Abstract
It is often proposed that the young unemployed are more likely to engage in political violence, conflicts, and protests. One problem in studying the unemployed – especially in the Global South – are the blurred lines between the unemployed, the employed, and those working in the informal sector. Further, the employed are a heterogeneous group so employment quality might also play an important role. To tackle these issues, this study uses a new quantitative dataset, which covers youth in five Middle Eastern and North African countries: Algeria, Egypt, Lebanon, Morocco, and Tunisia. These data provide considerably more fine-grained information about the employment situations of the respondents than the datasets previously used. The study investigates separately two forms of political participation: in political violence and in demonstrations. The regression analyses show that there is no clear difference between the young unemployed and the young employed in their likelihood to participate in the studied political activities. However, some features related to employment matter. Those whose employment status is ambiguous are substantially more likely to participate in demonstrations and political violence than the employed. Among those who work, those who are dissatisfied with their work and those who work fewer hours participate more often in these activities. Income on its own does not seem to have an effect; however, those who have more assets are more likely to participate, and compared to those who feel themselves middle income, those feeling rich or poor are more likely to engage in political violence and demonstrations. The results suggest that instead of thinking in terms of a dichotomy of the employed and unemployed, more emphasis should be placed on understanding the variety of employment situations and employment quality and their impact on political instability.
Introduction
‘Youth unemployment is a ticking time bomb’, stated Alexander Chikwanda, the finance minister of Zambia. That bomb was about to explode in Africa and shake the political stability of the continent (Ighobor, 2013). Tapscott (2011) forecast that if youth unemployment rates in North America, Europe, and elsewhere remained high, ‘a generational conflict’ would follow: a wave of youth radicalization, mass protests, and uprisings against the authorities.
Many predictions in the early 2010s were inspired by movements such as the Arab Spring and the Occupy movement in 2011, which were largely attributed to the young unemployed (see e.g. Kabbani, 2019; Van Stekelenburg, 2012: 225–226).
However, the idea of unemployment leading to protests and conflicts has echoed among journalists, academics, and practitioners time after time. At a more general level, Abulkalam Abdul Momen, the Chairman of the UN Peacebuilding Commission, has stated that ‘job creation, especially for young people, in all post-conflict countries is an essential part of peacebuilding and more importantly, conflict prevention or relapsing into conflicts’ (United Nations, 2012). The relative deprivation theory provides theoretical grounds to argue that unemployment and other disappointments in labour markets could cause protests and conflicts.
As Cramer (2015) summarizes, ‘The notion that unemployment is a strong probable cause or motivating factor behind violence and violent conflict is remarkably pervasive in international development. [. . .] [T]his idea is based more on intuition and assumption than on evidence’. Indeed, the empirical evidence linking unemployment to political violence or to protests remains scant and contradictory (Cramer, 2011; Paasonen, 2020: 4–5).
Even though unemployment is often considered a factor behind riots and rebellions, it is not an easy task to even define who is unemployed. This becomes especially difficult when studying the Global South. In the developing world, being unemployed in the sense understood in the West is often impossible, as state welfare benefits are largely non-existent (Cramer, 2011: 15; United Nations, 2004: 67). The prevalence of informal work is another challenge in defining unemployment. In almost all Arab countries, more than 50% of employment is informal (International Labour Organization, 2018: 13).
Those working informally are often young, have several and changing sources of income, and work irregularly. In short, from the research perspective, these people falling outside employment and unemployment challenge the definitions. In many survey datasets, the respondents are simply provided the options of being unemployed or employed. However, those who report themselves as unemployed might still work informally, and those who answer that they are employed might work only occasionally.
Based on theoretical reasoning and numerous empirical studies, Cramer (2011) argued that working conditions and labour markets, as a whole, have significance for people’s participation in organized violence, but mere unemployment is not enough to explain this participation.
Further, Glasius and Pleyers (2013) argue that the post-2010 protests in the Arab World and elsewhere share the common feature of the mobilization of youth in precarious working conditions. ‘Precarious work’ refers to the variety of temporary and uncertain employment.
This article contributes to the earlier research in this area by focusing on the complexity of employment and unemployment status. The article investigates whether employment quality or issues related to the employment matter in terms of the likelihood of individuals taking part in political violence and demonstrations. As far as I know, my article is the first to study these relationships statistically with this extensive and fine-grained survey data. At least, this has not been done previously in relation to Middle Eastern and North African youth.
The empirical part of the article is based on the Sahwa Youth Survey which was collected in five Middle Eastern and North African countries, Algeria, Egypt, Lebanon, Morocco, and Tunisia, in 2015–16. The region and the data provide an especially interesting case to study political instability and employment among the youth.
The survey was conducted a few years after the Arab Spring. In Tunisia and Egypt, the Arab Spring toppled rulers who had governed for decades, and large demonstrations emerged also in Algeria, Lebanon, and Morocco. In the neighbouring countries Syria and Libya, civil wars have flared up. In short, there was a great deal of political turmoil in the region in the 2010s.
The Middle East and North Africa has also been experiencing a so-called youth bulge; in 2015 the share of those aged 15–29 was 38% of the adult population. Of the world regions, the share was higher only in South Asia and sub-Saharan Africa. The youth unemployment rate of the Middle East and North Africa in 2015 was 27%; this in turn was higher than in any other world region. Youth unemployment is widespread also in the studied countries; in Lebanon and Morocco, the unemployment rate in 2015 was around 20%, and in Algeria, Egypt, and Tunisia, it was 30% or more (World Bank, 2021).
