
β
β
α
t
yityctititit CDemoIdg 3
6
1,,18,17 )Ε+++ ∑
=
−−−−
ββ
(18)
Here, s'
β
show the co-integration relationship for each variable under
investigation for each equation, and the s'
α
show the adjustment parameters.
sC'are the constant terms for each six variables on the right hand side of each
VECM equation, and s'Εare the respective error terms. As mentioned, the
general openness indicator, total export shares, total import shares and exports
to the other country of conflict for both India and Pakistan are utilized as 4
separate single country proxies of trade. Thus there are 4 separate
specifications for equation (16). This makes total number of VECM
specification rise to 6.
35
TABLE 7
VECM Regression Results for Fatal:
VECM Regression Equations for Fatal under multiple specifications of Trade, Education and Growth
VECM 1 VECM 2 VECM 3 VECM 4 VECM 5 VECM 6
Variables
α
β
α
β
α
β
α
β
α
β
α
β
Fatal -0.92* 1 -0.80* 1 -0.87* 1 -0.96* 1 0.70* 1 0.05 1
Popen 0.27 0.15*
Iopen 0.39* -0.38*
Pexpg 0.28*** 0.15
Iexpg 0.11**
-
0.36***
Pimpg -0.02 0.30*
Iimpg 0.29* -0.85*
Pxi 4.51 -0.007
Ixp 2.20 -0.015*
Pedu -0.02 2.27*
Iedu -0.002 -1.53*
Pgpc -0.11** 2.18*
Igpc -0.37* 2.57*
Pdg -0.027 0.034 -0.081 0.52* 0.031 -0.25 0.048 -0.024 -0.082 0.84* 0.02 2.62*
Idg 0.009 -0.866* -0.017 -0.32 0.030 -1.35* 0.002 -0.031 -0.006 -0.84** 0.002 -2.04
Demopi
17.83*
-
0.003**
*
17.8* -
0.004** 14.73** -0.002 14.76** -
0.0004 19.11 -0.008* 2.42*
-
0.017**
*
(MaximumVEC Rank)º (1) º (1) º (1) º (1) º (1)º (4) º
N 53 53 53 45 53 52
R2 0.53 0.57 0.48 0.52 0.42 0.04
VEC(p) VEC(1) VEC(1) VEC(1) VEC(1) VEC(1) VEC(1)
--*, **, *** shows significance at 1%, 5% and 10% level,
∆
denotes values first difference
-Here
α
captures adjustment coefficients for a co-integration equation and
β
are parameters for each variable in a co-integration equation. ºVEC Rank shows the maximum number of co-
integration equations for each VECM model significant at 5%. - VEC(p) reports lag-order for each VECM model based on final prediction error (FPE), Akaike’s information criterion (AIC),
Schwarz’s Bayesian information criterion (SBIC) and the Hannan and Quinn information criterion (HQIC)
36
0 2 4 6
050 100 150
150 200 250 300
2 3 4 5
1.5 22.5 3
-100 0100 200
2003 2003.5 2004 2003 2003.5 2004 2003 2003.5 2004
Forecast for fatal Forecast for p xi Forecast for pm i
Forecast for pdg Forecast for idg Forecast for demop i
95% CI forecast
Graph 1d (VECM 4)
0 2 4 6
20 25 30 35 40
20 22 24 26
1 2 3 4 5
1.5 22.5 3
-100 0100 200
2003 2003.5 2004 2003 2003.5 2004 2003 2003.5 2004
Forecast for f atal Forecast for popen Forecast for iopen
Forecast for pdg Forecast for idg Forecast for demopi
95% CI f orecast
Graph 1a (VECM 1)
0 2 4 6
10 12 14 16 18
910 11 12
1 2 3 4 5
1.5 22.5 3
-100 0100 200
2003 2003.5 2004 2003 2003.5 2004 2003 2003.5 2004
Forecast for fatal Forecast for pexpg Forecast for iexpg
Forecast for pdg Forecast for idg Forecast for demopi
95% CI fo recast
Graph 1b (VECM 2)
0 2 4 6
10 15 20
10 11 12 13 14
1 2 3 4 5
1.5 22.5 3
-100 0100 200
2003 2003.5 2004 2003 2003.5 2004 2003 2003.5 2004
Forecast fo r fatal Forecast for pimpg Fo recast for iimpg
Forecast for pdg Forecast for idg Forecast for demopi
95% CI f orecast
Graph 1c (VECM 3)
0246
1.8 22.2 2.4 2.6
3.4 3.6 3.8 44.2
1 2 3 4 5
1.5 22.5 3
0100 200
2003 2003.5 2004 2003 2003.5 2004 2003 2003.5 2004
Forecast for fatal Forecast for pedug Forecast for iedu g
Forecast for pdg Forecast for idg Forec ast for demopi
