ABSTRACT
The 2008/09 economic crisis was – unsurprisingly – paralleled by noteworthy perceptions of socio-economic insecurity among Europeans. Respondents in different countries reported more or less grave fears of employment and income insecurity depending on their respective countries’ performance during the crisis. Against this background, we are interested in the manifestations of perceived socio-economic insecurity and their macro contextual determinants. After formulating hypotheses regarding socio-economic and welfare-institutional factors, we test them by using two rounds of the European Social Survey (fielded in 2008/09 and 2016/17). Based on three-level regressions with between-within models of 23,000 individuals (11,611 individuals in 2008/09 and 11,389 individuals in 2016/17) nested in 34 country-years and 17 European countries, we find that the level of perceived socio-economic insecurity in 2016/17 was generally lower than in 2008/09. Additionally, we reveal differences in the effects of the socio-economic factors and welfare-institutional factors between and within countries. These findings suggest that the inconsistency of the results of contextual factors among previous studies might stem from jumbling between- and within-country effects.
Introduction
The recent economic crisis, with its climax in 2008, has been the subject of numerous empirical studies in the realm of social science research on people's perceptions and attitudes (Dotti Sani and Magistro 2016; Colloca 2018). More than a dozen years have passed since the outbreak of the crisis. Although some countries have recovered comparatively well (e.g. Germany), others have had more trouble recovering or are still struggling with the consequences of the crisis (e.g. Greece, Spain, Portugal). Therefore, investigating the current manifestations of perceived socio-economic insecurity and their determinants at the individual and country levels seems to be worthwhile.
Naturally, the majority of published studies on people's perceived socio-economic insecurity has focused on either pre-crisis or crisis data. Although many of them explored crucial macro-level determinants, such as socio-economic and welfare-institutional factors, the evidence is astonishingly inconsistent. For example, in terms of socio-economic factors, Mau et al. (2012) found that higher levels of economic performance are indeed associated with lower levels of perceived socio-economic insecurity, whereas the effect of the unemployment rate is not significant. In contrast, Erlinghagen (2008) reported that economic growth does not affect job insecurity, whereas the unemployment rate intensifies insecurity. Moreover, regarding welfare-institutional factors, Van Oorschot and Chung (2015) clarified the negative association between welfare state spending and people's perceptions of socio-economic insecurity, whereas Erlinghagen (2008) noted that the level of social security spending has no significant effect on job insecurity. Hence, the issue of why previous studies found partially inconsistent results warrants closer inspection. On this account, some studies have suggested that the inconsistent conclusions derived from previous findings may be caused by the confusion between the following two types of effects concerning macro-level variables: the effect of the levels of countries and the effect of changes within countries (Chung and Van Oorschot 2011; Chung 2019). However, because most studies adopted a cross-sectional view based on only one round of international comparative data, it has been difficult to precisely divide the two types of effects as previously stated (Chung and Mau 2014).1.
Against this background, this study utilizes data from the fourth and eighth rounds of the European Social Survey, which were conducted in 2008/09 and 2016/17 (hereafter ESS4 and ESS8), and employs a specific type of multilevel model known as a ‘between-within’ model (Fairbrother 2014; Schröder 2018).2 This approach enables us to more appropriately divide the effects of country-level variables into between- and within-country effects; between-country effects are related to the effects of the levels of countries, and within-country effects are pertinent to changes within countries (Fairbrother 2014). With this approach, by using the average values of multiple time points, it is possible to more properly measure the differences regarding the levels of country characteristics between countries than in a cross-sectional study with one wave. Additionally, by using the within-country effects in many countries, this approach provides more exact coefficients of the changes of country characteristics within countries than a study in one country (Schmidt-Catran 2016; Schröder 2018). Based on this approach, this analysis investigates the manifestations of perceived socio-economic insecurity among people in 17 European countries, their individual- and country-level determinants and the changes in insecurity perceptions.
Theoretical background
Perceived socio-economic insecurity
The debate on socio-economic insecurities goes further back than the recent crisis.3 Since approximately the 1980s, many advanced modern (Western) societies have witnessed heightened socio-economic insecurities (Kalleberg 2009; Ranci 2010). These insecurities have been caused by intensified globalization resulting in fundamental labour market changes, welfare state retrenchment and increased competitive pressures (Esping-Andersen et al.2002; Brown et al.2010). However, the recent crisis has provided fuel for this debate on socio-economic insecurities and growing (group-specific) social inequality (Heidenreich 2016).
Exploring socio-economic insecurities demands a clear-cut distinction between objective insecurity and subjective insecurity. Whereas objective conditions of insecurity indicate actual problems, such as financial strain and absolute deprivation (Ranci et al.2017), subjective states of insecurity refer to experienced threats, which may precede actual problems and deviate from objective conditions. This study focuses on subjective insecurity, namely, perceived socio-economic insecurity, for two reasons. First, although many studies have shown that subjective perceptions of socio-economic insecurity (e.g. job and income worries) are frequently more pronounced than objective conditions suggest (Helbling and Kanji 2018), they should be taken seriously. Second, previous studies have noted that perceived socio-economic insecurity is associated with a wide variety of outcomes, such as well-being, work commitment or attitudes towards the welfare state (Chung and Mau 2014).
Our research interest is directed at perceived socio-economic insecurity, which is made up of the two components of perceptions of employment insecurity and perceptions of income insecurity (Van Oorschot and Chung 2015). Perceived employment insecurity refers to the risk of becoming unemployed in the near future. In work-oriented societies, unemployment not only attacks social identities but also stands for potential financial stress. At the individual level, the findings are straightforward and point to typical vulnerable groups. Perceived employment insecurity tends to be particularly high among women, migrants, less educated individuals, lower occupational positions, persons in atypical (especially temporary) employment and employees with previous unemployment experiences (Böckerman 2004; Lübke and Erlinghagen 2014; Chung 2019). Regarding the impact of age, the findings are mixed. Although some researchers argue that younger people suffer more from perceived job insecurity (Chung 2019), other scholars consider older people as more subjectively insecure (Erlinghagen 2008).
