Low-pay higher pay and job satisfaction within the European Union empirical evidence from fourteen countries.

Low-pay higher pay and job satisfaction within the European Union empirical evidence from fourteen countries Luis Diaz-Serrano♣ National University of Ireland Maynooth, IZA and CREB Jose A. Cabral Vieira♦ University of the Azores CEEAplA Abstract: We examine differences in job satisfaction between lowand higher-paid workers within the European Union (EU). To do so The European Community Household Panel Data covering the period 1994-2001 is used. Then we test for differences in reported job satisfaction between lowand higher-paid workers. We also explain the existence of differences in the determinants of job satisfaction between these two types of workers and across countries. Our results indicate that low paid workers report a lower level of job satisfaction when compared with their higher paid counterparts in most countries, except in the UK. This supports the idea that low-wage employment in these countries mainly comprises low quality. The results also indicate that gap in average job satisfaction between lowand higher-paid workers is markedly wider in the Southern European countries than in the rest of EU. Finally, there are significant differences in the determinants of job satisfaction across countries. It seems then that a homogeneous policy may be inappropriate to increase satisfaction, and hence labour productivity, in the EU as a whole. Hence, an improvement of the quality of the jobs in the EU may require different policies. In particular, in some countries such as the United Kingdom removing low employment, namely through regulation, may worsen the workers’ well-being, although in other cases such a policy may lead to a totally different outcome. Running title: low-pay higher pay and job satisfaction Key words: Job satisfaction, job quality, low-wage employment JEL-Code: J28 ♣ Department of Economics (Rhetoric House), National University of Ireland Maynooth, Co. Kildare, Ireland. E-mail: [email protected]. ♦ Department of Economics and Management. University of the Azores, Rua Mãe de Deus, 9500 Ponta Delgada – Portugal. E-mail: [email protected] 1 Introduction Job satisfaction has become a matter of great interest in social sciences. It has been argued in the literature that answers to questions about how people feel toward their job are not meaningless but rather convey useful information on individual behavior such as job quits (Hamermesh, 1977, Freeman, 1978, McEvoy and Cascio,1985, Akerlof et al.,1988 and Shields and Price, 2002), absenteeism and productivity (Vroom, 1964, Mangione and Quinn, 1975, and Clegg, 1983). Moreover, job satisfaction is considered a strong predictor of overall individual well-being (Argyle, 1989 and Judge and Watanabe, 1993). As a consequence, several studies have attempted to identify the determinants of job satisfaction (see, Borjas, 1979, Miller, 1990, Meng, 1990, Idson, 1990, Clark, 1996 and 1997, Clark and Oswald, 1996 and Souza-Poza and Sousa-Poza, 2000). In addition, the incidence and the persistence of low-pay employment has become a topic of concern in many developed economies as a result of increasing wage inequality (see OECD, 1996 and 1997, Asplund et al., 1998, Lucifora and Salverda, 1998, Stewart and Swaffield, 1999, Cappellari, 2000, Cardoso et al., 2000). Moreover, low pay employment and job quality have become important policy issues, namely in the European Union (see European Commission, 2001 and 2002). Also Salverda et al. (2001) put forward the idea that policies towards low-wage jobs should center on their quality at least as importantly as on the level of pay which they provide. Job quality is a relative concept regarding objective characteristics of the jobworker match. It also involves the subjective evaluation of these characteristics by the respective worker, on the basis of his characteristics and expectations. The European Union Employment in Europe (2001) report suggests that in the absence of a single 2 composite indicator any analysis of job quality must be based on data on both objective and