The individual-level data used in this study came from two datasets. The first dataset contains both the combined World (WVS) and European Values Study (EVS) (WVS 2016). It includes a number of good proxies for the social norm to work and is unique in the large number of countries covered. Moreover, it covers that large number of countries over a long period of time, which increases the power of the study. The second dataset we used is the European Social Survey (ESS), covering only European countries over the period from 2002 to 2008 (ESS 2015). Although this dataset does not include individual proxies for the social norm to work, it contains a more refined measure of employment status. This makes a better examination of the impact on the economically inactive possible, and it allows us to make the distinction between the long- and short-term unemployed.
WVS and EVS
The WVS and the EVS share a common questionnaire and procedures for sampling and data collection. The data were collected through phone or face-to-face interviews on a sample generated through a multi-stage or stratified random sampling procedure. After we omitted country-years from the analysis because of a lack of data on a key variable, we had 134 country-year pointsFootnote 1 (48 countries for the period from 1990 to 2009) at our disposal. The analysis focused on people of working age (21–65 years old, N = 142,833), because the norm to work applies mainly to people in this age group.
The individual measurements of this study are as follows. To describe their subjective well-being, individuals were asked to describe their well-being both through items that measured both their affective and cognitive reflections on their well-being (Veenhoven 2008). The ‘affective’ item asked individuals how happy they felt, and the ‘cognitive’ item asked them to compare their current life with their desired life. This latter item captures life satisfaction and was measured with the question ‘All things considered, how satisfied are you with your life as a whole nowadays?’ Answers ranged from dissatisfied (1) to satisfied (10). The ‘affective item’ captures happiness, and was measured with the question: ‘Taking all things together, how happy would you say you are?’, with answer categories ranging from very happy (1) to not happy at all (4). For the analyses the mean of both measures was calculated after happiness was rescaled to range from 0 to 9. The suitability of the measure for cross-national purposes can be questioned because it can be influenced by cultural conditions, which differ among countries. However, the main interest here is in the well-being gap between the unemployed and the employed within countries, and how this well-being gap differs between countries. Therefore, the results of this study are unlikely to be affected by cross-national differences in people’s understanding of the item (descriptive statistics are shown in Table 1).
Table 1 Descriptive statistics on individual and country characteristics, WVS/EVS data
Employment status was measured as follows: respondents could indicate that they were a full-time employee, part-time employee, self-employed, unemployed, homemaker, student or retired, or specify their employment status as ‘other’.
The social norm to work of a society was calculated by aggregating the answers of all respondents within a given country-year on one item. The instructions were: ‘How would you place your views on this scale? 1 means you agree completely with the statement on the left, 10 means you agree completely with the statement on the right, or you can choose any number in between’, and the item was ‘People who are unemployed should have to take any job available or lose their unemployment benefits’ (1) to ‘People who are unemployed should have the right to refuse a job they do not want’ (10). The item was reverse-coded such that a high score (up to 10) represented a strong norm to work (‘they should take any job’) and a low score (toward 1) indicated a weak norm to work (‘right to refuse a job’). This measure thus captures the extent to which people in a society are expected by others to work. A higher share of people that impose this expectation on others will increase the vulnerability of the unemployed to external social sanctions. The norm to work is particularly weak in eastern European countries such as Montenegro (4.8) and Ukraine (5.1), and strong in the US (6.2), northern European countries such as Germany (6.5), and in Slovenia (7.3) and Italy (7.3).
We argue that this item measures a social norm to work because it reflects the tendency of individuals to opine that everybody should work if given the chance (‘should take any job’); it is an imperative that people impose on the generalized other (unemployed). This is what makes social norms social: they are behavioural expectations imposed on others (Elster 2007; Bicchieri 2006). Individuals who allow the unemployed to refuse any unwanted job they, do not impose a norm to work on others, while others may locate themselves somewhere between the two extreme answer options and adhere to a medium weak or strong social norm to work.
Additionally, the present study also looks at the individual endorsement of the social norm to work. This way, the influence of possible self-sanctioning among unemployed people can be taken into account. As a results, the effect of the social norm to work is independent of individual level preferences.
Across societies, the national measure correlates remarkably weakly with countries’ work ethic (which measure will be discussed later) (r = .05, p = .659) (using the same standard measure as Stam et al. 2016; Stavrova et al. 2011). This indicates that this measure taps into a distinct characteristic of the social norm to work and hence provided a completely new test of the social norm theory of unemployment in the present study. There are a number of reasons why the measure from the current study might tap a different dimension of the social norm to work.
First, the face validity of the current measure can be argued to be higher than that of work ethic due to its stronger incorporation of the aspect of social expectations. It questions whether others, i.e. the unemployed, should take any job.
The second reason this measure may be more appropriate is its availability over a longer time period and a larger number of countries than previous studies (e.g. Stavrova et al. 2011, who studied the period 1999–2009 and covered 55 country-years). This is crucial because a social norm to work is a national characteristic, and hence a large number of countries (over time) is needed to test its effects.
