Does aid cause conflict in Pakistan?

Abstract

This study provides evidence from Pakistan on how the delegated task of achieving strategic objectives of the donor can lead to incompatibility of aid objectives which then generates perpetual and multidimensional domestic conflict in the recipient society. We use count data method to estimate the relationship between aid and conflict. At the aggregate level, social sector spending, regime change and youth bulge are positively and significantly related with conflict. However, aid per capita gives ambiguous results. It is significant with conflict count in the terrorism data-set and insignificant for data on armed conflict. Inclusion of youth bulge and unemployment rate confirms the marginalization hypothesis of conflict. Inflation rate and the tax variables are insignificant. This confirms that aid erodes fiscal capacity. At project-level data, conflict is strongly related with aid commitment and purpose. Discrepancy in aid allocation and commitment may accentuate conflict.

Keywords:

• Conflict
• Aid
• Military expenditure
• Social sector
• Fungibility
• Inflation
• Pakistan

JEL Codes:

• C61
• D74

1. Introduction

In 2012 alone, 1649 events of terrorism were recorded in Pakistan. These events caused 315 deaths and 716 persons were injured. Of the total, 438 attacks were on business and private property (GTD 2013). Almost 80% of these attacks were clustered in the provinces of Khyber Pakhtunkhwa (KPK) and Balochistan as well as the Federally Administered Tribal Areas (FATA). These are poorer regions, and their income per capita is significantly lower than the rest of Pakistan. The loss of life and property has worsened the pre-existing inequalities and the overall economic prospects.

Conflict in Pakistan is not a recent phenomenon. It has its roots in history. However, the intensity and proximity of conflict have changed over time and space. Since independence in 1947, she has fought four wars with India. The first armed conflict was in 1948 over the Kashmir issue. The war of 1965 applied brakes to an economy that was a growth leader in the region (Noman 1988; Papanek 1967). Inequalities and social tensions resulting from the lop-sided growth in the 1960s led to civil war in the country and ultimately to the separation of Bangladesh in 1971 (Noman 1997). The 1980s witnessed the fallout of the Soviet occupation of Afghanistan – the rise of the transnational terrorism and the massive influx of Afghan refugees. In the 2000s, the war on terror resulted in non-state actors posing an existential threat to the state itself. These non-state actors include religious sects, ethnicities and nationalities (Lodhi 2011).

Economic and social indicators have deteriorated. Growth per capita in Pakistan is low (GOP 2013). The country fared below average on five out of six governance indicators in the South Asian region (World Bank 2014) and was ranked 139 out of 176 in terms of corruption (Transparency International 2012). Pakistan is also struggling with its poor development outcomes. Her HDI position was 146 out of 187 countries in 2013 (UNDP 2014). In general, these are considered sufficient conditions for a society to be in conflict (Fearon and Laitin 2003; Hegre and Sambanis 2006).

Pakistan received net ODA of USD 117.6 billion in 1970–2012, but spent almost the same amount on security and defence (WDI 2014). Huge inflows of ODA have neither reduced the intensity and occurrence of conflict nor improved the HDI. Why is the country in a state of continuing conflict? Is aid a reward for combating conflict in Pakistan or the prize money for rent seeking and rebels to capture the state? The findings of previous studies on aid effectiveness seem to be divided. One group (AhwAhwireng-Oben and McGowan 1998; Alesina and Dollar 2000; Balla and Reinhardt 2008; Collier 2007; Collier and Hoeffler 2004; Easterly 2006; Uvin 1996) argues that ODA is less effective as donors tie aid with their strategic and military objectives and prefer to provide significant aid to their former colonies. This type of aid inflows create economic inequalities by disturbing resource allocation mechanism and priorities in the society. Collier and Hoeffler (2007) see another negative dimension of aid in that it triggers arms race as an unintended consequence. Nielsen et al. (2011) view armed conflict as a consequence of aid shock. The other group of scholars views ODA as an instrument tailored to assist developing countries in mitigating poverty. Aid increases availability of resources for investment, technology and means to increase productivity. It increases saving and provides opportunities for growth. Without aid, the situation would be much worse (Hansen and Tarp 2000; Radelet 2006). However, aid works in an environment of good governance and policies (Burnside and Dollar 2000). The earliest evidence that aid works only in democratic and well-functioning societies was provided by the success of the Marshall Plan (De Long and Eichengren 1991). Scholars like Balla and Reinhardt (2008), Easterly (2006, 2008), and Nunn and Qian (2012) seem to agree that aid works in a good policy environment. Donors’ neglect of the preconditions leads to the failure of aid programmes in developing countries.

There is a shift from the view that aid for development per se reduces conflict. These authors (Azam and Delacroix 2006; Azam and Thelen 2008, 2010; Bapat 2011, Nielsen et al. 2011; Ree and Nillesen 2009; Young and Findley 2011) find a negative relationship between aid and transnational conflict and prescribe more aid as a means to reducing conflict. Azam and Thelen (2008, 2010) developed a model that aid is money given for protecting the economic and strategic interests of the donor, as Western democracies are the main target of the terrorists. Aid is the preferred option over military intervention. Bapat (2011) holds that military aid is not effective in disarming terrorist organizations but it wins support of the host country to negotiate with rebel groups which is important for achieving strategic objectives of military aid. The host country has no incentive to completely disarm the rebels, knowing that the end of conflict means the end of aid. Young and Findley (2011) support the argument that aid reduces conflict, but fungibility of aid and rent seeking makes aid ineffective. It is the prize money for capturing the state. Grossman (1992) also viewed aid as a reward for the rebels; increased foreign aid induces rent seeking behaviour and increases the occurrence and intensity of conflict. It is also a prize for capturing the state as aid inflows provide resources to rebels for financing their operations.

This study aims to explain how aid is responsible for increased conflict and violence in Pakistan. We see conflict as a long-term problem which emerges from the incompatibility between the objectives of donors and recipients. It changes the societal expectations and creates inequalities. Aid creates incentives for groups competing for resources. The donor delegates a task to the host country. The host has to perform to take care of the donor’s objectives and simultaneously achieve its own objectives which may not completely align with the donor’s task. This incompatibility of objectives creates distress between the donor and the recipient and encourages rebel groups who for their political gains try to capture the state. The donor may feel that the delegated task of controlling terrorism is not achieved and the recipient feels that the money is not enough for performing the delegated task.

