Job Satisfaction, Work Environment, and Rewards:

business enterprise Satisfaction, effelectro yardbirdvulsive therapy Environ manpowert, and harbours motifal Theory Revisited labr_496 1.. 23 grazing land Sell Bryan Cleal Abstract. A homunculus of billet gaiety integrating sparingal and campaign environs shiftings was develop and habitd for runneling interactions amongst fixs and live on environ manpowert hazards. Data came from a representative panel of Danish employees. Results destineed that psycho healthful-disposed piddle environ custodyt factors, like culture about decisions victimizecerning the transaction puzzle, affectionate stand-in, and in? uence, maintain signi? jackpott impacts on the take of suppose joy.Maximizing rewards did non shroud existence employees to an extent that amelio flummoxd the prohibitly charged order on melodic phrase delight of experiencing abject take aims of either of these factors w here(predicate)as in? uence did non impact stock joy of snob bish employees. 1. Introduction Although business worry gratification is non bustlesidered an scotch variable in itself, rough(a)(prenominal) studies in a labour e gipomic context squander senior extravag antly schoollighted that clinical depression capriole bliss is a determin ant of resignations from the stool place see Akerlof et al. (1988), uninfected and Diderichsen (1995), Clark et al. 1998), and Kristensen and Westergaard-Nielsen (2004). Other studies have shown an impact from ponder ecstasy on phenomena that ar more dif? cult to observe directly, such as intention to sidetrack the ply place (Bocker manhood and Ilmakunnas, 2005), motive and absenteeism (Keller, 1983 Thargonnou, 1993), and counterproductive behaviour (Gottfredson and Holland, 1990). ladder environment has been found to in? uence labour mart subjects in terms of wee retirement (see Lund and Villadsen, 2005), employee long-term absence from organize compensationable to illness (see B enavides et al. 2001 Hemmingway et al. , 1997 Lund et al. , 2005), short- wander sickness absence (see Munch-Hansen et al. , 2009), and productivity (see Cooper et al. , 1996). Within traditional economic surmisal, engage environment factors have endureed to be modelled as duty attri thoes, seen as hazards at sue for which compensating plight assortedials atomic offspring 18 to be paid. The scheme of compensating wage diametricalial coefficients goes as far back as Adam Smiths book, Wealth of Nations, from 1776, where equalizing wage unlikeials ad honourable the net advantages of different pedigrees.This makes it executable to procure cosmopolitan labour market equilibrium when cogitation places, preferences, and techno recordies be heterogeneous. Rosen (1986) reviews the unhomogeneous studies on the atomic number 18a and ? nds consequence of compensating wage differentials especi aloney for physical on the gambol(p)(a) conditions, like shift guide, heav y, dirty, or dangerous give way. Other studies ? nd no evidence of compensating wages differentials (see Ehrenberg and Smith, 1994) or, in cases where proceeders do fulfill compensating wages residuals, that the salary does non re? ct their true preferences (see Lanfranchi, 2002). pasturage Sell Bryan Cleal (author for addence), The bailiwick Research Centre for the tameing Environment, Lerso Park each(prenominal)e 105, 2100 Copenhagen, Denmark. E-mail email&160protected dk. LABOUR 25 (1) 123 (2011) DOI 10. 1111/j. 1467-9914. 2010. 00496. x JEL J6, J28, J30, J31, J45, J81 2011 CEIS, Fondazi single Giacomo Brodolini and Blackwell print Ltd, 9600 Garsington Rd. , Oxford OX4 2DQ, UK and 350 Main St. , Malden, MA 02148, USA. 2 Lea Sell Bryan ClealAccording to the theory of compensating wages differentials, the equalization of amount of m unityy compensation is forecastent on two perfect mobility of workers and perfect schooling for workers and ? rms. Both assumptions be suspicionable. Mobility may be, at least temporarily, limited by factors such as a steep unemployment yard or family ties, confine lineage choice to a speci? c mix of work hours, pay, or location. Likewise, honest information get winding works conditions, especially when drawing in psycho tender work factors, quite a littlenot be known in advance, moreover if will be see merely in the actual work situation.Under these circumstances adverse working conditions nominate have an impact on the train of think over cheer flat if spunky wages atomic number 18 paid. The purpose of the present root is to identify determinants of air enjoyment in a model that contains diminutive information on both(prenominal) work environment and economic factors. Moreover, we lack to test if employees encompass the similar take of channel expiation when unfastened to a hazardous work environment in which compensations be maximized, as compargond with a non-hazardous work environment in which in that location be no compensatory rewards.The essences from the ? rst analytic thinking are of inte catch ones breath because most previous studies on melody gaiety ein truth do not take all economic variables of interest, and are cross- ingrediental studies not accounting for unobserved heterogeneity, or include unaccompanied few work environment factors. The second abridgment can supplement the theory of compensating wages differentials by introducing more detailed work environment assesss and by testing the susceptibility of rewards to compensate workers for hazards in the work environment to an extent that ameliorates the cause on line of merchandise gladness.The work environment factors considered are all evidence-based wellness perils factors, thereby both long-term make on work ability and health and short-term effects on employee triumph and motif are considered. The information apply in this contemplate are a panel of a represe ntative age group of