Welfare states and terrorism : understanding the relationship between social welfare policy and terror

(1)

UNIVERSITY OF AMSTERDAM

Welfare States and

Terrorism

Understanding the relationship between social

welfare policy and terror

Mandy Malan 10000791 Bachelor thesis Welvaartstaten en hun gevolgen R.J. van der Veen M.L. Hofer Political Science (International Relations) 5-2-2012 mandymalan@outlook.com

Abstract - Plenty of research has been done to find out what the determining factors of terrorism are. A rather new idea is that generous welfare states are less prone to terrorism through their social welfare policy. This idea is evaluated in this statistical contribution. Using pooled cross-section time series estimation, the relationship between social welfare policy and the incidence of terrorism is determined in the 18 most developed welfare states in the world. Distinct from previous studies is the use of a more sophisticated measure of the welfare state; the decommodification-index. Results show that there is indeed a modest, but significant negative effect of the welfare state on terrorism. Furthermore, two causal mechanisms through which this effect may work are examined: poverty and inequality. Poverty does not seem to have a significant effect on terrorism, wheras inequality does.

(2)

1

1. Introduction

2 2. Theory

3 2.1. Welfare states and terrorism

4 2.1.1. Measuring the welfare state

5 2.1.2. Time-lag

6 2.2. Poverty

7 2.3. Inequality

9 3. Data and method

12 3.1. Dependent variable

12 3.2. Independent variable

13 3.3. Intermediate variables

14 3.4. Statistical method

14 4. Results

15 5. Conclusions and implications

18 6. Discussion

20 7. References

20 8. Appendix

23 8.1. Statistical analyses

23 8.2. Codebook

27 8.3. Do-file

29

(3)

2

1. Introduction

Following the 9/11 attacks in the United States, James Wolfensohn, president of the World Bank at that time, repeatedly called for poverty-reduction and economic development as a strategy to fight terrorism. For instance, he argues that:

‘Central to conflict prevention and peace-building must be strategies for promoting social cohesion and inclusion. Inclusion means ensuring that all have opportunities for gainful employment, and that societies avoid wide income inequalities that can threaten social stability. But inclusion goes well beyond incomes. It also means seeing that poor people have access to education, health care and basic services, such as clean water, sanitation and power. It means enabling people to participate in key decisions that affect their lives.’ (Wolfensohn 2008: 43).

If terrorism really is caused by economic conditions, then naturally, one would argue that more developed countries are less prone to terrorism. Furthermore, states that have a habit of taking care of their population through social welfare policy, thereby eradicating the economic conditions that fuel terrorism, should be even less prone to terrorism. However, this proposed relationship is not as obvious as presented here.

The idea has been elaborated in an article on welfare states and terrorism by Burgoon (2006), in which empirical evidence for a negative relationship between the extensiveness of the welfare state and the incidence of terrorism is found. With his study, he contributes to a large debate on the effects of social welfare policies and an equally large debate on the causes of terrorism. The mechanisms through which the welfare state may reduce terrorism, according to Burgoon (2006), are precisely the effects of the welfare state and the causes of terrorism that are most heavily debated: economic conditions. Much research has been done on the effects of social welfare policies on poverty and inequality. On the other hand, many studies have discussed the role of economic conditions in the development of terrorism. In both research fields, conclusions point to different directions.

Burgoon (2006) distinguishes five mechanisms through which social welfare policies may have an effect on terrorism. Social welfare policies lead to less poverty and higher development which in turn decreases terrorism. Social welfare policy also decreases economic insecurity, leading to less terrorism. Third, social welfare policies may reduce religious-political extremism (also through lower poverty and economic security). A fourth causal mechanism is the influence of social policies on inequality, if there is a negative effect of social policy on inequality, this might also lead to less terrorism. The last factor Burgoon (2006) mentions is a positive effect of social welfare policies on terrorism, through the effect on the ‘capacity for terror’. Burgoon (2006) argues that potential terrorists have more time and money to organize terrorist attacks.

(4)

3 The purpose of this article is to further analyze the relationship between the welfare state and terrorism. First, the relationship will be reexamined using a different measure for the extensiveness of the welfare state: decommodification. This measure was first used by Esping-Andersen in The Three Worlds of Welfare Capitalism (1990) and was later revised by Scruggs and Allan (2006a). Even though, social expenditure as a measure of the welfare state is heavily relied on by many studies, including Burgoon (2006), decommodification provides a better measure for welfare state extensiveness as it takes the effectiveness of social welfare policy into account (Scruggs and Allan 2006a: 56). Second, the mechanisms through which the welfare state influences terrorism will be researched. Specifically, the inequality and poverty connections are examined.

The study is of quantitative nature. Using pooled cross-section time series data, the first analysis examines the effect of the level of decommodification on the incidence of terrorism. Subsequently, the intermediate effects of inequality and poverty will be explored. The analysis is done for the 18 welfare states of Esping-Andersen. These countries are the most developed countries in the world. Using these countries to look at the effect on terrorism is useful because of the extensiveness of the welfare states and the low rate of terrorist incidents. Finding a relationship in these countries will strongly confirm Burgoon’s (2006) hypothesis. Results will also show whether the relationship between the welfare state and terrorism is maintained when using a more sophisticated measure for the welfare state. Furthermore, this study contributes to the discussion on the mechanisms through which welfare states affect terrorism. This will give more insight in how terrorism can best be prevented. If welfare states reduce terrorism through decreasing poverty and inequality, terrorism may be countered using economic policy.

In the first section the theoretical framework of this article will be presented. First, research on the link between the welfare state and terrorism will be covered. In the second part the poverty connection will be discussed. Both the relationship between the welfare state and poverty as the link between poverty and terrorism are addressed. Lastly, the inequality connection is dealt with. The second section deals with the methodology of this analysis, the variables will be introduced and the statistical method will be treated further. In the third section the empirical results of the analysis are presented and discussed. The fourth part consists of the concluding remarks and implications of this study. Finally, a discussion is presented in the last section.

2. Theory

To analyze the effects of social welfare policies on terrorism, it is important to understand how this effect takes place. As noted, Burgoon (2006) distinguishes five mechanisms through which the welfare state might reduce or increase terrorism. Two of these, poverty and inequality, are

(5)

4 1. Poverty (50% of median income) 2. Inequality (Gini index) Terrorism (total terrorist incidents by country) Welfare state (decommodification)

being examined here (figure 1). After a brief outline of Burgoon’s (2006) study and subsequent studies on the welfare state and terrorism, the poverty and inequality connection will be considered.

2.1. Welfare states and terrorism

Up until Burgoon’s 2006 study there was hardly any debate on the connection between the welfare state and terrorism. Since then, various studies have been done in which similar effects were found (Crenshaw et al. 2007; Krieger and Meierrieks 2009; Freytag et al 2009).

