Mediating Role of Institutional Quality on the Relationship between Political Violence and Inward Foreign Direct Investment

(1)

Mediating Role of Institutional Quality on the

Relationship between Political Violence and Inward

Foreign Direct Investment

Master Thesis

(2)

(3)

iii Master Thesis:

Mediating Role of Institutional Quality on the Relationship between Political Violence and Inward Foreign Direct Investment

Groningen, 2nd_{of December 2019} Word Count: 13.218

Student Information

Name: Siham Hari

Groningen Student Number: S3913481 Newcastle Student Number: B8060092

Groningen Email: s.hari@student.rug.nl

Newcastle Email: s.hari2@newcastle.ac.uk

MSc International Business Management (University of Groningen)

MSc Advanced International Business Management (Newcastle University) Supervisors

Dr. Mariko Klasing Dr. Tom McGovern University of Groningen Newcastle University

m.j.klasing@rug.nl tom.mcgovern@newcastle.ac.uk

Universities

University of Groningen

Faculty of Economics and Business Nettelbosje 2, 9747 AE

Groningen, Netherlands Newcastle University Business School (NUBS) 5 Barrack Road

(4)

iv

Abstract

The primary aim of the present study is to assess the relationship between political violence on foreign direct investments and the mediating role of institutional quality on this effect. By providing a Cross-Sectional Time-Series analysis of 20 countries affected by the revolutions of the Arab Spring crises, the findings display a negative effect of political violence on FDI and mixed results on the effect of institutional quality on FDI. The mediating role of

institutional quality on the relationship between political violence and foreign direct

investment is confirmed, although this effect is not found to be statistically significant in the selected sample.

_______________________________________________________________________

Keywords: Political Violence, Institutional Quality, Foreign Direct Investment, Cross-Section

(5)

v

LIST OF FIGURES

FIGURE 1 – Conceptual Framework P.12 FIGURE 2 – Distribution of FDI across the time span for each country P.23

LIST OF TABLES

(6)

vi

1 Introduction

During the late 2010, poor living conditions made the auto-immolation of a vegetable vendor from a small Tunisian town, the pretext of the start of one of the biggest crises in the annals of Arab history (Kumaraswamy, 2019). The countries constituting the Middle East and North African (MENA) region exposed for years their citizens to a prolonged denial of economic and political rights. While oil-rich Arab countries silenced their citizen with wealth, the others intimidated them with the external threat of the conflict with Israel (Kumaraswamy, 2019).

The power and control of the rulers over the populations provoked a common deterioration of the socio-economic conditions of the Arabs. This situation could not have lasted much longer as the globalization and the technological advancements gave the possibility to everyone to be aware of the reality and to fight for their personal expectations. Fostered by the usage of social media that characterizes the current century, this national revolution spread from Tunisia rapidly all around the region affecting the surrounding Arab republics and monarchies. Even if the involvement in the crisis includes almost every country in the region, the level of participation and the mode of protests differed enormously from country to country.

(9)

2 of 70

creates new job opportunities is strictly connected with FDI and it is essential for policy makers to understand which factors attract FDI flows in order to stimulate economic growth (Abdel-Latif and Ouattara, 2017).

As a consequence, this study aims at studying the relationship between political violence and FDI inflows and the mediating effect of institutional quality in the context of the Arab Spring revolutions. The effect of the Arab Spring crises on the attraction of inward investment has already been addressed in the prior literature (Mtar and Bannour, 2017; Abdel-Latif and Ouattara, 2017; Aziz, 2018). However, to the best of my knowledge, the present research represents the first study that attempts to analyse the mediating role of the quality of the institutional environment affecting the direct relationship between political violence and FDI. The governments of the countries that responded adopting reforming policies as a response to the crisis, are expected theoretically and empirically, to attract more FDI. Addressing this gap in the literature is particularly significant in light of the divergent outcomes that this situation of political instability and violence provoked in the affected countries, which demonstrated how effective responses can impact positively the economic growth of a country, allowing it to successfully overcome a period of crisis.

Hereinafter, the following analysis presents a Cross-Section Time-Series analysis of 20 countries (Algeria, Bahrain, Egypt, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Palestine, Qatar, Saudi Arabia, Sudan, Syria, Tunisia, Turkey, United Arab Emirates and Yemen) from the years 2005 to 2018 and tries to answer the following research question:

(10)

3 of 70

2 Literature Review

2.1 Foreign Direct Investment

(11)

4 of 70

manufacturing and the tourism industry. Furthermore, FDI in Jordan and Tunisia is found to be mainly efficiency-seeking (export opportunities and lower production costs) while in Egypt is found to have more market-seeking reasoning (exploiting a larger market and tariff-jumping).

The theoretical and empirical literature about FDI has identified several factors which determine the investments. When selecting foreign locations, decision makers take into account the foreign market structure, the market size and its growth (Goswami and Haider, 2014), political, economic and institutional stability (Brunetti and Weder, 1998), labour costs, exchange rate, management skills, and innovative product technologies (Busse and Hefeker, 2007), human capital and the business and financial environment (Jude and Levieuge, 2015). Alongside with being a great generator of employment, FDI is proven to increase productivity, technology transfer, competitiveness and access to heterogeneous markets (Denisia, 2010). As a result, FDI is recognised to contribute importantly to economic growth through tax revenue and foreign exchange (Smith, 1997; Quazi, 2007). This implication has been the focal driver of policy makers and citizens in developing countries that struggle to foster and advance their economies. An improvement in the institutions that support local and foreign private investment are necessary in order to recover from the weak economy of these heavily indebted developing countries (Rodrik, 1991).

(12)

5 of 70

become transnational: the advantage of ownership (O) of tangible and intangible assets, the advantage of internationalization (I) and the hierarchical control over cross-border activities and the advantage of location (L) and the resources of the host country (Dunning, 1988; 1993). In conclusion, FDI is proven to entail profit and benefits both for expanding companies and for receiving countries.

(13)

6 of 70

2.2 Political Violence

The Arab Spring crises is characterized by an overall climate of political uncertainty and its most extreme form: political violence. In some countries, civilians acted with peaceful demonstrations, whereas in others the manifestations were violent, even turning to civil wars. Foreign companies witnessing these violent wave are able to assess the entailing risks and repercussions on their investment and performance. These events can bring a desirable investment location to be a less attractive investment site, deterring possible future investments and potentially causing pre-emptive divestment (Li, 2006).