In the literature, the youth bulge, unemployment, and an increase in years of schooling have been presented as drivers of the Arab Spring uprisings. Following the argument, if the labour markets are not able to absorb the youth reaching working age this results in high unemployment and frustration which again increases the likelihood of protests and political violence. Arguably, higher education further raised the expectations of people before the Arab Spring and thus unemployment was an even bigger disappointment (Campante and Chor, 2012; Urdal, 2012).
The rest of the article proceeds as follows. I first discuss the theoretical framework and justify each hypothesis. I then present the operationalizations of the variables. Thereafter, the results and the conclusions are presented.
Theoretical approach and hypotheses
This section presents theoretical arguments which form the basis to understand how unemployment and employment quality could be connected to protests and political violence.
Over the decades, the relative deprivation theory has been developed and adapted by hundreds of scholars in various fields of the social sciences, and therefore no single definition exists. In their review on the development of the theory, Smith et al. (2012: 2–3) define relative deprivation according to certain key conditions. Relative deprivation only applies to when a person compares their current situation to someone else’s situation or to the situation in which the person has been or could be. In addition, the person who performs the comparison has to think that their own current situation is worse than the situation compared and they also have to feel that their situation is unjust. Indeed, as the name of the theory indicates, it does not matter whether one possesses something in ‘absolute’ terms or whether a bystander would think somebody is deprived.
In peace and conflict research, the relative deprivation theory is best known through the work of Gurr (1970), who presented relative deprivation as a driver of political violence in his award-winning book Why Men Rebel. He states that relative deprivation can lead to frustration which provokes aggression, and aggression, in turn, can erupt as collective violence. Gurr (1970: 37), however, argues that the basic idea of relative deprivation sparking revolutions was presented already by Aristotle in the 4th century BCE. Runciman (1967), a notable developer of the theory, separated relative deprivation into two groups: one based on personal comparisons and one based on comparisons between groups.
Some previous studies have also utilized the theory to examine the connection of unemployment to protests and political violence (Richardson, 2011; Walker and Mann, 1987). Following the theory, an unemployed person can either experience relative deprivation after comparing their position to some employed individual or can feel relatively deprived as a part of a group, the unemployed. The comparison can also be temporal; for example, if an individual expected to get a job but failed to do so, or if they had a job which they expected to keep but lost, there is a gap between expected and realized conditions.
Still, unemployment as such is not the reason we expect the unemployed to protest or engage in political violence. Instead, it is what follows from being jobless. We assume the unemployed to have smaller or non-existent income, and therefore less property. This is thought to inflict the feeling of poverty, which supposedly leads the unemployed to feel themselves relatively deprived and consequently participate in protests and political violence.
Based on the relative deprivation theory, we can also expect employment quality to matter for participation in political violence and protests. Presumably, those who are working compare themselves to other people in their social circles who are working and people in other occupations. If people feel deprived following this comparison, the theory suggests that they take to the streets.
Of course, unemployment or low-quality employment is not just about money. Without decent earnings from a job, it is hard for a young person to move from home, get married, and establish a family. This prolonged period of not being able to proceed from youth to adulthood is sometimes called ‘waithood’, and waithood specifically is suggested to drive youth to protests and revolutions in the Arab region and elsewhere (see for example Honwana, 2014). From the relative deprivation perspective, those stuck in waithood are comparing themselves to the future they expected or to people around them and feel deprived. The notion of waithood highlights why the absence of decent work and money earned from it could be such an important factor in explaining behaviour of youth.
On the other hand, the formations of political actions are complex processes. As Murshed and Tadjoeddin (2009) point out in relation to civil wars, they do not break out just because of relative deprivation or grievances citizens face. Other factors such as natural resources, the political system, sharing revenues and practices to settle grievances play a role, too. Arguably, also in case of other forms of political violence and protests, looking at the relationship of individual-level traits provides only partial explanations. In other words, individuals do not participate in political actions in a vacuum. For example, if the authorities respond to peaceful demonstrations with violence, the protesters are more likely to turn to violent actions. On some occasions, states have also infiltrated violent provocateurs among the protesters hoping that the protests would turn violent and that repressing them would then seem legitimate. Whereas demonstrations can escalate into violence, violence can also provoke demonstrations. Both demonstrations and political violence are means to advance political goals. Nevertheless, demonstrations are considered part of a functioning democracy whereas political violence should not be accepted. Bearing this wider picture in mind, this study will pay attention to participation at the individual level.
Finally, participation of an individual is also a question of resources and identity. The importance of identity and resources like time, social contacts, and money are discussed more later when presenting each hypothesis in more detail. However, the studied employment-related variables and their assumed relationships are summarized in Figure 1. The hypotheses which study the relationships are in parentheses. Hypotheses 1 and 2 compare the participation among the employed, unemployed and those with ambiguous employment status. Hypotheses 3 and 4 consider employment quality and Hypotheses 5–7 study poverty which, as noted, often follows from unemployment or insecure work.
Possible chains from employment quality to protests and political violence.