Perceived income insecurity relates to the risk of not having enough money to cover household necessities. Because being able to make ends meet depends largely on having adequate employment, perceived income insecurity is often intertwined with perceived employment insecurity. In addition to the risk of job loss, precarious (substandard) employment and in-work poverty are the main sources of perceived income insecurity (Kalleberg 2009; Standing 2014). Previous research has found typical determinants of perceived income insecurity identical to those of perceived employment insecurity (Reeskens and Van Oorschot 2014).
Country-level determinants of perceived socio-economic insecurity: market economy versus welfare state
The question of whether the market economy or the welfare state matters more for people's perceptions and attitudes, including perceived socio-economic insecurity, is probably one of the most powerful motives that drives social scientists to work at the interface of social inequality and welfare research. Thus, the present study focuses on the impacts of the macro variables related to both the socio-economic and welfare-institutional perspectives on perceived socio-economic insecurity.
Many studies have examined socio-economic factors to reveal the determinants of perceived socio-economic insecurity. First, previous research has found that the general economic condition of a given country affects perceived socio-economic insecurity because it affects the probability of unemployment and decreasing incomes (Chung and Van Oorschot 2011; Reeskens and Van Oorschot 2014). For example, international comparative analyses have shown that citizens in countries with a higher GDP per capita are more likely to report lower socio-economic insecurity than those in countries with a lower GDP per capita (Mau et al.2012; Chung 2020). Other studies have found that a growing GDP per capita in a given country also reduces subjective socio-economic insecurity (Lübke and Erlinghagen 2014; Van Oorschot and Chung 2015).
Second, economic inequality has been discussed as one of the most decisive macro-level determinants of perceived socio-economic insecurity. The reason is that the level of income inequality stimulates economic fears, such as income insecurity (Mau et al.2012). On this issue, Van Oorschot and Chung (2015) analysed data from the ESS for 2008/09 and reported that the Gini coefficient has a positive effect on both employment and income insecurities.
Third, numerous studies have discussed the unemployment rate being positively associated with perceived job and income insecurities (Chung and Van Oorschot 2011; Lübke and Erlinghagen 2014; Balz 2017). This association occurs because a significant number of citizens losing their jobs raises fear among other citizens about their future prospects (Chung and Van Oorschot 2011). Nevertheless, some studies have reported that the unemployment rate has no significant effect on this issue (Mau et al.2012; Chung 2020). Moreover, recent research has insisted that changes in the unemployment rate in a given country boost subjective job and income insecurities (Lübke and Erlinghagen 2014; Chung 2019).
In addition to socio-economic factors, many publications have emphasized welfare-institutional factors as crucial determinants of perceived socio-economic insecurity (Mau et al.2012; Chung and Mau 2014). First, this analysis focuses on the effect of welfare provisions on subjective insecurity because ‘[t]he issue of security is … highly topical in comparative welfare state research’ (Chung and Mau 2014: 307). Previous studies have pointed out that welfare protection is supposed to stabilize a society by decreasing perceived socio-economic insecurities (Mau et al.2012; Chung and Mau 2014). In line with these discussions, it has been revealed that total social expenditures are negatively associated with job and income insecurities (Reeskens and Van Oorschot 2014; Van Oorschot and Chung 2015), while some studies have suggested that the effect of social security spending on job insecurity is not significant (Anderson and Pontusson 2007; Erlinghagen 2008).
Second, active labour market policies (ALMPs) are regarded as an essential factor with respect to perceived insecurity (Chung and Mau 2014). By providing training programs to improve the skills of the unemployed and support in job search, ALMPs can enhance the chances of reemployment for the unemployed, reduce lay-off risks for the employed and, thus, job and income insecurities (Anderson and Pontusson 2007; Chung and Mau 2014). Accordingly, many studies have reported that ALMPs reduce job and income insecurities among European countries (Chung and Van Oorschot 2011; Lübke and Erlinghagen 2014; Van Oorschot and Chung 2015; Chung 2019, 2020). For example, Chung (2020) analysed data from ESS rounds 4 and 8 and identified that ALMP expenditures significantly decreased job insecurity in both rounds. Additionally, Lübke and Erlinghagen (2014) elucidated that changes in ALMP expenditures also reduce the perceived likelihood of job loss.
Third, previous research also treated employment protection legislation (EPL) as another substantial aspect of insecurity research because EPL is related to regulations in terms of hiring and firing workers (Chung and Mau 2014). Although strict EPL is supposed to raise employment security by making it difficult to dismiss employees, some studies highlight that EPL may unintendedly increase job insecurities because employers may hesitate to hire employees with permanent contracts when it is difficult to dismiss them; as a consequence, EPL can bring about longer unemployment durations, more temporary contracts and, thus, intensified job and income insecurities (Anderson and Pontusson 2007; Chung and Van Oorschot 2011). Although some researchers detect a significant effect of EPL on reducing perceived employment insecurity (Anderson and Pontusson 2007), other studies ascertain no significant impact of employment protection measures on this issue (Mau et al.2012; Lübke and Erlinghagen 2014; Chung 2019).
In summary, a host of studies have explored the macro-level determinants of perceived socio-economic insecurity. Although these studies broadened our understanding of the determinants of job and income securities, in terms of some macro-level factors, the empirical findings are not consistent. On this account, some authors underlined two fundamental limitations of previous studies in this field. First, until now, the majority of international comparative analyses of perceived insecurity have deployed cross-sectional data to explain variations between countries; only a few studies have adopted a longitudinal perspective. Second, although a few recent studies have given attention to the effect of changes in the macro-level variables in a given country, due to the first reason and the limitations of statistical models, it has been difficult to appropriately separate the effects of the levels of country characteristics from their changes within a country on an attitudinal feature, such as perceived insecurity (Fairbrother 2014; Schröder 2018).