(14)

7 of 70

organised activity to achieve political goals. This would apply to activities in which organisers initiate the use of violence, as well as peaceful or violent ones that a similar organised group (often the State or rival groups) reacts to by violent means, such as peaceful demonstrations. It also covers various types of military coups, for while military putsches generally involve the use of violence and killings, there are instances where they are described as ‘bloodless’. But this qualification means that their nature is to be ‘bloody’, thus implying potential violence with the act.” (UNDP, 2012)

(15)

8 of 70

phenomena is easy to argue and it is shared by the great majority of the literature on this topic (Jong A Pin, 2008; Khan and Akbar, 2013).

In the literature, political violence is studied as a subcategory of its more generic form, political uncertainty. Tosun et al. (2008) classifies the concept of political instability in three main divisions: weak government, myopia and polarization and social unrest. The first view refers to political tensions and uncertainty about the survival of a government, which is mostly common in countries where political parties mostly represent the interest of its supporters. Myopia and polarization is associated to the frequency of government changes that provoke inconsistent policies. Finally, social unrest is linked to socio-political tensions and recorded violent acts, such as strikes, riots or socio-politically motivated killings (Tosun et al., 2008). In the previous literature, this three views are either discussed jointly as a single entity or, like in this research with political violence, faced separately. In each case, the theoretical demonstration of the negative effect of political instability on economic growth holds for most studies (Levine and Renelts, 1992; Alesina and Perotti, 1996; Gupta et al., 1998; Jun and Singh, 1999; Asteriou and Price, 2001; Busse and Hefeker, 2007).

(16)

9 of 70

the host country an enormous regulatory burden and lack of commitment and protection from the local institutions (Daude and Stein, 2007; Dixit, 2011), excessive liquidity in the financial sector (Fielding and Shortland, 2005), growth of disputes in which domestic firms find themselves in a privileged negotiating point being familiar with the local institutions (Bhattacharya et al, 2011). It also causes massive killings and migration, social cleavages, economic recession and the damage, loss, destruction and disappearance of tangible assets and infrastructures (Li, 2006). All these factors are considered to be inward investment friendly (Kruja and Dragusha, 2014).

Based on prior research and on the international production theory which states that the propensity of a company to invest abroad is connected to the resource implications and advantages of locating in a foreign country compared to their home country (Dunning, 1980; Fayerweather, 1982), it is concluded that political instable and violent environments would only result in a competitive disadvantage for foreign multinationals. Thus, the first hypothesis states:

(17)

10 of 70

2.3 Institutional Quality

North (1991) affirms that “good institutions” establish an incentive structure that promotes efficiency and reduces uncertainty, hence causing better economic performance. National growth is highly associated and stimulated by the quality of institutions (North 1990) and represents one of the main driving forces of this crisis. The association between the role and quality of institutions and FDI flows has received commensurate attention in the literature (Buchanan et al., 2012; Kurul and Yalta, 2017; Peres et al., 2018). Several factors are found to have an actual influence on the quality of the institutions, and all of them constitute FDI determinants. Among these, good governance and economic freedom are increasingly becoming more important determinants of FDI for multinational companies that are shifting their attention from resource seeking to efficiency seeking objectives (Dunning, 2002). The presence of corruption or bad governance influence greatly the decision of an MNE to invest or not in a country and it is proven to decrease FDI flows (Wei, 2000). Daude and Stein (2007) claim that corruption increases uncertainty and Julio and Yook (2016) confirm that FDI is more likely in countries with relatively political stability and higher controls on corruption. Jude and Levieuge (2015) found that lower institutions quality is correlated to fewer investments which in turn entail lower productivity growth, lower per capita income and more generally slower output growth. Aziz (2018) confirmed that good institutions reduce transaction costs and improve productivity, which stimulates foreign investments. Furthermore, the author adds the importance of government’s effectiveness in enforcing and protecting intellectual property rights, that affect the degree to which foreign companies increase investments in a host country.

(18)

11 of 70

Krugell, 2007). The Middle East and North Africa region has been also considered: Mina (2012) found that religion in politics, internal and external conflicts and ethnic tensions affected FDI and that establishing bilateral treaties and government stability increased FDI. Helmy (2013) on its study of the determinants of FDI after the Arab Spring crises found that freedom and security of investments impacts positively on inward investments while Aziz and Mishra (2016) observe that FDI is mostly attracted by a higher degree of privatization, trade liberalization and an educated labour force. Higher political quality is found to increase FDI also in the study of Abdel-Latif and Ouattara (2017) on political shocks of the Arab Spring and their effects on inward investments.

Based on the FDI theory, institutional quality enables investors to exploit the resources of the host country and therefore gain competitive advantage. As a conclusion, the second hypothesis affirms:

H2: Better institutional quality in the host country positively influences inward foreign direct investments.

(19)

12 of 70

Spring crises were able to implement economic, political and social reforms and enhance their institutional environment and other countries subject to this political violence culminated in civil wars (Amin et al., 2012), the third and fourth hypotheses assert:

H3a: When political violence leads to an improvement of the institutions, the negative relationship between political violence and FDI is positively mediated by institutional quality.

H3b: When political violence causes a worsening of the institutional environment, the negative relationship between political violence and FDI is aggravated.

2.4 Conceptual Framework

The conceptual framework illustrated in the figure below exhibits the proposed interconnections.

FIGURE 1

(20)

13 of 70

3 Methodology

In the following section I describe the intended empirical measures to adopt in order to study the presented hypotheses about the relationship between political uncertainty and FDI inflows and how this relationship is mediated by the presence of institutional quality. Some of the proposed variables have been inspired from the literature and others have been directly calculated from the referential datasets that were regarded as relevant. All the measures added to the analysis have been considered to be the most pertinent in order to efficiently answer the initial research question.

3.1 Data

Quantitative analysis will be performed adopting secondary data on the country-level, collected from several different datasets which have been found all available online.

3.2 Sample

(21)

14 of 70

3.3 Dependent Variable – FDI Inflows

The variable of interest Fdiwdi has been derived from the “Wold’s Bank World Development Indicators” database. This database has been adopted for the same purpose in the previous studies of Abdel-Latif and Ouattara (2017) and Aziz (2018). For robustness reasons, the analysis will be repeated using the “UCTAD annual FDI inflows” data such as Daude and Stein (2007) represented by the variable named Fdiunctad. The variable Fdiwdi covers all the countries included in the sample and most of the timeframe, apart from Syria 2008-2018 and Libya the year 2011 and the years 2014-2018. The variable Fdiunctad presents much more missing data: Libya 2011, 2014-2018; Palestine 2005-2018; Sudan 2005-2011 and Syria 2001-2018.

3.4 Independent Variable – Political Violence

In order to assess political violence, an extreme form of political instability, there have been used and measured several variables: Csp, Acled, Ucdpdeaths, Ucdpconflicts. 3.4.1 Csp

(22)

15 of 70

and ethnic warfare involving that state in that year. This variable lacks data for Palestine and for Sudan between the years 2011 and 2018.