subjective evaluations of the worker-job match. In addition, the European Union Employment (2002) report reinforces this stance with the suggestion of the inclusion of job satisfaction in its definition of job quality. Indeed, Leontaridi and Sloane (2001) use job satisfaction as a proxy of job quality in the British labor market. Furthermore, they attempt to distinguish between two strands of the literature: the segmented labor market theory versus compensating wage differentials theories. The segmented labor market view, or, at least, the dual labor market version, claims the existence of two distinct labor markets with strong mobility barriers between them. In addition, this theory argues that we can classify jobs into good jobs and bad jobs, with bad jobs being those not only having worse working conditions, but also lower wages. As Leontaridi and Sloane (2001) argue, this contrasts with the compensating wage differentials theory according to which jobs with poor working conditions would be expected, ceteris paribus, to compensate for this with higher pay. Leontaridi and Sloane (2001) surprisingly conclude that low pay workers report a higher job satisfaction than their higher pay counterparts. In their opinion, this casts doubt on the view that low paid jobs are bad jobs and that high paid jobs are good jobs. This is reinforced by their finding that it is by no means the case that moving from a low paid to a higher paid job increases job satisfaction. In sum, the results do not support the view that low paid jobs are inherently of low quality, at least as far as British evidence is concerned. This seems in line with a view that low paid workers likely obtain compensating differences in the form of non-pecuniary benefits. Jones and Sloane (2003) and Leontaridi et al. (2004) also present this type of conclusion. 3 Apparently, there is a matching process in the labor market as a whole in which individuals seeking higher pay are allocated to higher-paying jobs and those seeking for non-pecuniary benefits are allocated to low-paying jobs. In such a case, removing low paid jobs, namely through regulation, would not necessarily improve worker’s well being. However, there is no reason to believe that such a type of findings hold for the other European labour markets. In this paper we also use job satisfaction as a proxy of job quality and attempt to shed further light on the issue, trough the examination of a large number of EU countries. For this purpose, we use seven waves of the European Community Household Panel (ECHP). The panel nature of the data allows us to use a random effects estimator in order to control for unobservable individual heterogeneity. The study takes advantage from the fact that period of analysis is the same for all countries. In addition, the data is based on a standardized questionnaire and the regressions use the same type of explanatory variables and estimation procedures. These features are suitable to perform comparisons among countries. The paper is organized as follows. The next section describes the data and provides evidence on reported levels of job satisfaction by low and higher paid workers within fourteen EU countries. Section 3 evaluates the determinants of job satisfaction for low and high paid workers separately. Finally, section 4 concludes. The Data and variables The data used in this paper come from the European Community Household Panel (ECHP). This is a yearly panel of the EU-15 countries carried out by the Statistical Office of the European Communities (Eurostat) in cooperation with the National 4 Statistical Office of each country. The data collection started in 1994 and was conducted over eight consecutive waves. We make use all waves of the ECHP, thus covering the 1994-2001 period, for ten of the EU-15 countries (Denmark, the Netherlands, Belgium, France, Ireland, Italy, Greece, Spain and Portugal). For Austria and Finland the available files only cover the period 1995-2001 and 1996-2001, respectively. During the period 1994-1996 the data for Germany and Luxembourg come from two different sources and the original ECHP files are mixed