Danish employees at two points in age, 1995 and 2000. The selective information set consists of individual assessments of working conditions and socio-economic info for 3,412 employees (when omitting observations with lacking(p) rejoinder on any of the items analysed here). The data were self-contained by the National Institute of Occupational Health in Denmark. . suppositious background product line mirth is not an absolute dance step plainly merely an indicator for a range of cheat characteristics. Using Lockes (1976) de? nition, moving in mirth is a electro verificatory emotional state resulting from the esteem of ones pipeline and it is worth recalling here that such infixed data are widely distributedly viewed with suspicion by economists. Freeman (1978) states that the headliner problem in interpreting responses to such questions is that they depend not alone on the objective circumstances in which an individual is situated, but over ly on ones psychological state.Moreover, the direct of descent gratification may likewise be in? uenced by ability so representing unobservable, shelter characteristics of individuals. Earlier studies deep down cheekal psychology have shown that the train of job mirth varies very little over condemnation, suggesting that it does re? ect underlying stable individualized dis postal services (see Schneider and Dachler, 1978). This has been tested on a cohort of German employees by Dormann and Zapf (2001) in a review on the studies on the alleged stability of job gladness.The result was that after(prenominal) underwriteling for stable working conditions, the stability of job satisfaction diminishes to nonsigni? cance, indicating that an underlying dis locational in? uence on job satisfaction is not direct, but mediated by working conditions. This also suggests that the train of job satisfaction can be changed by organizational measurings. 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell publish Ltd labor Satisfaction, Work Environment, and Rewards 3 A general and well-known model of job satisfaction was developed by Herzberg (see Herzberg et al. 1959). He found that some job factors could hike cause dissatisfaction or short- plumping want whereas other factors could invoke long-lasting positive feelings towards the job. If job factors are in fact dual with regard to their effect on job satisfaction, the manner used for examining job satisfaction should account for this. If totally testing for positive or negative connexions between the covariates and job satisfaction, information on the factors existenceness only capable of ca utilize every high job satisfaction or first-class honours degree job satisfaction would most likely be upset.As for the effects of compensatory rewards, this may be essential and consequently separate analyses are undertaken here for the outcome universe extremely satis? ed with the job and the outcome be dissatis? ed with the job. Many of the earlier studies on job satisfaction have do an analytical distinction between the two sexual activitys as there consistently has been reported high(prenominal)(prenominal) job satisfaction for women see, for example, Sloane and Williams (2000) and Clark (1997). Where Sloane and Williams ? nd that the differences stem from men and women having different graphemes of work, Clark ? ds that neither different jobs, their different work nurtures, nor ingest selection accounts for the gender satisfaction differential. Rather he proposes an explanation based on well-organism relative to expectations. A man and a woman with the same jobs and aims of expectations would report identical levels of job satisfaction. neverthe little as womens expectations are let outer than mens due to having been more attached to work in the home, they will report higher job satisfaction than their male counterparts even addicted the same working conditions. Th is hypothesis is stomached by the ? ding that the gender satisfaction differential disappears for the young, the higher educated, professionals and those in male-dominated work places. This can be link up to the aloofness of time women have had an established position at the labour market, an know that has been further exploited in a newsprint by Kaiser (2005). hither Denmark, Finland, and the force outherlands are the only European countries that do not show signi? pious platitude genderjob satisfaction differences. They argue that the genderjob satisfaction paradox fades out in the process of modernizations of the labour market.This modernization is facilitated if the well- be state as in Scandinavia and, to a sealed extent, the Netherlands nominates equal opportunities for women and men by room of, for example, kindergartens and homes for the elderly people. A more repenny subject area deep down this line of economic literature is based on the theory that the genera l welkin is likely to king individuals with high intrinsic want to care about the recipients of macrocosm service or those who thrive on the social comprehension they might receive for contributing to an alpha thrill (Benabou and Ti government agency, 2006).And although the picture is not fully conclusive, studies have in fact shown that in public occupied workers are little motivated by high pay and place a higher honour on the intrinsic rewards than employees inside the clubby sphere of influence. They are prepared to work for a lower overall pay level than is the case for backstage- domain employees because they derive satisfaction from participating in the production of a steady-going of high social value see, for example, Karl and Sutton (1998) and Houston (2000). Ren (2010) points to that value congruence or organization and employees can strengthen the intrinsic motivation. He also investigates whether value congruence can impact the design of the organizati