To the best of our knowledge, this study is the first to provide combined evidence on aid and conflict at the aggregate and the project level. At the aggregate level, our results are in line with previous studies, confirming the marginalization and the demographic implications of funnelling more aid into conflict-ridden societies. At the project-level data, we provide evidence on how aid is related with the incidence of conflict. It helps to identify the type of aid required in a conflict situation. Section 2 reviews the literature on the historical nexus between aid and conflict to see what it is that hinders the achievement of the stated development objectives of aid. Section 3 lays out the contents of our model. Section 4 explains the methodology and research design. Data sources and limitations are also given here along with descriptive analysis. Section 5 presents empirical results. The last section summarizes the findings and presents the conclusions.

2. Conflict-aid nexus

Countries can either use wars and military interventions or aid to achieve their strategic objectives. War increases transnational terrorism, whereas aid can be used to increase influence on recipient country for achieving strategic objectives of the donor. It also decreases transnational terrorism. According to Levy (1989), Meernik, Krueger, and Poe (1998), Meernik and Waterman (1996), Ostrom and Job (1986), Rummel (1963) and Smith (1996), belligerent politicians use wars as a diversionary tactic to consolidate their domestic position in domestic crises. These scholars hold that unpopular democratic governments can use active foreign policy (diversionary war) to win electoral support but there is no evidence that such a policy leads to re-election or the realization of strategic objectives.

External war is one of the policy options to solidify domestic economic gains. Meernik and Waterman (1996), Ostrom and Job (1986) and Wilkenfeld (1968) find that the decision to start war is determined by the nature of regime (polyarchic or authoritarian), international political issues and domestic concerns. Meernik and Waterman (1996) find little evidence that domestic conditions, trade and domestic support determine the use of force. However, Levy (1989) finds that there is no clear cut relationship that domestic factors influence foreign policy or international crises affect the choice to start a war. Since the gains associated with wars are not certain, foreign aid offers protection of economic and political gains with relatively more certain policy outcomes.

Scholars working on aid outline at least three reasons for allocation of aid. It is allocated in pursuit of economic, political and strategic interests. Economic and strategic interests shape the foreign aid distribution. While aid is allocated on the basis of trade and business opportunities, there is hardly any evidence that aid is associated with the promotion of human rights and democracy. Poe, Tate, and Keith (1999), Meernik, Krueger, and Poe (1998), Poe and Meernik (1995) and Poe and Tate (1994) are of the view that aid distribution and allocation differ in the cold war and post-cold war periods. In the cold war period, bipolar competition was responsible for security-driven imperatives of states, whereas in the post-cold war period, ideology-driven goals are more important. The United States is increasingly rewarding democratic states with foreign aid in the post-cold war period while reducing assistance to abusers of human rights. Braaten (2014) finds mixed results for the hypothesis that human rights are an important consideration for aid allocation. Performance-based lending is associated with trade relations and military relations with the United States. The United States’ voting rights in multilateral development banks (MDBs) significantly influence the aid allocations. Political rights violations are a significant factor for her abstentions in the MDBs while personal rights were not significant. The recipient must meet certain policy requirements before receiving aid.

How aid influences internal conflict is an area which is still a matter of controversy. A number of scholars (Collier and Hoeffler 2004, 2007; Fearon and Laitin 2003; Krueger 2008; Krueger and Jitka 2003; Piazza 2011) test the deprivation theory of conflict. These scholars are of the view that economic deprivation is the reason of conflict and suggest that aid could be used effectively to control conflict in a society. They find military action as counterproductive for controlling conflict and insurgency. The results of testing deprivation conflict hypothesis are mixed. Macro studies conclude that national poverty may indirectly cause conditions of civil war but it is not directly related to conflict. Micro-level studies show that terrorists are neither improvised nor uneducated. Krueger and Jitka (2003) contend that suicide bombers are not persuaded by their own economic gains. Wealthy people from poor countries could be motivated by the poverty condition in their countries to abet terrorism.

Aid received from donors to reduce poverty or for development creates an opportunity for rebels to capture resources or power. Rummel (1979) defines conflict in a society as an outcome of shrinking opportunities for the marginalized. It emerges when there is change in a particular structure of expectations. Once triggered, it generates an unending spiral that accentuates disruption of normalcy, economy, institutions, norms and values. Blattman and Miguel (2010) see conflict as an outcome of the absence of rent seeking opportunities for an isolated group and negotiation failure due to asymmetric information. Hirshleifer (1989) and Grossman (1992) see conflict as a contest where rebels and government are two parties in a predatory fight for resources. The winning party has full control over resources to consume. Bates (2001) and Powell (2007) advocate a modern theory of conflict by arguing that exclusion from political, economic and social power may lead to conflict. Fearon (1995) finds that conflict is always persistent in nature. It is a rational choice as state prefers war over peace because peace deals are not reliable and cheating may increase payoff. It is difficult to bargain with rebels. Wars are costly but nonetheless recur. Rationality lies in carefully weighing the incentive to compete. The advantage of first strike may secure the control over resources. Future expectations of economic gains, ideology and ethnicity motivate participation in conflict. In all these models, availability of resources, inequalities of wealth and income or poverty are reasons for conflict. The poor and corrupt governments have the tendency to indulge in war.

Some scholars test the rational choice model to estimate the economic consequences of terrorism (Enders and Sandler 1993; Enders, Sandler, and Gaibulloev 2011; Hudson and Mosley 2008; Sandler 2011; Sandler and Enders 2004; Sandler and Siqueira 2006). Economic gains associated with conflict are perceived as a reward for combating terrorism. They find that terrorism reduces the inflow of foreign resources significantly by making investment and tourism risky. States are focused on controlling domestic conflict because it significantly reduces economic benefits.

Raleigh and Hegre (2009) conclude that most armed conflicts are local. More aid to the governments can help in controlling domestic conflict. Findley et al. (2011), Nielsen et al. (2011) and Ree and Nillesen (2009) examine the relationship between aid and conflict. Based on these researches, one can conclude that significant increase in aid reduces conflict. Moreover, sudden withdrawal of aid causes violent armed conflict. It erodes the state capacity to combat conflict. Azam and Thelen (2008, 2010) advocate a choice model between military intervention and aid for reducing terrorism. In their model, aid appears as a payment for delegating the task to other countries for protecting donors’ economic and political interests. Their main finding is that aid is effective in reducing transnational terrorism.