3.4.2 Acled

The variable Acled stems from “The Armed Conflict Location & Event Data Project”(ACLED). The original dataset provides information on six types of events belonging to three different categories: violent events (battles, explosions/remote violence, and violence against civilians), demonstration events (protests and riots) and non-violent actions (strategic developments). In this research, a 3-points score has been assigned to the events included in the first category, a 2-points score for the events of the second category and a 1-point score for the last category. Consequently, all the scores for each year and country have been summed and represent the final score of the Acled variable. Hereinafter, the table below provides a description of the different events outlined in the ACLED dataset. The data for Acled is covering the whole timeframe for North African countries (Algeria, Egypt, Libya, Morocco, Sudan and Tunisia) while for Middle Eastern countries availability of data is limited only to more recent years: Saudi Arabia and Yemen are covered for the years 2015-2018; Bahrain, Iraq, Israel, Jordan, Kuwait, Lebanon, Palestine and Turkey for the years 2016-2018; Syria and United Arab Emirates for the years 2017-2018; Oman for 2016 and 2018 and Qatar only for 2017.

TABLE 1

Acled Events Descriptions.

General Event Type Sub-Event Type

Violent Events Battles – violent

interactions between two organised armed groups

 Armed clash  Government regains territory  Non-state actor overtakes territory Explosions/Remote violence – one-sided violence events in which the tool for engaging in

(23)

16 of 70 conflict creates asymmetry by taking away the ability of the target to respond

 Shelling/artillery/missile attack

 Remote explosive / landmine / IED  Grenade Violence against civilians

– violent events as organised armed group deliberately inflicts violence upon unarmed non-combatants

 Sexual violence  Attack

 Abduction / forced disappearance

Demonstrations Protests – a public demonstration against a political entity, government institutions, policy or group in which the participants are not violent

 Peaceful protest

 Protest with intervention  Excessive force against

protesters

Riots – violent events where demonstrators or mobs engage in disruptive acts or disorganised acts of violence against property or people

 Violent demonstration  Mob violence

Non-violent actions Strategic developments – often non-violent activity by conflict and other agents within the context of the war/dispute. Recruitment, looting and arrests are included

 Agreement  Arrests

 Change to group / activity

 Disrupted weapons use  Headquarters or base established  Looting / property destruction  Non-violent transfer of territory  Other

3.4.3 Ucdpdeaths and Ucdpconflicts

(24)

17 of 70

against civilians, resulting in at least 1 direct death at a specific location and a specific date”. The dataset provides three types of organised violence (state-based conflict, non-state conflict and one-sided violence) and three different estimates of deaths for each event identified in the source material: the low estimate represents the most conservative number, the best estimate contains the most reliable number and the high estimate consists of the highest reliable estimate. The observations are not reported at the country level, instead the dataset has recorded the number of deaths in a given geo-referenced location and year associated with a particular conflict. Thus, two constructed measures have been adopted in this research in order to calculate political violence: the number of deaths (showing the intensity of the violence) and the number of conflicts (displaying the extensive margin of violence).

The first variable, Ucdpdeaths, reflects the total number of deaths in a country in a given year. In order to estimate this measure, I only considered the “best” column represented in the dataset. As previously mentioned, this score illustrates the most likely (best) estimate of total fatalities resulting from one event and it is the sum of the best estimate of deaths sustained by side a and side b of a conflict plus the best estimate of dead civilians and of persons of unknown status. In conclusion, Ucdpdeaths displays the total sum across all conflicts within a given country, year and type of violence from the best estimate scores represented in the original dataset.

The second variable, Ucdpconflicts, represents the number of different conflicts within a given country and year calculated separately for each type of violence. Firstly, I have created an indicator variable for each first observation of a conflict in order to end up with only one line for each conflict and year and finally I have summed all the resulting scores for each year and country with Stata.

(25)

18 of 70

data for the year 2007; Lebanon for 2007, 2009-2011 and for the Ucdpdeaths score both for the years 2005 and 2018; Libya for 2005, 2007, 2009-2010; Saudi Arabia for 2008-2014 and Yemen for 2005-2008. Bahrain was covered only for the year 2011, Jordan only for the years 2005 and 2016, Kuwait only for 2015, Tunisia only for the years 2015-2018 and United Arab Emirates only for 2010.

3.5 Mediator – Institutional Quality

The mediator has been calculated employing different measures, some were inspired by other studies and others were determined first-hand from the relevant sources. The four variables that have been used for the sake of estimating institutional quality in this research are: Wgi, Icrg, Edb9 and Edb5.

3.5.1 Wgi

The variable Wgi derives from the “Worldwide Governance Indicators” database by The World Bank. Control of corruption, government effectiveness, political stability and absence of violence/terrorism, regulatory quality, rule of law and voice and accountability represent and measure the six dimensions of governance and their results communicate that higher values correspond to better institutions. I have calculated the average across all dimensions for each country and year using the percentile rank score. Availability of data is covered for the whole sample of countries and the timeframe from 2005 to 2017. Data for the year 2018 is missing for all the countries.

3.5.2 Icrg

(26)

19 of 70

in politics, religious tensions, law and order, ethnic tensions, democratic accountability and bureaucracy quality. Lack of data interests Palestine for the whole time period and the years 2017 and 2018 for all the other 19 countries of the sample.

3.5.3 Edb5 and Edb9

(27)

20 of 70

3.6 Control Variables

Several control variables can be applied to the research motivated by prior existing literature (Julio and Yook, 2016; Mtar and Bannour, 2017; Aziz, 2018; Peres et al., 2018; Chen et al., 2019). These measures include the annual percentage of GDP growth, GDP per capita (based on purchasing power parity) in international U.S. dollars, trade as the sum of exports and imports of goods measured as a share of GDP, inflation rate calculated as the annual percentage change in consumer price index and the official exchange rate of the local currency against the US dollar. As in the study of Aziz (2018) the variables

Telephone and Cellphone, both representing subscriptions per 100 people, are added as

a proxy of technological infrastructures and the variable Education (measured as the ratio of total enrolments in primary education) as a proxy of labour skills. Aziz (2018) includes in the analysis also the following dummy variables: Gfc (Global Financial Crises) represents a dummy variable with 1 as a value for the years 2009, 2010 and 2011 and 0 otherwise. Pta (Prefential Trade Agreements) takes the value of 1 in the years in which the countries in the sample were part of an agreement with either Turkey, the U.S.A. or the European Union. The variable Wto (World Trade Organisation) takes a value of 1 in the year in which a country joined the organisation, otherwise it is 0. Finally, the dummy variable Emu (Euro-Mediterranean Association Agreement) equals 1 in the year of the implementation of the agreement, otherwise 0. The source of data for Wto and Pta is the World Trade Organisation website and the European Commission website for the Emu agreements. The source of data for the variables GDPgrowth, GDPpercapita, Trade,

Inflation and Exchangerate, Telephone, Cellphone and Education comes from The World

Bank’s “World Development Indicators”.