95% CI forecast
Graph 1e (VECM 5)
0 2 4 6 8
-5 0 5 10
-5 0 5 10
2 3 4 5
1.5 2 2.5 3
0 100 200 300
2003 2003.5 2004 2003 2003. 5 2004 2003 2003.5 2 004
Forecast for fatal Forecast fo r pgpc Forecast for igpc
Forecast for pdg Forecast for idg Forecast for d emopi
95% CI forecast
Graph 1f (VECM 6)
FIGURE 7
Forecasting Simulations based on VECMs for Fatal
The results for each VECM equation are presented in table 7. The lag
length for each VEC equation is (1), based on final prediction error (FPE),
Akaike’s information criterion (AIC), Schwarz’s Bayesian information criterion
(SBIC) and the Hannan and Quinn information criterion (HQIC). There is one
co-integrating equation in each VECM, confirming the robustness of the
model specification. Note, that optimal value of conflict is zero in the long run,
37
meaning that our conflict measure, Fatal takes the value of 0. The results for
VECM 1 show that Fatal takes a negative value, and Popen and Iopen positive
values. This means that in the short-term both Pakistan and Indian trade shares
are negatively related with Fatal. However, only Indian trade is significant
enough to exert a negative pressure on hostilities in the short-term adjustment
period. In the long run both Pakistan’s and Indian trade shares with rest of the
world will adjust by moving in opposite directions. In the long run Pakistan
would witness a rise in its trade with the outside world and Indian trade would
decline to its optimal level. The long-term net result on the trade share of both
countries is expected to be positive as trade would be at higher levels with
peace than without peace. The long-term rise in Pakistan’s trade shares in order
to adjust to a fall in hostility levels also mean that the negative effects of India-
Pakistan conflict have thwarted Pakistan’s capacity to trade in international
markets more than in India’ case. Results on VECM 2 suggest that in short-
term both exports by India and Pakistan would rise to adjust to any fall in
conflict. However, in the long run Pakistani exports would remain unchanged,
whereas Indian exports will adjust downwards. Similar short-term adjustment
dynamics for imports are observed for India in VECM 3. However in the long
run Pakistan’s imports would rise as conflict moves to its optimal value of 0,
whereas imports by India will adjust downwards. The above discussion suggest
that Pakistan’s trading capability has been seriously hampered by the conflict
between both nations despite the fact Pakistan has been historically more open
economy when compared to India. As far as Indian trade with the outside
world is concerned, in the short-term it is destined to rise further especially if
hostilities with Pakistan abate. However, the long run trade share would adjust
downwards unless India follows a more open trade policy and further reduce
its tariffs to levels similar to its neighbour.
Bilateral trade would also respond to increased peace as shown by the
results of VECM 4. In the short-term there is a sign of increase in bilateral
trade between India and Pakistan, but the increase is not significant meaning
trade between India and Pakistan would remain low. However in the long run
Indian exports to Pakistan would go down to reach an optimal level. This is an
important finding. Some in Pakistan fear that peace initiatives like reducing
tariffs for Indian goods would mean greater dependency on Indian produce.