Burgoon’s (2006) study is of quantitative nature and examines via pooled cross section time series and cross sectional analysis whether government welfare expenditure is related to the incidence of terrorism. He uses three measures of terrorism incidence: 1) total incidents by country, 2) transnational incidents by country, 3) significant transnational incidents by country where terrorist is from (Burgoon 2006: 177). In addition to social expenditure, Burgoon (2006) looks at total spending and total welfare spending, the latter being social spending plus education spending. He controls for Left-party power, democracy, population, government capability, conflict and trade openness. The results are clear: all measures of spending are significantly negatively related to all three measures of terrorism (Burgoon 2006: 192-194).

A second study dealing with this topic is a replication and expansion of Burgoon’s study by Crenshaw et al. (2007). Crenshaw et al. express some concerns about Burgoon’s (2006) method and usage of variables. After adjusting the dependent variable and a few control variables they find that Burgoon’s (2006) results are overstated (Crenshaw et al. 2007: 13-14). First off, they find that when terrorist incidents are attributed to the nations where the terrorists are originated instead of to the nations where the attack takes place, the effect of social welfare policies is less strong (Crenshaw et al. 2007: 11). This doesn’t fit in with Burgoon’s (2006) theory. A potential terrorist should be more influenced by the country of origin and that country’s welfare state (Crenshaw et al. 2007: 6-7) Second, social welfare policy is only significant for Leftist terrorism, whereas religious identity terrorism is not influenced by the

(6)

5 welfare state (Crenshaw et al. 2007: 13). The results of this study are less strong and manifest than Burgoon’s (2006) research, however, social welfare policy still seems to work upon terrorism. It is therefore significant to further look into the connection between the welfare state and terrorism.

In their 2009 study, Krieger and Meierrieks examine if spending in certain policy fields leads to less terrorism. Their reasoning is similar to Burgoon (2006), arguing that social welfare policy ‘…reduce terrorism by (i) promoting short-run economic performance and (ii) ameliorating structural economic conditions, thus generally curtailing and impeding extremist influence in societies.’ (Krieger and Meierrieks 2009: 5). They also connect their findings to Esping-Andersen’s (1990) worlds of welfare capitalism by examining if certain worlds are less prone to terrorism (Krieger and Meierrieks 2009: 1). They find that higher spending on social welfare is significantly negatively related to terrorism. Health spending has the same effect, whereas unemployment and old age spending have no effect on terrorist activity. However, when taking into account the severity of terrorist attacks unemployment and old age spending are significant determinants (Krieger and Meierrieks 2009: 13). Support is not found for their hypothesis that certain worlds of welfare capitalism are less prone to terrorist attacks due to being more effective in improving socio-economic conditions (Krieger and Meierrieks 2009: 14).

Another support of Burgoon’s (2006) hypothesis comes from a study of Freytag et al. (2009) on whether socio-economic conditions are determinants for terrorism. They find government spending to be significantly negatively related to terrorist activity in Europe and the OECD. In the Islamic region of the world, government spending is irrelevant (Freytag et al. 2009: 18-19). As government spending largely consists of social spending, this study supports the idea that social welfare policy may reduce terrorism. Building on these various studies supporting the welfare state-terrorism connection, the first hypothesis of this study is as follows:

Hypothesis 1a: The extensiveness of the welfare state has a negative effect on the incidence of terrorism in a country.

2.1.1. Measuring the welfare state

The studies above all use a form of government spending to measure the extensiveness of the welfare state. In this study the extensiveness of the welfare state will be measured using the decommodification-scale. This concept was first used by Esping-Andersen in his 1990 book ‘The Three Worlds of Welfare Capitalism’. It is the antipode of the commodification of labor, which evolved along with the development of modern capitalism. Because labor as a commodity may be exhausted through sickness or deemed unnecessary during periods of economic stagnation, decommodification becomes necessary for the system’s survival (Esping-Andersen 1990: 37). As Esping-Andersen defines it:

(7)

6 ‘…de-commodification should not be confused with the complete eradication of labor as a commodity; it is not an issue of all or nothing. Rather, the concept refers to the degree to which individuals, or families, can uphold a socially acceptable standard of living independently of market participation.’ (Esping-Andersen 1990: 37).

Decommodification is a considerably better measure of social welfare programs than the often used measure of social spending as percentage of GDP. The latter is also used by Burgoon (2006) in his analysis. Decommodification takes many more features of welfare programs into consideration. An example of the differences in measurement is France, which has rather high social spending (OECD data 2012). When looking at decommodification scores, however, France scores around the average of the 18 welfare states and thus ends up in the middle (Scruggs and Allan 2006a: 68). This implies that social expenditure in France is not very effective in providing welfare and security. Thus, when social expenditure data are used, the extensiveness of the welfare state may be under- or overestimated. Measuring the welfare state with decommodification will therefore give a better view on the effects of the welfare state on terrorism.

Another benefit of using decommodification in this study is that it incorporates one of the causal mechanisms distinguished by Burgoon (2006), namely economic insecurity. He argues that economic insecurity will be lower in a larger welfare state and that this will reduce terrorism. Even though, Burgoon (2006) defines economic insecurity in subjective terms, one could reason that in a country where decommodification is high, people will generally worry less about the economy and their economic position. Consequently, measuring the welfare state by decommodification integrates economic security. If economic security truly is an intermediate variable, the effect of decommodification on terrorism should be larger than when using social expenditure.

2.1.2. Time-lag

Social welfare policy takes time to have effect on economic conditions and on the attitudes of the public. It may therefore take some time before changes in the extensiveness of social welfare policy will affect terrorism, but it is difficult to grasp exactly how long it takes. A few possibilities are explained. One can imagine that it may take years for a change in decommodication to actually be felt by the public. Moreover, the frustration of having a poor upbringing or living in poverty for years in a country that barely provides social security might be necessary for a person to actually commit a terrorist attack. On the other hand, it may only take a while for policy-changes to affect attitudes of people as they may be assured that better times are ahead because of the shift in policy. Two additional hypotheses will be added to the first hypothesis to find out how much time it takes for welfare states to affect terrorism.

(8)

7 Hypothesis 1b: Changes in the extensiveness of the welfare state take years to reduce the incidence of terrorism in a country.

Hyothesis 1c: Changes in the extensiveness of the welfare state quickly reduce the incidence of terrorism in a country.

2.2. Poverty

Whether generous welfare states are more effective in reducing poverty through their social policy has been the topic of debate for a long time. Even though, there are studies that indicate that social welfare policies do not eradicate poverty (Barro and Lee 1993; Lee 1987), most evidence supports the idea (Kenworthy 1999; Lindert 2004; Moller et al. 2003; Caminada and Goudszwaard 2009; Scruggs and Allan 2006b).

Lee (1987) finds there are numerous social programs designed to redistribute income that in the long term make the poor worse off (Lee 1987: 162). He argues that on balance social programs targeting certain groups do not reach those groups and therefore do little to improve their position (Lee 1987:162). He does however acknowledge that some social programs do benefit the poor in the long run and should thus be continued. Another argument against social programs is that social welfare policies impede economic growth due to a long run trade-off between equality and efficiency and thereby increase or fail to reduce poverty (Kenworthy 1999: 1121). Spending on social programs implies spending on the less productive and taxing the more productive. This lowers the overall productivity of the economy and can only damage economic growth (Lindert 2004: 16). Studies also suggest welfare policies stimulate dependence on welfare programs which may foster a poverty trap possibly leading to even more poverty (Kenworthy 1999: 1121).