It is still possible that there are factors other than poverty that link employment status and political participation. For example, if the results show that unemployment is connected to participation in studied actions but the feeling of poverty is not, then unemployment is arguably working through some other chain, maybe through a feeling of inferiority. In addition to lack of income, the unemployed can suffer from a feeling of inferiority because either they or the people around them consider them to be unsuitable for work, or useless to society.
In Figure 1, boxes with thicker outlines show what is actually tested in the article. Relative deprivation as well as frustration and aggression are factors which are assumed to link employment features and political participation, but unfortunately their role cannot be tested. As discussed above, following Smith et al. (2012: 2–3), relative deprivation occurs only when a person compares their current situation to some other situation, thinks that their current situation is worse than the situation compared, and feels their own situation unjust. Indeed, the respondents are asked for example whether they feel poor compared to fellow citizens. If some respondents do feel poor, they still do not necessarily feel the situation is unjust. Neither do we know whether the unemployed actually feel relatively deprived. This is a limitation of the empirical approach. I will come back to this issue when interpreting the results.
Each hypothesis will be tested separately in relation to participation in political violence and participation in demonstrations.
Hypotheses about employment status
Despite the above-described challenges, in the first hypothesis I do my best to define who are employed and who are unemployed and compare them. In addition to the relative deprivation theory, the motivation behind Hypothesis 1 is to test the conventional wisdom that the unemployed are more likely to engage in organized violence and protests.
Hypothesis 1: The unemployed are more likely to participate in political violence and demonstrations than the employed.
After this I compare the employed and those who fall outside employment and unemployment, i.e., those with ambiguous employment status. Indeed, collective identity and resources, like social contacts, organizational skills, and money, are theorized to be significant in the mobilization of collective actions, such as protests (Edwards and McCarthy, 2004; Klandermans and de Weerd, 2000). Consequently, the lack of collective identity and resources among the unemployed have – rightly or otherwise – been regarded as hindrances to their protests (Chabanet and Faniel, 2011). Collective identity is unlikely to develop among the unemployed, as they generally do not want to be unemployed and hope their unemployment is temporary. Presumably, those with ambiguous employment status are more likely to reach collective identity than the unemployed. They may also have more resources compared to the unemployed. This theoretical reasoning leads me to expect that those with ambiguous employment status are able to mobilize to a greater extent than the unemployed, as they do not face the obstacles for mobilization the unemployed face.
In many surveys, such as the World Values Survey or the Arab Barometer, the people falling between employment and unemployment were lumped together with the unemployed if they first answered that they were unemployed, or with students, if they first answered that they were students, etc. The survey used here asks about employment status considerably more thoroughly.
Hypothesis 2: Those with ambiguous employment status are more likely to participate in political violence and demonstrations than the employed.
In Hypotheses 1 and 2 the unemployed and those with ambiguous employment status are compared with the employed. One of the underlying goals of the study is to find out whether providing people with jobs would reduce their likelihood to take to the streets, as is commonly assumed. Comparisons with the employed serve this goal.
Hypotheses about employment quality
People’s time is also a central resource needed in protests; if people lack the time needed to organize and participate in protests, no protests emerge (Edwards and McCarthy, 2004: 116). Lichbach (1995: 42–45) postulates that those working part-time engage in protests more often than those working full-time simply because working part-time means more free time. If those working full-time took part in protests during their working hours, they would lose money. These arguments give rise to the next hypothesis.
Hypothesis 3: Those who work fewer hours per week are more likely to participate in political violence and demonstrations.
Having paid work does not necessarily make people satisfied. The International Labour Organization (2019a) has called poor working conditions the ‘main global employment challenge’. In addition to poor wages, work can be dangerous or unhealthy, working hours can be extreme, and the continuation of the work can be uncertain. Based on the relative deprivation theory, we can expect that those who are dissatisfied with their work will engage more in protests or political violence. Beinin (2009) reports that in 1998–2008, approximately two million workers in Egypt took part in demonstrations and other collective actions driven mainly by the dissatisfaction with and insecurity of their jobs. This characteristic of work leads to Hypothesis 4.
Hypothesis 4: Those who are less satisfied with their job are more likely to participate in political violence and demonstrations.
Hypotheses related to poverty
As noted earlier and presented in Figure 1, what follows from being out of decent work is crucial for relative deprivation, not the job per se. In a sense, following the relative deprivation theory, we should not expect higher participation in protests or political violence among those who have a low income or little property in absolute measures. Instead, we should look at this from the viewpoint of relative deprivation. In practice, absolute and relative poverty are supposedly linked. However, in Hypotheses 5–7 I will study each step of the proposed chain in Figure 1. Thus, I will also test whether low income or small property are connected to a higher likelihood of participation in these activities.
In real life, having a job does not necessarily mean earning a decent living. The International Labour Organization estimates that in the Arab Countries in 2018, about 8% of employed people were living in extreme poverty which means they had less than $1.90 per day in purchasing power parity (PPP). Throughout all the regions of the world, the youth suffer working poverty more often than older age groups (International Labour Organization, 2019b). If the employed do not protest less often than the unemployed, is it because employment does not always lead to the absence of economic hardship? This question leads to Hypothesis 5.
Hypothesis 5: Those with lower incomes are more likely to participate in political violence and demonstrations.