Taking into account the historically high hostility levels between two countries,
any peace initiative or confidence building measure which leads to more
market access to India is viewed with scepticism in Pakistan, as many fear that
dependence on India may expose Pakistan to unnecessary pressures from
India, and vulnerable to one sided solutions to the Kashmir dispute. Our
results show that in the long run the dependency on Indian cheap goods would
actually decline, and both countries would end up being equal trading partners.
Thus more bilateral trade, far from creating any power imbalance between
India and Pakistan, would equally distribute the gains. Pakistan may fulfil its
import needs more from the other developing countries such as China. The
results for VECM 5 suggest that education expenditure would increase in the
short-term to reduce conflict, and as conflict falls to its optimal level, Pakistan
would be able to simultaneously put more resources in education sector. High
growth rates also adjust positively to decrease hostility levels and in the long
run as the hostilities fall, both countries also witness a strong positive effect on
38
their growth rates. This means that peace would put India and especially
Pakistan on higher growth paths on a sustainable basis.
The results for the military burden show that in the short-term military
expenditure would continue to remain at high levels. However, in the long run,
as hostilities decline, Indian military expenditure would fall. However,
Pakistan’s military spending would adjust upward with declining trends in
hostility.8 In the short-term there is also evidence of higher democracy scores
as hostilities fall, but low values of the s'
β
show that conflict mitigation is
quite weakly related to conflict.
In order to further check the conclusions drawn from our VECM results
in table 7, we generated 6 different forecast schedules from 6 co-integrating
VECMs as a simulation exercise to predict how conflict would be affected to
changes in its determinants. Note that the data on Fatal are only up to 2002.
Thus the one year forecasts are generated for Fatal for 2003 period. Figure 7
shows the forecast graphs. Graph 1a, 1b and 1c suggest that if military
expenditures by both countries would remain at its current high levels, along
with trade with the outside world at their 2002 levels, a slight deterioration in
democracy scores will have a significant effect on the rise in hostility. However,
if India is able to export or import more, this would at least put a check on any
rise in the severity of conflict and hostilities would adjust to some average
level. Any decline in Indian trade will enhance hostilities. Current low levels of
bilateral trade between Pakistan and India is conflict enhancing so more trade
with increased exports by both sides to each other should be encouraged. More
access to Pakistani markets on the Indian side may not lead to conflict
mitigation if Pakistan is not able to also export more to India. A rise in
education expenditure puts a check on hostilities as seen in Graph 1e. Graph
1f is the standard representation of India-Pakistan conflict, and best fits
historical trends. The forecasts suggest that conflict will rise, even if there is a
significant increase in combined democracy scores, if growth rates plummet.
Both Pakistan and India have seen many such years, when hostilities between
both countries rose significantly when at least one of the countries is
performing poorly, but were channeling more resources on the military as a
proportion of their GDPs. The forecasts favour the liberal peace over
democratic peace. Thus one may look at current peace talks between both
countries with optimism as both are performing well on the economic front
and channeling fewer resources on military as a proportion of national income,
while at the same time having a divergent set of political institutions.
8 We have also run VECM regressions military personnel of each country as a proxy
of military burden. The results show that in the short to long-term there is a
significant decline in military personnel by Pakistani side, indicating lower levels of
militarization in the country. Thus high military spending by Pakistani side despite
decreasing hostility may indicate procurement of high end technology military
imports. Growth rates would rise as hostilities fall, Pakistan may have more resources
to channel to not only its development sector but also spend more to increase the
efficiency of its armed forces.
39
4 CONCLUSIONS
Conflict between India and Pakistan, which spans over most of last 60 years
since their independence from British rule, has significantly hampered bilateral
trade between the two nations. However, we also find that the converse is also
true; more trade between India and Pakistan decreases conflict and any
measures to improve the bilateral trade share is a considerable confidence
building measure. In the short term, greater Indian access to Pakistani markets
will help decrease hostilities between the two countries; whereas in the long
run as the peace is achieved, both countries could be exporting more to each
other. Lately, there has been a high demand of cheaper Indian raw materials in
Pakistani industries. A regional trade agreement along the lines of a South
Asian Free Trade Agreement (SAFTA) could enable freer access to the
markets of member countries, and has a high potential for the improvement of
relations between India and Pakistan on a long term basis. Pakistan and India’s
degree of openness to world trade is the dominant economic factor in conflict
resolution. One would imagine that in the counterfactual case of significant
mutual inward investment, that too would also decrease mutual belligerent
tendencies.