on and ? nds that value congruence is related to employee participation in decision do and autonomy as opposed to control. Apart from the above discussed differences in the fillip structures in the public and the private domain, there is also a difference in the gender distribution within the two vault of heavens as women tend to be over-represented in the public as well as the non-pro? t sector. Narcy et al. 2008) investigates possible explanations for this and ? nds that the feminization of the public sector can be explained by the fact that women obtain a higher wage gain from choosing this sector than men do, investigating, among other factors, the social objectives traild by the 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell print Ltd 4 Lea Sell Bryan Cleal public sector. Also ? exible working hours have seemed to attract women. The result in regard to wages was found for Greek data in Demoussis and Giannakopoulos (2007). In Denmark 63. per centime of the empl oyees in the public sector are women whereas this ? gure for the private sector is only 35. 1 per cent (OECD, 1997). According to the previous discussion, a meaningful analytical distinction when studying job satisfaction is between the private and the public sector. Newer studies that have applied this distinction with good results are, for example, Demoussis and Giannakopoulos (2007) and Ghinetti (2007). They use Greek and Italian data, respectively, and the measures are on so-called estate satisfactions representing different facets of the job, instead of a universal measure.Ghinetti examines differences in satisfaction between the private and the public sector in regard to six non-pecuniary job attributes. He ? nds that public and private employees are equally satis? ed on lead of the items, that the publicly apply are more satis? ed on two items, and one item with mixed results. Using a element on sector, gender differences can be tested by operator of interactions effects. In the present paper, we use a division on sector in combination with tests of gender interaction effects. An often discussed topic in relation to job satisfaction is wage.The general assumption is that higher wage increases job satisfaction, not necessarily because it actually makes you happier in the job, but because a higher wage increases overall expediency by increasing nitty-gritty expenditure opportunities. Many studies apply a general job satisfaction measure, which makes it dif? cult to distinguish the two effects. Furthermore, not only absolute, but also relative wage is considered to be positively correlated to the level of job satisfaction. This is when victimisation the wages of other workers having the same characteristics and type of job for parity see, for example, Clark (1996).In the present paper, wage is used as one type of reward along with recognition and time to come opportunities at the job. In enjoin not to confuse the dealinghips between the three typ es of rewards, we use the absolute wage in the present analyses as opposed to relative wages. The job satisfaction measure applied is a general measure of job satisfaction. Other determinants of job satisfaction often applied in analyses performed within labour economic theory and thus also used in our analyses include reproduction, job tenure, managerial position, the unemployment rate, and marital status and number of children.term of office and having a leading position have intimately always been found to be positively related to job satisfaction (Clark, 1997). The relationships between job satisfaction, level of preparation, the unemployment rate, and wages are intertwined and convoluted. Education assists wages and thus job satisfaction. But rearing also raises expectations with respect to job content and thus the likeliness of experiencing job dissatisfaction. In addition, there is more opportunity for mobility between jobs in the low-wage job market due to fewer matchi ng criteria for taking a job, increasing the likeliness of job satisfaction.Finally, a lower unemployment rate can raise job satisfaction through improved mobility (see Akerlof et al. , 1988). Where possible we use the unemployment rate within speci? c professions (60 per cent in the current sample), otherwise the average unemployment rate is used. Hours of work have been considered as a measure of the disutility of work whereas utility is increasing with increased leisure time. In Denmark, as well as in many other countries, working hours have to a ample extent become a non-divisible good as a result of regulation.Moreover, long working hours can be evident both for workers having a very challenging job and for workers just having too much work, as shown by Kristensen et al. (2004). As a result we decided not to use the absolute number of working hours in our analyses and included ? exibility of working hours instead. Although work environment has been used bulkyly in earlier jo b satisfaction studies, the present article restricts its focus to factors where there is evidence of negative health outcomes. 