Especially for people working irregularly, their income changes over time. Consequently, what people possess is probably a better measure of their wealth than income. In other words, more important than the income itself is arguably what one can get with the income. Income and price levels vary within countries. These reasons justify studying the relationship of assets to political participation, as expressed in Hypothesis 6.
Hypothesis 6: Those with fewer assets are more likely to participate in political violence and demonstrations.
As already mentioned, from the relative deprivation theory viewpoint, it is ultimately not about unemployment, income, or assets. It is about whether the individual feels deprived. Hypothesis 7 examines the leap from the feeling of economic hardship to rebellion.
Hypothesis 7: Those who feel themselves poor are more likely to participate in political violence and demonstrations.
Research design
My primary data come from the quantitative survey dataset of the Sahwa Youth Survey 2016 (Weber et al., 2017), which was collected as a part of the European Commission-funded project Sahwa: Researching Arab Mediterranean Youth: Towards a New Social Contract. 1 Hereinafter, this dataset will be referred to as the Sahwa dataset. The dataset covers close to 10,000 respondents aged between 15 and 29, from Morocco, Algeria, Tunisia, Egypt, and Lebanon. The respondents were interviewed face-to-face between October 2015 and February 2016.
There are pros and cons in using survey data. On the negative side, people can be dishonest in their answers; especially, they can underreport their participation in political actions. However, there is no obvious reason to assume that people with a certain employment status would underreport their participation clearly more often than people with other employment statuses. In other words, if underreporting affects responses about equally despite the respondents’ employment status, the results concerning employment-status-based differences should be sufficiently credible. On the positive side, with this extensive survey dataset, it is possible to test the relationships between employment status and political participation, which are not easy to assess through other means. It is possible that people themselves do not even recognize that their employment is connected to their participation in political activities. In addition, as contingency plays a major role in participation in political actions, analysing the data from a large number of respondents helps to see these connections. The method of analysis is the ordinary least squares linear regression.
Dependent variables
From the Sahwa data, I separately studied two forms of political action: participation in political violence and in demonstrations. This allowed me to compare the characteristics of participants in contentious and accepted political actions.
The dependent variable in each of the analyses was participation in either of the two political actions. More precisely, the respondents were asked (question POL62) how often they had participated in different forms of political activities in the past 12 months. The two items I studied are ‘Use forms of violent action for social or political ends’ and ‘Participate, attend or help demonstrations’. Later on, the former is referred to as ‘violent action’ or ‘political violence’ and the latter as ‘demonstration’ or ‘protest’. For each political action, the respondents were given six options to answer: never, a few times a year, about once a month, about once a week, more than once a week, and every day. These were coded on a scale from 1 (never) to 6 (every day).
Violent actions for the sake of political or social ends can be interpreted to include anything from throwing a stone in a demonstration to committing a suicide attack or fighting in a war. The Sahwa dataset does not specify the type of violent political action. Further, in each country there are always several political issues provoking political actions. The Sahwa dataset does not reveal what is the issue a respondent reacts to when participating in a political action. Despite these limitations, the data provide interesting insight, on both the societal and academic levels, as to who has engaged in protests and political violence.
According to the Sahwa data, in Algeria and Morocco, political participation has been at a remarkably higher level than in the three other countries. I have used two other datasets – ACLED, the Armed Conflict Location and Event Data Project (Raleigh et al., 2023) and SCAD, Social Conflict Analysis Database (Salehyan et al., 2012) – to explore demonstrations, protests, riots, and events of political violence in Algeria, Egypt, Morocco, and Tunisia from late 2014 to late 2015, during the 12-month period the respondents were asked about. These datasets do not cover Lebanon. In ACLED there are 2,215 recorded events and 209 in SCAD.
These protests have had numerous issues and aims. According to the categorization of SCAD, 21% of the events have been about religious discrimination or religious issues, 15% have been pro-government, 12% have been provoked by issues related to human rights or democracy, and 4% have raised ethnic discrimination or ethnic issues. Most of the events have been about some other or unknown issues.
Both datasets show that the number of events with over one thousand participants has been highest in Algeria and Morocco, concordantly with the Sahwa data. In Algeria the largest protests were about shale oil exploitation, economic, religious and language issues, and the violations of security forces. In Algeria there were also clashes between the Arab and Berber communities and the Algerian army launched a military operation against a militant Islamist group. In Morocco, the largest protests concerned government policies and high costs of living, gender equality, the status of Western Sahara, Palestine, mandatory civil service, actions of police, and topples protests by activists (Raleigh et al., 2023; Salehyan et al., 2012).
There are reasons why there was less political activity in Egypt, Tunisia, and Lebanon, which further explain the differences in the level of activity. Laine and Myllylä (2019: 164–165), who have studied the Sahwa data, note that in Tunisia, a state of emergency was in effect and, in Egypt, people had possibly lost their trust in political actions because nothing had changed after the elections. The state of emergency in Tunisia allowed the president to suspend freedom of expression and assembly. It was in place before and partly while the survey was conducted in Tunisia (Amnesty, 2019). When interpreting the results, it is worth remembering that they reflect more situations in Algeria and Morocco than in the other analysed countries.
Independent variables
As discussed earlier, one of the key issues in the study was how to operationalize the employed and the unemployed from the data. The process of classifying the respondents’ employment status is presented in the main points in Figure 2 and described thoroughly below.