Some of our results may appear to suggest that Pakistan’s relative military
expenditure is conflict enhancing, whereas Indian relative military expenditure
has a deterrent effect on conflict. This result, however, needs to be interpreted
with caution. It does not necessarily mean that Pakistan is the principal actor
initiating inter-state conflict with India. Rather it means that India, the regional
hegemon, has other domestic and international concerns to which its defence
spending is targeted, besides its disputes with Pakistan. India, for example, has
unilaterally massed troops on Pakistan’s borders in 1951 and 2002. Indeed,
there is some reverse causality between some of the military proxies and
conflict suggesting that Pakistan’s military build ups may be more reactive.
Overall military expenditures are still at high levels in both countries and are
diverting scarce resources away from social development spending, such as on
education, and poverty reduction. Education spending has been shown to be
good for both peace and economic progress.
In an ideal world increased dyadic democracy between pairs of nation
should reduce inter-state hostility according to the democratic peace
hypothesis; this relationship in our case is present but weak. Peace initiatives, it
should be remembered, are not the sole prerogative of democracies; they can
also be made by countries which are less than perfectly democratic out of
economic self-interest. Pakistan, at present, is making unilateral concessions on
many disputed issues with India. Our findings, however, veer towards the
liberal peace hypothesis. Economic progress and poverty reduction combined
with greater openness to international trade in general are more significant
drivers of peace between nations like India and Pakistan, rather than the
independent contribution of a common democratic polity. So it is more
economic interdependence rather than politics which is likely to contribute
towards peaceful relations between India and Pakistan in the near future. In
many ways, our results for an individual dyad echo Polcahek’s (1997) work
across several dyads, where it is argued that democracies cooperate not because
they have common political systems, but because their economies are
40
intricately and intensively interdependent. As pointed by Hegre (2000), it is at
these higher stages of economic development that the contribution of
common democratic values to peace becomes more salient. Meaningful
democracy cannot truly function where poverty is acute and endemic, even in
ostensible democracies such as India. In the final analysis, it may be that
democracy itself is an endogenous by-product of increased general prosperity,
as suggested nearly half a century ago by Lipset (1960). Then and only then,
will nations be able to fully comprehend Angell-Lanes’ (1910) arguments
regarding the futility of inter-state conflict.
REFERENCES
Angell-Lane, Ralph Norman (1910) The Great Illusion: A Study of the Relation of
Military Power in Nations to Their Economic and Social Advantage, London:
Heinemann
Azar, Edward E. (1980) ‘The Conflict and Peace Data bank (COPDAB) Project’,
Journal of Conflict Resolution, 24 (1): 143-152
Barbieri, Katherine (1996) Economic Interdependence: A Path to Peace or a Source
of Intersate Conflict?’, Journal of Peace Research, 33 (1): 29-49.
Chang, Yuan-Ching, Solomon W Polachek and John Robst (2004) ‘Conflict and
Trade: The Relationship Between Geographic Distance and International
Interactions’, Journal of Socio-Economics, 33: 491-509.
Deger, Saadet and Somnath Sen (1990) ‘Military Security and the Economy: Defence
Expenditure in India and Pakistan in Keith Hartley and Todd Sandler (eds) The
Economics of Defence Spending, London: Routledge: 189-227.
Dorussen, Han (1999) ‘Balance of Power Revisited: A Multi-Country Model of Trade
and Conflict’, Journal of Peace Research, 36(4): 443-462.
Edgeworth, Francis Y. (1881) Mathematical Psychics, London: C Kegan Paul.
Elliot, Graham, Thomas Rothenberg, and James Stock (1996) ‘Efficient Tests for an
Autoregressive Unit Root’, Econometrica, 64: 813-836.