GDPgrowth and GDPpercapita had missing data for the years 2008-2018 for Syria; Trade for the year 2018 for Israel, Kuwait, Qatar, Tunisia, for the years 2008-2018 for

(28)

2016-21 of 70

2018 for Sudan, 2005-2007 for United Arab Emirates, 2013-2018 for Syria and 2015-2018 for Yemen. Exchangerate lacks data only for Syria in the year 2015-2018, Telephone misses data for Yemen and Libya for the year 2018 and for Saudi Arabia in the year 2014 and Cellphone only for Libya and Yemen in the year 2018. Education, on the other hand lacks a lot of data, which is the reason why it has been included only in the last robustness check as a control variable.

3.7 Statistical Analysis

Here as follows there is a description of the statistical analysis implemented in order to determine how these politically instable events affected the MENA region’s countries economic growth and the quality of their institutions. This research aims at studying the following four models (see also Figure 1 and 2 in Appendix A).

TABLE 2

Description of the Empirical Models.

Model Description Path

(1) Effect of political violence on FDI inflows

c

(2) Effect of political violence on institutional quality

a

(3) Effect of institutional quality on FDI inflows

b

(4) Effect of political violence on FDI controlling for institutional quality

c’

*See Appendix A for paths.

(29)

22 of 70

of OLS regressions is that the observations need to be IID – independent and identically distributed. Given that observations regarding the dependent variable lack of independence, the dataset does not represent one of the main assumptions of OLS. Data is presented for the same countries going across multiple years and a key determinant of FDI in one year is FDI in the previous year. In consideration of this, OLS regression has been considered unsuitable for the type of retained data for which is considered to yield biased estimates. As a consequence, it will be implemented a Time Series Cross Sectional analysis. Two alternative models have been performed in order to study the four effects: a fixed and a random effects model. Another concern that has been raised during the analysis is the issue of reverse causality. Albeit to a certain extent it could be possible in all 4 Models, it is considered a stronger concern for Model 2 in which it is explored the effect of political violence on institutional quality that can also be affected the other way around. Also in this case, the time series nature of the data is believed to help getting around this question.

Panel data, also known as cross-sectional time-series or longitudinal data, presents observations of entities across time. Fixed effects and random effects models are both suited to research questions with complicated structure, in addition to temporal hierarchies, such as time-series cross-sectional (TSCS) data and panel data (Bell and Jones, 2014). Usually, TSCS data is distinguished by repeated remarks (often annual) on the same fixed political units (generally states or countries) and in the field of economics, these are mainly adopted in the study of economic growth (Beck, 2001).

(30)

23 of 70 FIGURE 2

Distribution of FDI across the Time Span for each Country.

The following analysis is constituted by multiple fixed effects regressions. In one robustness check, it will also be implemented a Hausman test in order to evaluate which model is more appropriated between the fixed and the random effects regressions. The multiple regression analysis is followed by a mediating analysis in order to test Model 4 and the assumed mediating effect of institutional quality.

3.7.1 Fixed-Effects Model

(31)

24 of 70

traits that may or may not affect the predictor variables (for instance, the quality of the institutions of a country would very likely affect trade or GDP). Using FE enables to control characteristics that are assumed to impact or bias the predictor: this logic for belief is at the foundation of the presumption of the correlation between the predictor variables and an entity’s error term. FE discharges time-invariant aspects, which are unique to each country and should not be correlated to other individual features captured by the constant. Following this uniqueness argument of each country, its error term and constant should not be correlated in the context of FE with the others.

As outlined in “Econometric Analysis of Panel Data” of Baltagi (2005) the fixed effects model could be represented by the following equation:

Yit = α+ X’itβ +Uit i =1,...,N; t =1,...,T (1) In this equation i is denoting countries (or companies, individuals, households, etc.) and the cross-section dimension of the sample and t is denoting time and the time-series dimension of collected data. Y represents the dependent variable (DV), α is a scalar or unknown intercept, β is the coefficient K ×1 of the independent variable and Xit is the itth observation on K explanatory variables (IV). The panel data application in this case utilizes a one-way error component model with

Uit = µi +νit (2) In this equation µi expresses the unobservable individual-speciﬁc effect (for example, managerial skills) and νit represents the remainder disturbance which changes with time and countries. In the fixed-effects model, µi represent fixed parameters and νit is assumed to be IDD- independent and identically distributed.

(32)

25 of 70 3.7.2 Random-Effects Model

In contrast to fixed effects models, the variation across entities in random effects (RE) models is presumed to be random and uncorrelated with the independent variables of the model and the predictor. Random effects are more appropriate in the case in which dissimilarities across entities are believed to have some effect on the dependent variable. These models expect the entity’s error term to not be correlated with the predictors and this grants for time-invariant variables to portray as explanatory variables. An issue could be that since some observations may not be available, the problem of bias in the model is introduced. The equation representing random effects models is:

Yit = βXit + α + uit + εit (3)

3.7.3 Hausman Test

(33)

26 of 70

4 Results

The following section presents a description of the findings of this study. Firstly, in order to have a general overview of the utilised data, it is granted a discussion of the descriptive statistics and the correlation matrix. These are followed by the regression results of the main analysis and finally by the results of the robustness checks.

4.1 Basic Assumptions

Before conducting the cross-sectional time-series analysis, the present study has tested some basic assumptions. The results of the variance inflation factors test for multicollinearity between the independent predictor variables did not present for the main analysis and for the three robustness checks the presence of multicollinearity, considering that all the results for the variables were lower than the tolerated limit of 10 (Table 1, Appendix B).

Similarly, the results of the Breusch-Pagan/Cook-Weisberg test for heteroscedasticity for the main analysis and for the robustness checks accepted the null hypothesis of homoscedasticity since the p-value results were higher that the significance level (Table 2, Appendix C).

4.2 Descriptive Statistics

As previously detailed, the sample of this study totals 20 countries across the period 2005-2018 but in terms of reliability, data for the countries within the sample are not all well represented. Some countries, with special conditions (lack of independence, Palestine) or extremely severe violent conditions (for example, Yemen or Syria), have many more missing observation compared to the rest of the sample.

(34)

27 of 70

(diffusion of data within the dataset). The Between value represents the number of observations in common for all the variables of the study, while the Overall observation represents the total number of values available for each measure. As it can be observed from the table, the number of the observations falls drastically when matched with the observations of the other variables, rendering missing data a relevant concern for validity of the results.