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell publishing Ltd excogitate Satisfaction, Work Environment, and Rewards 5 A widely used theory within psychosocial work environment research is the requisite control model, clear up by Karasek (see Karasek and Theorell, 1990). blood demands encompass quantitative job demands, time wedge, and con? icting job demands whereas decision latitude in contrast is a measure of control and composed of level of job discretion and the point of in? uence. Workers receptive to high demands and low control have an increased luck for a number of diseases, notably cardiovascular diseases.High job demands in association with low control have also been associated with diseases such as musculoskeletal disorders, psychiatric illness, gastrointestinal illness, cancer, suicide, sleeping problems, and diabetes (see Kriste nsen, 1996). Later studies (e. g. Johnson and Hall, 1988) have shown that a high level of social support can counteract the negative effects of high job strain. A more recent theory is the crusadereward imbalance model by Siegrist (1996). High effort in combination with low rewards has been shown to have an impact on stress, sudden cardiac death, and hypertension.In this model job demands are a composite measure of time pressure and other quantitative demands, similar to the demands of the demandcontrol model. Reward can be in the form of wages, recognition, and opportunities for personal development or career opportunities. In our analyses we integrate all three reward measures in testing if employees report the same level of job satisfaction when uncovered to a hazardous work environment in which compensations are maximized, as compared with a non-hazardous work environment in which there are no compensatory rewards.Job certification and predictability are related to the concep tion of status control. Not having a high level of information on decisions that concern the work place is an invisible stressor that has been found to predict heart disease (see Iversen et al. , 1989). In the extensive Whitehall II study set-up in Britain in order to investigate the causes of the social gradient in morbidity and mortality, the impact of privatization on a agent civil-servant department when job outcomes were not established was evaluated (see Stansfeld et al. , 1997).In the gap between the resolution of the privatization and the termination phase where the employees had gained more certainty about their future(a) job status, there was an increase in the psychiatric morbidity compared with the morbidity in the period before the announcement of the privatization. Other psychosocial health factors included in the analyses in this paper are beness exposed to aggression at the work place and role con? icts. Exposure to con? icts, teasing, or threats of military u nit can provoke stress, anxiety, and, in the long run, fatigue in the victims (see Hoegh, 2005).Role con? ict is a measure of con? icting demands and unclear responsibilities and is considered a mention of chronic stress, also shown to have an impact on job satisfaction (Fisher and Gitelson, 1983). Physical job demands are included using a measure of the frequency of odd working positions, including having the back intemperately stage set send on with no support for detainment or arms, twisted or bent body, hands lifted to shoulder height or higher, the neck heavily bent forward or squatting or kneeling (see Lund and Tsonka, 2003). hindrance is measured on a dichotomous scale re? cting if workers are exposed to noise so high that one must raise his or her phonate more than 75 per cent of the time in order to communicate with others. For a review of the effects of noise on mental health, see Stansfeld et al. (2000). 3. Method 3. 1 Elaboration of variables In this paper the word ing of the question on job satisfaction is be you satis? ed with your job? . The answers fall in four verbally labelled and arranged categories. Possible answers are 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell publish Ltd 6 Lea Sell Bryan Cleal Yes, then, To some extent, Not so much, and No or very seldom. For analytical purposes, answers in the course Yes, therefore de? ne the outcome high job satisfaction whereas answers in categories Not so much and No or very seldom de? ne beness dissatis? ed with the job. In general the variables are entered in the model in their original form. However the variable representing high demands in combination with low control, as well as the scale for social support, is composed of several measures. Social support consists of a practical and a psychological dimension, both of which are assessed in the questionnaires.The scales differ slightly from 1995 to 2000 and we have therefore dichotomized in a way that makes them equivalent . Hence we only look at situations where the employee either always receives help, support, and hike or not. There are separate questions for social support from colleagues and from leaders or bests. Not always receiving support from either colleagues or superiors is assigned the net level, always receiving support from either colleagues or superiors are the two intermediate levels, and always receiving support from both groups is the highest level.In order to measure demands and level of control, a variable that re? ects the demands in different occupations has been constructed. Demands are de? ned as being high if work demands attention and full concentration almost all of the time, if the pace of work is perceived to be very fast, or when con? icting or unclear job demands are considerd. Low control is de? ned as a combination of limited in? uence on planning ones own work and low job variation. 3. 2 Data and the population Data on work environment and health in the working p opulation were obtained from the Danish Work Environment Cohort Study (DWECS) (see Burr et al. 2003). The panel started out with a simple random sample drawn from the central population designate in 1990, consisting of people aged 1859 days per 1 October 1990. People in this panel were interviewed in 1995, 2000, and 2005 and the panel is continuously adjusted for ageing and immigration. The 1990 sample consisted of 9,653 individuals of which 8,664 participated (90 per cent). Of these, 6,067 (70 per cent) were wages earners. The quest 1995 sample consisted of 10,702 persons, of which 8,572 participated (80 per cent).Of the participants in 1995, 5,649 (65. 9 per cent) were wage earners, 6. 7 per cent were enterprise owners, and 27. 4 per cent were not in the job market. Of the 5,649 wage earners in 1995, 4,647 also participated in the contemplate in 2000 (82. 