The leisure-income trade-off and the (dis)utility of work

In standard labour supply models work is purely instrumental; only leisure and consumption, outside the work domain, can improve the welfare of individuals. These models include the extreme position that if individuals like working, it must be because they enjoy leisure at work (Reference DickinsonDickinson, 1999). In fact, the labour supply model is framed in terms of a leisure-income trade-off from which work actually disappears (Reference DerobertDerobert, 2001). However, in the 1980s, the theory of compensating wages (Reference Rosen, Ashenfelter and CardRosen, 1986) extended the conventional model of labour supply by including in the workers’ utility function a variable that captures their preferences for the several facets of work. The utility/disutility derived from work facets is revealed through the willingness to accept reduced or to demand increased wages. Work ends up resembling more consumption – of pleasurable and unpleasurable facets of work – than production, and the worker as a producer is eclipsed and replaced by the worker as a consumer (Reference LopesLopes, 2011).

In the economic literature, the conception of work as a source of disutility is related to a hedonic view of well-being, conceived as the satisfaction of individual preferences: the ‘hedonic price/wage theory states that individuals should accept lower wages for an occupation with personally desirable characteristics’ (Reference WellsWells, 2010: 296). It is assumed that the satisfaction of preferences yields utility, or more precisely experienced utility (Reference Kahneman and ThalerKahneman and Thaler, 1991). Experienced utility is related to instant or temporal hedonic, pleasurable, experiences and to pain avoidance (Reference Mitra, Jenkins and GuptaMitra et al., 2015), and effort is undisputedly considered painful.

As in standard economics, in popular representations work is often depicted as the antithesis of pleasure. That working is not pleasurable is documented by a large body of empirical work. Using the Day Reconstruction Method (DRM), Reference Kahneman, Kruege and SchkabeKahneman et al. (2004) show that time at work provides fewer positive feelings than leisure and time spent with family. Similarly, Reference Kopperud and VittersoKopperud and Vitterso’s (2008) DRM study shows that engaging in core work tasks provides lower pleasure than having breaks at work and commuting. However, various facets of work generate positive utility; in the economic literature, such facets are associated with ‘hedonic’ value, and job quality is defined as ‘well-being from working’ (Reference ClarkClark, 2015: 2), that is, the well-being derived from ‘good jobs’ that is associated with experiencing pleasure. The valued work facets include decision-making, skill usage, job security and pay (Reference WellsWells, 2010), a list to which Reference ClarkClark (2015) adds job content and interpersonal relations. These aspects are reported by workers as ‘being important’ and are shown to be highly related to job satisfaction.

To sum up, the discontinuous and contradictory image of well-being at work provided by the evidence collected by various social sciences suggests that the (dis)utility of work is much more complex than is commonly assumed. In our view, the apparently contradictory results can be reconciled, and a coherent account of work experience can be provided. This requires further elaboration of the concepts of utility and well-being regarding work. Viewing work as a meaningful activity and further elaborating its relationship with well-being call for a eudaimonic approach that conceives well-being as an active process in terms of human flourishing. This was well understood by institutional economists, such as Thorstein Veblen, who long ago pointed out that

The hedonic conception of well-being held by mainstream economists tends to disregard the role of active flourishing towards human welfare by merely focusing on consumption. The relation between work and human flourishing is also an aspect which is not given due attention in the job quality literature (Reference Findlay, Kalleberg and WarhustFindlay et al., 2013; Reference GallieGallie, 2013b). Even the literature on the economics of happiness that distinguishes evaluative from affective and eudaimonic well-being does not address work, nor its contribution to well-being (Reference Benjamin, Heffetz and KimbalBenjamin et al., 2014; Reference Clark, Adler and FleurbaeyClark, 2016).