Faten, Ghosn Glenn Palmer, and Stuart Bremer (2004) "The MID3 Data Set, 1993–
2001: Procedures, Coding Rules, and Description." Conflict Management and Peace
Science, 21:133-154.
Granger, Clive (1969) ‘Investigating Causal Relations by Econometric Models and
Cross-Spectral Methods’, Econometrica, 37: 424-438.
Harbom, Lotta, Stina Högbladh and Peter Wallensteen (2006) ‘Armed Conflict and
Peace Agreements’, Journal of Peace Research, 43 (5): 617-631.
Hegre, Håvard (2002) ‘Trade decreases Conflict More in Multi-Actor Systems: A
Comment on Dorussen’, Journal of Peace Research, 39(1): 109-114.
Hegre, Håvard (2000) ‘Development and the Liberal Peace: What Does It Take to Be
a Trading State’, Journal of Peace Research, 37 (1): 5-30.
Huntington, Samuel P (1996) The Clash of Civilizations and the Remaking of the World
Order, New York: Simon and Schuster.
International Financial Statistics (2006), International Monetary Fund, Washington
D.C.
Jones, Daniel M., Stuart A Bremer and David J Singer (1996) ‘ Militarised Inter-state
Disputes, 1816-1992: Rationale, Coding Rules, and Empirical Patterns’, Conflict
Management and Peace Science, 15 (2): 163-212.
41
Kant, Immanuel (1795) Perpetual Peace and Other Essays on Politics, History and Morals,
reprinted 1983, Indianapolis: Hackett Publishing.
Lacina, Bethany and Nils Petter Gleditsch (2005) 'Monitoring Trends in Global
Combat: A New Data set of Battle Deaths', European Journal of Population, 21(2–3):
145–166
Lipset, Seymour (1960) Political Man: The Social Bases of Politics, New York: Doubleday.
McClelland, Charles A. (1978) ‘Warnings in the International Events Flow: EFI and
Roz as Threat Indicators’, International Interactions, 5 (2-3): 135-204.
Murshed, S Mansoob (2002) ‘Civil War, Conflict and Underdevelopment’, Journal of
Peace Research, 39 (4): 387-393.
Oneal, John and Bruce Russett (1999) ‘Assessing the Liberal Peace With Alternative
Specifications: Trade Still Reduces Conflict’, Journal of Peace Research, 36 (4): 423-
32.
Polachek, Solomon W (1997) ‘Why Democracies Cooperate More and Fight Less: The
Relationship Between International trade and Cooperation’, Review of International
Economics, 5(3): 295-309.
Polachek, Solomon W and Carlos Seiglie (2006) ‘Trade, Peace and Democracy: An
Analysis of Dyadic Dispute’, IZA DO No. 2170.
Polity IV Project (Center for International Development and Conflict Management),
http://www.cidcm.umd.edu/polity.
Rawls, John (1999) The Law of Peoples, Cambridge, Mass: Harvard University Press.
Richardson, Lewis Fry (1960), Arms and Insecurity, A Mathematical Study of the Causes and
Origins of War, Pittsburgh: The Boxwood Press.
Robst, John, Solomon Polachek and Yuan-Ching Chang (2006) ‘Geographic
Proximity, Trade and International Conflict/Cooperation’, IZA DP no. 1988.
Sims, Chris (1980) ‘Macroeconomics and Reality’, Econometrica, 48: 1-48.
Singer, David J. and Melvin Small (1972), The Wages of War, 1816-1965: A Statistical
Handbook, New York: Wiley.
Schwert, G William, 1989. ‘ Why Does Stock Market Volatility Change over Time?’,
Journal of Finance, 44(5): 1115-53.
World Development Indicators (2006), World Bank, Washington D.C
Wright, Quincy (1942), A Study of War, Chicago: University Press.