Kenworthy (1999) examines the effects of social welfare policy on poverty rates for 15 OECD countries. He uses three measures for social-welfare policy extensiveness: government transfers, decommodification and social wage and finds significant negative effects on poverty-rates for all three variables (Kenworthy 1999: 1126-1131). Lindert (2004) shows that, in contradiction to what many economists argue, more spending on social welfare doesn’t damage economic growth. A cross-sectional time series study of the determinants of relative poverty by Moller et al. (2003) shows that welfare state generosity is an important explanatory factor of poverty reduction. They find that more generous welfare states are more effective in reducing poverty. Especially child and family allowances are important determinants of poverty reduction. Whereas unemployment benefits and maternity allowances are less significant and means-tested benefits are not significantly related to poverty reduction (Moller et al. 2003: 42-43). The results do however support the idea that overall social spending reduces poverty. Similar results are found by Caminada and Goudswaard (2009). They too find that child and

(9)

8 family allowances are most effective in reducing poverty. Their results do however show less strong effects for social spending on poverty than previous studies have found (Caminada and Goudswaard 2009: 2). Lastly, a study finding strong positive effects of social policy on poverty is by Scruggs and Allan (2006b). They show that sickness benefits lead to lower poverty rates using different measures of poverty. Whereas generous unemployment benefits have no effect on poverty (Scruggs and Allan 2006b: 899).

As shown, there is plenty of research suggesting welfare states have a poverty reducing effect. However, this is only half of the theoretical framework needed to support the expectation examined in this study. The other half concerns the connection between poverty and terrorism. As the link between the welfare state and poverty, this connection is not straightforward.

Discussion on this connection occurs within the debate about the determinants of terrorism. Of the determinants that have been distinguished most of them fall into the categories political determinants, social determinants and economic determinants (Krieger and Meierricks 2009; Burgoon 2006). Proponents of the idea that terrorism is (partially) driven by material factors argue that poverty and other structural economic conditions generate feelings of relative deprivation, scapegoating and discontent, which may lead to the development of conflict, which in turn fuels terrorism and political extremism (Burgoon 2006: 180). Opponents of economic determinants of terrorism claim economic conditions are irrelevant and that terrorism is alternatively sparked by political factors, for instance: political transformations (Abadie 2006), political development in terms of political rights and civil liberties (Kurrild-Klitgaard et al. 2006), authoritarian rule and state failure (Burgoon 2006: 177).

Evidence also points in both directions. Li and Schaub (2004) find that more economically developed countries are less prone to transnational terrorism (Li and Schaub 2004: 248). Chen (2003) examines through research on the Indonesian financial crisis whether religious intensity is increased by economic distress. He finds a significant positive link between economic distress and religious intensity. Moreover, religious intensity is highly associated with communal violence. Thus, economic conditions lead to an increase in religious intensity and ideological extremism, which in turn may lead to violent religious conflict (Chen 2003: 32). Blomberg et al. (2004) investigate the relationship between economic contraction and terrorism. Their argument is as follows:

‘Groups that are unhappy with the current economic status quo, yet unable to bring about drastic institutional changes, may find it rational to engage in terrorist activities. The result is a pattern of reduced economic activity and increased terrorism.’ (Blomberg et al. 2004: 463)

(10)

9 This argument is supported by their results, which show a positive relationship between economic contraction and terrorist activity (Blomberg et al. 2004: 477). Lastly, Freytag et al. (2009) find economic development to be related to terrorist activity.

On the other hand, research indicates that there is no connection linking poverty or economic conditions to terrorism. Krueger and Maleckova (2002) examine participants in Hezbollah’s militant wing and find that there is little connection between poverty, education and terrorism. Participants were found to have an equal chance of coming from rich families as from impoverished families. Additionally, education was not a determining factor of participation in terrorism (Krueger and Maleckova 2002: 29). As mentioned, other studies point to other factors being much more important in determining the incidence of terrorism.

Now that both parts of the causal mechanism of poverty have been explained through a literature review the second hypothesis can be devised as follows:

Hypothesis 2a: The effect of the welfare state on terrorism runs through the negative effect the welfare state has on poverty and the positive effect poverty has on terrorism. As mentioned before the effects of policy take time to work through on other factors. Accordingly, changes in the welfare state take time to affect poverty and poverty takes time to affect terrorism. Once again, it is difficult to know much time is needed. It might take years for social welfare policy to reduce poverty and for poverty to increase terrorism. But effects may also be seen quickly. Two additional hypotheses are added to hypothesis 2:

Hypothesis 2b: Changes in poverty take years to affect the incidence of terrorism. Hypothesis 2c: Changes in poverty quickly affect the incidence of terrorism.

2.3. Inequality

The second intermediating variable examined here is inequality. The relevant debate is twofold. On the one hand, the relationship between the welfare state and terrorism is of importance. On the other hand the effect of inequality on the incidence of terrorism needs to be highlighted. As with the poverty connection, the inequality connection is not straightforward and is even more so, questionable. This section will give an overview of the scholarly debate surrounding the welfare state, inequality and terrorism.

The first question that needs to be answered is whether welfare states reduce income inequality. Apart from the relevance for this study, this question is also important for the debate on the effectiveness and consequences of the welfare state itself, as one of the main functions of the welfare state is to reduce income inequality. Income inequality arises because of differences in capacities, socio-economic background, opportunities and because of the market process (Caminada and Goudszwaard 2009b: 2). Governments try to accomplish the goal of inequality

(11)

10 reduction through the tax system and through social welfare policies. But whether a reduction of income inequality is actually achieved by the welfare state is not evident. Some scholars conclude welfare states are effectively eradicating income inequality (Caminada and Goudszwaard 2009b; Förster and Pearson 2002; Korpi and Palme 1998). Whilst some argue social policy actually increases inequality in developing countries (Birdsall and James 1990; Castro-Leal et al. 1999).

In response to a shift from public to private social spending in the OECD-countries, Caminada and Goudszwaard (2009b) examine the difference in income redistribution between public social expenditure and private social expenditure. For public social expenditure they find a significant positive relationship with income redistribution. Countries in which social expenditure is higher income redistribution is also higher. Income inequality measured by the Gini-coefficient is reduced by 8 to 46 percent in all OECD-countries. Private social expenditure has a (weak) significant negative effect on income distribution. Thus, in countries with more public social expenditure income inequality reduction is less (Caminada and Goudszwaard 1990b: 17). This research does however confirm the link between welfare states and inequality. Furthermore, OECD research by Förster and Pearson (2002) has shown that although market income inequality has risen in OECD countries, social welfare policies have become more effective in reducing inequality and thus disposable income inequality has not increased (Förster and Pearson 2002: 36). Korpi and Palme (1998) investigate whether different welfare state institutions have different effects on inequality and poverty. They find that welfare states that target benefits at the poor or use flat-rate benefits are less efficient in reducing poverty and inequality. Encompassing welfare states that provide universal benefits and earnings-related benefits are better at redistributing income as to create more income equality (Korpi and Palme 1998: 681). This finding also supports the idea that welfare states achieve inequality reduction through their redistributive social policy.