Determining the unemployed, those with ambiguous employment status, the employed, and others for Hypotheses 1 and 2.
In the Sahwa survey, the respondents were first asked (HM27_1) whether they were employed (including military service), unemployed, students, retired, housewives, or other inactive persons. This question was the point of departure for the categorization process and forms the topmost row of Figure 2. Only a very few of those who answered as employed were doing military service, so I included them among the other employed. Depending on how the respondents answered the first question about their employment status, they were asked further questions about their employment. Based on these follow-up questions, I moved some respondents from the unemployed and the employed and formed a group of those with ‘ambiguous employment status’. In the bottom of the figure, the groups which are compared in Hypotheses 1 and 2 are shown in boxes. The precise coding of these groups to variables is presented in Table 1 and explained later.
Coding of employment status variables in Hypotheses 1 and 2.
More specifically, I defined as employed those who first answered that they were employed and in the follow-up questions (EMP317, EMP317b) reported that, in the past week, they had worked more than half-time, i.e. at least 20 hours. The coding of the employed is presented rightmost in Figure 2. I think that counting as employed all those who work even a few hours a week would be problematic.
In the survey, those who first answered that they were not employed, in other words, those who for question HM27_1 answered something other than being ‘employed’, were then asked (EMP32) if they had still performed defined activities ‘during the past 7 days for at least an hour in exchange for remuneration in money or in kind’. They were further asked (EMP34) whether they were prepared to start working if they were offered a job and (EMP331) whether they had looked for work in the past month. In Hypothesis 1, I considered as unemployed those who in question HM27_1 had said that they were unemployed, had not carried out activities for remuneration, had not worked more than 0 hours a week, had not identified main job performed in the past week (EMP39), had been looking for work, and had been available for work. The operationalization reflects the international standard definition of the unemployed. The separation of the ‘real’ unemployed from the ‘unemployed’ who had worked anyway or who were not looking or available for work is illustrated to the left in Figure 2.
Indeed, some respondents first answered that they were not employed but then reported that they had, during the past week, carried out defined activities for remuneration or worked more than 0 hours or identified main job performed in the past week. These respondents who first reported they were not employed, but obviously had been working in the past week, are studied in Hypothesis 2 and are labelled as having an ambiguous employment status. The formation of this group is also shown in Figure 2.
Question EMP32 specifies work activities, such as the sale of goods in a market or on the street, agricultural work, and preparing food products, which are typical jobs in the informal sector. So, apparently many respondents who first answered that they did not work but had, however, carried out defined activities for remuneration, were working in the informal sector.
Table 1 shows the coding of employment status to three variables used to study Hypotheses 1 and 2. The respondents are divided into four categories, as described above. There are 2,277 employed respondents, 685 unemployed respondents, 545 respondents with ambiguous employment status and 6,341 respondents in the group of others.
The employed are coded 0 in each of the variables. This makes the employed the reference category, so the variables show the difference in relation to the employed. Thus, the model will make it possible to compare the unemployed to the employed, and those in ambiguous employment status to the employed. In the Others variable, so many different people are coded as 1 that their coefficients or statistical significance do not reveal much.
An alternative way to study Hypotheses 1 and 2 would have been to exclude in both hypotheses all other respondents than those I wished to compare; in the first hypothesis, for example, only the unemployed and the employed would have been included. In this article I have chosen to use the set of binary variables as presented in Table 1. This is a justified solution as it allows for including more respondents in the analyses, which in turn means that the control variables are better adjusted.
In Hypothesis 3, Hours per week is counted by multiplying the number of working days during a week (EMP317b) by the number of hours per day (EMP317).
When studying Hypothesis 4, the main independent variable is Satisfied with job, which is asked in the survey (EMP321). On a four-step scale, those very dissatisfied are coded 1 and those very satisfied are coded 4.
For Hypothesis 5, I created a variable Income quartile. Those respondents who work were asked to estimate their income (EMP324). Based on this, I divided these respondents into quartiles so that, in each country, those in the lowest income quartile are coded as 1 and those with the highest income as 4.
In Hypothesis 6, the Assets variable is a combination of 23 different questions in the survey. From all the respondents, there is information on whether they have certain assets in their household, for example a computer, washing machine, refrigerator, passenger vehicle, internet connection, or bathroom. For each asset the respondent’s household has, the respondent got one ‘point’ for having it or zero if they did not have it. So, the asset variable could vary between 0 and 23. Those with low values are living in the poorest households and those with the highest values in the richest. Especially as many of the respondents are very young, the assets of their household illustrate more their actual situation than their individual assets.
In Hypothesis 7, the feeling of being rich is taken from question POL628, in which the respondents were asked to compare themselves to people of their age, and to classify themselves on a scale where 0 is the poorest and 10 is the richest.
Control variables
As discussed above, the political contexts of the societies matter for the occurrence of political actions and in the countries under scrutiny the level of political activities has varied. Thus, the models are controlled for country. The occurrence of protests and political violence also varies within a country, so a control variable for urban residence is applied. In the Urban variable, those living in an urban area (ID5) are coded 1 and those living in a rural area are coded 0. Unfortunately, the Sahwa dataset does not provide administrative regions or other more detailed information about location. As men are often more likely to participate in protests, I have included a Male variable where males are coded 1 and females 0 (HM23_1).