DATA AND SOURCES
Single Country Variables
Idg: India’s Defence Expenditure as a %age India’s GDP at current market prices,
Years: 1950-2005, Sources: Correlates of war data set version 3.02, World
Development Indicators 2006 (World Bank), Government Finance Statistics
Year Book 2006 (IMF) and Economic Survey of Pakistan
Iedug: India’s education expenditure as a %age of India’s GDP at current market
prices, Years: 1950-2005, Sources: Indian Economic Survey, Education Statistics
(Department of Education, India) and Education Statistics 2006 (World Bank)
Iexpg: India’s total exports as a percentage of India’s GDP, Years: 1950-2005, Source:
Indian Economic Survey, International Financial Statistics 2006 (IMF)
Ig: Annual growth rate of India’s per capita gross domestic product (GDP) at
constant prices, Years: 1950-2005, Source: Indian Economic Survey
42
Igpc: India’s real per capita growth rate: Years: 1950-2005, Source: Indian Economic
Survey, International Financial Statistics 2006 (IMF), World Development
Indicators 2006 (World Bank)
Iimpg: India’s total imports as a percentage of India’s GDP, Years: 1950-2005,
Source: Indian Economic Survey, International Financial Statistics 2006 (IMF)
Imilopop: India’s number of military personnel as a percentage of Indi’s total
population. Years: 1950-2003, Source: COW Inter-State War Data, Version 3.02,
Faten et al (2004), International Financial Statistics 2006 (IMF)
Iopen: India’s exports plus imports as a %age India’s GDP at current market prices,
Years: 1950-2005, Source: International Financial Statistics 2006 (IMF)
Ixp: Indian exports to Pakistan, Years: 1960-2005, Source: Direction of Trade
Statistics yearbook, IMF
P2i: Polity 2 Score for India, numeric range from -10 (high autocracy) to 10 (high
democracy), Years: 1950-2003, Source: Polity IV Project (Center for
International Development and Conflict Management)
P2p: Polity 2 Score for Pakistan, numeric range from -10 (high autocracy) to 10 (high
democracy), Years: 1950-2003, Source: Polity IV Project (Center for
International Development and Conflict Management)
Pedug: Pakistan’s education expenditure as a percentage of Pakistan’s GDP at current
market prices, Years: 1950-2005, Sources: Pakistan Economic Survey and
Education Statistics 2006 (World Bank)
Pexpg: Pakistan’s exports as a percentage of Pakistan’s GDP, Years: 1950-2005,
Source: International Financial Statistics 2006 (IMF)
Pdg: Pakistan’s Defence Expenditure as a percentage Pakistan’s GDP at current
market prices, Years: 1950-2005, Sources: Correlates of war data set version 3.02,
World Development Indicators (2006), Government Finance Statistics Year
Book 2006 (IMF) and Economic Survey of Pakistan
Pg: Annual growth rate of Pakistan’s GDP per capita at constant prices, Years: 1950-
2005, Source: Pakistan Economic Survey
Pgpc: Pakistan’s real GDP per capita Growth rates, Years: 1950-2005, Source:
International Financial Statistics 2006 (IMF), Pakistan Economic Survey
Pimpg: Pakistan’s imports as a percentage of Pakistan’s GDP, Years: 1950-2005,
Source: International Financial Statistics 2006 (IMF)
Pmilpop: Pakistan’s number of military personnel as a percentage of Pakistan’s total
population. Years: 1950-2003, Source: COW Inter-State War Data, Version 3.02,
Faten et al (2004), International Financial Statistics 2006 (IMF)
Popen: Pakistan’s exports plus imports as a percentage Pakistan’s gross domestic
product at current prices, Years: 1950-2005, Source: International Financial
Statistics 2006 (IMF)
Pxi: Pakistan’s exports to India, Years: 1960-2005, Source: Direction of Trade
Statistics yearbook, IMF
Dyadic Variables
Cnfpi: Intensity of Conflict between Pakistan and India, Scores 1 (Minor) when 25 to
999 battle-related deaths and 2 (War) when at least 1000 battle-related deaths in a
given year, Years: 1950-2003, UCDP/PRIO Armed Conflict Data set Version
IV, Harbom et al (2006)
43
Demopi: Pakistan and India’s combine democracy score (by adding 10 to India and
Pakistan’s Polity2 values for each year and then taking the product of these
values in order to covert the variable in dyadic form), Years; 1950-2003
Dur: Number of days a conflict lasts in a year between Pakistan and India, Years:
1950-2003, Source: COW Inter-State War Data, Version 3.02, Faten et al (2004).