Hedonism, eudaimonism and work

Psychologists advance two contrasting perspectives on well-being. The hedonic approach supports the idea that pleasure is the only good in human life, and the eudaimonic approach defends that realising one’s potential is the ultimate human goal. Hedonism associates well-being with a subjective experience constituted by high levels of positive affect, low levels of negative affect and high satisfaction with life. In contrast, eudaimonism considers that while pleasantness is a part, it is not enough to fully realise human well-being; the latter ‘is not so much an outcome or end state as it is a process of fulfilling or realising one’s daimon or true nature’ (Reference Deci and RyanDeci and Ryan, 2008: 2, our italics). Eudaimonia was first invoked by Aristotle as an answer to the fundamental question of how we should live (Reference RyffRyff, 2014). Thus, it has an ethical dimension which is associated with the purpose and meaning of life.

Eudaimonism claims that well-being is more than a psychological or affective state; it is a way of living that is focused on what is intrinsically worthwhile and on acting virtuously. This entails that nurturing intrinsic aspirations has more beneficial effects on psychological health than nurturing extrinsic goals. A growing body of empirical evidence shows that pursuing intrinsic goals (e.g. self-development, affiliation and health) is more positively related to well-being indicators (e.g. positive affect, vitality and self-actualisation), and less negatively related to measures of ill-being (e.g. depression and anxiety), than the pursuit of extrinsic goals (e.g. accumulation of money, fame and social image; Reference Niemiec, Ryan and DeciNiemiec et al., 2009; Reference Ryan, Veronica and DeciRyan et al., 2008). Moreover, the direct pursuit of hedonic activities has been shown to increase pleasure (positive affect) in the short run, but the pursuit of eudaimonic activities has revealed a more enduring effect on self-realisation (Reference Ryan, Veronica and DeciRyan et al., 2008).

The self-determination theory, which endorses a eudaimonic perspective, argues that three kinds of ingredients (considered basic psychological needs) are required to achieve self-realisation (Reference Deci and RyanDeci and Ryan, 2008; Reference Ryan and DeciRyan and Deci, 2001): (1) control or self-determination over one’s life, (2) competence and (3) social respect and care. These ingredients correspond to the facets of work reported in the economic literature that are associated with positive utility: discretion, skill usage, job security and interpersonal relations (Reference ClarkClark, 2015; Reference WellsWells, 2010). Likewise, these are also the factors that were identified as the main ingredients of job quality in sociology-oriented literature (Reference Findlay, Kalleberg and WarhustFindlay et al., 2013; Reference GallieGallie, 2013b; Reference Green, Felstead and GallieGreen et al., 2016).

Drawing on Aristotle, within the framework of lifespan theory of human flourishing, Reference Ryff and SingerRyff and Singer (2000) identify six ingredients of self-realisation: autonomy, personal growth, self-acceptance, life purpose, mastery and relatedness. More recently, Reference RyffRyff (2014) claimed that only a few studies have directly analysed the relation between work and eudaimonia. Job quality studies, in turn, use measures involving affective states/feelings, such as job satisfaction (i.e. measures associated with hedonia), while focusing on well-being at work rather than on overall well-being. However, contrary to mainstream economics, work intensity (a measure of effort) is not associated with low well-being at work; rather, only when high work intensity is associated with low scope of discretion is the situation revealed to be detrimental to well-being (Reference Lopes, Lagoa and CalapezLopes et al., 2014; Reference Theorell, Hammarström and AronssonTheorell et al., 2015). Jobs combining high work intensity and high level of discretion are termed ‘active jobs’ in the literature, which may be associated with both eudaimonia and hedonia. Moreover, the fact that volunteer work is shown to be positively related to eudaimonic well-being and not at all related to hedonic well-being suggests that work does involve eudaimonia (Reference Son and WilsonSon and Wilson, 2012).

Finally, while work provides fewer positive emotions when compared with other daily activities, work core tasks are nonetheless associated with perceptions of being in control and feelings of engagement, which are two basic components of eudaimonia (Reference Kopperud and VittersoKopperud and Vitterso, 2008). Indeed, having control over one’s life and engaging in activities one considers intrinsically worthwhile are often proxies for eudaimonia in the empirical literature (Reference Ryan, Veronica and DeciRyan et al., 2008).