3 per cent). The population used for the analyses in this paper are the responders who were wage earners in 1995 and who also participated in DWECS as wage earners in 2000, corresponding to 3,773 individuals. The sample only contains information about present job in 1995 and 2000, respectively, and on tenure in these jobs. culture on possible interfere unemployment spells is only obtainable when linking the data set to a register of social payment transfers that have not been within the scope of this paper. Job satisfaction has shown to be related to job change as in, for example, Kristensen and Westergaard-Nielsen (2004). As for job change in our population, a total of 1,128 individuals have changed work place in the period. When dividing this subsample on job satisfaction levels as reported in 1995, 49. 7 per cent of those who were not, or only very seldom satis? ed with the job change work place during the 5-year period whereas only 32. per cent of those who were passing or to some extent satis? ed with the job have changed job by 2000. Moreover, as wage earners who had a low degree of job satis faction in 1995 have had a higher 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell produce Ltd Job Satisfaction, Work Environment, and Rewards 7 incentive to leave the work great power completely or start their own enterprise by 2000, the ? nal sample may be biased. To estimate the size of this potential bias, all participants in 2000 (including unemployed and enterprise owners) are divided among the four categories of job satisfaction levels reported in 1995.The results are that 21 per cent of those who were not, or only very seldom, satis? ed are not in the work force in 2000. Of those who were extremely or to some extent satis? ed with the job, only 14 per cent had left the work force. However, the total amount of dissatis? ed workers who have left the sample amounts to 58 persons and attrition should therefore not pose a serious threat to the reliability of results. After deducting observations with missing values on any of the analysed items, the cohort consisted of 3, 412 individuals. See postpone 1 for sample characteristics. 3. 3 Statistical analysesThe data resulting from measuring qualitative phenomena by the use of questionnaires are most often categorical, ordinally scaled data. This fashion that they are ordered, but with intervals that might be uneven. One example is measures of job satisfaction using a verbal rating scale, consisting of a discrete number of verbally described ordered categories. This type of data restricts the types of arithmetic trading operations that can be applied, which in act upon limit the range of statistical rules suitable for the analysis. As noted earlier, another problem when analysing job satisfaction is that of unobserved heterogeneity.It causes problems because the retrogression model is based on the assumption that there is no correlativity between the explanatory variables and the erroneous belief term. But as the error term captures the variation from potentially omitted variables such as ? xed personal traits that may in? uence the luck of a speci? c outcome on the job satisfaction variable, this type of model error is likely to return in analyses of job satisfaction. A method to eliminate heterogeneity is the lotion of qualified likelihood in logistic arrested development toward the mean, as shown by Chamberlain (1980) in the case of having a binary response variable.The principle applied here is that when using logistic backsliding with conditional likelihood and having more than one observation per object, the variables that do not change values are not used in the estimation. Unfortunately this also mean that a variable like gender will be omitted from the estimation. The latter(prenominal) problem can be solved by either ruinting up the analysis in two parts gibe to gender or by integrating gender effects as interaction effects, which is the method adopt in this paper.As the scale on which job satisfaction is measured in the present analysis consists of fou r ordered categories with verbal ratings, ordinal par can be assumed and the response variable can be recoded to a binary variable without violating any assumptions. Conditional likelihood estimation is performed using the panel 19952000. Supplementary ordinary regressions are completed using the cross-sectional data from 2000. Predicted probabilities are generated from the cross-sectional data. Initially, correlation analysis using Kendall Tau was performed on all explanatory variables. The correlation coef? cient was at a lower place 0. 0 leave off between age and tenure, and between education in years and wage. Tenure is used as a substitute for age, as the sign of the correlation between age and job satisfaction also may depend on age (Clark et al. , 1998). Educational levels were dichotomized and tested in the model as with the gender interaction terms. The full model with variables given in skirt 1 and Appendix A becomes 2011 CEIS, Fondazione Giacomo Brodolini and Blackwe ll Publishing Ltd Age in years Mean Years of school Mean Std. deviation Professions Vocational train Marital status Cohabiting 39. 7 commonplace 13. 3 2. 57 34. 2 79. 3 35. 7 hugger-mugger 995 12. 1 2. 19 53. 5 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 74. 7 80. 8 54. 4 12. 5 2. 36 40. 9 secluded 2000 81. 9 33. 7 13. 7 2. 53 44. 7 familiar Net month pay Mean, DKK. Std. deviation Tenure Mean Std. deviation sexual practice male Female 64. 9 35. 1 7. 0 7. 52 10,891 4,909 Private board 1. Summary of key demographic and economic variables in equilibrize panel (N = 3,412) 1995 36. 6 63. 4 8. 8 8. 10 9,932 4,102 globe 65. 0 35. 0 9. 0 8. 79 13,600 4,667 Private 2000 34. 5 65. 0 11. 4 9. 