Fatal: Annual fatality level of conflict between Pakistan and India, scores from 0 to 6
0 None
1 1-25 Deaths
2 26-100 Deaths
3 101-250 Deaths
4 251-500 Deaths
5 501-999 Deaths
6 6>999 Deaths
Years: 1950-2003, Source: COW Inter-State War Data, Version 3.02, Faten et al
(2004)
Gpi: Weighted average of real GDP per capita growth rates for Pakistan and India,
Years: 1950 to 2005. Sources: Pakistan Economic Survey, Indian Economic
Survey, International Financial Statistics 2006 (IMF)
Hiact: Highest action by Pakistan and India in annual corresponding dispute
[bracketed numbers refer to corresponding hostility level]
0 No militarised action [1]
1 Threat to use force [2]
2 Threat to blockade
3 Threat to occupy territory [2]
4 Threat to declare war [2]
5 Threat to use CBR weapons [2]
6 Threat to join war
7 Show of force [3]
8 Alert [3]
9 Nuclear alert [3]
10 Mobilisation [3]
11 Fortify border [3]
12 Border violation [3]
13 Blockade [4]
14 Occupation of territory [4]
15 Seizure [4]
16 Attack [4]
17 Clash [4]
18 Declaration of war [4]
19 Use of CBR weapons [5]
20 Begin inter-state war [5]
21 Join inter-state war [5]
Years: 1950-2003, Source: COW Inter-State War Data, Version 3.02, Faten et al
(2004)
Hstlev: Annual hostility levels reached by India and Pakistan in each annual
corresponding dispute
1 No militarised action
2 Threat to use force
3 Display of force
4 Use of force
5 War
Years: 1950-2003, Source: Faten et al (2004)
44
Ledupi1: Log GDP weighted average of India and Pakistan’s per capita education
expenditures, Years: 1950 to 2005 Sources: Pakistan Economic Survey, Indian
Economic Survey, Education Statistics 2006 (World Bank), International
Financial Statistics 2006 (IMF)
Ledupi2: Log mean average of India and Pakistan’s per capita education
expenditures, Years: 1950 to 2005 Sources: Pakistan Economic Survey, Indian
Economic Survey, Education Statistics 2006 (World Bank), International
Financial Statistics 2006 (IMF)
Ledupi3: Log of Pakistan plus India’s education expenditures as a ration of Pakistan
plus India’s GDPs, Sources: Pakistan Economic Survey, Indian Economic
Survey, Education Statistics 2006 (World Bank), International Financial Statistics
2006 (IMF)
Ledupi4: Log of average of Pakistan’s education expenditure over GDP plus India’s
education expenditure over GDP, Years: 1950 to 2005, Sources: Pakistan
Economic Survey, Indian Economic Survey, Education Statistics 2006 (World
Bank), International Financial Statistics 2006 (IMF)
Lmilbrd1: Log of Pakistan’s defence expenditure over Pakistan’s GDP as a ratio of
India’s defence expenditure over India’s GDP, Years: 1950-2005, Sources:
Correlates of war data set version 3.02, World Development Indicators 2006
(World Bank), Government Finance Statistics Year Book (IMF) and Economic
Survey of Pakistan
Lmilbrd2: Log of India’s defence expenditure over India’s GDP as a ratio of
Pakistan’s defence expenditure over Pakistan’s GDP, Years: 1950-2005, Sources:
Correlates of war data set version 3.02, World Development Indicators 2006
(World Bank), Government Finance Statistics Year Book (IMF) and Economic
Survey of Pakistan
Lmilbrd 3: Log of Pakistan’s defence expenditure over Pakistan’s GDP as a ratio of
Pakistan’s defence expenditure over Pakistan’s GDP plus India’s defence
expenditure over India’s GDP, Years: 1950-2005, Sources: Correlates of war data
set version 3.02, World Development Indicators 2006 (World Bank),
Government Finance Statistics Year Book (IMF) and Economic Survey of