There is, however, evidence that social welfare policy may have a positive effect on inequality in developing countries. For instance, Castro-Leal et al. (1999) find that health care and education spending in African countries benefit only the rich, thereby increasing income inequality. This problem arises because the poor don’t have proper access to health and education (Castro-Leal et al. 1999: 68-69). However, this reversed effect of social welfare policy is not an issue for this study as a different measure of the welfare state is used, which controls for the coverage of social policy. Also, this study only looks at the 18 most developed countries.

The second question addressed here is whether inequality may have a stimulating effect on terrorism. Once again, research has not provided definitive answers to this question. The debate also overlaps the debate on poverty and terrorism and the more comprehensive debate on whether economic factors influence terrorism.

(12)

11 Inequality may, even more than poverty, spark feelings of relative deprivation. Relative deprivation arises when people see that they have less than those around them but feel that they are entitled to more (Yitzhaki 1979: 321). Thus, when income inequality is large, feelings of relative deprivation may be more present. As argued before, feelings of relative deprivation or discontent about the economic circumstances may fuel terrorism.

There is not much evidence on the relationship between inequality and terrorism. Li and Schaub (2004) find that income inequality does have a significant positive effect on terrorism (Li and Schaub 2004: 251). Ehrlich and Liu (2002) argue that transnational terrorism may occur because of socio-economic factors like inequality. Here, country differences are more important than domestic socio-economic conditions. Another study by Krieger and Meierrieks (2010) concentrating only on income inequality and terrorism shows robust positive relationships between income inequality, measured in two different ways, and the incidence of terror (Krieger and Meierrieks 2010: 4). They also find that the intensity of terrorist attacks is positively influenced by the level of income inequality in a country (Krieger and Meierrieks 2010: 4). Moreover, there is much evidence that income inequality leads to discontent, political instability and finally violence (Burgoon 2006: 181). Thus, there are reasons to expect a positive relationship between income inequality and terrorism.

This overview has shown the empirical evidence and the theory behind the relationship between the welfare state, inequality and terrorism. Although, Burgoon argues the effect of social policy on inequality is not clear, more recent research by Caminada and Goudszwaard (2009b) has shown that there is a negative effect between the two variables. Also, by using a more sophisticated measure of the welfare state rather than merely looking at social expenditure, possible positive effects of social policy on inequality because of targeting of the rich are accounted for. Therefore, the hypothesis formulated here does suggest a causal direction in the relationship between the welfare state and inequality:

Hypothesis 3a: The effect of the welfare state on terrorism runs through the negative effect the welfare state has on inequality and the positive effect poverty has on terrorism. Once more, the time it takes for policy to take into effect is an issue. Additional hypothesis are added to take this into account:

Hypothesis 3b: Changes in inequality take years to affect the incidence of terrorism. Hypothesis 3c: Changes in inequality quickly affect the incidence of terrorism. A schematic overview of the theoretical framework presented here is shown in figure 2.

(13)

12

3. Data and method

The analysis is done for 18 industrialized countries in the period 1971 and 2002. The countries being: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Sweden, Switzerland, United Kingdom, United States. These are the 18 countries for which data on decommodification is available. This data is available for the period 1971-2002 as are data on terrorism, poverty and inequality.

3.1. Dependent variable

In his study, Burgoon (2006) uses three different measures for terrorist incidence: total terrorist incidents occurring in a country, transnational incidents occurring in a country and transnational terrorist incidents by country where the terrorist comes from. The effects of social welfare are significant on all three measures. In this study terrorism will be measured solely by total terrorist incidents, as this is the most encompassing measure. Data from the National Consortium for the Study of Terrorism and Responses to Terrorism (START); the Global Terrorism Database (GTD) is used. This data set has collected data on all terrorist incidents since

Figure 2: Theoretical framework showing the relationship between the welfare state and terrorism. Welfare states are thought to reduce terrorism through having an effect on poverty and inequality. These are economic determinants of terrorism. Scholars have however argued that political determinants may be more decisive in causing terrorist activity.

(14)

13 1970. The definition of terrorism used by GTD is ‘the threatened or actual use of illegal force and violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation.’ (Global Terrorism Database 1970)1_{. Figure 3 shows the total terrorist} incidents in the 18 welfare states from 1971 to 2002.

Figure 3: Total terrorist incidents in 18 countries, 1971-2002. Source: National Consortium for the Study of Terrorism and Responses to Terrorism(START); the Global Terrorism Database (GTD).

3.2. Independent variable

The main independent variable in this study is the welfare state, measured as decommodification. The decommodification index used in this study has been revised and

1_{A second data set is used to control for differences in measurement. This is the RAND Database of Worldwide}

Terrorism Incidents (RDWTI), also used by Burgoon (2006). RAND defines terrorism as follows: ‘…the deliberate

creation and exploitation of fear through violence or the threat of violence in the pursuit of political change.’ (Hoffman

1986). The largest difference between this definition and that of GTD is that the latter only includes incidents of political nature, whereas the former includes political, economic, religious, and social goals and is therefore more encompassing. This dataset gives the similar results as the GTD-dataset, results of this analysis will thus not be reported further. 0 100 200 300 T o ta l te rr o ri st i n cid e n ts 1970 1980 1990 2000 Year

(15)

14 recalculated by Scruggs and Allan (2006a). However, the key characteristics of the Scruggs and Allan-index are similar to Esping-Andersen’s index. De index is based upon three social welfare programs, namely pensions, sickness and unemployment benefits (Scruggs and Allan 2006a: 57). The most important feature considered in the index is the net replacement rate for the three programs. The replacement rate is ‘the ratio of the after-tax benefit payable to a typical (single) worker…to their after-tax income.’ (Scruggs and Allan 2006a: 57). The second feature of the index are the conditions that apply for qualification, also included are waiting time and benefit duration (Scruggs and Allan 2006a: 58). The third feature of the index is the coverage rate. This is the fraction of the relevant population covered by pensions, sickness or unemployment programs (Esping-Andersen 1990: 49)2_.

3.3. Intermediate variables

The first intermediate variable is poverty, data from the OECD is used. Poverty is measured as the percentage of the population with a posttax/posttransfer income below 50 percent of the current median income and is thus a relative measure of poverty. Absolute poverty (percentage of median income in a set country) might be a better measure of poverty as it takes into account indirect dynamic effects of social welfare policy. When using a relative poverty line (50% of median income in home country) effects on overall economic growth or decline are not included (Kenworthy 1999: 1123-1124). However, data on absolute poverty is not available for the all years examined in this study. Data is only available for periods of five years; missing years are therefore linearly interpolated and extrapolated. Switzerland and New Zealand are omitted, as data is not sufficient to inter- and extrapolate.