Another group of authors find that a ‘carrot policy’ of aid allocation is more effective. Aid should be spent on winning the hearts and minds of the local population. The ODA could be effectively used to reduce the influence of insurgents on local population and minimize conflict in future. Iyengar, Monten, and Hanson (2011) find that labour-intensive development projects are a good counterinsurgent strategy as it increases the opportunity cost of those participating in insurgencies and also helps win the hearts and minds of the local population instead of treating them harshly using military force which creates a more sever response in future. They conclude that Commandar’s Emergency Response Programme (CERP) in Iraq and Afghanistan decreased violence effectively. However, the success of this strategy depends on effective management of reconstruction activities. However, Bohnke and Zurcher (2013) and Chou (2012) found that development aid in Afghanistan was not effective in reducing insurgency. It is not the money spent but effective service delivery which matters for reducing attacks. Scoones (2013) and Scoones and Child (2013) find that conflict is a manifestation of dissatisfaction of society with the development efforts. The society did not value all reconstruction efforts positively. The CERP and non-CERP projects were not significantly different in reducing conflict in society. Child (2014) finds that overall development projects have no meaningful effect on violence. He used sectoral-level data of development project in various districts in Afghanistan and concluded that health- and economy-related projects are negatively related with conflict, whereas education and infrastructure projects are positively related with conflict. Scoones (2013) found military measures counterproductive in winning the support of local community against insurgents. Conflict may arise because of disagreement over government policies, but it adversely affects government’s ability to provide public goods. The government starts military action against insurgents which hinders the process of providing public goods. Counter insurgency measures of reconstruction positively affect the community support for economic agenda. Using the rational choice model of conflict (Fearon 1995; Powell 2007), Crost, Felter, and Johnston (2014) conclude that strategic collaboration between the partners may intensify conflict in a society because it shifts power structure in the area and the insurgents have the ability to hinder the project. They show that conditional cash transfers caused a substantial decrease in conflict in Philippines.

3. The model

Our point of departure is Azam and Thelen (2010). In their model, a foreign power delegates the task of protecting its economic and political interests to another country. Home country has two options: either go for military intervention or give aid to achieve its objectives. The host country has either to face war or carry out the task assigned by the foreign power. To avoid war, the host country gives its support to the foreign power and signs a contract. It seems like a win–win situation for both. More aid means more commitment and reduction in terrorism. Conflict is the opportunity cost of protecting the interests of foreign power and avoiding war. This is our point of departure from the assumptions of Azam and Thelen (2010). Declining the offer of aid for the delegated task results in an all-out war. Acceptance of aid means rebels would declare war against the host country. For the host country, the choices are not aid or military intervention but war or aid with conflict. As in the wake of 9/11, Pakistan was offered a Hobson’s choice by the United States: ‘Either you are with us, or you are with the terrorists’ (Hirsh 2002). In its own interest, the host agrees to sign on the contract and shows willingness to reform without counting the cost of cooperation. No allowance is made for the indirect implications of aid in the form of conflict in the society.

The schematic diagram describes our model. There are three players – donor, recipient and terrorist organizations. Terrorist organization includes all types of rebels and violent activities. Donor country pays for the cost of combating transnational terrorism which targets and threatens Western democracies. The donor maximizes its self-interest. The recipient, besides being an ally, has to deliver to its people. Its agenda is to extract maximum resources from the donor to spend on development. The rebel organizations seek to create terror and disturb peace.¹ With more aid in the system, rebels just need to change their strategy from transnational to domestic violence. On the one hand, the recipient government faces domestic conflict in order to protect economic and strategic interests of the foreign government. On the other hand, it faces foregone development by diverting resources to combating terrorism. Low development outcomes in the economically deprived regions suffering from terrorism exacerbate social conflict (Diagram 1).

Incompatibility of goals based on self-interest creates conflict. The donor government is clearly concerned with the protection of its interest and rebels want to create terror. The multiplicity of goals of the recipient government creates a gap between the foreign and the home country. Strategic and political alliance with the donor entails external conflict, and its fallout is internal conflict. In the next period, the inflow of aid changes societal expectations. People demand more responsive government that can ensure their security and development needs. Donors demand diligent behaviour and strong support for combating terrorism and rebels who are indulging in rent seeking and capturing resources to finance their operations. The recipient government wants to maximize the welfare of its people by minimizing conflict and maximizing social and human capital . Maximization of welfare depends on maximization of expenditure on development, debt servicing and defence. We assume a balanced budget that depends on total availability of resources.

Total resources in any given period depend on internal sources and external sources. denotes total resources available, t₀ stands for domestic resources, α = weight assigned on the basis of the tax rate, t₁ is official development assistance (ODA) and (1– α) is weight assigned to ODA rate. For simplicity, we assume that all ODA appears as budgetary support.

Normally, expenditure side of the government consists of current expenditure and development expenditure . We are including conflict-related expenditure in this equation. is expenditure that the government has to incur on defence, social sector and reforms required to get aid from donors. is social sector expenditure, is defence expenditure and is reforms expenditure. Defence expenditure is further subdivided into conventional and unconventional defence expenditure. We also assume that is always positive and greater than zero in the presence of conflict. This is the reason why than . To get aid, the recipient government has to introduce a set of reform . A simple cost benefit analysis of the situation would show that if net discounted gains are positive, it is not a bad outcome.

Subject to the constraints

Lagrangian function to maximize welfare

when (a negative value shows social deficit), (a positive value shows elite capture of the state).

Residual value of Lagrangian function depicts incompatibility of aid agenda. A decreasing welfare function shows social deficit that will increase in each time period and the recipient has to incur more expenditure in each time period to win the trust of the donor. The increasing welfare function generates a bureaucratic system and elite capture of the state. We treat conflict as minimizing the welfare of the society. The cost of combating terrorism will increase in each time period, and its dimensions will also change. It will have an exponential function with demands for more and more resources. Conflict is not a one-period game, which ends with the dispensation of aid. According to Enders and Sandler (1993), once the donor and the recipient agree on a development agenda, rebels substitute their strategy. Unlike Azam and Thelen (2010), we show that the presence of foreign power increases military- and security-related expenditure of the government as it has to face a stronger resistance form rebels. The country moves up the supply curve of attacks and loses the trust of the foreign power. The failure to effectively protect foreign interest reduces aid in the next period. Conflict increases but aid decreases. It appears as a change in social expectations and a more unequal society facing a different type of conflict. So far in this model, we have not assumed a predatory government. Inclusion of such a government would mean that it neither addresses the needs and demands of its people nor shows due diligence in the implementation of the contract agreed with the foreign power. Predatory government will delay the necessary reforms like higher expenditure on social sector, taxing the rich and shun the use of inflation as a means to collect revenue. The presence of such a government aggravates the conflict situation. In contrast, a democratic system allows groups and individuals non-violent recourse as an alternative to the costly pursuit of terrorist activity (Li 2005).