Thirdly, the measure correlates as expected with other individual and national characteristics. We find that it correlates with a right-wing political orientation (as can be expected since right-wing parties are stricter on the unemployment benefits (r = .08, p < .001; r = .18, p < .001 in West European countries in which right wing is more connected to liberal labour market policies); and that it correlates with the belief that immigrants are a strain on the welfare state (r = .12, p < .001). Furthermore, there are strong correlations with countries’ GDP (r = .40, p < .001), social expenditures (r = .35, p = .075) and unemployment level (r = −.26, p < .01). Whereas countries’ work ethic also correlates highly with social expenditures (r = .64, p = .115), it correlates with a lower GDP (r = −.41, p < .001) and a higher unemployment level (r = .30, p < .05). We find it more plausible that a social norm to work is stronger in wealthier countries and countries with low unemployment levels because individual attributions of unemployment may become more dominant during times of low unemployment (Buffel et al. 2017), although one could also argue that the social norm to work would be stronger in times of higher unemployment because of a higher strain on the welfare state. Moreover, the current measure correlates most strongly with the items underlying people’s work ethic that most directly relate to the unemployed, i.e. ‘people who do not work turn lazy’ (r = .14, p < .001 on the individual level and r = .01, p = .982 on the macro-level) and ‘work is a duty towards society’ (r = .15, p < .001 on the individual level and r = .41, p < .001 on the macro-level), while it correlates most weakly with the item that least relates to the unemployed, i.e. ‘to develop talents you need to have a job’ (r = .06, p < .001 on the individual level and r = −.05, p = .659 on the macro-level), as well as with the entire work ethic scale (r = .17, p < .001 on the individual level and r = .05, p = .699 on the macro-level).Footnote 2
A clear weakness of the current measure is that it is only based on one item, as compared to five items representing countries’ level of work ethic. This also makes it difficult to assess its cross-cultural equivalence.
We used a number of individual and national-level control variables that are commonly known to be related to unemployment and well-being (Russell et al. 2013; Stavrova et al. 2011): age and age squared; education (ranging from 0, meaning uncompleted primary education, to 7, meaning university with degree); self-rated income level (running from 0, the lowest category, to 10, the highest category); marital status (married/living together, single, divorced/separated, widowed); and religiosity (‘how often do you attend religious services?’—with answer categories ranging from (0) ‘never’ to (7) ‘more than once a week’). Moreover, a control variable was added for people’s personal endorsement of the social norm to work. At the national level, the analyses controlled for GDP per capita and unemployment levels derived from World Bank data. Like some previous studies (e.g. Sjöberg 2010), the analyses controlled for a state’s welfare generosity by including the income replacement rate of unemployment insurance benefits, which was derived from OECD data.
The European Social Survey
The ESS is a repeated cross-sectional survey in which individuals 15 years of age and older are interviewed face-to-face. In general, the ESS is considered to be of high quality, which is reflected in its relatively high response rates and many reliable and valid measurements. Additionally, samples are representative of the countries from which they are drawn (ESS 2016). Ultimately, the dataset we used contained information on 125,233 respondents living in 31 countries for 2002, 2004, 2006 and 2008. In sum, we had 74 country-year points at our disposal
We attempted to align the data as much as possible with the WVS/EVS, except for the more detailed measure of employment status (see Table 2 for descriptive statistics). For subjective well-being, again the mean score of people’s happiness and life satisfaction was used. For the item ‘Taking all things together, how happy would you say you are?’, respondents ranked their happiness on a scale from 0 (extremely unhappy) to 10 (extremely happy). Another item asked respondents: ‘All things considered, how satisfied are you with your life as a whole nowadays?’, with answer options ranging from 0 (extremely dissatisfied) to 10 (extremely satisfied). In this way, our analyses using the ESS dataset also provide a robustness check for the sensitivity of the findings since we converted a 4-points measure (happiness in the WVS/EVS) to a 10-point measure.
Table 2 Descriptive statistics on individual and country characteristics, ESS data
For employment status, people were asked: ‘Which of these descriptions applies to what you have been doing for the last 7 days?’. The following categories were distinguished: ‘employed (or self-employed)’, ‘unemployed for less than 2 years’, ‘unemployed more than 2 years’, ‘in education’, ‘permanently sick or disabled’, ‘retired’, ‘doing housework’, and ‘other’.
The social norm to work values derived from the EVS/WVS data were added to the ESS data as a contextual variable and assigned to the corresponding country-years. We entered the same control variables in the analyses of the ESS as in the EVS/WVS analyses. Educational attainment was measured on the ISCED scale, ranging from 1 (‘less than lower secondary’) to 7 (‘higher tertiary education’). Furthermore, we controlled for self-ranked income level (standardized score ranging from −1.49 to 1.73), marital status (married, separated, divorced, widowed, and never married), and church attendance (ranging from (1) ‘never’ to (7) ‘every day’). Unfortunately, it was not possible to control for people’s personal endorsement of the norm because the item used for the social norm to work in the EVS/WVS was not available in the ESS. On the national level, we again controlled for GDP, unemployment rate and replacement rate.
Work ethic was measured using the same measurement as in Stam et al. (2016) and Stavrova et al. (2011), using people’s endorsement of the following five statements: ‘Work is a duty towards society’, ‘People who don’t work turn lazy’, ‘It is humiliating to receive money without having to work for it’, ‘Work should always come first, even if it means less spare time’, and ‘To fully develop your talents, you need to have a job’.