The second independent variable is inequality, defined as income inequality. Inequality is measured using the Gini-index, an often used index ranging from 0 (absolute equality) to 100 (absolute inequality). Data comes from the Estimated Household Income Inequality Database (EHIID) composed by Galbraith and Kum (2004). This dataset has the largest empirical coverage over time but still doesn’t cover the entire period. Mainly, the last few years in the period are missing for some countries and data from Switzerland is completely missing. This variable thus contains one country less and also misses observations for a few country-years.

3.4. Statistical method

Because decommodification data is only available for the 18 welfare states this minimizes the number of observations available, which in turn makes it harder to find significant relationships.

2_{Scruggs and Allan (2006a) also devised another scale: benefit generosity. This measure is similar, but according to}

Scruggs and Allan, more sophisiticated than decommodification. For validity-purposes the statistical analyses were done for benefit generosity as well. Using this scale of yields similar results as decommodification and thus will not be reported on.

(16)

15 However, using a pooled cross-section time series data set greatly increases the number of observations as every country-year becomes one observation. The effect of the welfare state on terrorism is estimated with negative binomial regression. As Burgoon explains, standard ordinary least squares (OLS) regression is not reliable due to the dependent variable being a count-variable (Burgoon 2006: 190). Negative binomial regression adjusts for overdispersion in the count of terrorist incidents. Overdispersion takes place when the variance of the count is greater than the mean (Crenshaw et al. 2007: 8)

To account for the time it takes for social welfare policy to affect terrorism (and poverty and inequality) all variables are lagged. Lags of 1, 5 and 10 years are used because it may take a years before a person decides to commit an assault due to discontentment about economic conditions. Trending patterns in terrorism are also accounted for by including autoregressive effects. This is achieved by adding 1-year lagged terrorism incidents as independent variable. Hence, the effects on terrorist incidents by previous incidents are taken into account. Furthermore, year dummies are added to account for variance in particular years. Although, there is no consensus on the value of adding year dummies, they will be used here to check the maintainability of the models. Terrorist incidence might be exceptionally high in some years and year dummies resolve this. Year dummies will be added in a separate model. Regional dummies are not used because the countries analyzed are not as dispersed over different regions than larger samples are. Thus, there are no theoretical reasons to take regional differences into account.

4. Results

Table 1 shows the results of the cross sectional pooled time series regression analysis. Models 1 to 3 contain only the main effect ‘decommodification’ each model using a different lag (1, 5 and 10 years). In models 4 to 6 poverty rates are added in the analysis, also using different lags. Models 7 to 9 look at the effect of inequality at different lags. Model 10 consists of the main effect and both intermediate variables.

Decommodification does not have a significant effect on the incidence on terrorism when lagged 1 or 5 years. However, a significant negative effect is found when lagging decommodification by 10 years. This supports hypothesis 1a and 1b: the extensiveness of the welfare state has a negative effect on the incidence of terrorism and it takes years for changes in the extensiveness of the welfare state to affect terrorism. As for the interpretation of the coefficients, negative binomial regression estimates the effects of independent variables on the logged number of events of the dependent variable. Table 2 shows all significant models estimated in ‘incident rate ratios’ which easier to interpreted. Using these coefficients one may conclude that for every 1-point increase on the decommodification scale terrorist incidents

(17)

16 occur 0.967 times as much. Thus, the welfare state does seem to reduce terrorism, but the effect is modest. Because the effect of decommodification is only significant when lagged 10 years, it takes a long time before a change in a country’s decommodification-score (i.e. the extensiveness of the welfare state) actually affects terrorism.

Table 1

Cross-sectional pooled time-series of total terrorist incidents in countries (multiple lags)

1 2 3 4 5 6 7 8 9 10 Terrorist incidentst-1 .008*** (.001) .009*** (.001) .012*** (.001) .012*** (.001) .012*** (.001) .012*** (.001) .011*** (.001) .011*** (.001) .011*** (.001) .011*** (.001) Decommodificationt-1 .004 (.014) Decommodificationt-5 -.015 (.014) Decommodificationt-10 -.034* (.014) -.020 (.017) -.023 (.017) -.021 (.017) -.021 (.016) -.029 (.015) .031* (.015) -.005 (.017) Povertyt-1 .045 (.024) .080* (.033) Povertyt-5 .037 (.025) Povertyt-10 .045 (.024) Inequalityt-1 .039* (.019) .000 (.026) Inequalityt-5 .017 (.019) Inequalityt-10 .008 (.019) Constant .083 (.356) .562 (.379) 1.07** (.380) .276 (.582) .437 (.580) .333 (.561) -.575 (.829) .366 (.787) .691 (.747) -.347 (.897) Year dummies No No No No No No No No No No Log-likelihood -1507 -1335 -1080 -1023 -1019 -1013 -939 -998 -1027 -932 Mean variance

inflation factor (VIF)

1.05 1.04 1.03 1.78 1.60 1.39 1.16 1.13 1.11 1.77

N 558 486 396 370 365 360 340 365 374 335

NOTE: Negative binomial regression. Dependent variable: Total terrorist incidents in a country, 1971-2002 (Global

Terrorism Database). Independent variables: Terrorist incidentst-1, total terrorist incidents in a country lagged by 1

year (Global Terrorism Database) Decommodificationt-x, by country, 1971-2002, lagged by 1, 5 and 10 years

(Scruggs and Allan 2006). Povertyt-x, percentage of the population with a posttax/posttransfer income below 50

percent of the current median income, lagged by 1, 5 and 10 years, 1971-2002 (OECD). Inequalityt-x, gini index,

lagged by 1, 5 and 10 years (EHIID, Gailbrath and Kum 2004). *p<0,05; **p<0,01; ***p<0,001, two-tailed tests

Model 4 to 6 show whether the effect of decommodification on the terrorism actually runs through poverty. This seems to be the case, as the effect of decommodification becomes

(18)

17 insignificant when poverty is added to the model. However, all three lagged poverty-variables do not have a significant effect on terrorism. In model 4, the p-value of poverty is 0.067 and in model 6, poverty has a p-value of 0.059, thus only one-tailed tests would give significant effects. The coefficients of poverty are in the hypothesized direction: the higher poverty rates, the higher terrorist incidents. But as all effects are insignificant, all three hypotheses concerning the poverty connection (2a, 2b and 2c) must be rejected.

Table 2

Cross-sectional pooled time-series of total terrorist incidents in countries (incident rate ratio) 3 4 7 10 Terrorist incidentst-1 1.01*** (.001) 1.01*** (.001) 1.01*** (.001) 1.01*** (.001) Decommodificationt-10 .967* (.014) .981 (.016) .979 (.015) .995 (.017) Povertyt-1 1.05 (.026) 1.08* (.035) Inequalityt-1 1.04* (.019) 1.00 (.026) Constant 2.91** (1.11) 1.32 (.766) -.563 (.467) .707 (.634) Log likelihood -1080 -1023 -939 -932 Mean variance inflation factor (VIF)

1.05 1.78 1.16 1.77

N 396 370 340 335

NOTE: Negative binomial regression. Dependent variable: Total terrorist incidents in a country,

1971-2002 (Global Terrorism Database). Independent variables: Terrorist incidentst-1, total terrorist

incidents in a country lagged by 1 year (Global Terrorism Database) Decommodificationt-10, by

country, 1971-2002, lagged by 10 years (Scruggs and Allan 2006). Povertyt-1, percentage of the

population with a posttax/posttransfer income below 50 percent of the current median income, lagged by 1 year, 1971-2002 (OECD). Inequalityt-1, gini index, lagged by 1 year (EHIID, Gailbrath and

Kum 2004).