Research questions and hypotheses
To overcome these limitations, the present study uses data from two rounds of the ESS and adopts the ‘between-within’ model to divide the variation of the macro-level determinants into variations between and within countries (Fairbrother 2014). In this way, we can include both the between (cross-sectional) and within (longitudinal) effects of the macro-level determinants and compare the effects of the levels of the macro contexts and their changes within countries (Schmidt-Catran 2016; Schröder 2018). Building on this approach, our research interest is directed at the manifestations of perceived socio-economic insecurity and their country-level determinants. Specifically, we raise two questions: (1) To what extent do objective contextual factors explain the differences between and within countries regarding perceived socio-economic insecurity? (2) How do the between- and within-effects of these contextual factors differ?
On this basis, we formulate hypotheses with respect to the socio-economic and welfare-institutional factors. This study examines the effects of the levels of the macro context (between-effect) and those of the changes in the macro-level determinants within countries (within-effect) as follows.
Hypotheses H1-a and H1-b: (a) The levels of / (b) changes (increases) in economic conditions are negatively related to perceived socio-economic insecurity.
Hypotheses H2-a and H2-b: (a) The levels of / (b) changes (increases) in income inequality are positively related to perceived socio-economic insecurity.
Hypotheses H3-a and H3-b: (a) The levels of / (b) changes (increases) in the unemployment rate are positively related to perceived socio-economic insecurity.
Hypotheses H4-a and H4-b: (a) The levels of / (b) changes (increases) in social expenditures are negatively related to perceived socio-economic insecurity.
Hypotheses H5-a and 5-b: (a) The levels of / (b) changes (increases) in the expenditures for ALMPs are negatively related to perceived socio-economic insecurity.
Hypotheses H6-a and 6-b: (a) The levels of / (b) changes (increases) in employment protection regulation (EPL) are negatively related to perceived socio-economic insecurity.
Because individual-level determinants are also likely to influence perceived socio-economic insecurity, we consider them to be relevant control variables in the analysis.
Empirical setup
Data
Our empirical analysis is based on individual-level data from two rounds of the ESS that contained the rotating module ‘welfare attitudes’ and that were fielded in 2008/09 (ESS4) and 2016/17 (ESS8).4 We limit our analysis to countries that participated in both rounds of the ESS. A second restriction refers to respondents aged 15–65 who have paid work for a minimum of 30 hours per week.5 These limitations and a listwise exclusion of missing individual-level data leave a pooled dataset with 23,000 cases nested in 34 country-years and 17 countries (ESS4: ni = 11,611; ESS8: ni = 11,389; see Table A1 in the Online Appendix for the list of countries).
Perceived socio-economic insecurity, our dependent variable, is measured by using the following two survey items:
The risk of becoming unemployed: ‘Please tell me how likely it is that during the next 12 months you will be unemployed and looking for work for at least four consecutive weeks?’
The risk of not being able to make ends meet: ‘ … how likely is it that there will be some periods when you don't have enough money to cover your household necessities?’
Respondents could choose among four answers: (1) ‘not at all likely’; (2) ‘not very likely’; (3) ‘likely’; (4) ‘very likely’. Both items are recoded to the value range 0 (‘not at all likely’) to 3 (‘very likely’) and taken to create an additive insecurity index. Thus, the index ranges from 0 (‘not at all likely’) to 6 (‘very likely’). To avoid discarding information, we treat it as a metric (continuous) variable,6 which allows linear multilevel regression analysis.
Determinants that may matter according to theory are taken as independent variables in our multilevel regressions. At the country- and country-year-levels, we employ six macro variables belonging to two analytical dimensions (see Table A1 in the Online Appendix for details).
The first dimension captures the socio-economic factors and reflects the market conditions and outcomes. This dimension comprises three macro variables:
Gross domestic product (GDP) (per capita) serves as a measure of a country's economic prosperity (Organisation for Economic Co-operation and Development 2017). We used the GDP per capita/1,000 (in US-$ 2010) adjusted for purchasing power parity (PPP) in our analysis.
The Gini coefficient reflects a country's unequal income distribution (European Commission 2017).
The unemployment rate (as a percentage of the total labour force) serves as a further measure of inequality and indicates a welfare state's main target group (European Commission 2017).
The second dimension encompasses welfare-institutional factors and stands for market interventions. This dimension consists of three additional macro variables:
Social expenditures (as a percentage of GDP) indicate the magnitude of a country's social protection, often termed ‘welfare state effort’ (European Commission 2017).
The expenditures for ALMPs (as a percentage of GDP) serve as an indicator of a country's inclination towards adopting social investment and activation strategies (Organisation for Economic Co-operation and Development 2017).
The OECD Employment Protection Index (EPI; Code EPRC_V3) is an indicator that incorporates 13 data items to assess a country's strictness of employment protection concerning individual and collective dismissals (regular contracts). Its values range from 0 to 6 (Organisation for Economic Co-operation and Development 2017).
In all models, we use seven individual-level predictors for perceived socio-economic insecurity as follows: (1) gender (1 = female); (2) age in years in five groups (15 to 25 = reference category; 26 to 35; 36 to 45; 46 to 55; 56 to 65); (3) education in years; (4) class affiliation in a six-class version of the EGP scheme (EGP I; EGP II; EGP III; EGP IV; EGP V&VI; EGP VII = reference category)7; (5) previous unemployment experience of at least three months (1 = yes); (6) work contract (unlimited = reference category; limited; no contract); and (7) equivalized household net income (in deciles).8 Table A3 in the Online Appendix provides a descriptive overview of these individual-level control variables.9
Method and analytical strategy
All results are based on calculations using Mplus (version 8). To tackle the problem of the relatively small number of countries, we conducted an analysis using Bayesian estimation (Bryan and Jenkins 2016).