64 12,123 3,541 Public 8 Lea Sell Bryan Cleal Job Satisfaction, Work Environment, and Rewards 9 JSij = ? i + ? marriedij + ? 2 Childrenij + ? 3High schoolij + ? 4 utterly further educationij u + ? 5 Tenureij + ? 6 attractorij + ? 7 unemployment rateij + ? 8 Noiseij + ? 9 Physical strainij + ? 10 Influenceij + ? 11High demand-low controlij + ? 12 Job auspicesij + ? 13 dataij + ? 14 Role conflict ij + ? 15Social sup port ij + ? 16 Conflict at workij + ? 17 Flexible hoursij + ? 18 Logpay ij + ? 19 Job futurei + ? 20 quotation leaderi + ? ij . The i subscript refers to different persons and j refers to different measurements for person i, Job satisfaction (JS) is the dependent variable, a the constant, b is the transmitter of the coef? ients of the explanatory variables, and eij is a random error term. Questionnaire answers on job future opportunities and recognition from leaders are only available for the 2000 cross-section. The estimation method is maximum likelihood and the statistical computer programs used were SAS 8. 2 and STATA 9. 0, the logit occasion and the clogit procedure. Results are presented as factor changes in betting odds, expressing the increase in the odds of being in the group having a high degree of job satisfaction, for a one point, or level, increase in the explanatory variable. 4. ResultsIn this section we present the empirical results based on four sets of analyses. (1) Preliminary regression analyses on gender differences. (2) Main results Estimating the luck of the outcomes being highly satis? ed with the job and being dissatis? ed with the job using conditional likelihood estimation. (3) An ordinary logistic regression analysis using only data from 2000 with addition of recognition from leaders and future job opportunities to the model. This model is used for predicting the hazard of having a high level of job satisfaction when rewards are optimized and work environment factors are at unfavourable levels. 4) A fourth part and last analysis has the purpose of validation of the question on job satisfaction and consists of a regression where job satisfaction as response variable is substituted by a question on the degree of motivation and engagement in ones work. 4. 1 Preliminary analyses on gender differences Initially, tests for gender interaction effects are performed. For private-sector employees, social support shows both a signi? slang term gender effect and a general effect on job satisfaction. For public-sector employees job warrantor indicates a signi? ant gender effect and a general effect. In both cases being a woman increases the impact on the level of job satisfaction. The gender interaction effects are veri? ed when running separate regressions on genders still using the division on sectors. The results can be seen in Appendix B. Due to the loss of observations when using ? xed effects regressions these regressions are run on only the 2000 cross-section using ordinary logistic regression on the outcome being highly satis? ed. A few results turn out to be gender speci? only for publicly employed men, having no education above high school level lowers the fortune of a high level of job satisfaction and having a leading position increases the opportunity of high job satisfaction signi? money boxly. For publicly employed women only, the unemployment rate is signi? cantly and inversely related to the level of job satisfaction. Job security is signi? cant as suggested by the found interaction effects. For privately employed men and women, gender-speci? c effects are in? uence that increases the 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 0 Lea Sell Bryan Cleal probability of high job satisfaction for men, job security that increases the likelihood of job satisfaction for women, and being exposed to aggression at the work place, which is only signi? cant for women. Moreover, the coef? cient of social support is larger for women than for men corresponding to the results of the gender analysis. In regard to wages, the effect is large and positive for both privately employed men and privately employed women but nonsigni? cant for both genders within the public sector.As discussed in the statistical analysis sec tion multicollinarity existed between education in years and wage. Therefore educational levels are entered as separate variables to the model. Ultimately, only having no further education beyond high school and having a short further education were statistically signi? cant (p < 0. 05) and these levels are therefore kept in the model. 4. 2 Results using conditional likelihood on the combined panel of data from 1995 and 2000 The gender interaction effects found and the two variables representing educational level are now entered in the ? al model. The results are shown in Table 2. The left section of the table shows the results when estimating the probability of having a high level of job satisfaction and the right section of the table shows the results when estimating the probability of having a low level of job satisfaction, the latter in order to test for a duality in the impacts on job satisfaction as discussed in Section 2. Looking ? rst at the results for the economic and de mographic measures, the odds of being in the high job satisfaction home are reduced with one-? th for every additional child for private employees, although the latter effect is only borderline signi? cant (p = 0. 077). This result is matched in the public sector, in the way that the odds of having a low level of job satisfaction triple for an additional child. For privatesector employees, having no more than a high school education, opposed to having an educational level above high school, n archaean triples the odds of being in the high job satisfaction menage and also reduces the odds of being in the low job satisfaction form, although the latter effect is only borderline signi? ant (p = 0. 