4. Research design and methodology

Conflict, in this study, is seen as a consequence of incompatibility of aid objectives between the donor and the recipient government. When a foreign government delegates a task to a local government to protect its economic and strategic interest, the latter may act to reduce transnational terrorism, as shown in the literature, or help its people by avoiding direct military intervention. However, the action can also trigger domestic conflict. This study hypothesizes that conflict is an outcome of aid that appears as an intervention in societal expectations. Further, not only does aid appear as prize money for rebels but it also creates an elite group. Furthermore, there is the issue of fungibility. Fungibility allows the recipient government to divert aid for purposes other than development or priorities different from those set by the donor (Feyzioglu, Swaroop, and Zhu 1998). The substitution effect of aid decreases social sector development and increases military expenditure. Wrong policy and misallocation of resources result into unending episodes of conflict. Donor objectives and fungibility in aid allocation generate negative externalities in a developing society like Pakistan, which lacks the capacity to internalize negative effects. The cost of conflict is less economic development, and the reliance on aid creates less efficient governments that delay reforms in the hope of getting more aid.

We attempt to model the relationship between conflict, aid, military spending, and tax to GDP ratio, inflation rate, regime instability and lower economic growth. Our hypothesis is that conflict in Pakistan is a function of aid, which increases military expenditure, and is a source of less responsive and weak government. We used tax to GDP ratio and inflation rate as proxy for government capacity. It has adversely affected GDP growth rate, infrastructure and human security. Aid appears as easy money that increases the scope of military interventions and escalation of conflict in Pakistan.

To determine the relationship between conflict and aid, we use conflict count data because variance of conflict has increased over time. Hence, the use of Poisson distribution, which is defined as one parameter distribution, specified by mean only. Its variance equals mean. There are days with more than one conflict and others with no conflict. If the mean number of conflicts is λ, then the probability of observing conflict per year is given by

when .

We can compute successive probability simply by multiplying with the mean and dividing by x. We also experimented with a negative binomial regression, which accounts for the over-dispersion in conflict count data. Count data are in integers, and the residuals for counts can only be normally distributed. If errors are Gaussian, then the usual time series modelling technique can be applied. If not, then there is need to write down conditional density function to proceed with identification and estimation (Brandt and Williams 2001).

We also check the time dependence of conflict issue in Pakistan. By inspection of mean and variance of conflict data, and by employing unit root tests, we can clearly say that conflict series is non-stationary.

4.1. The Data

The focus variable of this study is conflict in Pakistan, as coded in Uppsala Conflict Data Program (UCDP) armed conflict data-set, version 4-2011 (Themner and Wallensteen 2011) and Global Terrorism Database (GTD 2013). These data sets are not comparable. The UCDP data-set provides a longer series and measures intensity, occurrence and purpose of conflict. From this data-set, we use armed and non-armed conflict, that is interstate, religious, ethnic and external support conflict. It covers the period from 1961 to 2011.GTD data record terrorist incidents from 1970 to 2012. Before 1996, however, it was reported as not the most reliable source of information. This data-set gives comprehensive information on targets, weapons and number of killings.

Aid data were obtained from Aiddata.org compiled by Tierney et al. (2011). This data-set gives comprehensive information on donors, recipients, aid released and project type. However, there is no information on military aid. To overcome this limitation, information on ODA per capita and ODA in constant USD was taken from WDI (2014). Military expenditure series is from various SIPRI (2011) yearbooks. Data on GDP and exchange rates were taken from PBS (2014). Social sector spending data were taken from the Pakistan Economic Survey. Data on youth bulge which represents population above 15–64 years were obtained from WDI (2014), and data on unemployment rate were taken from various Labour Force Surveys of Pakistan with gaps filled from WDI (2014). Information on Consumer Price Index was obtained from International Financial Statistic (IFS 2014) and PBS (2014).

4.2. Descriptive Analysis

This subsection is devoted to an overall analysis of the relationship between conflict, ODA per capita, military expenditure and GDP growth.

Figure 1 shows the relationship between conflict and aid in Pakistan. With every episode of conflict, Pakistan received increased allocations of ODA. Simultaneously, the military expenditure also increased. It can be seen from Figure 1 how aid allocation suffered from volatility. Since 1993, conflict increased sharply, accompanied by increased ODA per capita. Another negative consequence of fluctuations in aid appears in volatile GDP growth. Fungibility of aid leads to declining defence expenditure, despite increasing conflict.

Figure 1 Conflict, foreign economic assistance and military expenditure in Pakistan

Diagram 1 The model structure

4.2.1. Aid distribution

Pakistan received increased allocations of ODA during 1960–2012. Initially, most aid allocation was from bilateral sources, with the multilateral organizations giving more aid in later years. As a percentage of GNI (Gross National Income), aid declined but the total amount in absolute and per capita terms has been increasing. From US$ 416 per capita from bilateral sources in the 1960s, it rose to US$ 785 per capita in 2011.The composition of aid has changed from predominantly bilateral to multilateral and IMF loans, US$ 2250 million and US$ 3324 million, respectively, in 2011. In 1958–1969, Pakistan received most aid from the United States, a bilateral source. It was almost 6.62% of the GNI. Multilateral and non-consortium sources provided only 0.87% of GNI. In 1969–1971, bilateral aid declined by half to 3.44% of the GNI. Bilateral aid as percent of GNI was further squeezed to 2.46% during 1971–1977, but it received a boost again during 1977–1988 to reach 3.87% of GNI. During 1988–1999, bilateral aid fell sharply to 1.08% of GNI. In the period 1999–2008, Pakistan received only 0.58% of GNI from bilateral sources for fighting the war on terror. The decline continued in the period 2008–2011, and the allocation stood at a mere 0.46% of GNI. Smaller bilateral aid means fewer grants and more loans in the total inflow.

4.2.2. Conflict distribution

During 1970–2012, Pakistan had 7157 incidents of conflict of varying degrees (GTD 2013). The country has spent almost 14 years in wars with India and the rest in dealing with ethnic and religious conflicts. Pakistan’s wars with India have led to conflict in the society. There are certain years when the state had to deal with more than one type of conflict. For example, as Pakistan faces the threat of the Taliban, insurgency in Balochistan has picked up at the same time. One is a religiously motivated conflict, and the other is an ethnic conflict. Both seek to capture resources. Balochis feel marginalized, and Taliban has ideological objectives.