Empirical findings
Descriptive section
Perceived socio-economic insecurity in 2008/09 and 2016/17 with 95% confidence intervals.
Notes: The scales range between 0 and 6, with higher values indicating higher levels of perceived socio-economic insecurity. The vertical line indicates the overall mean of the insecurity index. All country-specific means are calculated by using weights.
Source: ESS4 (ni = 11,611; unweighted); ESS8 (ni = 11,389; unweighted); nj = 17.
Perceived socio-economic insecurity in 2008/09 and 2016/17 with 95% confidence intervals.
Notes: The scales range between 0 and 6, with higher values indicating higher levels of perceived socio-economic insecurity. The vertical line indicates the overall mean of the insecurity index. All country-specific means are calculated by using weights.
Source: ESS4 (ni = 11,611; unweighted); ESS8 (ni = 11,389; unweighted); nj = 17.
Across all countries, the mean of the insecurity index amounts to 1.92 (SD: 0.55) in 2008/09 and 1.62 (SD: 0.46) in 2016/17. Therefore, the level of perceived socio-economic insecurity was generally lower in 2016/17 than in 2008/09. If we divide the country rankings into three segments (lowest: lower than mean – SD; middle: between mean ± SD; highest: higher than mean + SD) and start by looking at the first survey date, we see that the country ranking approximately follows a familiar pattern of welfare regimes. The Scandinavian countries of Norway and Sweden, together with the Netherlands, exhibit particularly low levels of perceived socio-economic insecurity. At the other end of the scale (with high levels of perceived socio-economic insecurity), we find three former communist countries, specifically, Estonia, Hungary and the Czech Republic. The middle segment of the scale encompasses countries belonging to diverse welfare regime families: Finland as a further member of the Scandinavian countries, Continental welfare regimes (Switzerland, Germany, Belgium, France), Anglo-Saxon welfare regimes (Great Britain, Ireland), Mediterranean welfare regimes (Spain, Portugal) and post-communist welfare regimes (Slovenia, Poland). If we turn to the second survey date, we realize some changes worth mentioning. (1) Germany shows a lower level of perceived socio-economic insecurity in 2016/17 than in 2008/09 and moved up into the privileged segment of countries with low levels of perceived socio-economic insecurity. (2) Three countries (Finland, Switzerland, France) experienced degradation within the middle segment of the scale and are the only countries that show higher levels of perceived socio-economic insecurity in 2016/17 than in 2008/09. (3) Estonia and Hungary were able to leave the deprived segment with high levels of perceived socio-economic insecurity and moved into the intermediate segment. (4) Although Poland shows a lower level of perceived socio-economic insecurity in 2016/17 than in 2008/09, it moved down into the most deprived segment (as the general level of perceived socio-economic insecurity decreased within the period under review).
Multilevel analysis
In Table 1, Model 0 examines random effects without any explanatory variables. The random part indicates that variances at the country level (0.237) and the country-year level (0.094) are significant, which signals that three-level modelling is appropriate. Moreover, the intraclass correlation coefficient (ICC) is 0.118 at the country level and 0.047 at the country-year level, which means that 11.8 per cent of the variance with respect to perceived socio-economic insecurity is attributable to differences across the 17 countries under study, whereas 4.7 per cent can be explained by the variations among country-year units.12
| Model 0 | Model I | Model II | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coeff. | S.D. | Sig. | 95% C.I. | Coeff. | S.D. | Sig. | 95% C.I. | Coeff. | S.D. | Sig. | 95% C.I. | ||||
| 2.5% | 97.5% | 2.5% | 97.5% | 2.5% | 97.5% | ||||||||||
| Fixed Part | |||||||||||||||
| Intercept | 1.756 | 0.138 | *** | [1.49 to 2.04] | 2.750 | 0.113 | *** | [2.53 to 2.98] | 2.907 | 0.119 | *** | [2.67 to 3.14] | |||
| Gender (1 = female) | 0.045 | 0.018 | ** | [0.01 to 0.08] | 0.045 | 0.017 | ** | [0.01 to 0.08] | |||||||
| Age | |||||||||||||||
| 15–25 | Ref. | Ref. | |||||||||||||
| 26–35 | 0.026 | 0.032 | [−0.04 to 0.09] | 0.020 | 0.031 | [−0.04 to 0.08] | |||||||||
| 36–45 | 0.007 | 0.032 | [−0.05 to 0.07] | 0.002 | 0.031 | [−0.06 to 0.06] | |||||||||
| 46–55 | −0.045 | 0.032 | [−0.11 to 0.02] | −0.051 | 0.031 | [−0.11 to 0.01] | |||||||||
| 56–65 | −0.199 | 0.035 | *** | [−0.27 to −0.13] | −0.204 | 0.035 | *** | [−0.27 to −0.14] | |||||||
| Education (in years) | −0.027 | 0.003 | *** | [−0.03 to −0.02] | −0.027 | 0.003 | *** | [−0.03 to −0.02] | |||||||
| EGP class position | |||||||||||||||
| EGP I | −0.317 | 0.032 | *** | [−0.38 to −0.26] | −0.317 | 0.032 | *** | [−0.38 to −0.25] | |||||||