The meaning of work and human flourishing

Major classical political economists have pointed out the link between work and human flourishing. Marx emphasised that work, if organised in non-alienating circumstances, can provide people with the opportunity for self-realisation and a meaningful life. In turn, Marshall stressed that ‘it is the science of activities and not that of wants [… which] may claim to be the interpreter of the history of man’ (Reference MarshallMarshall, 1966 (1890): 116). According to Marshall, certain types of activity are pursued not for ulterior motives, but because they are ends in themselves. He refuses to make the theory of consumption the ‘scientific basis of economics’ and insists that the primary task of economists is to develop a theory of activities.

Reference HinchcliffeHinchcliffe (2004) contends that the still prevailing instrumental and quite devalued view of work originates in Aristotle’s distinction between techne (technical activities whose ends are external to themselves, like producing use-values) and arete (virtuous activities motivated by ethical dispositions). According to Aristotle, it is arete which is associated with eudaimonia: ‘humankind realizes its value through activities far apart from anything as vulgar as the creation of use-value’ (cited in Reference HinchcliffeHinchcliffe, 2004: 538). Work and virtuous activities would then belong to separate domains, which would mean that work and eudaimonia would be antithetical.

Recently, Reference Rosso, Dekas and WrzesniewskiRosso et al. (2010) built an integrative conceptual model of meaningful work, which ‘highlights the importance of the reciprocal dynamics between individuals and groups [… whereby] the individual works to benefit the self and the collective, and the fruits of this work enhance both self and collective’ (Reference Steger, Dik and DuffySteger et al., 2012: 324). Thus, in this view, the meaningfulness of work by and large rests on the perceptions of the impact of one’s work on others, on which the sense of one’s worth is significantly based. Work, as experienced by workers in contemporary societies, may hence be said to be ethically driven (Reference LopesLopes, 2018).^{Footnote 1} Once we accept that work has an ethical dimension, as it is now becoming consensual in studies on the subjective experience of work, it is legitimate to consider work as an opportunity for human flourishing and it is by no means incompatible with the instrumental nature of work. The two domains are intertwined rather than separate.

The data and the hypotheses

Our study is based on micro-level data collected by the FESSUD Finance and Well-Being Survey, a cross-sectional survey designed to assess the impact of the financial crisis on well-being (Reference Santos, Lopes and CostaSantos et al., 2016). The survey was carried out in Germany, Poland, Portugal, Sweden and the UK, countries selected to account for different types of financial systems and Welfare State regimes in the EU. It consisted of telephone interviews (landline and mobile phone) carried out in November/December 2014 with nationally representative samples of households. The questionnaire was applied to the individual (aged 18 or older) in the household who declared he or she knew about and was co-responsible for making decisions about household finances. The sample size ranged from 1300 in Portugal to 1501 in Poland and Sweden, with a total of 7009 interviews conducted. The formula used to calculate the sample size took into account the desired margin of error (2.5%) for a 95% confidence level, as well as population size. The data set was weighted to correct for demographic biases, namely for household type, household size and household income. The overall response rate (completed interviews to cases reached) was on average 13.3%, which is relatively low but reasonable for a fairly lengthy telephone interview on a difficult topic (i.e. household finances). Since we are interested in work issues, the empirical analysis considers only respondents who declared they were employed, gathering 1936 observations with non-missing values in all variables.^{Footnote 2}

In the countries of the study, Portuguese employees report their households were severely hit by the financial crisis, with an average score of 7.0 (on a 1–10 scale, with 1 being ‘not bad’ and 10 ‘extremely bad’), followed by the Polish (4.5), the British (4.4), the German (3.4) and the Swedish (2.6) (Reference Santos, Lopes and CostaSantos et al., 2016: 90–92). Notwithstanding these contrasting perceptions, country divergences are less pronounced, though still very marked, in responses about changes occurring between 2009 and 2014 in the households’ financial situation (Table 1) and work conditions (Table 2).

Table 1. Responses to questions on household financial situation.

Question: Please tell me whether or not each of the following has happened to you in the past 5 years. Percentage of yes responses among employees, excluding DK/NA answers.