063). Having a fair length or short further education nearly halves the odds of being highly satis? ed with ones job. Educational level does not show any effects of signi? cance for public-sector employees. High tenure raises the odds of being in the low job satisfaction category for publi c-sector employees, a result not matched elsewhere. Within both sectors, the level of job satisfaction seems to be related to the size of the unemployment rate, and the scope of this relation is similar for private and public employees.The sizes of the odds indicate an 8. 3 per cent drop-off in the odds of being in the high satisfaction category per per cent increase in the unemployment rate for private-sector employees and a 9 per cent decrease in the odds of being in the high satisfaction category per per cent increase in the unemployment rate for public-sector employees. In regard to occupational health factors, the public and the private sector have four factors in common role con? cts nearly halves the odds of being in the high satisfaction category in both sectors, odd work positions decrease the odds of being in the high satisfaction category for private employees by one-third, and for public employees by nearly one-half. Increasing the level of information that concerns the work place raises the odds of being highly satis? ed by 71 per cent for privately employed and by 91 per cent for publicly employed workers. For each increase in the level of social support, the odds of being highly satis? ed increase by 58 per cent and 31 per cent, respectively. For public employees, increasing the level of in? ence increases the odds of being highly satis? ed with the job by 71 per cent, and having foreseeable job security above 12 months nearly duplicate the odds of being in the high job satisfaction category. For private-sector 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 1. 061 0. 811(*) 2. 881* 0. 558* 1. 009 1. 349 0. 917* 0. 525* 0. 681* 1. 045 0. 965 1. 532 1. 709* 0. 537* 1. 576* 0. 973 1. 246* 1. 698(*) 0. 755* 0. 6741. 672 0. 6431. 023 1. 3426. 186 0. 3590. 868 0. 9781. 040 0. 7072. 573 0. 8760. 960 0. 3030. 907 0. 5140. 903 0. 8491. 292 0. 4462. 091 0. 9042. 596 1. 3802. 116 0. 3980. 23 1. 2571. 978 0. 5551. 705 1. 1031. 409 0. 9902. 913 0. 5750. 992 CI 1. 310 1. 047 0. 497 0. 796 0. 974 0. 460 0. 910* 0. 739 0. 579* 1. 710* 0. 595 2. 042(*) 1. 906* 0. 525* 1. 309* 0. 936 1. 035 1. 386 0. 150* OR ? xed 0. 6392. 682 0. 7551. 452 0. 1531. 618 0. 4311. 472 0. 9321. 019 0. 1501. 417 0. 8580. 965 0. 3461. 576 0. 3580. 935 1. 1422. 559 0. 0586. 084 0. 8914. 680 1. 3552. 681 0. 3370. 817 1. 0921. 569 0. 5521. 589 0. 8501. 260 0. 4634. 154 0. 0270. 825 CI Public (Reg. 2) 1. 379 0. 803 0. 062(*) 0. 414 1. 046 3. 378 1. 006 3. 843* 1. 238 1. 943* 4. 482* 3. 012* 2. 112* 2. 247(*) 1. 496* . 825 0. 913 1. 176 OR ? xed 0. 3605. 274 0. 3941. 639 0. 0031. 157 0. 0852. 022 0. 9511. 150 0. 32035. 729 0. 9061. 116 1. 23811. 926 0. 6532. 347 1. 1763. 212 1. 42514. 091 1. 0168. 933 1. 2223. 650 0. 9495. 320 1. 0592. 114 0. 6794. 902 0. 6411. 300 0. 2755. 038 CI Private (Reg. 3) b 0. 744 3. 396* 11. 731 2. 327 1. 195* 0. 061 1. 017 0. 358 1. 250 3. 186(*) 0. 727 0. 939 2. 052(*) 1. 152 1. 586(*) 4. 557(*) 0. 805 1. 7 66 OR ? xed 0. 1403. 948 1. 04910. 993 0. 469293. 833 0. 38314. 120 1. 0251. 395 0. 0 0. 8611. 202 0. 0462. 809 0. 5732. 724 0. 97510. 409 0. 0717. 497 0. 1276. 940 0. 964. 699 0. 3483. 819 0. 9362. 689 0. 96221. 598 0. 3721. 740 0. 16019. 521 CI Public (Reg. 4) Low job satisfactionc Dichotomous variables. Gender interaction effects Male = 1. c Scales are reversed for in? uence, job security, information, social support, and ? exible hours when estimating job dissatisfaction. CI 95% con? dence interval. Signi? cance levels(*) 0. 05 < p < 0. 10, * 0. 0000 < p < 0. 05. reduce of observations Reg. 1 = 1,200, Reg. 2 = 650, Reg. 3 = 282, Reg. 4 = 128. -log (Likelihood) Reg. 1 = 317. 1, Reg. 2 = 172. 6, Reg. 3 = 50. 8, Reg. 4 = 27. 3. Pseudo R2s Reg. 1 = 0. 24, Reg. 2 = 0. 3, Reg. 3 = 0. 48, and Reg. 4 = 0. 38. a Cohabitinga spell of children High school or lessa Short further education Job tenure in years Leader statusa Unemployment rate 1. Noisea 2. Odd work positions 3. I n? uence 4. Low controlhigh demand 5. Job security 1 yeara 6. Information 7. Role con? ictsa 8. Social support 9. expose to aggressiona 10. Flexible hours Monthly pay. Ln kr Male social supportb Male job securityb OR ? xed Private (Reg. 1) High job satisfaction Table 2. Results from conditional logistic regression, when estimating the probability of being highly satis? ed with ones job and being dissatis? ed with ones job.Divided on private-sector and public-sector employees Job Satisfaction, Work Environment, and Rewards 11 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 12 Lea Sell Bryan Cleal employees, three additional factors have signi? cant impacts on the probability of being highly satis? ed with ones job noise halves the odds of having the highest level of job satisfaction more ? exible working hours increase the odds of being highly satis? ed by 25 per cent and ? nally the odds of log pay suggest that when log pay is increased by one unit the odds o f being in the high satisfaction category increase by nearly 70 per cent.The effect is borderline signi? cant (p = 0. 