TABLE 3

Descriptive Statistics.

(35)

(36)

(37)

30 of 70

Table 3 provides the correlation matrix between the variables. This table suggests that it would be better not to employ the variables with a high correlation value in the same regression, considering that they would estimate a highly similar concept or its opposite (depending on the sign of the value) and this would result in multicollinearity. High and statistically significant correlation is found between the variables: Fdiwdi and Fdiunctad (0.963***); Csp and Icrg 0.707***); Acled and Csp (0.607***); Ucdpdeaths and Icrg (-0.546***); Ucdpdeaths and Csp (0.604***); Ucdpdeaths and Acled (0.705***); Ucdpconflicts and Icrg (-0.548***); Ucdpconflicts and

Csp (0.555***); Edb9 and Icrg (0.600***); Edb5 and Edb9 (0.940***); Wto and Edb5 (0.628***).

(38)

31 of 70

4.4 Regressions Results

The stepwise regression analysis is constituted of four models in total, which are illustrated in the following table. This represents the main analysis and it employs the variables Fdiwdi (FDI inflows), Acled (political violence) and Edb5 (institutional quality). In this analysis, the control variable Education has been removed because of missing data and the dummy variables Wto and Emu are omitted because of collinearity in all the four models.

TABLE 5 Regression Results.

(1) (2) (3) (4)

VARIABLES Model 1 Model 2 Model 3 Model 4

(39)

32 of 70 (0.138) (0.192) Constant 98.67*** 181.6*** 243.3*** 215.4*** (36.04) (23.28) (61.79) (50.72) Observations 101 94 203 94 R-squared 0.314 0.430 0.201 0.392 Nr.of countries 17 17 17 17

Country FE YES YES YES YES

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Model 1 studies the effect of political violence on FDI and addresses the first hypothesis derived from the literature (see path c in Appendix A). The main finding is that political violence has a slightly negative yet significant effect on FDI (β = -0.0048, p<0.1). The only other significant result of the model is the control variable Gdpgrowth with a positive coefficient (β = 0.253, p<0.01). In addition to Wto and Emu, also Pta is omitted. Model 2 analyses the effect of political violence on institutional quality(see path a in Appendix A). Interestingly, it is found that political violence positively and significantly affects institutional quality (β = 0.00258, p<0.1). Along with the main independent variable, several control variables occur to be significant in this model: Gdpgrowth (β = -0.1177, p<0.05), Inflation (β = -0.1037, p<0.1), Telephone (β = 0.1227, p<0.1) and

Cellphone (β = -0.0975, p<0.05). The dummy variable Pta is also omitted in this model

because of collinearity.

Model 3 explores the influence of institutional quality on FDI (see path b in Appendix A). Surprisingly, it results a significant negative effect of the predictor on the outcome (β = -0.592, p<0.01). The control variable Inflation is also found to be significant (β = 0.1589, p<0.05).

(40)

33 of 70

but non-significant (β = -0.0029). The control variables have non-significant p-values and the dummy variables Pta, Wto and Emu are omitted because of collinearity. The negative effect of political violence when the relationship is controlled for institutional quality is weaker also if the result becomes statistically non-significant. The effect of the mediator on the model is equal to -0.0019 [ab (indirect effect) = c (total effect) – c’(direct effect) = -0.0048 – (-0.0029)]. This value represents the indirect effect that institutional quality has on the relationship between the predictor and the outcome. After the inclusion of institutional quality, political violence affects FDI slightly lesser, proving partial mediation in the model although the result loses its statistical significance.

4.5 Robustness Tests Results

The following robustness tests are conducted in order to check whether the findings of the research are structurally valid. The measurements of the dependent variable, independent variable and of the mediator change in each test. Time dummies are added to the equation in the second test. In the third robustness check, a Hausman test is conducted for each regression in order to choose the most appropriate approach between fixed and random effects and the control variable Education is added to the analysis. 4.5.1 Robustness Check 1

The first robustness test employs the variables Fdiunctad for FDI inflows, Ucdpdeaths for political violence and Edb9 for institutional quality. The control variable Education is excluded from the analysis and the dummy variables Wto and Emu are omitted in all the models. Table 1 in Appendix E illustrates the results of the first robustness test. Model 5 studies the total effect between the independent variable and the outcome. In this model Fdiunctad suffers a significant and negative influence from Ucdpdeaths (β = - 0.00555, p<0.05). Four control variables in the model are found to be significant: Trade (β = 0.2578, p<0.05), Gdpgrowth (β = 0.1268, p<0.1), Cellphone (β = 0.1696, p<0.1) and

(41)

34 of 70

Model 6 analyses the effect of the independent variable on the mediator. In this model political violence is not found to have a significant influence on institutional quality (β = 0.000970). Trade (β = -0.323, p<0.01), Exchangerate (β = 0.3046, p<0.5) and Cellphone (β = 0.1117, p<0.5) are the significant variables. In this model also Pta is omitted for collinearity.

Model 7 explores the influence of institutional quality on FDI inflows. This model presents a significant negative coefficient for Edb9 (β = -0.556, p<0.01). In addition,

Trade (β = 0.1377, p<0.5) and Inflation (β = 0.1588, p<0.05) display significant

coefficients. Pta is omitted because of collinearity.

Model 8 examines the impact of institutional quality on the relationship between political violence and FDI inflows. In this model institutional quality is negative and significant (β = -1.138, p<0.01) and political violence is also negative but non-significant (β=-0.0027). Gdppercapita (β = -0.3293, p<0.05) is the only significant control variable. Pta is omitted for collinearity. The indirect effect of the mediator is equal to -0.0028 [c-c’=-0.0055-(-0.0027)]. Only the result for the total effect is statistically significant (Model 5) while the coefficient for the direct effect is not found to be statically significant (Ucdpdeaths coefficient in Model 8). Also in this analysis, the effect of political violence is weaker after the intervention of institutional quality proving partial mediation, although not all the effects are statistically significant.

4.5.2 Robustness Check 2

(42)

35 of 70

Model 9 studies the effect of political violence on FDI inflows. In this model,

Ucdpconflicts is negative but non-significant (β = -2.7018). The significant variables are: Gdpgrowth (β = 0.1295, p<0.1), Trade (β = 0.2840, p<0.1) and Gfc (β = -46.801, p<0.1).

Model 10 analyses the effect of the dependent variable on the mediator. In this test, this effect is found to be positive and significant (β = 1.87869, p<0.1). The significant control variables in this model are: Gdppercapita (β = -0.3095, p<0.01), Trade (β = -0.2069, p<0.01) and Telephone (β = -0.3040, p<0.01).