5. Empirical analysis

To test the hypothesis that aid appears as easy money, changes societal expectations, makes people impatient and becomes a source of predatory government, we test how aid causes conflict at macro level. We use the conflict count as dependent variable and ODA per capita (odapercapi), inflation rate (cpi), social sector spending (sss), military expenditure (mexp), ODA at constant USD (odacons), total taxes as percentage of GDP (ttax), regimes (regimes dummy 1,0), youth bulge and unemployment rate as independent variables and GDP growth (gdpg) as exposure variable. Further, we use Poisson count model and negative binomial regression model for conflict which is the count variable. Negative binomial regression models are used because conflict is an over-dispersed count outcome variable. All the estimated models showed a nonzero alpha value which indicated over-dispersion in variance.

Table I displays the results of our macro model of how aid causes conflict. The estimates of negative binomial regression model and Poisson regression model using ODA per capita and ODA at constant USD are given at Appendix Tables I and II, respectively. To account for potential heterosecdasticy, we used robust standard errors.

Table I Conflict and Aid Relationship Based on GTD 2013 (Negative Binominal Regression)

	(1)	(2)	(3)	(4)	(5)
Variables	Conflict	Conflict	Conflict	Conflict	Conflict
Odapercapi	0.138^**	0.104^*	0.110	0.0439	0.167^***
	(0.0630)	(0.0563)	(0.0789)	(0.0495)	(0.0403)
Sss	3.889^***	3.128^***	4.291^***	2.952^***	2.354^***
	(0.709)	(0.754)	(0.626)	(0.528)	(0.603)
Cpi	0.0445
	(0.0424)
Mexp	−0.721^***	−0.622^***	−0.370
	(0.209)	(0.150)	(0.420)
Regimes		1.297^***		1.200^***	1.629^***
		(0.452)		(0.376)	(0.438)
Ttax			−0.317
			(0.370)
Youthbulge				0.313^***
				(0.0736)
Unemployment				0.376^***	0.669^***
				(0.105)	(0.144)
Constant	-1.533	-0.564	0.345	−21.82^***	−6.185^***
	(2.203)	(2.057)	(3.430)	(3.739)	(1.107)
Lnalpha	0.632^***	0.478^*	0.631^***	0.0173	0.324
	(0.216)	(0.244)	(0.215)	(0.291)	(0.265)
Observations	43	43	43	43	43

Note: Robust standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

On the basis of GTD data-set 2013, our results show that aid caused conflict in Pakistan during 1970–2011 along with youth bulge and unemployment. We further tried to test how conflict count varies along with social sector spending, military expenditure, taxes, unemployment rate and youth bulge. In model 1, we tested the hypothesis that due to increased military expenditure, there is a reduction in social sector spending and the government prefers inflation as a means to collect tax. Social sector spending and military expenditure are significant but give opposite sign, whereas inflation rate turns out insignificant. In model 2, we dropped inflation rate and introduced regime instability to test that it causes conflict along with ODA per capita, military expenditure and social sector spending. All the variables turned out significant. In model 3, total taxes were used to verify the fiscal capacity. With aid money available, the government prefers not to increase the tax burden. We dropped regime change and introduced taxes. Taxes turned out to be insignificant.

An interesting finding from models 1–3 is that total military expenditure as percent of GDP is significant but negatively related with conflict in Pakistan. Increased military expenditure can decrease conflict. In model 4, we tested the impact of youth bulge, unemployment, ODA per capita, social sector spending and regime change on conflict. Unemployment rate and youth bulge made ODA per capita insignificant. Social sector spending and regime change remained significant.

This made us to finalize model 5 which uses ODA per capita, social sector spending, and regime change and unemployment rate to explain conflict in Pakistan during 1970–2012. All these variables are significant and positively related with conflict. The log of alpha calculated for this model is positive which indicates that negative binomial model is different from Poisson regression model.

To test the conflict–aid relationship at macro level, we used UCDP’s armed conflict data-set, version 4-2011 (Themner and Wallensteen 2011). Table II displays the results. In this data-set, we used a longer series from 1961 to 2012 to overcome the problem of degrees of freedom in our earlier analysis. Data on armed conflict were used to test that conflict generates a spiral and results in intolerant behaviour which generates a series of internal conflict.

Table II Conflict and Aid Relationship Based on UCDP Armed Conflict Data Set, Version 4-2011 (Themner and Wallensteen 2011)

	(6)	(7)	(8)	(9)	(10)
Variables	Conflict	Conflict	Conflict	Conflict	Conflict
Mexp	−0.307^**	−0.359^***	−0.248
	(0.121)	(0.102)	(0.249)
Sss	0.799^***	0.754^**	1.007^***	0.637^*	0.590^*
	(0.294)	(0.343)	(0.390)	(0.345)	(0.307)
Odapercapi	0.0584	0.0462	0.0827^*	0.00170	0.00167
	(0.0475)	(0.0451)	(0.0499)	(0.0345)	(0.0344)
CPI	0.0844^***
	(0.0255)
Regimes		0.922^**		0.731^**	0.745^**
		(0.362)		(0.323)	(0.300)
Ttax			−0.00846
			(0.179)
Unemployment				−0.0198
				(0.0938)
Youthbulge				0.192^***	0.187^***
				(0.0426)	(0.0449)
Constant	−1.122	−0.343	−1.156	−11.70^***	−11.41^***
	(0.813)	(0.826)	(1.089)	(2.254)	(2.463)
Lnalpha	−2.139^*	−2.333	−1.546	−14.79^***	−16.96^***
	(1.160)	(2.180)	(1.772)	(1.551)	(0.401)
Observations	50	50	50	52	52

Note: Robust standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

We used negative binomial regression count model. In all models, log-transformed value of alpha is negative which shows that these are not over-dispersed models and their results are not different from Poisson regression model. In all these models, ODA per capita is insignificant but positively related with conflict. Inflation rate is positively related but insignificant. Social sector spending is positively and significantly related with conflict, whereas taxes are insignificant over the time period. In this data-set, conflict is positively and significantly related with social sector spending, regime change and youth bulge, whereas unemployment rate and ODA per capita are insignificant. These results are similar to the GTD data-set, but ODA per capita and unemployment rate are not significant. Taxes and military expenditure are again insignificant. Details are at Appendix Tables III and IV.

5.1. Micro Data

After testing the overall relationship between aid and conflict, we tried to test how project level aid is related with conflict. For this purpose, we used data from GTD (2013) and aiddata.org. Aid data give us information on all aid allocated for development. Table III explains the results of aid–conflict relationship in Pakistan. The dependent variable is conflict, enumerated from the conflict data count. We set the count data by year, weapon type, attack type and commitment of aid. Further, we assume that conflict is triggered by the purpose of aid and commitment. A certain type of aid and withdrawal of commitment can trigger conflict. We used Poisson distribution because the conflict data and aid data are positively skewed. The mean for conflict count is 363, and standard deviation is 1154. The mean for aid data is 195, and standard deviation is 356.