with the German Socioeconomic Panel (GSOEP) and the Luxembourg Household Panel (PSELL), respectively, whereas for the remaining waves covering the period 1997-2001 all the data come exclusively from the ECHP files. These ECHP files for Germany and Luxembourg do not provide valid answers about the question on job satisfaction, whereas the GSOEP and the PSELL do. Therefore, for these two countries we can only use the information covering the period 1994-1996. For the UK the data comes from two different sources, the ECHP for the period 1994-2001 and the British Household Panel Survey (BHPS) for the period 19942001. Both data sources provide valid answers on the question on job satisfaction. In this study we use both surveys. Finally, the Swedish ECHP files do not provide information on job satisfaction in any wave. Therefore, this country is omitted from our analysis. The ECHP files contain information not only at household, but also very detailed data at individual level such as demographic information, employment status, job, education, training, earnings, hours worked, type of contract and a set of variables about the activity, size and sector of their employers. To our purposes the survey also provides answers on the level of satisfaction about some aspects of the individual’s life. Of these questions, one refers to the individual’s level of satisfaction with his/her job, 5 where individuals are asked to report on a six-point scale how satisfied they were with their work or main activity. The lowest level of the scale stands for workers who were not satisfied at all whereas the highest stands for fully satisfied workers. To carry out our analysis we select males and females aged bellow 65 that are salary earners working in either the public or private sector. Thus, non-salary earners and self-employed dropped from the final sample. As usual in most the literature, a low-wage worker is defined as an individual who earns less than two thirds of the median hourly wage. For this purpose, hourly wages were calculated as gross monthly wages divided by the number of hours worked per month. The corresponding values were transformed to 2001 prices through the consumer price index. Raw differences job satisfaction between low and higher paid workers Table 1 includes the sample mean values of job satisfaction broken by the lowpay threshold. There we also include the results concerning the test of the hypothesis of equality in the average job satisfaction between low-pay and higher pay workers. The results indicate that the mean level of job satisfaction is lower for low pay workers as compared with the one reported by their higher pay counterparts in most of the countries under analysis. --------- insert Table 1 about here --------- This supports the idea that low wage employment in these countries mainly comprises low quality jobs and is consistent with the segmented labor market theory, which claims the existence of good and bad jobs. The only clear exception to this 6 pattern is the UK where low pay workers significantly report a higher level of satisfaction[1]. This is in consistent with the other results reported in previous studies for this country and in line with the notion that British low pay workers may obtain compensating differences in the form of non-pecuniary benefits (see Leontaridi and Sloane, 2001, and Leontaridi et al., 2004). In addition, the difference in job satisfaction between low and higher pay workers is much more pronounced in Southern European countries such as Portugal, Spain, Greece and Italy. This may indicate that labor market segmentation and therefore differences in job quality, is less prominent in the other countries than in Southern of Europe. The determinants of job satisfaction The Econometric model This section describes the econometric model to be used in order to assess the determinants of job satisfaction for low paid jobs and for higher paid jobs separately. As we have mentioned, the level of satisfaction is reported on a six-point ordinal scale. Therefore, a suitable estimator for our purposes is the ordered probit model. However, a common