* Significantly different from Germany at 10% applying a chi-square test to the equality of proportions between countries, at **5% and at ***1%.

Table 2. Responses to questions on work conditions.

Question: Please tell me whether or not each of the following has happened to you in the past 5 years. Percentage of yes responses among employees, excluding DK/NA answers.

* Significantly different from Germany at 10% applying a chi-square test to the equality of proportions between countries, at **5% and at ***1%.

Tables 1 and 2 indicate that between 2009 and 2014 the households’ financial situation and work conditions deteriorated the most in Portugal, followed by Poland. On average, the Swedish, followed by the Germans, were the least affected. The UK occupies an intermediate position, even if closer to Germany than Poland.^{Footnote 3}

Perceived changes in family life and community involvement (Table 3) are less pronounced across these countries, while changes in life satisfaction again display a wide variation, with a higher percentage of the Portuguese respondents (66.4%) declaring a deteriorating overall life satisfaction (Table 4).

Table 3. Responses to questions on family life and community involvement.

Question: Please tell me whether or not each of the following has happened to you in the past 5 years. Percentage of yes responses among employees, excluding DK/NA answers.

* Significantly different from Germany at 10% applying a chi-square test to the equality of proportions between countries, at **5% and at ***1%.

Table 4. Change in life satisfaction.

Question: Compared to 5 years ago, has your overall life satisfaction deteriorated, stayed more or less the same or improved? Percentage of yes responses among employees, excluding DK/NA answers.

As it can be seen from the phrasing of the questions reported in Tables 1 –4, respondents were asked about (their perceptions of) changes in their households’ financial situation, working conditions, family life and community involvement and overall life satisfaction. This represents an important advantage regarding the absolute levels of well-being. Indeed, the analysis of absolute levels of well-being poses added interpretative difficulties. Reference BruniBruni (2008), for example, highlights the role of personality, the gap between aspirations and achievements (i.e. reports of well-being resulting from low aspirational levels), and social comparisons. Since the data we use refer to perceived changes rather than absolute levels, the measurement of well-being is deemed to be less affected, though still influenced, by these factors. This may be even more relevant in the context where the financial crisis produced different impacts on individual aspirations (Reference Boffo, Brown and SpencerBoffo et al., 2017).

Based on the reasoning presented in Section ‘Questioning the disutility of work and the related concept of well-being’ and the available data, we formulate four hypotheses. First, considering that hedonia (i.e. pleasure-seeking) and eudaimonia (i.e. human flourishing) are distinct but related types of well-being (Reference Benjamin, Heffetz and KimbalBenjamin et al., 2014; Reference Clark, Adler and FleurbaeyClark, 2016; Reference Ryan and DeciRyan and Deci, 2001), Hypothesis 1 (H1) states that hedonia and eudaimonia are distinct though related phenomena. Second, supposing that work fosters human flourishing that brings about a type of well-being (i.e. eudaimonic well-being) qualitatively different from that delivered by income and consumption (i.e. hedonic well-being), Hypothesis 2 (H2) states that depressed work conditions and FD are associated with a decline in both types of well-being, hedonia and eudaimonia. Hypothesis 3 (H3) states that depressed work conditions are more associated with a decline in eudaimonia than increased FD. Hypothesis 4 (H4) states that depressed work conditions have a larger impact on eudaimonia than on hedonia.

Building composite indicators of eudaimonia, FD and deteriorated work conditions

While self-reported satisfaction with life is already established as a reliable measure of hedonic well-being, measuring eudaimonia is more complex. Indeed, all available measures involve the construction of indices composed of a diversity of indicators, where control over one’s life is a key component in all of them (Reference BruniBruni, 2008; Reference Deci and RyanDeci and Ryan, 2008; Reference RyffRyff, 2014).

Accordingly, our measure of hedonia is obtained from the survey question on the evolution of overall life satisfaction (Table 4). As all the other questions refer to negative evolutions, we recoded this variable to facilitate interpretation, attributing the highest value (three) to a negative change, the intermediate value (two) to unchanged satisfaction and the lowest value (one) to a positive evolution in overall life satisfaction.