Table III Micro Data Analysis (GTD 2013 and Aiddata)

	(11)	(12)	(13)	(14)	(15)
Variables	crit1	attacktype1	weaptype1	Target	Nkill
Purpose	2.27e-07	1.95e-07	−1.20e-07	−3.04e-07^**	−8.90e-07^***
	(4.81e-07)	(2.78e-07)	(1.90e-07)	(1.48e-07)	(3.35e-07)
	−0	−2.26e-10^**	7.47e-11	9.86e-11^**	−5.15e-10^***
	(1.52e-10)	(1.09e-10)	(5.28e-11)	(0)	(1.57e-10)
Constant	−0.0338	1.043***	1.847***	2.356***	0.808***
	(0.0336)	(0.0195)	(0.0132)	(0.0103)	(0.0230)
Observations	7133	7133	7133	7133	7056

Note: Standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

In model 11, we used the criteria whether an act is conflict or not and regressed it on purpose and commitment of aid. All were positively related with conflict but insignificantly. We used integer variable attack type in model 12 as dependent variable and again regressed on purpose and aid commitment. The Poisson regression results were negatively related with attack type. Model 13 was used to find the relationship between weapon type with purpose and commitment. Our results show that purpose of aid negatively affects the selection of weapon and are positively related with donor’s aid commitment. In model 14, we focus on the relationship between target type with purpose and commitment. Target selection is positively related with aid commitment and negatively related with purpose of aid. Model 15 uses number of killings in a conflict with purpose and commitment of aid. All these are negatively related but are highly significant. On the basis of our results, we can say that commitment of aid amount is a highly crucial variable in the case of conflict. Most of the conflict events are significantly correlated with it.

Next, we regressed conflict as a dependent variable on various exposure variables. In model 1a, we used purpose of aid as an exposure variable and regressed conflict count on donor and commitment of aid. Aid commitment is positively related and significant with conflict count. This variable allowed us to find the purposes which are positively related with the occurrence of conflict in Pakistan. When a donor committed aid for education policy and administration management (code1110 of GTD data), it caused conflict as it is positively related with conflict count. However, the relationship is negative when aid is given for education facilities and training (code11120), or for teacher training (code 11,130), early childhood education (code 11,240), advanced technical and managerial training (code 11,430), medical education/training (code 12,181), health personnel development (code 12,281), population policy and administration management (code 13,010), family planning (code13030), standards controls and HIV/AIDS (code 13,040), water resource administration and management, water supply and sanitation (code 14,010, 14,015, 14,020), river development (code 14,040), human rights (code 15,162), women’s equality organizations and institutions (code15164), conflict prevention and security system management and reform (code 15,205, 15,210). (Details are at Appendix Table V.)

6. Conclusion

Our aim in this study was to explain how aid causes conflict in Pakistan. Aggregative-level and project-level evidence is combined for the first time to show that aid may not reduce domestic conflict. Aid appears as easy money, weakening the government resolve to tax and reform. At the micro level, conflict is strongly correlated with donor’s aid commitment and purpose in all of our models. Importantly, we find that conflict is strongly related with the purpose of aid and aid commitment. Further, aid related to education, teacher training, women and human rights, water and sanitation is negatively related with conflict. Thus, aid consequences vary with commitment and purpose. Pakistan is a typical case of high volatility of aid allocations and frequent changes in donors’ aid commitment. At aggregate level, we conclude that aid causes conflict. It appears as easy money that changes expectations of the people. The government becomes less responsive to the local needs at a time when people expect more. Social sector spending is positively related with conflict in our model, a result consistent with Deger and Somnath (1991) and Murshed and Sen (1995).

Our study confirms the earlier study of Fearon and Laitin (2003) and Findley et al. (2011) that ethnicity and population diversity explain conflict in Pakistan. Our empirical results show that youth bulge and unemployment rate are also positively related with conflict. It calls for more spending on social sector which creates employment opportunities. This result is consistent with Iyengar, Monten, and Hanson (2011). Another important finding is that regime instability, measured as a dummy variable with zero indicating dictatorship and 1 indicating democracy, is positively related with conflict in Pakistan. Military expenditure in our model is negatively related with conflict. This is contrary to the intuitive arguments that military expenditure increases as a result of easy aid money. A partial explanation may be that Pakistan’s defence spending has been declining as a percentage of GDP as well as government expenditure since 1990 but domestic conflict has increased expenditure on internal security and the justice system. More aid with lower growth, decreasing defence expenditure and regime instability cause conflict. A limitation of the study is that aid database does not provide information on military aid. The variable, therefore, could not be included in our micro models. Another limitation is that we have not explicitly measured the degree of fungibility of aid. On the basis of our data, however, it is possible to infer that social sector spending in Pakistan has not increased in the presence of aid, but conflict has.

Acknowledgments

The study has benefited from the useful comments and constructive suggestions by anonymous referees. The usual disclaimer applies. The research was completed while the author was a Visiting Senior Lecturer at the Department of Political Economy, King's College London in June–July, 2014. Useful comments by Dr Humeira Iqtidar and Dr Pervez Tahir are gratefully acknowledged.

Notes

¹ ‘to use or threat to use violence against non-combatants in order to achieve political and social objectives’ (Enders, Sandler, and Gaibulloev 2011, 321).

Appendix

^{Table I}

Table I Conflict and Aid Relationship (Excluding Non-conflict Years of GTD Data-set)

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	Conflict	Conflict	Conflict	Conflict	Conflict	Conflict
Odapercapi	0.176^***	0.180^***	0.172^***	0.114^*	0.109^**	0.120^**
	(0.0576)	(0.0577)	(0.0595)	(0.0636)	(0.0556)	(0.0517)
Sss	1.696^***	1.723^***	1.646^***	0.769^**	0.671^*
	(0.420)	(0.396)	(0.385)	(0.357)	(0.343)
Cpi	−0.0303
	(0.0564)
Mexp	−0.728^***	−0.625^***	−0.497^*	0.185
	(0.188)	(0.144)	(0.289)	(0.130)
Regimes		−0.500			0.405
		(0.448)			(0.406)
Ttax			−0.210
			(0.228)
Youthbulge				0.511^***	0.482^***	0.445^***
				(0.103)	(0.0611)	(0.0888)
Unemploy						0.129^*
						(0.0768)
Constant	3.680^**	3.020^**	5.024^**	−26.44^***	−23.83^***	−21.04^***
	(1.559)	(1.377)	(2.068)	(5.535)	(2.891)	(4.394)
Observations	38	38	38	38	38	38