problem in ordinal scales is that surveyed individuals may have a different perception of the same scale. On the one hand, we might expect that individual’s unobserved factors such as the emotional state or mood, which may vary across individuals, be also relevant for the outcome. On the other hand, we also might expect the observed and unobserved factors be correlated, which in turn may lead to inconsistent estimates. Given the panel nature of our data, one way for dealing with this problem would be the fixed-effects ordered probit model. Unfortunately, there is no 7 available formulation yet to estimate such a model. Hence, one alternative option is the random-effects ordered probit model, which possesses the attractive feature that allows us to control for this individual’s unobserved heterogeneity, although it does not allow for correlation between observed and unobserved determinants of satisfaction. This model is described below. Assume that the propensity of individual i to report a certain level of satisfaction in period t is driven by the following structure: *' 1,..., ; 1,..., it it it SXvi Nt T β =+ = = (1) where * it S is the latent outcome, Xit are the determinants of the outcome, iitit uv += ε with 222 1uuit )vvar( σσσ ε +=+= and 2 u 2 u v1 σ σ ρ + =. The term ui picks up the individual’s unobserved heterogeneity, which is assumed normally distributed with zero mean and variance u σ , and ε it is a time-varying error term distributed as ),(N 10 . Note that we do not observe * it Sbut observe instead an indicator variable of the type: ⎪ ⎪ ⎪ ⎩ ⎪ ⎪ ⎪ ⎨ ⎧ µ> =µ≤<µ µ≤ =− 4 * it j * it1j 0 * it it Sif6 4,3,2,1j,Sifj Sif1 S (2) The log-likelihood function reads: ∑ = =N 1i iT1i )S,...,S(PlogLogL (3) 8 Defining itjit X'a β µ −= −1and itjit X'b β µ − = we can write (3) as follows: [] 1 1 1 1 111 1 1 ( ,..., ) ... ( ,..., ) ... ... ( | ) ( ) ... () ( |) ( |) iiT iiT iiT iiT bb iiT iiTiiT aa bb it i i i iT i aa T iitiitii t PS S v v dv dv uudud d ubuaudu φ φ εφ εε φ +∞ −∞ +∞ = −∞ == == =Φ−Φ ∫∫ ∫∫∫ ∏ ∫ (4) where φ and Φ denote the density function and the cumulative distribution function of the normal distribution, respectively. Therefore, the log-likelihood for this model can be generalized from the arguments made by Butler and Moffit (1982). Heterogeneity is handled by using the Gauss-Hermite quadrature to integrate out the joint density (see Frechette, 2001, for further details). Estimation results In order to identify the determinants of job satisfaction we relied on available evidence on the issue, which suggests that wages are important but do not explain the whole variation in reported levels of job satisfaction. For instance, Clark (1997) finds that after controlling for wages and for a large set of other covariates, females are happier at work than males. Moreover, it has been found that reported satisfaction depends on variables such the age of the worker, comparison wage rates, level of education, employer size, industry, union membership status, region, health status, type of employment contract, hours of work and educational mismatches, among others (see, for instance, Borjas, 1979, Miller, 1990, Meng, 1990, Idson, 1990, Clark, 1996 and 1997, Clark and Oswald, 1996, Leontaridi and Sloane, 2001, Sousa-Poza and SousaPoza, 2000 and Sloane and William, 2000 and Jones and Sloane, 2003). For the purpose 15 European Commission (2001), Employment in Europe 2001: Recent Trends and Prospects, in Employment and Social Affairs, Chapter 4, Luxembourg. European Commission (2002) Employment in Europe 2002: Recent Trends and Prospects, in Employment and Social Affairs, Chapter 3, Luxembourg. Frechette, G. (2001), “Random-effects ordered probit”, Stata Technical Bulletin, Vol. 59, January, pp. 23-27. Freeman, R. (1978) “Job satisfaction as an economic variable”, American Economic Review Papers and Proceedings, Vol. 68 No.2, pp. 135-141. Hamermesh, D. (1977), “Economic aspects of job satisfaction”, in Ashenfelter, O. and Oates, W. (Ed.), Essays in Labor Market Analysis, John Wiley, New York, pp. 5372. Idson, T. (1990), “Establishment size, job satisfaction and the structure of