*p<0,05; **p<0,01; ***p<0,001, two-tailed tests

Model 7 to 9 show that inequality has a significant positive effect on the incidence of terrorism when lagged for 1 year. With lags of 5 and 10 years the effect of inequality is insignificant. The effect is not very strong, when inequality increases by 1-point on the gini-index, terrorism occurs 1.04 times as much. The effect of decommodification becomes insignificant when inequality is added to the model indicating an intermediate role for inequality. This supports the third main hypothesis of this study; the effect that the

(19)

18 extensiveness of the welfare state has on terrorism runs through inequality. As for the time it takes to affect terrorism, only inequality lagged by 1 year is significant. Hypothesis 3c can thus be accepted, whereas hypothesis 3b can be rejected.

Model 10 shows the effects of all three explanatory variables. The effect of decommodification is insignificant. Poverty has a significant positive effect on the incidence of terrorism, whereas inequality has an insignificant effect on terrorism. Multicollinearity between the explanatory variables is not present as the variance inflation factor (vif) is 1.16 (the boundary is 10 or higher). It is difficult to find a theoretical explanation for poverty suddenly having a significant effect and inequality becoming insignificant. The cause of this might be found in inaccuracies in data of poverty and inequality. But this is not certain.

The analyses have also been done including year dummies (appendix table 1), however most effects become insignificant when using year dummies. This is a familiar problem as year dummies absorb a lot of variance out of the data. In this study, this effect is even stronger because of the relatively few countries used. There is much methodological debate on the use of dummies in time-series analyses. Some scholars even argue that year dummies should not be used. Plümper et al. (2005) argue that ‘the inclusion of a lagged dependent variable and/or period dummies tends not only to absorb large parts of the trend in the dependent variable, but likely biases estimates’ (Plümper et al. 2005:). The results found when using year dummies may not necessarily mean the hypotheses of this study are wrongly accepted. However, they should not be ignored and therefore the robustness of the remaining evidence is weakened.

Lastly, separate regressions are done for the link between the welfare state and poverty, welfare state and inequality, poverty and terrorism and inequality and terrorism. The results are shown in the appendix in table 5, 6 and 7. From these results we can conclude that the welfare state does not significantly affect poverty and inequality. Poverty does not affect terrorism, which is in line with the results found in models 4, 5 and 6. Inequality does seem to have a significant positive effect on terrorism.

5. Conclusions and implications

The purpose of this article was to find more evidence regarding Burgoon’s (2006) proposed relationship between the welfare state and terrorism. Several findings have been made. Most importantly, the negative relationship between the welfare state and terrorism holds up when using a more sophisticated measure of the welfare state than social expenditure. However, the relationship is far less strong than argued by Burgoon (2006). An explanation for this can be found in the use of a much smaller sample size. Only the 18 most developed countries were researched. The fact that there is a significant but modest relationship between the welfare state and terrorism in these countries does offer support for Burgoon’s theory, as it may be harder to

(20)

19 find significant effects in these countries. However, as mentioned, Burgoon argues economic insecurity may be a source of terrorism and a causal link through which the welfare state affects terrorism. Decommodification, used to measure the welfare state, incorporates economic insecurity; the higher decommodification is, the lower economic insecurity in a country should be. If economic insecurity indeed fuels terrorism, the effect of decommodification should be stronger than the effect of social expenditure. At least three explanations can be given for this not being the case. First, economic insecurity is not a determinant of terrorism. Second, higher decommodification does not lead to lower economic insecurity and third, the link between the welfare state and terrorism is just not as strong and universal as Burgoon (2006) argues. The latter explanation is also supported by the models in which year dummies were used; the effect of decommodification becomes insignificant.

Two other causal mechanisms, distinguished by Burgoon (2006), were studied, namely poverty and inequality. The theory behind these mechanisms is that welfare states reduce poverty and inequality, which are determinants of terrorism. Poverty results were all insignificant indicating that poverty doesn’t seem to be a determinant of terrorism. However, decommodification did become insignificant when poverty was added. The effect of decommodification does seem to run through poverty but is presumably not strong enough.

Regarding the inequality link, expectations were met. Inequality has a significant positive effect on the incidence of terrorism. Futhermore, the effect of decommodification on terrorism does seem to run through income inequality. This means that a more generous welfare state may reduce terrorism by reducing income inequality. Theories on how inequality may fuel terrorism state that inequality causes relative deprivation which in turn leads to conflict. As for the time lag; it takes years for changes in the welfare state to affect inequality, but terrorism quickly adjusts to changes in income inequality.

What are the implications of these results? As discussed earlier, this study covers three debates: (1) the debate on the relationship of welfare states and terrorism, (2) the debate on whether welfare states reduce or increase poverty and inequality and (3) the debate on whether economic conditions like poverty and inequality fuel terrorism. The results have implications for each of these. First off, a contribution is made to the growing interest in the link between the welfare state and terrorism. This study has shown that even when using a refined measure of the welfare state, a modest, but significant negative relationship with terrorism holds up. It is thus more plausible that welfare states indeed reduce terrorist activity. This implies that social welfare policy may be an effective strategy to fight terrorism. As for the second debate, the results are less clear. The analyses done in this study have not given any significant relationships between welfare states and poverty and welfare states and inequality. But this doesn’t necessarily mean there is no relationship, as many other studies have found welfare states to be

(21)

20 poverty and inequality reducing. Moreover, the poverty and inequality data used were not ideal, given the fact that interpolaion was necessary to reduce missing data. Thirdly, economic conditions do seem to matter for terrorism. Naturally, economic conditions do not solely determine terrorism. Political factors and sociological factors presumably also have a share in fueling terrorism. But economic conditions are not, as some scholars have argued, completely irrelevant.

6. Discussion

As with all studies in social sciences, some remarks are in place. Although, the effects found are significant they are also mostly weak. Implying that terrorism has far more important determinants than social welfare policy. The effects also became insignificant when using year dummies, which also points to weak relationships. As for the data, inequality and poverty data were incomplete. Missing inequality data was kept but poverty data was interpolated and extrapolated linearly and this affects the reliability of the relationships found. This may of course also be a reason for the fact that no significant results for the poverty connection were found. Much more research is thus needed to provide strong evidence for the relationships discussed here. For example, this study could be extended to more countries instead of the 18 most developed welfare states. More research into the remaining causal mechanisms Burgoon (2006) distinguishes will also provide more understanding. Furthermore, qualitative research on the motives of terrorists will give more insight into the determinants of terrorism and may ultimately provide strategies for fighting terrorism. However, this study does provide a reason to seriously look into social welfare policy as a terrorism-reducing strategy.