Note: Robust standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

^{Table II}

Table II Conflict and Aid Relationship (Excluding Non-conflict Years of GTD Data-set)

	−1	−2	−3	−4	−5	−6	−7	−8
Variables	Conflict	Conflict	Conflict	Conflict	Conflict	Conflict	Conflict	Conflict
Odacons	6.20e-10^**	6.25e-10^**	5.71e-10^*	3.17E-10	3.38E-10	3.08E-10	2.10E-10	2.69E-10
	−2.78E-10	−2.81E-10	−2.93E-10	−2.48E-10	−2.41E-10	−2.23E-10	−1.69E-10	−1.65E-10
Sss	2.329^***	2.340^***	2.289^***	0.983^**	1.098^**	0.892^**
	−0.483	−0.442	−0.433	−0.439	−0.433	−0.38
Cpi	−0.0224
	−0.0536
Mexp	−0.755^***	−0.678^***	−0.528^*	0.207
	−0.188	−0.144	−0.273	−0.129
Regimes		−0.394				0.503	0.76
		−0.475				−0.43	−0.464
Ttax			−0.208
			−0.245
Youthbulge				0.546^***	0.451^***	0.516^***	0.600^***	0.486^***
				−0.102	−0.0821	−0.0541	−0.0416	−0.0864
Unemploy								0.126^*
								−0.0722
Constant	2.897	2.457	4.339^*	−28.43^***	−22.49^***	−25.72^***	−28.37^***	−22.66^***
	−1.857	−1.647	−2.58	−5.4	−3.7	−2.498	−2.295	−4.481
Observations	38	38	38	38	38	38	38	38

Notes: ODA at constant USD is used instead of ODA per capita; Robust standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

^{Table III}

Table III Conflict and Aid Relationship (UCDP Armed Conflict Data-set, Version 4-2011 (Themner and Wallensteen 2011) Including Wars

	(1)	(2)	(3)	(4)	(5)
Variables	Conf	Conf	Conf	Conf	Conf
Odacons	2.53e-10	2.29e-10	3.19e-10	1.89e-10	1.20e-10
	(2.72e-10)	(2.36e-10)	(2.79e-10)	(2.83e-10)	(2.19e-10)
Sss	1.071^***	1.047^**	1.557^***	0.935^**	0.717^*
	(0.296)	(0.454)	(0.392)	(0.428)	(0.412)
Cpi	0.0815^***
	(0.0235)
Mexp	−0.283^**	−0.360^***	−0.206	0.0964
	(0.116)	(0.101)	(0.233)	(0.196)
Regimes		0.925^***			0.780^**
		(0.332)			(0.312)
Ttax			−0.0777
			(0.170)
Youthbulge				0.249^***	0.191^***
				(0.0955)	(0.0381)
Unemployment					0.0107
					(0.0853)
Constant	−1.763	−1.005	−1.540	−15.98^***	−12.23^***
	(1.144)	(1.172)	(1.483)	(5.409)	(2.208)
Observations	50	50	50	50	51

Notes: ODA at constant USD is used instead of ODA per capita; Robust standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

^{Table IV}

Table IV Conflict and Aid Relationship (UCDP Armed Conflict Data-set, Version 4-2011 (Themner and Wallensteen 2011) Including Wars

Dependent variable	Conf	Conf	Conf
Cpi	0.068616 (0.029704)^**	0.082566 (0.050488)^*	0.063073 (0.040349)^*
gdpg	−0.052728 (0.072106)	−0.098837 (0.103028)
mexp	−0.496144 (0.210655)^**	−0.414890 (0.179291)^**
regimes	0.301764 (0.430212)	0.346783 (0.697694)	0.434017 (0.446604)
sss	0.811851 (0.476155)^*	1.202264 (0.698051)^*	0.855307 (0.373922)^**
tfea	5.843862 (7.873624)	5.984411 (12.81765)
ttax	0.109335 (0.156133)
c	−0.969560 (1.224694)	−0.755120 (1.939156)	−2.084668 (0.800138)^***

Notes: Total Foreign Economic Assistance (TFEA) is used instead of ODA per capita; Robust standard errors in parentheses.

*** p < 0.01

** p < 0.05

* p < 0.1.