As for eudaimonia, we constructed a composite indicator by applying the Principal Component Analysis (PCA) to the three items that address dimensions usually associated with eudaimonia: overall control over life, time spent with family and friends, and involvement in unpaid voluntary work (Table 3). Although involving just a few questions, our measure contemplates three out of the six dimensions considered to fully capture eudaimonia (Reference Ryff and SingerRyff and Singer, 2000), namely environmental mastery, relationships with others and autonomy. The PCA revealed only one underlying dimension (eigenvalue 1.56 – accounting for 52.1% of the variance), on which the three items had large and positive loadings (Table 9 in the Appendix). Individuals’ scores for that component are our indicator of eudaimonia. The 0.00 score corresponds to the average level of the eudaimonia indicator across individuals, positive values show that the indicator is above average (indicating a higher deterioration of eudaimonia) and negative values indicate a below average level.^{Footnote 4}

To construct the deteriorated work conditions indicator, we applied the PCA to the eight questions related to several dimensions of work (Table 2). Although FD are captured in a separate indicator, we kept the question about changes in pay in this indicator because pay is a relevant dimension of work. Indeed, while job pay is important for individuals’ financial situation, it is also an important form of recognition for the work performed (Reference Méda and VendraminMéda and Vendramin, 2013) and may thus be also considered an intrinsic feature of work. The PCA analysis of work-related questions generated two components with an eigenvalue greater than one. After Varimax rotation, questions related to job security, job interest and pay have high loadings on the first factor, while questions on work intensity and longer working hours have positive loadings on the second factor (Table 10 in the Appendix). Given the items involved, we named the first component ‘Deteriorated Work Conditions’ (DWC) and the second one ‘More Intensive Work’ (MIW).

Finally, the indicator of FD was constructed by applying the PCA to the five questions about households’ financial situation (Table 1). The analysis revealed only one underlying dimension on which all items have high factor loadings (Table 11 in the Appendix). Since the ‘Deteriorated Work Conditions’ indicator has a high factor loading on the question about reduction of pay, we computed the correlation between DWC and the FD indicator. The resulting correlation coefficient is 0.45; consequently, it does not prevent the inclusion of both variables in the regression analysis.

Table 5 displays the positions of the five countries for the indicators constructed, where an increase denotes a worsening of the situation between 2009 and 2014, as perceived by the respondent.

Table 5. Composite indicators by country.

These indicators correspond to the respective scores resulting from the PCA (see text).

* Significantly different from Germany at 10% using a t-test, at **5% and at ***1%.

Portugal and Poland present above average deteriorated work conditions (DWC), the UK remains around average, and Germany and Sweden display below average values (DWC, MIW, FD and eudaimonia have an average of zero when observations for all countries are taken into account). As for work intensity (MIW), Portugal once again is the country where the situation deteriorated the most, but Poland is now close to the average, as is the UK, while the remaining two countries report an increase in work intensity below average. Finally, the relative position of the countries regarding FD is very similar to that of DWC, with Portuguese and Polish respondents reporting a worsening of their financial situation, but not the respondents in the UK, Germany and Sweden. As for eudaimonia, Portugal registers the largest decline, while Sweden displays the smallest. The countries’ behaviour in terms of hedonia is similar to that of eudaimonia (Table 5). However, Sweden stands out in that a substantial percentage of respondents (46.6%) report a positive evolution in overall life satisfaction, followed by Germany and the UK (35% each) and Poland (33%), with Portugal (7%) exhibiting an extremely low value (Table 4).

Testing the hypotheses

H1 states that eudaimonia and hedonia are separate but related phenomena. The Spearman’s correlation coefficient between the two well-being indicators is 0.23 and is statistically significant, meaning that, as expected, a decrease in eudaimonia is associated with a decrease in hedonia. We take this as supporting evidence for H1; the two measures of well-being are correlated but the correlation is weak, suggesting that the two indicators capture different aspects of well-being.