Model 11 explores the influence of the mediator on the outcome variable. This influence results in a negative and significant coefficient (β = -0.2854, p<0.01). Trade (β = 0.1401, p<0.1), Inflation (β = 0.1359, p<0.1) and Gfc (β = -41.169, p<0.1) are significant. Model 12 examines the mediating effect of institutional quality on the relationship of political violence on FDI. The results of this model are not significant for both predictors: institutional quality (β = 0.0639) and political violence (β = -2.913). The significant variables in the model are: Gdpgrowth (β = 0.1283, p<0.1), Gdppercapita (β = 0.2650, p<0.1),Trade (β = 0.6130, p<0.01) and Gfc (β = -54.617, p<0.05). The indirect effect of the mediator is: 0.2112 [c-c’=-2.7018-(-2.913)]. In this case, when controlled for the mediator the negative effect of political violence is bigger against the theoretical predictions but - considering that both the total effect and the direct effect are not statistically significant - the result is not considered reliable.

4.5.3 Robustness Check 3

The last robustness check includes the Hausman test for each model. Looking at the Hausman test’s results, random effects resulted to be the best approach only for Model 2 (Table 1, appendix D). Model 1, 3 and 4 used fixed effects regressions like in the main analysis and in the other robustness checks. The variables employed in this test are:

Fdiwdi for FDI inflows, Csp for political violence and Icrg for institutional quality. The

(43)

36 of 70

Model 13 studies the total effect for the independent to the dependent variable. In contrast to what has been predicted, the result for political violence is positive but also non-significant (β = 12.492). The only non-significant variables in the model are: Trade (β = 0.1797, p<0.05) and Inflation (β = 0.1746, p<0.05). Wto and Emu are omitted because of collinearity.

Model 14 studies the effect of the predictor on the mediator. This model is the only one that adopts a random effects regression in the analysis. The result of political violence on institutional quality is significant and negative (β = -14.313, p<0.01). Four other control variables in the model are found to be significant: Gdpgrowth (β = 0.0627, p<0.01), Trade (β = 0.0403, p<0.05), Exchangerate (β = -0.1898, p<0.05) and Cellphone (β = 0.0926, p<0.01). When tested with random effects, none of the dummy variables are omitted in the regression.

Model 15 explores the effect of the mediator on the outcome. In this model the influence of institutional quality on FDI inflows is highly positive and statistically significant (β = 0.9333, p<0.01). Trade and Inflation are also significant: respectively β = 0.2346 (p<0.01) and β = 0.1737 (p<0.05). The result of Wto and Emu are is omitted in the regression’s outcome.

(44)

37 of 70

4.6 Mediation Analysis Results

Hereinafter are outlined the results of the mediation analysis for each of the analysis conducted in this research.

4.6.1 Main Analysis Results

The ACME result (0.0015) represents the average effect of the predictor on the dependent variable which operated through the mediator. The estimate for the direct effect equals -0.0030 and the estimate for the total effect is -0.0045. In this analysis the percentage of the total effect mediates is 0.32284. The results of the mediation analysis are restricted by the smaller lack of observations compared to the simulations ran by the statistical software. In order to get results for the mediation analysis, the following countries had to be excluded from the equation: Palestine, Qatar, Syria, Turkey, Yemen and Iraq. Here as follows are outlined the results of the mediation analysis calculated with the statistical software Stata.

TABLE 6

Mediation Analysis Results. (1)

Effect Mean

ACME -0.0015

Direct Effect -0.0030

Total Effect -0.0045

(45)

38 of 70 4.6.2 First Robustness Check

In the first robustness check, the ACME estimate is 0.0010, the direct effect equals to -0.0028, the estimate for the total effect is -0.0038 and the percentage of the mediated total effect amount at 0.2310. These countries were excluded from the analysis in order to gain results: Israel, Jordan, Libya, Morocco, Oman, Palestine, Qatar, Syria and Yemen. The results are reported in Table 1 of Appendix F.

4.6.3 Second Robustness Check

The mediation analysis results for the second robustness test are outlined as follows: the ACME result is 0.1013, the direct effect amounts to -3.0277, the total effect to -2.9264 and the percentage of total effect mediated is -0.0295. Libya, Morocco, Oman, Palestine, Qatar, Syria, Turkey, Yemen are excluded from the analysis. The results are outlined in Table 2 of Appendix F.

4.6.4 Third Robustness Check

The last analysis displays the following findings: the estimate for ACME is -8.947801, for the direct effect it is 22.0411, the total effect amounts to 13.0933 and finally the percentage of total effect mediated is -0.3833. Iraq, Jordan, Palestine, Syria, Yemen are removed from the analysis. Results are displayed in Table 3 in Appendix F.

4.7 Overview of Findings

The empirical analysis provides the following outcomes for each of the hypotheses developed in the literature review.

(46)

39 of 70

Second, hypothesis 2 has been rejected and supported in the first three analyses, because resulted in a negative influence of the measures of institutional quality on FDI. In contrast, in the last robustness check the outcome confirms the hypothesis predicted in the literature review.

Lastly, hypotheses 3a and 3b present the two different scenarios that could result from the empirical analysis, therefore each hypothesis in each analysis can give result in either hypothesis 3a or 3b. Given this, hypothesis 3a is confirmed but not statistically supported in the main analysis and in the first robustness check, while it is rejected in the third robustness check but the result is not supported. Hypothesis 3b is rejected and not supported in the last robustness check.

TABLE 7

Overview of Findings.

Hypotheses Results

H1 Political violence negatively affects inward foreign investments.

[1] Confirmed [2] Confirmed

[3] Confirmed but not supported

[4] Rejected and not supported

H2 Better institutional quality in the host country positively influences inward foreign direct investment.

[1] Rejected

[2] Rejected [3] Rejected [4] Confirmed

H3a When political violence leads to an improvement of the institutions, the negative relationship between political violence and FDI is positively mediated by institutional quality.

(47)

40 of 70

[2] Confirmed but not supported

[4] /

H3b When political violence causes a worsening of the institutional environment, the negative relationship between political violence and FDI is aggravated.

[1] /

[2] / [3] /

[1] Main Analysis, [2] Robustness Check 1, [3] Robustness Check 2, [4] Robustness Check 3. Confirmed- The result confirms the hypothesis.

(48)

41 of 70

5 Discussion

The following section provides a discussion of the empirical findings of this study linking them to the hypotheses formulated in the literature review. The examination of the results is followed by their limitations and implications for future research, presented in this passage before the final display of the conclusion.