I have calculated the approximate age of the respondent based on the year of birth (HM24_1). The calculation gives the age correctly for most of the respondents as some will appear a year too young or old. From other surveys we know that the adults and elderly all less likely to protest than the youth. In this case, however, as the surveyed were 15–29 years old, I expected that the youngest and the oldest respondents were less likely to participate in political activities than those in the middle. Therefore, I also introduced the squared term of age (age*age).
Higher education generally increases the likelihood for participation in all kinds of political activities (Campante and Chor, 2012: 168). Thus, I controlled the results for Education (HM25_1), which is coded as a five-step variable running from 1, no education or pre-school education, to 5, higher than secondary education. My last control variable is Single (HM26_1), in which single, divorced, and widowed respondents are coded 1 and married respondents 0. I assume those who are married have more responsibilities towards their families and consequently, avoid protests and political violence which might get them to trouble.
The Appendix includes the descriptive statistics of all the variables in Table A1. The correlation matrix is also available in the Online Appendix.
Results
During the previous year 11% of the respondents have participated in demonstrations. Of all respondents, 9% have participated in demonstrations ‘a few times a year’ or ‘about once a month’ and the remaining 2% more often. These and the percentages above are averages of the five countries. The share of those who have used violence for social or political ends during the previous year is 7%. Above 5% have participated in political violence at least once a month while less than 2% more often. Overall, the share of those who have participated in these actions is relatively high. This is likely to be explained at least partly by the age of the respondents. As noted above, previous studies have shown that the young are active in participating in protests.
Among the unemployed and the employed the shares of participants are very close. Of the employed, 10% have participated in demonstrations; for the unemployed the figure is also 10%. Violent actions are participated in by 7% of the employed and 6% of the unemployed. Those with ambiguous employment status have been considerably more active – 19% of them have participated in demonstrations and 13% in political violence.
The results of the regression analyses in Table 2 confirm that between the unemployed and the employed there is no difference in their participation in the studied forms of political action. However, those with ambiguous employment status are indeed considerably more likely to take part in both studied forms of political action than the employed.
Ordinary least squares linear regression, Hypotheses 1–4: Political participation, employment status, and employment quality.
Models controlled for country (not reported). *p < 0.05, **p < 0.01, ***p < 0.001. Standard errors in parentheses.
Reference category are the employed.
Employment quality also plays a role. Those who work fewer hours per week are clearly more likely to participate in both studied political actions than those who work more. The same pattern appears also when looking at mere percentages. Of those who work less than 40 hours per week 18% have participated in demonstrations and 13% in political violence. For those who work 40 hours per week or more the corresponding shares were 11% and 8%. Also, as hypothesized, those who are unsatisfied with their work are clearly more active in engaging in demonstrations and political violence than those who are satisfied with their work. Of those who are dissatisfied or very dissatisfied with their work, 16% have participated in demonstrations while 13% have participated in political violence. Among those who are satisfied or very satisfied with their work, the shares were 12% and 9% respectively.
The results in Table 3 show that, unlike what was assumed, income does not have an influence. Even more contrary to the hypothesis, those with more assets in their household are clearly more likely to participate both in political violence and in demonstrations. However, in line with the relative deprivation theory, those who feel themselves rich are less likely to take part in these studied actions than those who feel themselves poor.
As a whole, it seems that the same dependent variables explain participation in both violent actions and demonstrations. This indicates that – maybe not surprisingly – often those who have participated in political violence have also participated in demonstrations. One possible interpretation is that much of the political violence the survey captured happened during the demonstrations.
Control variables mainly take expected directions. As expected, the youngest and the oldest respondents are less active participants than those in the middle. As assumed, the males are more active in demonstrating but, perhaps surprisingly, when studying participation in violent actions, this difference is statistically significant only in some models. In general, those living in urban areas have not been more active participants. When interpreting these results, it is necessary to keep in mind that the results reflect more events in Algeria and Morocco where the level of political activities was higher. In Algeria, some of the main political activities during that time took place in the far south of the country, which is mainly rural, and it has also been reported that females participated actively (Dris, 2016).
Unlike what was expected, higher education has generally no effect on engagement in demonstrations and in several models it decreases the likelihood of the respondent participating in political violence. In some cases, singles are more likely to participate as assumed, but many models show that single status also has no effect.
Robustness checks
I ran a set of additional tests to study the robustness of the results. Instead of ordinary least squares linear regression, I applied both ordinal logistic regression and binary logistic regression. In binary logistic regression, those who had never during the previous year participated in the studied action were coded 0 and those who had were coded 1. With both ordinal logistic regression and binary logistic regression, higher income becomes a statistically significant predictor of demonstrating. Otherwise, the results show similar outcomes as the results of linear regressions, and even this shift in statistical significance is not a major change. As noted earlier, another option to code the employment status in Hypothesis 1 was to just have one binary variable for the two groups under scrutiny and exclude the others. Hypothesis 2 could have been studied in the same way. As expected, doing this gives similar results as the coding used in the article.
To be regarded as unemployed in the operationalization used in this article it was required that – in addition to considering themselves unemployed and not working – the respondent was looking for work and available for work. I tried three alternative operationalizations of unemployment. In these operationalizations either one or both of the latter requirements was not applied. If people are regarded as unemployed whether they were looking for work or not, the unemployed are more active protesters than the employed. In the two other alternative operationalizations, no difference between the unemployed and the employed appears.