work”, Applied Economics, Vol. 22 No.8, pp. 1007-19. Jones, R. and Sloane, P. (2003), “Low pay higher pay and job satisfaction in Wales”, WELMERC Discussion paper Nº. 2003-03, University of Wales Swansea. Judge, T. and Watanabe, S. (1993), “Another look at the job satisfaction - life satisfaction relationship”, Journal of Applied Psychology, Vol. 78 No. 6, pp. 93948. Leontaridi, R. and Sloane, P. (2001), “Measuring the Quality of Jobs: Promotion Prospects, Low Pay and Job Satisfaction”, LoWER Working Paper Nº. 07, University of Amsterdam. Leontaridi, R., Sloane, P. and Jones, R. (2004), “Are Low Paid Jobs of Low Quality? Some British Evidence”, WELMERC Discussion paper Nº. 2004-02, University of Wales Swansea. 16 Lucifora, C. and Salverda, W. (1998), Policies for Low Wage Employment and Social Exclusion in Europe, FrancoAngeli, Milan. Magione, T. and Quinn, R. (1975), “Job satisfaction, counterproductive behaviour and drog used at work”, Journal of Applied Psychology, Vol. 60, pp. 114-16. McEvoy, G. and Cascio, W. (1985), “Strategies for reducing employee turnover: a meta analysis”, Journal of Applied Psychology, Vol. 70 No.2, pp. 342-53. Meng, R. (1990), “The relationship between trade unions and job satisfaction”, Applied Economics, Vol. 22 No.12, pp. 1635-48. Miller, P. (1990), “Trade unions and job satisfaction”, Australian Economic Papers, Vol. 29 No.55, pp. 226-48. OECD (1997), “Earnings mobility: taking a longer run view”, in Employment Outlook 1997, pp. 27-61, Paris. OECD (1996), “Earnings inequality, low-paid employment and earnings mobility”, in Employment Outlook 1996, pp. 59-108, Paris. Salverda, W., Bazen, S. and Gregory, M. (2001), The European-American employment gap, wage inequality, earnings mobility and skill: a study for France; Germany, the Netherlands, the United Kingdom and the United States, European Low-wage Employment Research Network (LOWER), Final Report, University of Amsterdam. Shields, M. and Price, S. (2002), “Racial harassment, job satisfaction and intentions to quit: evidence from the Bristish nursing profession”, Economica, Vol. 69 No.274, pp. 295-362. Sloane, P. and William, H. (2000), “Job satisfaction, comparison earnings and gender”, Labour, Vol. 14 No.3, pp. 473-501. 17 Sousa-Poza, A. and Sousa-Poza, A. (2000), “Taking another look at the gender/job satisfaction paradox”, Kyklos, Vol.53 No.2, pp. 135-52. Stewart, M. and Swaffield, J. (1999), “Low pay dynamics and transition probabilities”, Economica, Vol.66 No.261, pp. 23-42. Vroom, V. (1964), Work and Motivation, Wiley, New York. Wooldridge, J.M. (2005), “Simple solutions to the initial conditions problem in dynamic, nonlinear panel data models with unobserved hetegeneity”, Journal of Applied Econometrics, Vol.20 No.1, pp. 39-54. 18 Table 1 - Sample statistics and test for the equality of means on reported low satisfaction between low-pay and higher-pay workers Sample size Mean Standard deviation Mean Difference t-statistic Low-pay Higher-pay Low-pay Higher-pay Low-pay Higher-pay Germany 1,944 10,840 4.15 4.42 1.22 1.05 0.28 10.39 Denmark 1,486 18,424 4.96 4.94 1.09 0.97 -0.02 -0.60 The Netherlands 3,580 30,632 4.73 4.75 0.98 0.87 0.02 1.11 Belgium 1,341 17,741 4.30 4.47 1.34 1.15 0.17 5.24 Luxembourg 537 2,177 4.43 4.84 1.26 1.01 0.41 8.05 France 6,026 35,073 4.25 4.45 1.19 1.02 0.20 13.97 UK 6,444 36,145 4.41 4.32 1.33 1.23 -0.08 -4.96 Ireland 3,010 13,715 4.38 4.65 1.33 1.15 0.26 11.04 Italy 3,813 37,039 3.31 4.11 1.44 1.24 0.81 37.65 Greece 3,134 17,243 3.13 4.02 1.21 1.18 0.89 38.91 Spain 5,919 29,048 3.83 4.35 1.41 1.21 0.52 29.16 Portugal 4,232 30,508 3.68 4.07 1.04 0.89 0.40 26.71 Austria 1,922 17,459 4.75 4.97 1.19 0.96 0.22 9.16 Finland 1,492 16,334 4.49 4.59 1.11 0.96 0.10 3.79 19 Annex Random effects ordered probit estimates Germany Denmark Low-wage Higher-wage Low-wage Higher-wage Coeff. APE z-value Coeff. APE z-value Coeff. APE z-valu e Coeff. APE z-value log(hourly wage) 0.021 0.006 0.22 0.393 0.083 6.63 -0.133 -0.039 -1.67 0.263 0.041 4.14 log(weekly hours) -0.061 -0.017 -0.38 0.277 0.059 2.67 -0.038 -0.011 -0.32 0.363 0.057 4.40 Age -0.015 -0.004 -0.70 -0.061 -0.013 -4.99 -0.038 -0.011 -1.77 -0.032 -0.005 -3.31 Age Squared/100 0.019 0.006 0.69 0.085 0.018 5.73 0.066 0.019 2.34 0.051 0.008 4.40 Gender (Male) -0.097 -0.028 -1.05 -0.215 -0.045 -4.57 0.090 0.026 1.06 0.008 0.001 0.22 Secondary Education 0.021 0.006 0.18 0.085 0.018 1.81 -0.043 -0.013 -0.39 0.006 0.001 0.20 Primary or lower Education 0.111 0.032 0.83 0.091 0.019 1.57 0.150 0.044 1.23 0.012 0.002 0.28 Good Health Status 0.423 0.122 5.34 0.427 0.090 12.19 0.291 0.086 2.89 0.364 0.057 11.04 Previously unemployed -0.167 -0.048 -2.10 -0.148 -0.031 -3.01 0.064 0.019 0.77 -0.023 -0.004 -0.64 Job characteristics Overskilled -0.256 -0.073 -3.48 -0.251 -0.053 -7.19 -0.300 -0.088 -3.78 -0.216 -0.034 -7.97 Use of languages 0.390 0.112 3.29 -0.038 -0.008 -0.89 0.070 0.021 0.78 -0.057 -0.009 -1.97 Traning at the moment 0.129 0.037 1.67 0.036 0.008 1.17 0.045 0.013 0.58 0.042 0.007 1.93 Permanent contract 0.183 0.052 1.99 0.121 0.026 2.25 0.200 0.059 2.37 -0.006 -0.001 -0.16 Full-time -0.142 -0.041 -1.06 -0.129 -0.027 -1.60 -0.077 -0.023 -0.68 -0.153 -0.024 -2.73 Public worker -0.018 -0.005 -0.18 0.037 0.008 0.75 0.134 0.039 1.47 0.019 0.003 0.57 µ1 -2.700 -4.03 -1.923 -4.43 -3.355 -6.57 -1.651 -4.55 µ2 -1.971 -2.96 -1.185 -2.74 -2.724 -5.45 -1.057 -2.92 µ3 -1.007 -1.52 -0.140 -0.32 -2.116 -4.29 -0.381 -1.05 µ4 0.000 0.00 0.956 2.21 -1.263 -2.58 0.642 1.77 µ5 1.552 2.35 2.738 6.32 -0.068 -0.14 2.183 6.03 ρ 0.414 11.75 0.412 32.29 0.291 5.94 0.409 41.02 Log-likelihood -2,858 -13,789 -1,919 -21,558 LR Chi-test 159 596 112 612 # of observations 1,944 10,484 1,486 18,218 Note: Estimates also include dummy controls for employer facilities (insurance, housing, training and leisure), employer size, occupations, industry, region and year. 20 Random effects ordered probit estimates The Netherlands Belgium Low-wage Higher-wage Low-wage Higher-wage Coeff. APE z-value Coeff. APE z-value Coeff. APE z-valu e Coeff. APE z-value log(hourly wage) 0.059 0.014 1.63 0.143 0.043 5.15 0.145 0.035 1.29 0.357 0.046 6.92 log(weekly hours) -0.063 -0.015 -0.71 0.035 0.010 0.70 0.580 0.140 3.76 0.539 0.070 7.48 Age -0.065 -0.016 -4.60 -0.062 -0.018 -8.43 -0.112 -0.027 -3.88 -0.070 -0.009 -6.83 Age Squared/100 0.088 0.021 4.52 0.077 0.023 8.57 0.151 0.037 3.99 0.090 0.012 7.05 Gender (Male) -0.026 -0.006 -0.49 0.071 0.021 3.52 -0.067 -0.016 -0.71 -0.145 -0.019 -5.40 Secondary Education 0.099 0.024 1.04 0.053 0.016 2.12 0.174 0.042 1.68 -0.003 0.000 -0.08 Primary or lower Education 0.043 0.010 0.46 0.042 0.013 1.53 0.210 0.051 1.81 0.072 0.009 1.92 Good Health Status 0.311 0.074 5.32 0.347 0.104 17.24 0.322 0.078 3.26 0.404 0.052 14.59 Previously unemployed -0.098 -0.023 -1.72 -0.026 -0.008 -1.14 0.086 0.021 0.94 -0.030 -0.004 -0.91 Job characteristics Overskilled -0.157 -0.038 -3.62 -0.103 -0.031 -6.72 -0.427 -0.103 -5.14 -0.203 -0.026 -8.99 Use of languages -0.032 -0.008 -0.58 0.001 0.000 0.04 -0.076 -0.018 -0.76 0.028 0.004 1.06 Traning at the moment -0.090 -0.022 -1.58 -0.053 -0.016 -2.33 0.042 0.010 0.45 0.009 0.001 0.39 Permanent contract 0.145 0.035 3.01 0.007 0.002 0.25 0.050 0.012 0.52 -0.011 -0.001 -0.33 Full-time -0.029 -0.007 -0.42 -0.124 -0.037 -3.61 -0.404 -0.097 -2.92 -0.256 -0.033 -5.06 Public worker -0.014 -0.003 -0.22 0.022 0.006 1.11 0.050 0.012 0.48 0.108 0.014 3.90 µ1 -3.917 -10.78 -3.632 -16.33 -1.436 -1.90 -0.874 -2.60 µ2 -3.242 -9.12 -3.026 -13.71 -0.881 -1.17 -0.300 -0.89 µ3 -2.511 -7.13 -2.284 -10.38 -0.007 -0.01 0.496 1.48 µ4 -1.493 -4.26 -1.133 -5.15 0.810 1.08 1.493 4.45 µ5 0.008 0.02 0.597 2.72 2.028 2.71 2.944 8.76 ρ 0.258 10.58 0.286 36.11 0.383 9.59 0.370 36.36 Log-likelihood -4,580 -35,238 -2,022 -23,516 LR Chi-test 120 695 142 802 # of observations 3,580 30,627 1,341 17,425 Note: Estimates also include dummy controls for employer facilities (insurance, housing, training and leisure), employer size, occupations, industry, region and year. 21 Random effects ordered probit estimates Luxembourg France Low-wage Higher-wage Low-wage Higher-wage Coeff. APE z-value Coeff. APE z-value Coeff. APE z-valu e Coeff. APE z-value log(hourly wage) -0.198 -0.053 -0.73 0.651 0.149 4.09 -0.146 -0.047 -3.25 