5.1 Discussion of Results

The primary aim of this paper was to assess whether the quality of the institutions played a mediating role on the effects that political violent events inflict to inward foreign investments. More precisely, it was intended to study how disparate levels of political violence differently affect institutional quality and how this in turn – if taken into account - has an indirect impact on FDI. In order to study the mediating effect, it is needed to conduct a multiple regression analysis, which has been arranged in four models (see Table 2) that study the direct, indirect and total effects of each analysis (see Figure 1, 2 in Appendix A).

(49)

42 of 70

on FDI (Globerman and Shapiro, 2003; Li, 2006), it can be concluded that the explanation for these contrasting findings are the measurement errors of the empirical analysis. Hypothesis 2 – studied in models 3, 7, 11 and 15 – theorised a positive impact of institutional quality on inward foreign investments. Surprisingly and against the theoretical predictions, this hypothesis has been rejected in the first three models with a significant and negative coefficient. Considering that the sample is composed of developing countries, these results point out that the quality of the institutions in the majority of these countries could still be premature in order for it to constitute a main determinant of foreign investments in these locations. Furthermore, another explanation could be that multinationals move their operations to developing countries in order to escape the commitments - and the responsibilities for higher levels of Corporate Social Responsibility (CSR) - that the institutions and the main stakeholders in the home country constrain on them. As a consequence, firms that have been exposed and pressured may seek countries where the institutions are weaker and therefore less powerful (Cuervo-Cazurra, 2006). Differently, model 15 confirmed the hypothesis with a highly significant and positive coefficient. This result is in line with the wide majority of the previous research which highlights the importance of effective institutions that undoubtedly enable businesses to achieve the best performance under the minimum wasted expenses and effort. The mixed findings for this hypothesis reflect the multifaceted aspect of institutional quality, that represents a wide concept and includes several factors that may differ from one context to the other.

(50)

43 of 70

major limitation, none of the four analyses provided statistically supported results. Nevertheless, the main analysis and the first robustness check confirmed Hypothesis 3a. In both analyses, the negative effect of political violence on FDI is weaker after the regression is controlled for institutional quality. In both models political violence affected positively institutional quality - recalling the examples explained in the literature review - in which some countries (i.e. Tunisia, Morocco, Jordan) reacted to the protests with social, economic and political reforms contributing to improvement of the institutions weakening the negative impact of political violence on FDI. Contrarily, the second and third robustness check not only did not support the third hypotheses, but the analyses rejected both Hypothesis 3a and Hypotheses 3b. In both tests the effect of political violence is greater after the inclusion of institutional quality, but the two models differ regarding the mediation role of institutional quality. In the second robustness check the effect of political violence on institutional quality resulted in a positive and significant coefficient (Model 10), but since the negative effect of political violence on investments is greater if controlled for institutional quality, Hypothesis 3a is rejected. The last robustness test shows a severe negative and significant coefficient for the impact of political violence on institutional quality and this effect raises the impact of political violence on FDI as predicted in Hypothesis 3b, but the hypothesis is rejected because in this analysis the result for the direct effect of political violence on FDI is positive in contrast to what the predicts. In conclusion, the findings for this hypotheses are mixed and not supported, highlighting the big issue of this research that will be addressed in the following section.

5.2 Limitations and Future Research

(51)

44 of 70

and structural traits, which made the reasons of the revolutions a common ground in these societies. As a consequence, in order to study the different effects of the phenomenon of political violence in different realities, the focus of the research has been addressed only to these 20 countries. However, in order to avoid the issue of sample selectivity bias, the analysis could be replicated in a different geographic area or with an international sample. Future studies could also employ the system GMM (Arellano-Bover/Blundell-Bond) technique which takes into account of issues such as endogeneity, measurement errors, simultaneity, omission of relevant variables and sample selectivity (Aziz, 2018).

(52)

45 of 70

Thirdly, a further concern of the present analysis matters the adopted measurements of political violence. During the stage of data collection, it has been extremely arduous to gather satisfactory variables which could fairly represent the situation of the Arab Spring revolutions. The variable Acled is the only one in the study that adequately estimates different levels of violence (from real battles to riots even to simple strategic developments), which differently affected the quality of the institutions in the 20 countries. Nonetheless, this measure covered completely only six North African countries for the entire dataset, while for the Middle Eastern region “The Armed Conflict Location & Event Data Project” only displays information for the last years (approximately for the timespan 2015-2018). Conversely, the other variables (Ucdpdeaths, Ucdpconflicts and Csp) mainly focused on the presence of violence, which I consider not entirely appropriate in order to assess if the lack of violence could lead to an improvement or to a worsening of the institutional environment. Future researchers should employ better up-to-date information, which will probably cover also the timeframe of the crisis and study the phenomenon better.

(53)

46 of 70

6 Implications and Conclusion

This paper analyses empirically the impact of institutional quality on the relationship between political violence and FDI, studying the recent crises exploded in the MENA region at the beginning of this decade. The research is performed implementing a mediation analysis on Time Series Cross Section data for 20 countries in the time frame 2005-2018. By studying the research question “What is the impact of political violence

on the quality of the institutions of the countries affected by the Arab Spring crises and which effects does it entail for the direct relationship between political violence and inward investments?”, this research examines the impact of political violence on FDI,

(54)

47 of 70

References

Abdel-Latif, H.O., Bazoumana 2017, 'FDI response to political shocks: What can the Arab Spring tell us?', Journal of Behavioral and Experimental Finance, vol. 24, p. 100233.

Aisen, A. & Veiga, F.J. 2013, 'How does political instability affect economic growth?',

European Journal of Political Economy, vol. 29, pp. 151-67.

Alesina, A., Özler, S., Roubini, N. & Swagel, P. 1996, 'Political instability and economic growth', Journal of Economic growth, vol. 1, no. 2, pp. 189-211. Alesina, A. & Perotti, R. 1996, 'Income distribution, political instability, and

investment', European economic review, vol. 40, no. 6, pp. 1203-28.

Amin, M., Assaad, R., Al-Baharna, N., Dervis, K., Desai, R.M., Dhillon, N.S., Galal, A., Ghanem, H. & Graham, C. 2012, After the spring: Economic transitions in

the Arab world, Oxford University Press.

Asteriou, D. & Price, S. 2001, 'Political instability and economic growth: UK time series evidence', Scottish Journal of Political Economy, vol. 48, no. 4, pp. 383-99.

Aziz, O.G. 2018, 'Institutional quality and FDI inflows in Arab economies', Finance

Research Letters, vol. 25, pp. 111-23.

Aziz, O.G. & Mishra, A.V. 2016, 'Determinants of FDI inflows to Arab economies',

The Journal of International Trade & Economic Development, vol. 25, no. 3,

pp. 325-56.

Baltagi, B.H. 2005, 'Econometrics Analysis of Panel Data', University Press. , vol. 2nd Edition.