| EGP II | −0.314 | 0.029 | *** | [−0.37 to −0.26] | −0.314 | 0.029 | *** | [−0.37 to −0.26] | |||||||
| EGP III | −0.208 | 0.028 | *** | [−0.26 to −0.16] | −0.212 | 0.028 | *** | [−0.27 to −0.16] | |||||||
| EGP IV | −0.513 | 0.083 | *** | [−0.68 to −0.36] | −0.520 | 0.085 | *** | [−0.69 to −0.35] | |||||||
| EGP V + VI | −0.156 | 0.029 | *** | [−0.21 to −0.10] | −0.159 | 0.028 | *** | [−0.21 to −0.10] | |||||||
| EGP VII | Ref. | Ref. | |||||||||||||
| Unemploym. experience (1 = yes) | 0.373 | 0.018 | *** | [0.34 to 0.41] | 0.374 | 0.018 | *** | [0.34 to 0.41] | |||||||
| Work contract | |||||||||||||||
| Unlimited | Ref. | Ref. | |||||||||||||
| Limited | 0.607 | 0.025 | *** | [0.56 to 0.66] | 0.605 | 0.025 | *** | [0.56 to 0.65] | |||||||
| No contract | 0.267 | 0.046 | *** | [0.17 to 0.36] | 0.267 | 0.045 | *** | [0.18 to 0.35] | |||||||
| Household net income | −0.145 | 0.006 | *** | [−0.16 to −0.13] | −0.145 | 0.006 | *** | [−0.16 to −0.14] | |||||||
| Year (1 = 2016) | −0.290 | 0.070 | *** | [−0.43 to −0.16] | |||||||||||
| Random Part | |||||||||||||||
| Country-level variance | 0.237 | 0.146 | *** | [0.09 to 0.63] | 0.131 | 0.092 | *** | [0.03 to 0.37] | 0.157 | 0.088 | *** | [0.07 to 0.39] | |||
| Country-year-level variance | 0.094 | 0.043 | *** | [0.05 to 0.21] | 0.084 | 0.038 | *** | [0.04 to 0.19] | 0.036 | 0.020 | *** | [0.02 to 0.09] | |||
| Individual-level variance | 1.678 | 0.016 | *** | [1.65 to 1.71] | 1.450 | 0.014 | *** | [1.42 to 1.48] | 1.450 | 0.013 | *** | [1.43 to 1.48] | |||
| DIC | 77211.804 | 73870.957 | 73866.506 | ||||||||||||
| Ncountry | 17 | 17 | 17 | ||||||||||||
| Ncountry-year | 34 | 34 | 34 | ||||||||||||
| Nindividual | 23,000 | 23,000 | 23,000 | ||||||||||||
| Model 0 | Model I | Model II | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coeff. | S.D. | Sig. | 95% C.I. | Coeff. | S.D. | Sig. | 95% C.I. | Coeff. | S.D. | Sig. | 95% C.I. | ||||
| 2.5% | 97.5% | 2.5% | 97.5% | 2.5% | 97.5% | ||||||||||
| Fixed Part | |||||||||||||||
| Intercept | 1.756 | 0.138 | *** | [1.49 to 2.04] | 2.750 | 0.113 | *** | [2.53 to 2.98] | 2.907 | 0.119 | *** | [2.67 to 3.14] | |||
| Gender (1 = female) | 0.045 | 0.018 | ** | [0.01 to 0.08] | 0.045 | 0.017 | ** | [0.01 to 0.08] | |||||||
| Age | |||||||||||||||
| 15–25 | Ref. | Ref. | |||||||||||||
| 26–35 | 0.026 | 0.032 | [−0.04 to 0.09] | 0.020 | 0.031 | [−0.04 to 0.08] | |||||||||
| 36–45 | 0.007 | 0.032 | [−0.05 to 0.07] | 0.002 | 0.031 | [−0.06 to 0.06] | |||||||||
| 46–55 | −0.045 | 0.032 | [−0.11 to 0.02] | −0.051 | 0.031 | [−0.11 to 0.01] | |||||||||
| 56–65 | −0.199 | 0.035 | *** | [−0.27 to −0.13] | −0.204 | 0.035 | *** | [−0.27 to −0.14] | |||||||
| Education (in years) | −0.027 | 0.003 | *** | [−0.03 to −0.02] | −0.027 | 0.003 | *** | [−0.03 to −0.02] | |||||||
| EGP class position | |||||||||||||||
| EGP I | −0.317 | 0.032 | *** | [−0.38 to −0.26] | −0.317 | 0.032 | *** | [−0.38 to −0.25] | |||||||
| EGP II | −0.314 | 0.029 | *** | [−0.37 to −0.26] | −0.314 | 0.029 | *** | [−0.37 to −0.26] | |||||||
| EGP III | −0.208 | 0.028 | *** | [−0.26 to −0.16] | −0.212 | 0.028 | *** | [−0.27 to −0.16] | |||||||
| EGP IV | −0.513 | 0.083 | *** | [−0.68 to −0.36] | −0.520 | 0.085 | *** | [−0.69 to −0.35] | |||||||
| EGP V + VI | −0.156 | 0.029 | *** | [−0.21 to −0.10] | −0.159 | 0.028 | *** | [−0.21 to −0.10] | |||||||
| EGP VII | Ref. | Ref. | |||||||||||||
| Unemploym. experience (1 = yes) | 0.373 | 0.018 | *** | [0.34 to 0.41] | 0.374 | 0.018 | *** | [0.34 to 0.41] | |||||||
| Work contract | |||||||||||||||
| Unlimited | Ref. | Ref. | |||||||||||||
| Limited | 0.607 | 0.025 | *** | [0.56 to 0.66] | 0.605 | 0.025 | *** | [0.56 to 0.65] | |||||||
| No contract | 0.267 | 0.046 | *** | [0.17 to 0.36] | 0.267 | 0.045 | *** | [0.18 to 0.35] | |||||||
| Household net income | −0.145 | 0.006 | *** | [−0.16 to −0.13] | −0.145 | 0.006 | *** | [−0.16 to −0.14] | |||||||
| Year (1 = 2016) | −0.290 | 0.070 | *** | [−0.43 to −0.16] | |||||||||||
| Random Part | |||||||||||||||
| Country-level variance | 0.237 | 0.146 | *** | [0.09 to 0.63] | 0.131 | 0.092 | *** | [0.03 to 0.37] | 0.157 | 0.088 | *** | [0.07 to 0.39] | |||
| Country-year-level variance | 0.094 | 0.043 | *** | [0.05 to 0.21] | 0.084 | 0.038 | *** | [0.04 to 0.19] | 0.036 | 0.020 | *** | [0.02 to 0.09] | |||
| Individual-level variance | 1.678 | 0.016 | *** | [1.65 to 1.71] | 1.450 | 0.014 | *** | [1.42 to 1.48] | 1.450 | 0.013 | *** | [1.43 to 1.48] | |||
| DIC | 77211.804 | 73870.957 | 73866.506 | ||||||||||||
| Ncountry | 17 | 17 | 17 | ||||||||||||
| Ncountry-year | 34 | 34 | 34 | ||||||||||||
| Nindividual | 23,000 | 23,000 | 23,000 | ||||||||||||
Notes: Results of Bayesian estimation. S.D. are standard deviations. C.I. are credible intervals. DIC is deviance information criterion. Sig.: *** p < 0.01; ** p < 0.05; * p < 0.10 (two-tailed tests).