0.414 0.027 14.05 log(weekly hours) 0.627 0.168 1.54 0.116 0.026 0.33 0.010 0.003 0.13 0.490 0.033 10.01 Age -0.052 -0.014 -1.01 -0.087 -0.020 -2.44 -0.017 -0.005 -1.36 -0.033 -0.002 -4.28 Age Squared/100 0.062 0.017 0.90 0.107 0.025 2.41 0.018 0.006 1.10 0.035 0.002 3.62 Gender (Male) -0.599 -0.160 -2.99 -0.292 -0.067 -2.53 -0.049 -0.016 -1.14 -0.091 -0.006 -4.05 Secondary Education -0.534 -0.143 -1.40 -0.144 -0.033 -1.14 0.027 0.009 0.51 0.002 0.000 0.07 Primary or lower Education -0.407 -0.109 -1.10 0.072 0.016 0.52 0.058 0.019 1.07 0.082 0.005 2.77 Good Health Status 0.373 0.100 2.50 0.272 0.062 3.09 0.446 0.143 12.36 0.439 0.029 26.37 Previously unemployed -0.447 -0.120 -1.93 -0.361 -0.082 -1.85 -0.107 -0.034 -2.43 -0.050 -0.003 -1.59 Job characteristics Overskilled -0.617 -0.165 -3.83 -0.301 -0.069 -3.46 -0.275 -0.088 -8.09 -0.181 -0.012 -11.28 Use of languages 0.171 0.046 0.76 -0.045 -0.010 -0.28 0.133 0.043 2.25 0.047 0.003 2.07 Traning at the moment -0.053 -0.014 -0.26 0.031 0.007 0.38 0.190 0.061 3.99 0.062 0.004 2.94 Permanent contract -0.212 -0.057 -0.96 0.031 0.007 0.16 0.020 0.006 0.46 0.076 0.005 2.29 Full-time -0.380 -0.102 -1.20 -0.012 -0.003 -0.05 0.060 0.019 0.92 -0.148 -0.010 -3.83 Public worker 0.288 0.077 0.94 0.370 0.084 2.83 0.139 0.045 2.72 0.240 0.016 9.81 µ1 -2.286 -1.38 -2.629 -1.96 -2.019 -5.81 -0.442 -1.95 µ2 -1.791 -1.08 -1.955 -1.46 -1.584 -4.56 0.067 0.30 µ3 -1.079 -0.65 -1.205 -0.90 -0.919 -2.65 0.818 3.61 µ4 -0.103 -0.06 -0.046 -0.03 0.082 0.24 2.067 9.12 µ5 1.701 1.04 1.772 1.32 1.740 5.01 4.025 17.71 ρ 0.458 7.42 0.503 18.69 0.280 14.91 0.364 48.06 Log-likelihood -730 -2,601 -8,424 -37,928 LR Chi-test 76 156 481 1807 # of observations 537 2,172 6,026 31,750 Note: Estimates also include dummy controls for employer facilities (insurance, housing, training and leisure), employer size, occupations, industry, region and year. 22 Random effects ordered probit estimates UK Ireland Low-wage Higher-wage Low-wage Higher-wage Coeff. APE z-value Coeff. APE z-value Coeff. APE z-valu e Coeff. APE z-value log(hourly wage) -0.109 -0.011 -1.83 0.239 0.063 8.32 0.272 0.052 2.78 0.261 0.076 5.34 log(weekly hours) -0.052 -0.005 -0.70 -0.093 -0.025 -2.17 0.347 0.066 2.99 0.030 0.009 0.40 Age -0.047 -0.005 -4.30 -0.039 -0.010 -6.44 -0.032 -0.006 -1.87 -0.042 -0.012 -4.52 Age Squared/100 0.070 0.007 4.79 0.056 0.015 7.32 0.048 0.009 2.06 0.062 0.018 5.27 Gender (Male) -0.251 -0.026 -4.83 -0.281 -0.074 -12.18 -0.153 -0.029 -2.20 -0.184 -0.053 -4.79 Secondary Education 0.009 0.001 0.15 0.059 0.016 2.61 0.144 0.027 1.80 0.047 0.014 1.23 Primary or lower Education 0.102 0.010 2.00 0.193 0.051 8.05 0.271 0.052 3.01 0.076 0.022 1.55 Good Health Status 0.247 0.025 6.16 0.231 0.061 13.42 0.239 0.046 2.71 0.402 0.117 9.10 Previously unemployed -0.125 -0.013 -2.46 -0.046 -0.012 -1.72 -0.257 -0.049 -3.80 -0.202 -0.059 -5.00 Job characteristics Overskilled -0.351 -0.036 -4.31 -0.233 -0.061 -7.54 -0.325 -0.062 -6.36 -0.284 -0.082 -11.52 Use of languages -0.566 -0.058 -3.30 -0.001 0.000 -0.02 0.160 0.030 1.27 0.106 0.031 2.31 Traning at the moment -0.017 -0.002 -0.41 -0.008 -0.002 -0.56 0.007 0.001 0.10 -0.083 -0.024 -3.02 Permanent contract 0.283 0.029 4.90 0.074 0.020 2.34 0.183 0.035 3.03 0.108 0.031 2.83 Full-time -0.239 -0.024 -3.67 -0.156 -0.041 -4.23 -0.139 -0.026 -1.42 0.089 0.026 1.57 Public worker 0.271 0.028 4.16 -0.048 -0.013 -2.04 0.060 0.011 0.60 0.120 0.035 3.20 µ1 -3.558 -11.08 -2.956 -16.09 -1.014 -2.03 -2.417 -7.82 µ2 -2.950 -9.23 -2.271 -12.38 -0.498 -1.00 -1.806 -5.86 µ3 -2.352 -7.37 -1.701 -9.28 0.284 0.57 -1.026 -3.33 µ4 -1.358 -4.27 -0.628 -3.43 1.226 2.45 -0.007 -0.02 µ5 0.297 0.93 1.248 6.81 2.243 4.47 1.260 4.08 ρ 0.534 37.11 0.500 84.03 0.383 14.18 0.394 34.81 Log-likelihood -10,711 -57,298 -4,592 -18,644 LR Chi-test 529 1233 174 548 # of observations 7,526 42,783 3,010 13,700 Note: Estimates also include dummy controls for employer facilities (insurance, housing, training and leisure), employer size, occupations, industry, region and year. 23 Random effects ordered probit estimates Italy Greece Low-wage Higher-wage Low-wage Higher-wage Coeff. APE