^{Table V}

Table V Poisson Regression Model with Purpose of Aid as a Rate Variable

Purpose Code	Conflict	Standard Error	Purpose Code	Conflict	Standard Error	Purpose Code	Conflict	Standard Error
11,000	−0.265^***	−0.0258	16,081	−0.262^***	−0.0258	32,110	−0.188^***	−0.0136
11,100	1.681^***	−0.0258	21,000	−0.235^***	−0.0204	32,120	0.0614^***	−0.0132
11,110	0.0133	−0.0131	21,005	−0.0628^***	−0.0183	32,130	0.0999^***	−0.0135
11,120	−0.036^***	−0.0132	21,010	−0.0681^***	−0.0147	32,140	0.0943^***	−0.0171
11,130	−0.0349^**	−0.0137	21,020	−0.0176	−0.0132	32,161	0.399^***	−0.0146
11,182	0.409^***	−0.0155	21,030	−0.135^***	−0.0134	32,162	0.0401^**	−0.0183
11,220	0.00878	−0.0131	21,040	0.233^***	−0.0137	32,163	0.443^***	−0.0139
11,230	0.0154	−0.0135	21,050	0.0472^***	−0.0148	32,164	0.442^***	−0.0149
11,240	−0.0354^**	−0.0138	21,061	1.271^***	−0.0259	32,165	−0.272^***	−0.0152
11,320	0.130^***	−0.014	22,000	0.290^***	−0.0259	32,166	−0.163^***	−0.0149
11,330	0.220^***	−0.0134	22,010	0.487^***	−0.0142	32,167	0.580^***	−0.0171
11,420	0.0688^***	−0.013	22,020	−0.143^***	−0.0133	32,168	0.136^***	−0.0183
11,430	−0.069^***	−0.0135	22,030	−0.0148	−0.014	32,169	−0.0113	−0.0158
12,000	0.169^***	−0.0155	22,040	−0.0264^*	−0.0139	32,171	−0.338^***	−0.0141
12,005	−0.274^***	−0.0259	23,000	−0.0569^***	−0.0142	32,172	−0.176^***	−0.0171
12,110	−0.0152	−0.0133	23,005	−0.248^***	−0.0158	32,182	0.00763	−0.0143
12,181	−0.043^***	−0.0139	23,010	−0.077^***	−0.0132	32,191	−0.172^***	−0.0138
12,182	0.198^***	−0.0171	23,020	−0.067^***	−0.0136	32,210	0.0243^*	−0.0146
12,191	0.125^***	−0.0134	23,030	−0.00304	−0.0136	32,220	−0.017	−0.0136
12,220	0.0957^***	−0.0131	23,040	0.0122	−0.0132	32,261	0.289^***	−0.0205
12,230	0.132^***	−0.0134	23,050	−0.150^***	−0.0148	32,262	−0.249^***	−0.0145
12,240	0.224^***	−0.0134	23,055	−0.145^***	−0.0183	32,265	−0.268^***	−0.0259
12,250	0.0396^***	−0.0131	23,061	−0.115^***	−0.0205	32,310	0.109^***	−0.0152
12,261	0.302^***	−0.0142	23,062	0.519^***	−0.0183	33,105	1.683^***	−0.0258
12,262	0.530^***	−0.0183	23,064	−0.585^***	−0.026	33,110	0.245^***	−0.0139
12,263	0.0494^***	−0.0171	23,065	0.0457^***	−0.0135	33,120	0.449^***	−0.0158
12,281	−0.269^***	−0.0148	23,066	1.267^***	−0.0259	33,130	2.374^***	−0.0258
13,000	−0.274^***	−0.0259	23,067	−0.285^***	−0.0171	33,140	0.00872	−0.0183
13,005	1.675^***	−0.0259	23,068	−0.264^***	−0.0258	33,181	0.872^***	−0.0204
13,010	−0.0249^*	−0.0132	23,070	0.291^***	−0.0204	33,210	0.154^***	−0.0147
13,030	−0.0227^*	−0.0132	23,082	−0.501^***	−0.0171	41,005	0.359^***	−0.0204
13,040	−0.071^***	−0.0131	24,000	−0.274^***	−0.0259	41,020	−0.0872^***	−0.0171
13,081	1.683^***	−0.0258	24,005	−0.460^***	−0.0204	41,030	0.159^***	−0.0143
14,000	−0.267^***	−0.0258	24,010	0.0791^***	−0.0133	41,040	−0.212^***	−0.0146
14,005	0.982^***	−0.0259	24,020	0.361^***	−0.0204	41,050	−0.424^***	−0.0183
14,010	−0.046^***	−0.0135	24,030	0.252^***	−0.0133	41,081	−0.185^***	−0.0145
14,015	−0.264^***	−0.0258	24,040	−0.157^***	−0.0143	41,082	0.548^***	−0.0164
14,020	−0.115^***	−0.0132	24,081	−0.264^***	−0.0258	42,010	0.228^***	−0.0134
14,030	0.0102	−0.0133	25,010	0.177^***	−0.0133	43,010	0.0594^***	−0.013
14,040	−0.057^***	−0.0135	25,020	−0.335^***	−0.0258	43,030	−0.037^***	−0.0136
14,050	0.149^***	−0.0145	31,100	−0.274^***	−0.0205	43,040	0.00927	−0.0131
14,081	0.910^***	−0.0183	31,105	0.539^***	−0.0171	43,081	0.211^***	−0.0136
15,100	0.190^***	−0.014	31,110	0.0118	−0.0133	43,082	−0.117^***	−0.0139
15,105	0.0331^**	−0.0152	31,120	−0.048^***	−0.0133	51,010	−0.067^***	−0.0133
15,110	0.108^***	−0.013	31,130	0.0179	−0.0133	52,010	0.0596^***	−0.013
15,120	0.0221	−0.0135	31,140	0.00127	−0.0131	53,030	−0.180^***	−0.0134
15,130	0.0526^***	−0.0133	31,150	0.0218	−0.0133	53,040	−0.152^***	−0.0133
15,140	0.0752^***	−0.0131	31,161	−0.124^***	−0.0137	53,050	2.372^***	−0.0258
15,150	0.00872	−0.013	31,162	0.427^***	−0.0152	60,010	0.243^***	−0.014
15,151	0.178^***	−0.0141	31,163	0.0294^**	−0.0141	60,020	0.0703^***	−0.0133
15,153	0.367^***	−0.0146	31,164	−0.267^***	−0.0258	60,040	0.0133	−0.0133
15,160	0.0184	−0.0133	31,165	0.176^***	−0.0258	60,061	−0.264^***	−0.0258
15,162	−0.272^***	−0.0259	31,166	−0.274^***	−0.0259	70,000	−0.00893	−0.0147
15,164	−0.405^***	−0.0155	31,181	−0.039^***	−0.0145	72,000	−0.049^***	−0.0134
15,170	0.0456^***	−0.0136	31,182	0.174^***	−0.0138	72,010	−0.0089	−0.013
15,205	−0.110^***	−0.0204	31,191	−0.0299^**	−0.0136	72,020	0.0235^*	−0.0132
15,210	−0.0569^***	−0.0152	31,192	0.418^***	−0.0183	72,030	0.0587^***	−0.0155
15,220	0.872^***	−0.0152	31,193	−0.031	−0.0205	72,040	0.119^***	−0.0131
15,230	0.607^***	−0.0155	31,194	0.169^***	−0.0183	72,050	0.0803^***	−0.0132
15,240	−0.428^***	−0.0204	31,195	−0.0255^*	−0.0155	73,010	0.203^***	−0.0132
16,010	0.0306^**	−0.0131	31,205	−0.364^***	−0.0183	74,010	−0.036^***	−0.0135
16,020	0.203^***	−0.0133	31,210	−0.166^***	−0.0146	91,010	0.0754^***	−0.0135
16,030	0.0306^**	−0.014	31,220	−0.115^***	−0.0138	92,005	0.222^***	−0.0142
16,040	0.230^***	−0.0205	31,282	0.296^***	−0.0258	92,010	−0.117^***	−0.0146
16,050	0.0530^***	−0.0132	31,291	−0.133^***	−0.0171	92,020	0.418^***	−0.0143
16,061	0.0328^**	−0.0131	31,300	−0.276^***	−0.0259	92,030	0.207^***	−0.0146
16,062	0.149^***	−0.0135	31,310	−0.439^***	−0.0183	93,010	0.0202	−0.0133
16,063	−0.072^***	−0.0137	31,320	−0.245^***	−0.0148	99,810	0.112^***	−0.013
16,064	0.255^***	−0.0183	32,105	0.367^***	−0.0183	99,820	−0.013	−0.0149
Observations	7146					Constant	5.227^***	−0.013

Note: Italic value shows positive and significant relation with conflict, bold italic value shows positive but insignificant relation with conflict and bold value shows negative and significant relation with conflict.

*** p < 0.01

** p < 0.05

* p < 0.1.