Pakistan
Lmilbrd 4: Log of India’s defence expenditure over India’s GDP as a ratio of
Pakistan’s defence expenditure over Pakistan’s GDP plus India’s defence
expenditure over India’s GDP, Years: 1950-2005, Sources: Correlates of war data
set version 3.02, World Development Indicators 2006 (World Bank),
Government Finance Statistics Year Book (IMF) and Economic Survey of
Pakistan
Lmilbrd5: Log of Mean average of India’s defence expenditure over GDP and
Pakistan’s defence expenditure over GDP, Years: 1950-2005, Sources: Correlates
of war data set version 3.02, World Development Indicators 2006 (World Bank),
Government Finance Statistics Year Book (IMF) and Economic Survey of
Pakistan
Lmilbrd6: Log GDP weighted average of Pakistan and India’s defence expenditures,
Years: 1950-2005, Sources: Correlates of war data set version 3.02, World
Development Indicators 2006 (World Bank), Government Finance Statistics
Year Book (IMF), Economic Survey of Pakistan, Economic Survey of India
Lmilppi: Log of Pakistan’s military personnel over Pakistan’s total population as a
ratio of India’s military personnel over India’s total population, Years: 1950-2001,
Sources: Correlates of war data set version 3.02 and International Financial
Statistics 2006 (IMF)
45
Lmilpip: Log of India’s military personnel over India’s total population as a ratio of
Pakistan’s military personnel over Pakistan’s total population. Years: 1950-2001,
Sources: Correlates of war data set version 3.02 and International Financial
Statistics 2006 (IMF)
Lmpi1: Log GDP weighted average of Pakistan and India’s total imports, Years:
1950-2005, Source: International Financial Statistics 2006 (IMF)
Lmpi2: Log mean average of Pakistan’s total imports as a proportion of Pakistan’s
GDP and India’s total imports as a ratio of India’s GDP, Years: 1950-2005,
Source: International Financial Statistics 2006 (IMF)
Lxpi1: Log GDP weighted average of Pakistan and India’s total exports, Years: 1950-
2001, Source: International Financial Statistics 2006 (IMF)
Lxpi2: Log mean average of Pakistan’s total exports over Pakistan’s GDP and India’s
total exports over India’s GDP. Years: 1950-2001, Source: International Financial
Statistics 2006 (IMF)
Poppi: Average of Pakistan’s total population and India’s total population, Years:
1950-2001, Source: International Financial Statistics 2006 (IMF)
Tpitp: Bilateral trade between Pakistan and India as a ratio of Pakistan’s total trade,
Years: 1950-2001, Source: Direction of Trade Statistics yearbook, IMF
International Financial Statistics 2006 (IMF)
Tpiti: Bilateral trade between Pakistan and India as a ratio of India’s total trade, Years:
1950-2001, Source: Direction of Trade Statistics yearbook, IMF International
Financial Statistics 2006 (IMF)
Xmpi: Pakistan’s total trade (exports + imports) as a ratio of India’s Total trade
(exports + imports), Years: 1950-2001, Source: International Financial Statistics
2006 (IMF)
Xmip: India’s total trade (exports + imports) as a ratio of Pakistan’s total trade
(exports + imports). Years: 1950-2001, Source: International Financial Statistics
2006 (IMF)
VolFatal: Precise volume of fatality in each annual corresponding dispute, Years:
1950-2003, Sources: COW Inter-State War Data, Version 3.02 (Faten et al, 2004),
CSCW/PRIO Battle Deaths data (Lacina, 2005), CSP Data set on Major
Episodes of Political Violence 1946-2006
http://members.aol.com/cspmgm/warlist.htm