In order to test H2–H4, we regressed eudaimonia and hedonia with indicators of work conditions and FD, controlling for socio-economic determinants of well-being.^{Footnote 5} The controls selected from the survey are: age and its square, a dummy for females, income quintile, a dummy for part-time workers, a dummy for temporary contract, dummies for occupation (high-skilled white collars, high-skilled blue collars, low-skilled blue collars, the military, other occupations; low-skilled white collars as base category), dummies for the composition of households (single, single parent, couple with children, extended family, and other households; couple as base category), dummies for the education level (no formal education, primary education, tertiary education; secondary education as base category), and dummies for countries (Germany as base country). Note that the controls refer to the present individual situation and not to the change occurred in the previous 5 years.

Since the eudaimonia indicator is a continuous variable, ordinary least squares (OLS) is the appropriate regression tool.^{Footnote 6} The coefficient of MIW (0.29) is slightly larger than the coefficient of FD (0.26), and that of DWC (0.19), (Model 1 in Table 6). The results of DWC, MIW and FD support H2 that states that depressed work conditions, like increased FD, are associated with a decline in eudaimonia. In line with our theoretical framework, this means that deterioration in the quality of work is associated with a negative effect on eudaimonia, which supports the view that work is actually related to human flourishing.

Table 6. Regressions for changes in eudaimonia and hedonia.

Model 1 was estimated with OLS and Model 2 with ordered logit. R2 is for the OLS and pseudo R2 is for the ordered logit. Gender, family composition and education level are controlled for, but they are not presented to save space. Base categories are: low-skilled white-collar and Germany. The standard deviations (in brackets) were obtained assuming correlation of errors within countries (clusters by country).

* Significant at 10%, at **5% and at ***1%.

Regarding H3, although the average impact of MIW on eudaimonia is larger than that of FD, the difference is not statistically significant. In addition, the average value of the coefficient of DWC is smaller than that of FD, but the F-test to the equality of these coefficients does not rule out that they are equal (p value = 0.2243). Overall, we are not able to confirm H3, which states that depressed work conditions are more associated with a decline in eudaimonia than are FD. Moreover, DWC is less grievous for eudaimonic well-being than MIW,^{Footnote 7} that is, workers with less interesting work, less job security and less pay experienced a lesser decline in eudaimonia when compared with those who had to work more intensively and for longer hours.

In what concerns the control variables, the type of contract (permanent or temporary) and job (part-time or full-time) do not relate to a change in eudaimonia, nor do the occupational categories.

Given that hedonia is assessed with a categorical scale, ordered logit regression was used. This logit model estimates the probability of an individual having experienced a given change in well-being. As expected, increased FD and worse work environments (measured by both MIW and DWC) considerably reduce hedonic well-being (Model 2 in Table 6). Overall, our findings indicate that, even after controlling for FD, increased work intensity (MIW) and deteriorated working conditions (DWC) reduce hedonic well-being (supporting H2).

Model 2 shows that an increase of one unit in FD, DWC or MIW increases the odds ratio of being on a lower level of hedonic well-being by a factor of 0.46, 0.17 or 0.13, respectively.^{Footnote 8} Hence, financial circumstances, related to income and consumption, have a greater average impact on hedonia than work conditions.^{Footnote 9} The marginal effects convey the same qualitative information, but their size varies along with the values of the explanatory variables and with the change of well-being.^{Footnote 10}

Country-specific effects on hedonia are relevant: Poland and mainly Portugal had a greater decline in hedonia than Germany, whereas Sweden and the UK registered a smaller decline. The fact that the effects are different for eudaimonia and hedonia shows that the two types of well-being are different (H1).

Finally, we test the impact of work conditions on eudaimonia and hedonia (H4). This task is difficult because the two indicators are expressed in different scales. In this case, we can look at the proportionality of the coefficients of the two models (Reference Hausman and RuudHausman and Ruud, 1987), analysing the relative sizes of the coefficients of the work-related variables (DWC and MIW) and the FD coefficients in the hedonia and eudaimonia equations.^{Footnote 11} Looking at the average values, we observe that the FD coefficient is relatively bigger in the hedonic equation than in the eudaimonia equation, and that, conversely, the work indicators (DWC and MIW) have bigger coefficients in the eudaimonia regression than in the hedonia regression (Table 7). However, the Wald test cannot reject the proportionality of coefficients (Table 7), which means that H4 is not statistically supported when using this technique.