Source: ESS4 (ni = 11,611; unweighted); ESS8 (ni = 11,389; unweighted); nj = 17.
Model I investigates the effects of our individual-level variables. Here, we obtain well-established results. Women, the respondents with previous unemployment experience and those with substandard work arrangements (a limited contract or no work contract) show significantly more perceived socio-economic insecurity, whereas the oldest age group (56 to 65), the higher educated respondents in higher EGP class positions (relative to the lowest EGP class VII) and those with higher household net incomes display significantly less perceived socio-economic insecurity. The explained variance is 0.171, which means that 17.1 per cent of the total variance can be explained by individual-level variables.13
In Model II, the year dummy has a significantly negative effect, which informs us that from 2008/09 to 2016/17, the general level of perceived socio-economic insecurity decreased. Even after controlling for individual-level variables and the year dummy, both the country-level variance (0.157) and the country-year-level variance (0.036) remain significant. Hence, we have reason to inspect the effect of macro variables at the country and country-year levels.
In terms of our analysis at the country-year level (within-effect) and the country level (between-effect), we find that four of our six macro variables exert a noteworthy impact on people's perceived socio-economic insecurity (Table 2).
| Model | Variable | Coeff. | S.D. | Sig. | 95% C.I. | DIC | |
|---|---|---|---|---|---|---|---|
| 2.5% | 97.5% | ||||||
| Socio-Economic Factors | |||||||
| III | GDP per capita (within) | −0.023 | 0.015 | [−0.05 to 0.01] | 73869.589 | ||
| GDP per capita (between) | −0.027 | 0.007 | *** | [−0.04 to −0.01] | |||
| IV | Gini coefficient (within) | −0.057 | 0.039 | [−0.13 to 0.02] | 73869.513 | ||
| Gini coefficient (between) | 0.041 | 0.033 | [−0.02 to 0.10] | ||||
| V | Unemployment rate (within) | 0.049 | 0.025 | ** | [0.00 to 0.10] | 73869.509 | |
| Unemployment rate (between) | 0.028 | 0.042 | [−0.05 to 0.11] | ||||
| Welfare-Institutional Factors | |||||||
| VI | Social expenditures (within) | 0.084 | 0.024 | *** | [0.04 to 0.13] | 73868.954 | |
| Social expenditures (between) | −0.059 | 0.020 | *** | [−0.10 to −0.02] | |||
| VII | Expenditures for ALMPs (within) | −0.428 | 0.333 | [−1.08 to 0.25] | 73869.579 | ||
| Expenditures for ALMPs (between) | −0.923 | 0.350 | ** | [−1.61 to −0.23] | |||
| VIII | OECD EPI (within) | 0.124 | 0.362 | [−0.60 to 0.84] | 73869.853 | ||
| OECD EPI (between) | −0.009 | 0.297 | [−0.60 to 0.56] | ||||
| Ncountry | 17 | ||||||
| Ncountry-year | 34 | ||||||
| Nindividual | 23,000 | ||||||
| Model | Variable | Coeff. | S.D. | Sig. | 95% C.I. | DIC | |
|---|---|---|---|---|---|---|---|
| 2.5% | 97.5% | ||||||
| Socio-Economic Factors | |||||||
| III | GDP per capita (within) | −0.023 | 0.015 | [−0.05 to 0.01] | 73869.589 | ||
| GDP per capita (between) | −0.027 | 0.007 | *** | [−0.04 to −0.01] | |||
| IV | Gini coefficient (within) | −0.057 | 0.039 | [−0.13 to 0.02] | 73869.513 | ||
| Gini coefficient (between) | 0.041 | 0.033 | [−0.02 to 0.10] | ||||
| V | Unemployment rate (within) | 0.049 | 0.025 | ** | [0.00 to 0.10] | 73869.509 | |
| Unemployment rate (between) | 0.028 | 0.042 | [−0.05 to 0.11] | ||||
| Welfare-Institutional Factors | |||||||
| VI | Social expenditures (within) | 0.084 | 0.024 | *** | [0.04 to 0.13] | 73868.954 | |
| Social expenditures (between) | −0.059 | 0.020 | *** | [−0.10 to −0.02] | |||
| VII | Expenditures for ALMPs (within) | −0.428 | 0.333 | [−1.08 to 0.25] | 73869.579 | ||
| Expenditures for ALMPs (between) | −0.923 | 0.350 | ** | [−1.61 to −0.23] | |||
| VIII | OECD EPI (within) | 0.124 | 0.362 | [−0.60 to 0.84] | 73869.853 | ||
| OECD EPI (between) | −0.009 | 0.297 | [−0.60 to 0.56] | ||||
| Ncountry | 17 | ||||||
| Ncountry-year | 34 | ||||||
| Nindividual | 23,000 | ||||||
Notes: Results of Bayesian estimation. S.D. are standard deviations. C.I. are credible intervals. DIC is deviance information criterion. Sig.: *** p < 0.01; ** p < 0.05; * p < 0.10 (two-tailed tests).