Barro, R.J. 1991, 'Economic growth in a cross section of countries', The quarterly

journal of economics, vol. 106, no. 2, pp. 407-43.

Beck, N. 2001, 'Time-series–cross-section data: What have we learned in the past few years?', Annual review of political science, vol. 4, no. 1, pp. 271-93.

Bell, A. & Jones, K. 2015, 'Explaining fixed effects: Random effects modeling of time-series cross-sectional and panel data', Political Science Research and Methods, vol. 3, no. 1, pp. 133-53.

Bhattacharya, U., Galpin, N. & Haslem, B. 2007, 'The home court advantage in international corporate litigation', The Journal of Law and Economics, vol. 50, no. 4, pp. 625-60.

Brunetti, A. & Weder, B. 1998, 'Investment and institutional uncertainty: a comparative study of different uncertainty measures', Weltwirtschaftliches Archiv, vol. 134, no. 3, pp. 513-33.

Buchanan, B.G., Le, Q.V. & Rishi, M. 2012, 'Foreign direct investment and institutional quality: Some empirical evidence', International Review of

financial analysis, vol. 21, pp. 81-9.

(55)

48 of 70

institutional quality: Some empirical evidence', International Review of

financial analysis, vol. 21, pp. 81-9.

Busse, M. & Hefeker, C. 2007, 'Political risk, institutions and foreign direct

investment', European journal of political economy, vol. 23, no. 2, pp. 397-415. Chen, K., Nie, H. & Ge, Z. 2019, 'Policy uncertainty and FDI: Evidence from national

elections', The Journal of International Trade & Economic Development, vol. 28, no. 4, pp. 419-28.

Cuervo-Cazurra, A. 2006, 'Who cares about corruption?', Journal of international

business studies, vol. 37, no. 6, pp. 807-22.

Cukierman, A., Edwards, S. & Tabellini, G. 1992, Seigniorage and political instability, National Bureau of Economic Research.

Daude, C. & Stein, E. 2007, 'The quality of institutions and foreign direct investment',

Economics & Politics, vol. 19, no. 3, pp. 317-44.

Denisia, V. 2010, 'Foreign direct investment theories: An overview of the main FDI theories', European journal of interdisciplinary studies, no. 3.

Dixit, A. 2011, 'International trade, foreign direct investment, and security', Annu. Rev.

Econ., vol. 3, no. 1, pp. 191-213.

Dragusha, B. & Kruja, A. 2014, 'The Impact of Political Risk on Foreign Direct Investment', ILIRIA International Review, vol. 4, no. 2, pp. 73-92.

Dunning, J. 2002, 'Relational assets, networks, and international business activity: Extended thoughts on Dunning 2002a', Reading and Rutgers University. Mimeo,

August.

Dunning, J.H. 1973, 'The determinants of international production', Oxford economic

papers, vol. 25, no. 3, pp. 289-336.

Dunning, J.H. 1980, 'Toward an eclectic theory of international production: Some empirical tests', Journal of international business studies, vol. 11, no. 1, pp. 9-31.

Dunning, J.H. 1988, 'The eclectic paradigm of international production: a restatement and some possible extensions', Journal of International Business Studies, pp. 1-31.

Dunning, J.H. 1993, 'Multinationals Enterprises and the Global Economy'Addison Wesley, London.

Edwards, S. & Tabellini, G. 1991, Political instability, political weakness and inflation:

an empirical analysis, National Bureau of Economic Research.

Fayerweather, J. 1982, International business strategy and administration, Ballinger Pub Co.

Fielding, D. & Shortland, A. 2005, 'Political violence and excess liquidity in Egypt',

Journal of Development Studies, vol. 41, no. 4, pp. 542-57.

Fischer, J.A. 2010, 'Accounting for unobserved country heterogeneity in happiness research: Country fixed effects versus region fixed effects'.

(56)

49 of 70

empirical analysis', Studies in Comparative International Development, vol. 33, no. 3, pp. 3-17.

Globerman, S. & Shapiro, D. 2003, 'Governance infrastructure and US foreign direct investment', Journal of International Business Studies, vol. 34, no. 1, pp. 19-39. Goswami, G.G. & Haider, S. 2014, 'Does political risk deter FDI inflow?', Journal of

Economic Studies.

Gupta, D.K., Madhavan, M. & Blee, A. 1998, 'Democracy, economic growth and political instability: an integrated perspective', The Journal of Socio-Economics, vol. 27, no. 5, pp. 587-9.

Helmy, H.E. 2013, 'The impact of corruption on FDI: is MENA an exception?',

International Review of Applied Economics, vol. 27, no. 4, pp. 491-514.

Hymer, S.H. 1976, International operations of national firms, MIT press. Jaimovich, D. & Panizza, U. 2007, 'Procyclicality or reverse causality?'. Jong-A-Pin, R. 2008, 'Essays on political instability', Mémoire de Thèse.

Ju, J. & Wei, S.-J. 2007, 'Domestic institutions and the bypass effect of international capital flows', NBER Working Paper, vol. 13148.

Jude, C. & Levieuge, G. 2015, 'Growth effect of FDI in developing economies: the role of institutional quality'.

Julio, B. & Yook, Y. 2016, 'Policy uncertainty, irreversibility, and cross-border flows of capital', Journal of International Economics, vol. 103, pp. 13-26.

Khan, M.M. & Akbar, M.I. 2013, 'The impact of political risk on foreign direct investment'.

Kirwin, M.F. & Cho, W. 2009, Weak states and political violence in Sub-Saharan

Africa, AfroBarometer.

Kobrin, S.J. 1979, 'Political risk: A review and reconsideration', Journal of

international business studies, vol. 10, no. 1, pp. 67-80.

Krug, E.G., Mercy, J.A., Dahlberg, L.L. & Zwi, A.B. 2002, 'The world report on violence and health', The lancet, vol. 360, no. 9339, pp. 1083-8.

Kumaraswamy, P. 2011, 'The Arab Spring', India International Centre Quarterly, vol. 38, no. 1, pp. 52-62.

Kurul, Z. & Yalta, A. 2017, 'Relationship between institutional factors and FDI flows in developing countries: New evidence from dynamic panel estimation',

Economies, vol. 5, no. 2, p. 17.

Lane, P.R. & Milesi-Ferretti, G.M. 2007, 'The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970–2004',

Journal of international Economics, vol. 73, no. 2, pp. 223-50.

Le, Q.V. & Zak, P.J. 2006, 'Political risk and capital flight', Journal of International

Money and Finance, vol. 25, no. 2, pp. 308-29.

Mediating Role of Institutional Quality on the Relationship between Political Violence and Inward Foreign Direct Investment