I tested Hypothesis 1, but instead of 20 hours per week, I tried putting the limit of being employed as 10 or 30 hours. This does not change the outcome. I also tested regarding as employed all those who reported working at least one hour per week, in other words those in ambiguous employment status were put together with the employed. With this operationalization the employed were more likely to participate in both studied forms of political action than the unemployed.
It has been supposed that if those with a higher education level remain unemployed, they become especially frustrated and eager to revolt (Campante and Chor, 2012; Murshed and Tadjoeddin, 2009: 97). I compared the highly educated unemployed with the highly educated employed, but no difference between them appears.
Possibly employment status does not have the same meaning for those who are still maintained by their parents, and, consequently, the younger respondents could disturb the analysis. I studied Hypotheses 1 and 2 including only respondents aged 22–29. The results are consistent with the original results.
In the models studying Hypotheses 3 and 4, all the respondents who answered the question about job satisfaction and working hours were included. Thus, these analyses include those with ambiguous employment status and those who first answered they were employed but worked less than 20 hours. I reran the models so that I excluded the respondents with ambiguous employment status. In other words, I included only those who answered they were employed – despite how many hours they worked. Now, dissatisfaction with job loses its statistical significance as a predictor of participation in political violence. Working hours per week remains a statistically significant predictor of participation in both studied forms of political action, even though the coefficients decrease.
The data show that those with ambiguous employment status often work part-time, seem to be disappointed in their work, and are especially active in political actions. Altogether, this implies that those with ambiguous employment status explain some of the strength of the relationships found in Hypotheses 3 and 4.
I modified Hypothesis 5 so that I tested whether those who earn less than expected considering their education level participate more often. In Hypothesis 6 studying household assets, I tested including only those respondents who are themselves heads of a household or spouses of the heads of a household. These modifications give results consistent with the original models.
Kandil (2012) and Diwan (2013) have argued that the middle class has played an important role in the revolutions of the region. I added squared assets to the model studying Hypothesis 6 and the squared term of the income quartile to the model studying Hypothesis 5. These modifications do not lend support to the argument of a particularly active middle class. Further, I tried adding the square of feeling of richness to the model studying Hypothesis 7. This exercise shows that those who feel themselves poor and those who feel themselves rich are both more active participants in demonstrations and political violence than those feeling to be middle class.
I tested another dataset, round 5 of the Afrobarometer (2015) which was conducted in early 2013 in Algeria, Egypt, Morocco, Sudan, and Tunisia. I analysed only the young respondents, those aged 18–29. The Afrobarometer survey does not provide equally detailed data about employment status, but it shows similarly that when only the young unemployed and employed are compared, there is no statistically significant difference between them in how often they have attended a demonstration or used force or violence for a political cause. In addition, the Afrobarometer shows similarly that those with more assets are more likely to demonstrate, even though this relationship does not appear when studying political violence.
The results of all robustness checks can be found in the Online Appendix. Also, the country coefficients for the analyses presented in Tables 2 and 3 are available there.
Ordinary least squares linear regression, Hypotheses 5–7: Political participation and poverty.
Models controlled for country (not reported). *p < 0.05, **p < 0.01, ***p < 0.001. Standard errors in parentheses.
Conclusions
A key finding is that there was no clear difference between the young employed and the young unemployed in their likelihood of engaging in political violence or demonstrations in the five studied Arab Countries from late 2014 to early 2016. However, the robustness checks showed that to some extent, this result is contingent on who exactly is regarded as unemployed and employed. On the other hand, comparing just the unemployed and employed still does not reveal substantial differences compared to the results which show that employment quality and some factors related to employment clearly affect participation in the studied political activities.
As work satisfaction among those who have a job, the feeling of richness and other factors related to employment quality can explain protests and political violence, economic fluctuations could cause societal instability but modelling this instability on a larger scale is challenging. Detailed data about employment quality and the characteristics and attitudes of people with different employment status are rarely available, especially from the Global South. Statistics about the unemployment rates and wages are much better available but, based on these results, they should not be expected to be good predictors of political instability. Moreover, the results highlight the complexity of the relationships; for example, income does not affect participation, but higher household assets increase the likelihood of participation, whereas those feeling rich or poor are more likely to participate than those feeling middle income.
Thus, the results of this study indicate that while the manifold claims about unemployment as an important causal factor behind political violence and protests are perhaps not entirely mistaken, they are, however, clearly too simplistic.
From the relative deprivation perspective, the results are contradictory. Those dissatisfied with their job and those working fewer hours are active participants in political violence and protests, as expected based on the theory. But in contrast with expectations, the results show that the unemployed are, compared to the employed, not more active protesters. The finding that those feeling poor are more active participants in political violence and protests than those feeling middle class is in line with the deprivation perspective theory, but the high involvement among those who feel rich and those who have more assets is not. On the other hand, it must be remembered that the questions used do not perfectly measure relative deprivation. We can assume that those unemployed and those working fewer hours are relatively deprived, but they are not asked whether they feel their situation unjust, and feeling of injustice is a requirement of the relatively deprivation theory.