Source: ESS4 (ni = 11,611; unweighted); ESS8 (ni = 11,389; unweighted); nj = 17.
In Model III, GDP per capita shows significantly negative effects at the country level: higher levels of economic situation are associated with lower perceived socio-economic insecurity. Thus, in fact, markets can be considered producers of welfare in a broader sense. Therefore, we confirm hypothesis H1-a at the country level (but reject hypothesis H1-b at the country-year level).
In contrast, in Model IV, the effect of the Gini coefficient is not significant at either the country or the country-year level. Accordingly, hypotheses H2-a and H2-b are not supported.
In Model V, we find that the unemployment rate exerts a positive effect on perceived socio-economic insecurity (only) at the country-year level. As expected, rising unemployment within a country is associated with more perceived socio-economic insecurity. This finding is in accordance with our hypothesis H3-b. At the same time, we reject hypothesis H3-a.
Model VI presents an ambivalent picture. Social expenditures at both the country-year level and the country level wield significant effects but with different signs. Counterintuitively, increasing social expenditures (within country) go along with more perceived socio-economic insecurity, whereas at the country level, higher levels of social expenditures are accompanied by less perceived socio-economic insecurity, as expected. This contradictory finding allows for at least two interpretations. First, it creates the assumption that more objective social security (X) does not necessarily translate into more subjective security (Y) because people become accustomed to a welfare state's safety net, which, in turn, reduces the capacity of the safety net to make people feel secure (‘security paradox’). However, this approach can explain more a ‘null effect’ of X on Y, not the decrease in subjective security that we find. For this reason, a second interpretation known as ‘reverse causality’ seems to be more convincing. Objective social security (X) and subjective security (Y) are associated but not in the way that we might suspect at first glance. Instead of X causing a change in Y, it may actually be the other way around, and Y causes a change in X. Specifically, when subjective insecurity (Y) rises in a country, the welfare state responds by increasing social protection spending (X). In view of these mixed findings, we can accept hypothesis H4-a but reject hypothesis H4-b while assuming that a decrease in social expenditures is accompanied by more perceived socio-economic insecurity.
Model VII indicates that higher levels of expenditures for ALMPs are associated with less perceived socio-economic insecurity (only) at the country level, which supports hypothesis H5-a but not hypothesis H5-b.
Finally, in Model VIII, we see that the OECD Employment Protection Index shows no significant impact on perceived socio-economic insecurity at either level. This finding is clearly contrary to our hypotheses H6-a and 6-b, which assert that more unregulated labour markets are echoed by more perceived socio-economic insecurity.
Conclusion
In this study, we employed data from two rounds of the ESS and the ‘between-within’ model to distinguish the effects of the crucial macro-level determinants of the between- and within-effects of these country characteristics on perceived socio-economic insecurity. For this analysis, we derived hypotheses based on socio-economic and welfare-institutional factors and performed linear (three-level) multilevel regression analysis.14
Our empirical analyses yield four main conclusions. (1) We found notable differences within and between countries regarding people's perceived socio-economic insecurity. In a cross-time comparison, the overall level of perceived socio-economic insecurity proved to be lower in 2016/17 than in 2008/09. From a cross-country perspective, we detected a fairly familiar pattern of welfare regimes. Scandinavian countries have been performing well, whereas former communist countries have shown higher levels of perceived socio-economic insecurity. Intermediate positions have been predominantly occupied by countries that belong to the Continental, Anglo-Saxon or Mediterranean welfare regimes. (2) In particular, a country's economic power (as measured by GDP per capita) did not have a significant impact on perceived socio-economic insecurity within countries but proved to be effective in explaining the variation in the perceived socio-economic insecurity levels between countries. This result suggests that the inconsistent findings concerning the impact of GDP per capita on perceived socio-economic insecurity in previous studies may originate from a failure to separate the between-effect (effect of levels) and the within-effect (effect of changes). (3) The unemployment rate has a positive impact on perceived socio-economic insecurity at the country-year level. A possible interpretation of this finding is that people react sensitively to changes in the unemployment rate in their own country; therefore, subjective insecurity increases when more citizens are laid off than before (Chung and Van Oorschot 2011; Chung 2019). (4) Interestingly, we revealed different directions of the within- and between-country effects regarding social expenditures on perceived socio-economic insecurity. Surprisingly, increasing social expenditures within a certain country go along with higher levels of perceived socio-economic insecurity. In contrast, higher levels of social expenditures across countries are expectedly associated with less perceived socio-economic insecurity. The former finding – which is surprising at first – may hint at a reverse causality effect; that is, instead of objective social security causing changes in subjective security, increasing subjective insecurity causes changes in objective social protection spending. In fact, previous research has suggested that insecure workers and employees are more likely to support the welfare state than secure workers; therefore, socio-economic insecurity may facilitate welfare provisions (Paskov and Koster 2014).
Some of our study's limitations invite future research. First, although we are confident that we have chosen pivotal objective measures of socio-economic and welfare-institutional factors, alternative measures may also help in understanding people's perceived socio-economic insecurity levels. In particular, the proxies for welfare policies are open to debate because it is divisive whether macro expenditure variables, such as total social expenditures and expenditures for ALMPs, effectively capture important aspects of welfare, such as access to and generosity of welfare states (Otto 2018).15 As an example, a large share of total social expenditures is related to the elderly (i.e. pensions, health and long-term care). Hence, we recommend that future studies think of alternative measures of welfare policies. Second, due to data availability, we addressed only two components of perceived socio-economic insecurity. Provided that suitable cross-national data are available, further research could choose a more complex measure of perceived socio-economic insecurity. Moreover, although our analysis assumes the linearity of the perceived socio-economic insecurity index, previous research has occasionally employed a dichotomous variable of perceived socio-economic insecurity (Chung and Van Oorschot 2011; Van Oorschot and Chung 2015; Chung 2019). To check the robustness of our findings, we also conducted a further analysis by utilizing a dummy for respondents with at least one of the two items of perceived socio-economic insecurity to be ‘likely’ or ‘very likely’ as a dependent variable (Table A5 in the Online Appendix shows the result of the ‘between-within’ model with a multilevel probit analysis and reports essentially the same results as given in Table 2). Third, the reliability of our statistical analyses would be improved by using a sample that contains more than 17 countries and more survey years. It is also beneficial to more appropriately calculate the average levels of the macro-level variables of countries and estimate the variation within each country. Therefore, this research recommends further studies and investigations of perceived socio-economic insecurity that adopt additional rounds of ESS in the future. Fourth, because our analysis concentrates on a sample of the working population with paid work and, thus, the unemployed and non-employed (i.e. early retirees and respondents on maternity or childcare leave) are excluded, there might be a selection bias. Hence, we recommend that future research take into account this issue.