z-value Coeff. APE z-value Coeff. APE z-valu e Coeff. APE z-value log(hourly wage) 0.473 0.051 5.10 0.894 0.081 23.31 0.492 0.008 5.02 0.831 0.010 21.99 log(weekly hours) 0.668 0.073 5.56 0.786 0.071 15.33 0.761 0.012 6.99 0.837 0.010 15.05 Age -0.037 -0.004 -2.15 -0.046 -0.004 -6.74 -0.035 -0.001 -2.48 -0.017 0.000 -2.13 Age Squared/100 0.038 0.004 1.67 0.052 0.005 6.19 0.038 0.001 2.03 0.015 0.000 1.61 Gender (Male) -0.077 -0.008 -1.18 -0.112 -0.010 -4.31 -0.036 -0.001 -0.68 -0.139 -0.002 -6.02 Secondary Education -0.161 -0.018 -1.23 0.011 0.001 0.32 -0.121 -0.002 -1.61 -0.040 0.000 -1.39 Primary or lower Education -0.217 -0.024 -1.71 -0.039 -0.004 -1.05 -0.204 -0.003 -2.40 -0.154 -0.002 -4.22 Good Health Status 0.295 0.032 5.17 0.266 0.024 16.40 0.089 0.001 0.98 0.100 0.001 2.62 Previously unemployed -0.059 -0.006 -0.97 -0.114 -0.010 -4.00 -0.096 -0.002 -1.85 -0.139 -0.002 -5.25 Job characteristics Overskilled -0.059 -0.006 -1.14 -0.076 -0.007 -4.89 -0.287 -0.005 -5.89 -0.180 -0.002 -9.23 Use of languages 0.112 0.012 0.86 0.049 0.004 1.67 0.181 0.003 2.09 0.091 0.001 3.11 Traning at the moment 0.117 0.013 1.27 0.088 0.008 4.15 0.191 0.003 2.28 0.210 0.003 6.35 Permanent contract 0.187 0.020 3.39 0.201 0.018 7.69 0.469 0.007 9.44 0.413 0.005 14.82 Full-time 0.024 0.003 0.22 -0.057 -0.005 -1.56 -0.016 0.000 -0.14 0.156 0.002 3.08 Public worker 0.475 0.052 4.72 0.156 0.014 6.74 0.410 0.007 4.75 0.372 0.005 14.20 µ1 0.237 0.45 1.087 4.56 1.446 2.53 1.745 6.93 µ2 1.258 2.40 1.904 7.98 2.451 4.28 2.587 10.28 µ3 2.288 4.36 2.883 12.07 3.600 6.27 3.710 14.70 µ4 3.200 6.08 4.002 16.74 4.636 8.05 4.805 18.97 µ5 4.358 8.23 5.279 22.03 5.590 9.65 5.938 23.36 ρ 0.484 21.56 0.411 62.17 0.264 10.49 0.203 22.46 Log-likelihood -5,896 -51,525 -4,548 -23,947 LR Chi-test 592 2717 538 2784 # of observations 3,813 36,238 3,134 17,210 Note: Estimates also include dummy controls for employer facilities (insurance, housing, training and leisure), employer size, occupations, industry, region and year. 24 Random effects ordered probit estimates Spain Portugal Low-wage Higher-wage Low-wage Higher-wage Coeff. APE z-value Coeff. APE z-value Coeff. APE z-valu e Coeff. APE z-value log(hourly wage) 0.296 0.100 4.72 0.422 0.099 14.20 0.077 0.014 1.23 0.684 0.084 19.93 log(weekly hours) 0.119 0.040 1.56 0.311 0.073 5.85 0.571 0.101 4.89 0.670 0.082 10.15 Age -0.029 -0.010 -3.17 -0.062 -0.015 -10.91 -0.026 -0.005 -1.98 -0.015 -0.002 -2.26 Age Squared/100 0.030 0.010 2.57 0.078 0.018 11.12 0.031 0.005 1.83 0.012 0.001 1.49 Gender (Male) -0.128 -0.043 -3.44 -0.126 -0.030 -6.10 -0.039 -0.007 -0.60 -0.009 -0.001 -0.31 Secondary Education 0.078 0.026 1.35 0.011 0.003 0.44 -0.694 -0.122 -2.49 -0.007 -0.001 -0.16 Primary or lower Education 0.255 0.086 4.52 0.105 0.025 4.03 -0.736 -0.130 -2.73 -0.010 -0.001 -0.21 Good Health Status 0.207 0.070 5.07 0.253 0.059 12.76 0.008 0.001 0.16 0.096 0.012 5.05 Previously unemployed -0.206 -0.070 -6.11 -0.089 -0.021 -4.46 -0.231 -0.041 -3.49 -0.164 -0.020 -4.81 Job characteristics Overskilled -0.190 -0.064 -5.88 -0.140 -0.033 -9.32 -0.172 -0.030 -3.56 -0.109 -0.013 -6.28 Use of languages -0.100 -0.034 -1.18 0.021 0.005 0.72 0.487 0.086 3.15 -0.024 -0.003 -0.70 Traning at the moment 0.024 0.008 0.54 0.010 0.002 0.56 -0.024 -0.004 -0.27 0.049 0.006 1.66 Permanent contract 0.135 0.046 3.60 0.110 0.026 5.60 0.313 0.055 5.82 0.245 0.030 10.67 Full-time -0.017 -0.006 -0.24 -0.008 -0.002 -0.17 0.174 0.031 1.41 0.129 0.016 2.12 Public worker 0.227 0.077 3.29 0.130 0.031 5.74 -0.006 -0.001 -0.05 0.223 0.027 8.14 µ1 -1.505 -5.04 -1.351 -6.46 -1.361 -2.61 -0.121 -0.47 µ2 -0.792 -2.66 -0.683 -3.27 -0.298 -0.57 0.683 2.68 µ3 -0.086 -0.29 0.083 0.40 0.902 1.74 1.789 7.02 µ4 0.604 2.03 0.925 4.43 2.853 5.48 3.916 15.32 µ5 1.665 5.57 2.249 10.75 3.812 7.29 5.414 21.11 ρ 0.177 10.15 0.242 32.52 0.467 22.96 0.435 58.58 Log-likelihood -9,674 -41,872 -5,435 -33,184 LR Chi-test 496 1872 320 1962 # of observations 5,919 29,000 4,232 30,318 Note: Estimates also include dummy controls for employer facilities (insurance, housing, training and leisure), employer size, occupations, industry, region and year.