The study found that those whose employment status is ambiguous – who first say they are not employed but who are nevertheless working – seem to be especially active in engaging in political action. This is interesting, as this group is often missed in other surveys. As theorized, the unemployed are possibly willing to demonstrate and participate in violent actions, but because of obstacles, such as a lack of resources and collective identity, they often fail to do so. Those with ambiguous employment status might also find their situation unjust and be willing to participate, and as they do not have the same obstacles as the unemployed, they also do. This indicates that various theoretical approaches need to be used simultaneously. Relative deprivation theory appears to explain most of the results either partly or entirely, but it alone is not sufficient to explain results entirely.
To dig deeper into the relationship of employment and political instability, more survey data describing employment status as detailed as Sahwa is needed. It would also be helpful for the analysis if the types of political violence the respondent has participated in were better specified. Further, individual-level participation in political activities depends also on the issue at stake. Thus, it would be useful to have information about the issues of the political activities the respondent participated.
The simple presumption that the young unemployed revolt is perhaps based on the belief that work means more to people than it actually does. Instead of mere employment and money, people might want to work to belong to a group or to have significance for society. However, belonging and significance are not always found at work but may be found at activities outside work. The finding that satisfaction with work has an impact on participation in the studied political actions supports this reasoning.
Finally, as discussed earlier, people’s employment status does not alone determine whether demonstrations or political violence emerge. Civic organizations and labour unions play a role, the ways the authorities respond to the political actions are important, and the political and societal circumstances as a whole have an effect. Therefore, these results cannot be generalized to the whole world, not even to the studied countries at all times. Obviously, these results tell about activities in these countries around 2015.
However, the outcome that no difference appears when the unemployed are simply compared to the employed strengthens similar findings from earlier studies. Furthermore, the conclusion that the nuances in employment and unemployment status matter for political participation is probably also something that we can expect to find in other times and in other places.
In terms of policy recommendations, the results of the study indicate that just giving people jobs does not improve societal stability. More attention should be paid to employment quality. Finally, this echoes the UN Agenda 2030 Goal 8, which highlights the importance of having decent work. Perhaps the decency of work plays a role in achieving another Agenda goal, namely the promotion of peaceful societies outlined in Goal 16.
Footnotes
Appendix A
Descriptive statistics of the variables.
| N | Minimum | Maximum | Mean | Std. deviation | |
|---|---|---|---|---|---|
| Urban | 9,860 | 0 | 1 | 0.611 | 0.487 |
| Male | 9,860 | 0 | 1 | 0.535 | 0.499 |
| Age | 9,860 | 15 | 29 | 21.834 | 4.190 |
| Age squared | 9,860 | 225 | 841 | 494.261 | 184.711 |
| Education | 9,857 | 1 | 5 | 3.788 | 1.039 |
| Single | 9,858 | 0 | 1 | 0.823 | 0.382 |
| Unemployed† | 9,851 | 0 | 1 | 0.070 | 0.254 |
| Ambiguous employment status† | 9,851 | 0 | 1 | 0.055 | 0.229 |
| Others† | 9,848 | 0 | 1 | 0.644 | 0.479 |
| Unemployed (compared to the employed)‡ | 2,962 | 0 | 1 | 0.231 | 0.422 |
| Ambiguous employment status (compared to the employed)‡ | 2,822 | 0 | 1 | 0.193 | 0.395 |
| Income quartile | 2,756 | 1 | 4 | 2.413 | 1.103 |
| Assets | 9,785 | 2 | 23 | 14.601 | 3.364 |
| Feeling rich | 9,851 | 0 | 10 | 4.717 | 1.854 |
| Working hours | 2,963 | 0 | 126 | 44.110 | 18.698 |
| Satisfied with job | 2,988 | 1 | 4 | 2.859 | 0.868 |
| Participation in demonstration | 9,858 | 1 | 6 | 1.196 | 0.651 |
| Participation in violent action | 9,858 | 1 | 6 | 1.146 | 0.584 |
Country variables used in the models are not presented in this table.
Employment is coded into three variables to study Hypotheses 1 and 2 (variables Unemployed, Ambiguous employment status, and Others). Thus, descriptive statistics of these variables do not reveal much when the variables are presented separately. The coding of these variables is presented in Table 1.
These variables are not used in the analyses presented in the article. They are used in the Online Appendix to illustrate correlations between employment status and other variables. In the variables the unemployed and those with ambiguous employment status are coded 1 and the employed are coded 0.
Acknowledgements
I am grateful for the anonymous reviewers and the editor at JPR for their helpful comments. I thank Marko Lehti, Päivi Honkatukia, Siri Aas Rustad, and all others who have given feedback on earlier versions of the manuscript.
Replication data
Replication data and code for the empirical analysis in this article, along with the Online Appendix, can be found at 
.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The research presented in the article has been conducted as a part of the project What works? Youth transitions from education to employment in the Middle East and North Africa, funded by the Research Council of Finland, decision no. 320449. Work with the article has also been supported by the Emil Aaltonen Foundation, the Waldemar von Frenckell Foundation, the Research Grants Committee of the City of Tampere, and the People’s Education Fund.
Notes
KARI PAASONEN, MSc in Human Geography (University of Helsinki, 2017); Leading Researcher, Helsinki City Rescue Department (2023–present); Doctoral Researcher, Tampere Peace Research Institute, Tampere University (2019–present).