Despite these possible improvements and extensions, our study contributes to the growing number of investigations on socio-economic insecurity perceptions among Europeans in turbulent times. Our study emphasizes the power of the economy and sheds differentiated light on the effect of welfare institutions on perceived socio-economic insecurity.
Acknowledgements
The authors are grateful to the participants at the annual conference of the ‘social policy’ section of the German Sociological Association (GSA) in Cologne, 26–27 April 2018, and at the 4th International ESS Conference (‘Turbulent times in Europe: Instability, insecurity and inequality’) in Mannheim, 15–17 April 2019, for their valuable comments on the presentations. Moreover, they thank Michalis Lianos and the two anonymous reviewers for their insightful suggestions and comments. All remaining deficiencies and errors are the authors’ alone.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Footnotes
Precisely, some previous studies have attempted to examine both the levels of some country characteristics and the changes in these variables within countries in the analysis of the between-country effects that adopts cross-sectional data with one wave (Anderson and Pontusson 2007; Chung and Van Oorschot 2011; Lübke and Erlinghagen 2014).
In some countries (i.e. Germany, the Netherlands, Norway, Spain and Switzerland), the field work for the fourth round of the ESS began a few weeks before 15 September 2008, the day the banking crisis commenced (Meuleman et al.2020). Therefore, it may not be appropriate to treat the ESS 2008/09 as a survey that fully captures the damages of economic crises.
It can be assumed that the COVID-19 pandemic will also be reflected in people's perceived socio-economic insecurity. However, because our latest data are from 2016/17, the crisis we refer to is the recent economic crisis (2008/09).
ESS4 edition 4.5 was released on 1 December 2018, and ESS8 edition 2.1 was released on 1 December 2018. The data for all rounds of the ESS, detailed information and documentation are made available for download for free on the ESS website (ESS-ERIC 2018).
According to the OECD, this 30-hour threshold serves as a common demarcation line for differentiating between part-time and full-time employment (Van Bastelaer et al.1997: 6).
Cronbach's alpha is 0.668 in the pooled dataset, meaning that the satisfactory value of this measure for internal consistency (generally > 0.7) was not quite met. However, these two items are significantly correlated – r = 0.502 (p = 0.000) and rho = 0.517 (p = 0.000) – which encourages us to use this index and perform linear multilevel regression analysis.
To compute the EGP class affiliation, we refer to the Trento version, which is available as SPSS syntax in Leiulfsrud et al. (2010).
Precisely, this is called ‘square root scale’ by Förster and D’Ercole (2009: 7). The missing values for equivalized household net income are imputed by using the multiple imputation method (2,618 cases).
A replication package for all analyses in this study is available online (Akaeda and Schöneck 2022).
We also created bivariate scatter plots to provide insights into each country's position regarding the relation of socio-economic and welfare-institutional factors on the one hand and country-specific means of perceived socio-economic insecurity on the other hand (see Figures A1 to A6 in the Online Appendix, where an overview of the bivariate correlations of our six indicators can also be found in Tables A2a and A2b).
To clearly show the three segments (‘lowest: lower than mean – SD; middle: between mean ± SD; highest: higher than mean + SD’) in each ESS round and countries’ trajectories, we present the results for 2008/09 and 2016/17 in two separate graphs. However, to visualize changes in each country, a combined graph is provided in Figure A7 in the Online Appendix.
According to Hox et al. (2017: 21), the ICC at the country level is , and the ICC at the country-year level is .
In accordance with Snijders and Bosker (2012: 113), we calculate the explained variance R2 as follows: the total variance in Model 0 is 0.237 + 0.094 + 1.678 = 2.009, and the total unexplained variance in Model I is 0.131 + 0.084 + 1.450 = 1.665. Thus, the explained variance R2 is 1 – (1.665/2.009) = 0.171.
For a robustness check, we conducted an additional analysis of each of the two rounds of the ESS. Table A4 in the Online Appendix reports similar results to Table 2 regarding the between-effects of the macro variables on perceived socio-economic insecurity.
On this issue, Otto (2018) reported that the unemployment cash benefit expenditure is positively associated with access rate and the cash benefit right generosity with unemployment cash benefit.
References
Nadine M. Schöneck is a Professor of Sociology and Empirical Social Research at Niederrhein University of Applied Sciences in Germany. Her current research focuses on cross-national comparisons and diagnoses of contemporary (advanced modern) societies in relation to social inequality and welfare states, the sociology of time and work-life balance and job-related spatial mobility. Recently, she (with Florian R. Hertel) published the paper, ‘Conflict perceptions across 27 OECD countries: The roles of socio-economic inequality and collective stratification beliefs,’ in Acta Sociologica.
Naoki Akaeda is an Associate Professor in the Faculty of Sociology at Kansai University in Japan. He conducts research on how social capital – including social trust, family, relatives, neighbours and friends – and well-being vary depending on the characteristics of countries through international comparative analyses. His recent work includes ‘Social Contact with Family and Relatives and Happiness: Does the Association Vary with Defamilialization?’ in the European Sociological Review.