7. References

Abadie, A. (2006) ‘Poverty, Political Freedom, and the Roots of Terrorism’, American Economic Review, 96(2), 50-56.

Barro, Robert, and Jong-Wha Lee (1993) ‘Losers and winners in economic growth’, National Bureau of Economic Research, Working paper no. 4341.

Burgoon, Brian (2006) ‘On Welfare and Terror’, Journal of Conflict Resolution, 50(2), 176-203. Caminada, Koen and Goudszwaard, Kees (2009a) ‘Effectiveness of Poverty Reduction in the EU:

A Descriptive Analysis’, Poverty & Public Policy, 1(2), 1-49.

Caminada, Koen and Goudszwaard, Kees (2009b) ‘The redistributive effect of public and private social programmes: A cross-country empirical analysis’, International Social Security Review, 63, 1-19.

(22)

21 Castro-Leal, Florencia; Dayton, Julia and Mehra, Lionel Demery Kalpana (1999) ‘Public Social

Spending in Africa: Do the Poor Benefit?’ World Bank Research Observer, 14(1), 49-72. Chen, Daniel L. (2003) ‘Economic Distress and Religious Intensity: Evidence from Islamic

Resurgence During the Indonesian Financial Crisis’, Harvard University, PRPES Working Paper No. 39.

Crenshaw, E.M.; Robison, K.K. and Jenkins, J.G. (2007) ‘The ’Roots’of Terrorism: A Replication and Extension of Burgoon’, Paper presented at the Annual Meetings of the Sociological Association, New York.

Ehrlich, Paul R. and Liu, Jianguo (2002) ‘Some roots of terrorism’, Population and Environment, 24 (2), 183-92.

Esping-Andersen, G. (1990)‘De-Commodification in Social Policy’ in The Three Worlds of Welfare Capitalism, Cambridge: Polity Press, 35-54.

Förster, Michael and Pearson, Mark (2002) ‘Income Distribution and Poverty in the OECD Area: Trends and Driving Forces’, OECD Economic Studies, 34, 1-39.

Freytag, Andreas; Krüger, Jens J.; Meierrieks, Daniel and Schneider, Friedrich G. (2009). ‘The Origins of Terrorism: Cross-Country Estimates on Socio-Economic Determinants of Terrorism’, Jena Economics Research Papers, no. 2009-009

Galbraith, James K. and Kum, Hyunsub (2004) ‘Estimating the Inequality of Household Incomes: A Statistical Approach to the Creation of a Dense and Consistent Global Data Set’, UTIP, Working paper no. 22.

Hoffman, Bruce (1986) ‘Defining Terrorism’, Social Sciences Record, 24(1), 6-7.

Kenworthy, Lane (1999) ‘Do Social Welfare Policies Reduce Poverty?’, Social Forces 77(3), 1119-39.

Korpi, W. and Palme, J. (1998), ‘The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries’, American Sociological Review, 63, 661- 687.

Krieger, Tim and Meierrieks, Daniel (2009) ‘Terrorism in the Worlds of Welfare Capitalism’ Center for International Economics, Working paper no. 2009-04.

Krieger, Tim and Meierrieks, Daniel (2010) ‘Does Income Inequality Lead to Terrorism?’, University of Paderborn, Working paper.

Kurrild-Klitgaard, P.; Justesen, M.K. and Klemmensen, P. (2006). ‘The political economy of freedom, democracy and transnational terrorism’, Public Choice, 128, 289-315.

Li, Quan and Schaub, Drew (2004) ‘Economic Globalization and Transnational Terrorism: A Pooled Time-Series Analysis’, Journal of Conflict Resolution, 8(2), 230-258.

Lindert, Peter H. (2004) ‘Patterns and Puzzles’ in Growing Public: Social Spending and Economic Growth Since the 18th_{Century, Cambridge University Press, 3-19.}

(23)

22 Lee, Dwight R. (1987) ‘The Tradeoff between Equality and Efficiency: Short-Run Politics and

Long-Run Realities’, Public Choice, 53(2), 149-165.

Moller, Stephanie; Huber, Evelyne; Stephens, John D.; Bradley David and Nielsen, François (2003) ‘Determinants of Relative Poverty in Advanced Capitalist Democracies’, American Sociological Review, 68(1), 22-51.

National Consortium for the Study of Terrorism and Responses to Terrorism (START), (2013) Global Terrorism Database. Retrieved from http://www.start.umd.edu/gtd

OECD (2013) Dataset: Income distribution and Poverty Income distribution – Poverty. Retrieved from: http://stats.oecd.org/Index.aspx?DatasetCode=POVERTY

Plümper, Thomas; Troeger, Vera E. and Manow, Philip (2005) ‘Panel data analysis in comparative politics: Linking method to theory’ European Journal of Political Research 44, 327–354.

RAND (2013) Database of Worldwide Terrorism Incidents (RDWTI). Retrieved from: http://www.rand.org/nsrd/projects/terrorism-incidents.html

Scruggs, Lyle and Allan, James (2006a) ‘Welfare-state decommodification in 18 OECD countries: a replication and revision’, Journal of European Social Policy, 16(1), 55-72.

Scruggs, Lyle and Allan, James (2006b) ‘The Material Consequences of Welfare States Benefit Generosity and Absolute Poverty in 16 OECD Countries’, Comparative Political Studies, 39(7), 880-904

Wolfensohn, James D. (2008) ‘Fight Terrorism by Ending Poverty’, New Perspectives Quarterly, 19(2), 42-44.

Yitzhaki, Shlomo (1979) ‘Relative Deprivation and the Gini Coefficient’, The Quarterly Journal of Economics, 93(2), 321-324.

(24)

23

8. Appendix

8.1. Statistical analyses

Table 1

Cross-sectional pooled time-series of total terrorist incidents in countries

1 2 3 4 Terrorist incidentst-1 .012*** (.001) .012*** (.001) .011*** (.001) .011*** (.001) Decommodificationt-10 -.339* (.014) -.020 (.017) -.021 (.016) -.005 (.017) Povertyt-1 .045 (.024) .080* (.033) Inequalityt-1 .039* (.019) .000 (.026) Constant 1.069** (.380) .276 (.582) -.575 (.829) -.347 (.897) Log likelihood -1080 -1023 -939 -932 Mean variance inflation factor (VIF)

1.05 1.78 1.16 1.77

N 396 370 340 335

Kum 2004).

(25)

24

Table 2

Cross-sectional pooled time-series of total terrorist incidents in countries (incident rate ratio) 1 2 3 4 Terrorist incidentst-1 1.01*** (.001) 1.01*** (.001) 1.01*** (.001) 1.01*** (.001) Decommodificationt-10 .967* (.014) .981 (.016) .979 (.015) .995 (.017) Povertyt-1 1.05 (.026) 1.08* (.035) Inequalityt-1 1.04* (.019) 1.00 (.026) Constant 2.91** (1.11) 1.32 (.766) -.563 (.467) .707 (.634) Log likelihood -1080 -1023 -939 -932 Mean variance inflation factor (VIF)

1.05 1.78 1.16 1.77

N 396 370 340 335

Kum 2004).