Table 7. Proportionality of key coefficients using models 1 and 2 of Table 6.

DWC: deteriorated work conditions; MIW: more intensive work; FD: financial difficulties.

The statistical test is based on a non-linear Wald test.

An alternative test of H4 is to convert the two indicators into a binary scale. The hedonia scale was transformed into a dummy variable that takes value one when life satisfaction deteriorates and zero otherwise. As the sense of control over one’s life is a key component of eudaimonia, we only use this question to measure eudaimonia, taking value one if overall control decreases and zero otherwise.^{Footnote 12}

Given the binary nature of the dependent variables, the regressions for eudaimonia and hedonia are estimated using a logit regression. They show that both MIW and DWC have a larger effect on eudaimonia (0.48 and 0.47, respectively) than on hedonia (0.306 and 0.308 respectively), which supports H4 (Models 1 and 2 in Table 8). The Wald test rejects the equality of the DWC coefficients of the hedonia and eudaimonia equations (p value = 0.0380), while the equality of the MIW coefficients cannot be rejected (p value = 0.1553). Consequently, in line with H4, there is evidence that an increase in DWC affects more eudaimonia than hedonia.

Table 8. Regressions for changes in eudaimonia and hedonia as binary variables, changes in hedonia standardised and robustness checks.

Model 1 and 2 were estimated with logit, Models 3 and 4 estimated with OLS, and Model 5 was obtained with ordered logit. In Models 4 and 5 the variables DWC and MIW were obtained without the item on reduction in pay. The other control variables present in Table 6 are also included in all regressions. The standard deviations (in brackets) were obtained assuming correlation of errors within countries (clusters by country).

* significant at 10%,

** 5%, and

*** 1%.

A third way of testing H4 is to standardise the hedonia variable (note that the eudaimonia variable is already standardised). Then, we applied the OLS to re-estimate the hedonia regression, and compare it with the regression for eudaimonia (model 1 of Table 6). We observe that the impact of the two work-related variables is larger in eudaimonia than in hedonia – now the impacts are comparable as they are measured in units of dependent variable’s standard deviation (Table 8). For example, a unitary increase in DWC leads to a decline in eudaimonia equivalent to 0.18 of its standard deviation, whereas the effect on hedonia is a decline equal to 0.07 of this variable standard deviation. The equality of the coefficients of DWC or MIW for eudaimonia and hedonia is statistically rejected through a Wald test (p-values of 0.0004 and 0.0000, respectively). This test shows the strongest evidence in favour of H4.

As robustness checks, we removed the question on pay from the work conditions indicators to assign financial issues only to the FD variable. The PCA on the seven work-related questions continues to produce two factors with broadly similar interpretations to those previously obtained. Estimation results are broadly the same, with the work-related variables having statistically significant effects on eudaimonia, but now only MIW affects significantly hedonia (Models 3 and 4 in Table 8).

The control variables part-time work and fixed term contract capture important work characteristics, and the variable of present income level is related to FD. Consequently, one may ask: do these controls hamper the effects of our variables of work conditions and FD on well-being outcomes? In order to answer this question, we excluded the three controls from the equations of eudaimonia and hedonia. The results obtained are similar to those including the controls,^{Footnote 13} the main change being that the work indicators are now statistically significant for eudaimonia at only 10%. This indicates that the initial inclusion of these controls does not lessen the impact of worse work conditions and FDs on well-being.

The measure of eudaimonia encompasses quite diverse components of life, which can react differently to work conditions and FD. We then ran a separate regression on the determinants of each of the three components of the eudaimonia indicator. The dependent variables are now binary, with value one for affirmative responses to each of the respective questions (see Table 3). Overall, the results are qualitatively similar to those using the composite indicator: MIW, DWC and FD have a statistically significant and positive effect on the deterioration of the three different aspects of eudaimonia.