(26)

25

Table 3

Cross-sectional pooled time-series of total terrorist incidents in countries (multiple lags)

1 2 3 4 5 6 7 8 9 10 Terrorist incidentst-1 .008*** (.001) .009*** (.001) .012*** (.001) .012*** (.001) .012*** (.001) .012*** (.001) .011*** (.001) .011*** (.001) .011*** (.001) .011*** (.001) Decommodificationt-1 .004 (.014) Decommodificationt-5 -.015 (.014) Decommodificationt-10 -.034* (.014) -.020 (.017) -.023 (.017) -.021 (.017) -.021 (.016) -.029 (.015) .031* (.015) -.005 (.017) Povertyt-1 .045 (.024) .080* (.033) Povertyt-5 .037 (.025) Povertyt-10 .045 (.024) Inequalityt-1 .039* (.019) .000 (.026) Inequalityt-5 .017 (.019) Inequalityt-10 .008 (.019) Constant .083 (.356) .562 (.379) 1.07** (.380) .276 (.582) .437 (.580) .333 (.561) -.575 (.829) .366 (.787) .691 (.747) -.347 (.897) Year dummies No No No No No No No No No No Log-likelihood -1507 -1335 -1080 -1023 -1019 -1013 -939 -998 -1027 -932 Mean variance

1.05 1.04 1.03 1.78 1.60 1.39 1.16 1.13 1.11 1.77

N 558 486 396 370 365 360 340 365 374 335

(27)

26

Table 4

Cross-sectional pooled time-series of total terrorist incidents in countries with year dummies (multiple lags) 1 2 3 4 5 6 7 8 9 10 Terrorist incidentst-1 .008*** (.001) .009*** (.001) .009*** (.001) .008*** (.001) .009*** (.001) .009*** (.001) .009*** (.002) .008*** (.001) .008*** (.001) .009*** (.002) Decommodificationt-1 .015 (.017) Decommodificationt-5 .015 (.018) Decommodificationt-10 -.001 (.018) -.035 (.023) .025 (.024) -.029 (.023) .017 (.021) .015 (.020) .013 (.020) .038 (.024) Povertyt-1 .053 (.028) .059* (.030) Povertyt-5 .031 (.028) Povertyt-10 .050 (.029) Inequalityt-1 .026 (.023) .018 (.024) Inequalityt-5 .011 (.022) Inequalityt-10 .006 (.021) Constant -.499 (.424) .114 (.429) .939* (.423) -.399 (.694) .018 (.711) -.238 (.674) -.455 (.983) .165 (.895) .361 (.841) -1.19 (1.09)

Year dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Log-likelihood -1447 -1285 -1032 -977 -974 -967 -904 -953 -981 -898 Mean variance

1.90 1.89 1.87 2.04 1.97 1.91 1.81 1.86 1.85 1.95

N 558 486 396 370 365 360 340 365 374 335

(28)

27

Table 5: Cross-sectional pooled time-series of poverty Poverty Povertyt-1 .986*** (.008) Decommodificationt-10 -.008 (.005) Constant .422* (.185) Mean variance inflation factor (VIF)

2.13

N 369

NOTE: *p<0,05; **p<0,01; ***p<0,001, two-tailed

tests

Table 6: Cross-sectional pooled time-series of inequality Inequality Inequalityt-1 .959*** (.018) Decommodificationt-10 -.018 (.012) Constant 2.11** (.780) Mean variance inflation factor (VIF)

1.20

N 320

NOTE: *p<0,05; **p<0,01; ***p<0,001, two-tailed

tests

Table 7: Cross sectional pooled time-series of total terrorist incidents

Poverty Inequality Terrorist incidentst-1 .009*** (.001) .009*** (.001) Povertyt-1 .013 (.017) Inequalityt-1 .032* (.015) Constant .051 (.191) -.921 (.557) Mean variance inflation factor (VIF)

1.04 1.02

N 514 493

NOTE: Negative binomial regression

*p<0,05; **p<0,01; ***p<0,001, two-tailed tests

8.2 Codebook

Variable Description

year Year 1971-2002

country_id Country abbreviation

country Country (18) Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Great Britain, Ireland, Italy, Japan, Netherlands,

(29)

28 New Zealand, Norway, Sweden, Switzerland, United States

gtd_totalincidents The total number of terrorist incidents that occurred in one country from 1971-2002.

Source: National Consortium for the Study of Terrorism and Responses to

Terrorism (START), (2013) Global Terrorism Database. Retrieved from http://www.start.umd.edu/gtd

gtd_killed_total The total number of persons killed by a terrorist incident in one country from 1971-2002.

Source: National Consortium for the Study of Terrorism and Responses to

Terrorism (START), (2013) Global Terrorism Database. Retrieved from: http://www.start.umd.edu/gtd

rand_totalincidents The total number of terrorist incidents that occurred in one country from 1971-2002.

Source: RAND (2013) Database of Worldwide Terrorism Incidents (RDWTI).

Retrieved from: http://www.rand.org/nsrd/projects/terrorism-incidents.html

decom Decommodification scale per country from 1971-2002

Source: Scruggs, Lyle (2004) Welfare State Entitlements Data Set: A Comparative

Institutional Analysis of Eighteen Welfare States. Scruggs, Lyle and Allan, James (2006a) ‘Welfare-state decommodification in 18 OECD countries: a replication and revision’, Journal of European Social Policy, 16(1), 55-72.

generosity Benefit generosity scale per country from 1971-2002

Source: Scruggs, Lyle (2004) Welfare State Entitlements Data Set: A Comparative

Institutional Analysis of Eighteen Welfare States. Scruggs, Lyle and Allan, James (2006a) ‘Welfare-state decommodification in 18 OECD countries: a replication and revision’, Journal of European Social Policy, 16(1), 55-72.

oecd_poverty Percentage of the population below the poverty line of 50% of the current median income per country per 5 years from 1975-2000

Source: OECD Dataset: Income distribution and Poverty Income distribution –

Poverty. Retrieved from:

http://stats.oecd.org/Index.aspx?DatasetCode=POVERTY

ipoverty Interpolated version of oecd_poverty

poverty Interpolated and extrapolated version of oecd_poverty. Percentage of the population below the poverty line of 50% of the current median income per country per year from 1971-2002.

Missing values: interpolated and extrapolated

gini Gini-index, income inequality in a country on a scale from 0 to 100 (100= complete inequality) per country per year from 1971-2002.

Missing values: kept

Source: Galbraith, James K. and Kum, Hyunsub (2004) ‘Estimating the Inequality

of Household Incomes: A Statistical Approach to the Creation of a Dense and Consistent Global Data Set’, UTIP, Working paper no. 22.

Welfare states and terrorism : understanding the relationship between social welfare policy and terror