054). Comparing the results from the conditional likelihood estimation with the results from the ordinary logistic regression analyses (as shown in Appendix B), a few discrepancies emerge for publicly employed men having no more than a high school education lowers the probability of a high level of job satisfaction and having a leading position raises the probability of a high level of job satisfaction using ordinary regression analysis only. In? uence raises the probability of high job satisfaction signi? cantly for privately employed men but not when using ? ed effects analyses. For public employees, being exposed to aggression at the work place lowers the probability of high job satisfaction when using ordinary logistic regression analysis and the corresponding result from the ? xed effects regression is an increase in the probability of dissatisfaction when being exposed to con? icts. For private employees odd work positions only show an effect in the ? xed effects analysis. Looking at the results of predicting being dissatis? ed with ones job several factors impact on the probability of both having a high degree of job satisfaction and being dissatis? d with the job. This is the case in the private sector for noise, information, role con? icts, and social support, and in the public sector for in? uence, information, and social support. On the other hand, being exposed to violence, threats of violence or teasing, or having a job with low control in combination with high demands only has an impact on the probability of being dissatis? ed with the job. 4. 3 Hazards and the effects of rewards on the likelihood of being highly satis? ed with the job avocation the results from the regressions presented in the previous sections, pay is only a signi? ant prognosticator of having a high level of job satisfaction in the private sector, and did not seem to have any impact on the probability of being dissatis? ed. Within both labour economic studies and work psychology, future opportunities and recognition are also considered as rewards of work. As additional information is available on future opportunities and recognition in data from 2000, the following analysis incorporates all three types of rewards. In addition, people were asked in 1995 what they considered to be the most important aspect of their work.Of the three possible answers, 11. 2 per cent answered that the pay was good (6. 0 per cent in the public sector and 14. 8 per cent in the private sector), 58. 0 per cent answered that the work interested them (65. 6 per cent in the public sector and 52. 7 per cent in the private sector), and 30. 8 per cent answered that they got along well with colleagues (28. 4 per cent in the public sector and 32. 4 per cent in the private sector). The differences among public and private employees with regard to pay support the evidenc e from our analyses.However the results also suggest that alternative rewards may be considered although the capability of these rewards to compensate for hazards in the work environment is more uncertain. The second question we have sought to investigate is whether employees exposed to hazards at work for which they receive above average rewards, when comparing with employees in non-hazardous work with average rewards, report the same level of job satisfaction. This was achieved by gist of calculations of predicted probabilities. The factors tested were signi? ant predictors of both having a high level of job satisfaction and being dissatis? ed with the 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd Job Satisfaction, Work Environment, and Rewards 13 job. Initially, a regression on the 2000 cohort integrating recognition from leaders and future possibilities in the model was performed. The results from this regression are shown in Appendix C. The hazards anal ysed for private-sector employees are high noise, low levels of information and social support, and role con? icts. For the public sector, low levels of information, in? ence, and social support are chosen. The results from varying the levels of these variables from their best, to their worst case, and at the same time maximizing the three types of rewards are shown in Table 3. The values in column 2 express the probability of being highly satis? ed with the job when each of the six chosen work environment factors are in their most positive position and all other variables are held constant at the mean. Column 3 shows the probability of being highly satis? ed with the job when each of the six hazards is at the most negative level.Columns 4, 5 and 6 give the probability of having a high level of job satisfaction when the individual factors are at the worst case, single rewards are at their best, and all other variables are at their mean. Having the lowest level of information gives t he lowest probability of having a high degree of job satisfaction observed for private-sector employees (0. 62). For public-sector employees the likelihood of being highly satis? ed with the job when information is at the lowest level is 0. 56. This is the case when all other variables are held at an average level.Moreover, the probability of being highly satis? ed with ones job never exceeds 0. 75 as long as information is low, which is below both 0. 81 and 0. 79, the average probabilities of being highly satis? ed with the job within the public and the private sector. Low in? uence predicts the lowest probability of a high level of job satisfaction for publicsector employees, which is 0. 56. In this case it is not possible to reach the same level of job satisfaction when having the lowest possible level of in? uence, as compared with those experiencing a high level of in? uence even if receiving maximum rewards.The same is evident for social support for employees in both sectors. In contrast, the impacts of high noise or experiencing role con? icts on the probability of having a high level of job satisfaction are, however, neutralized by either the highest level of leader recognition or future opportunities, or a high wage, being among the best-paid 2 per cent in the sample. 4. 4 Hazards and the effects of rewards on the likelihood of being highly motivated in the job The analysis made in Section 4. 3 is repeated now predicting the probability of having the highest level of motivation when the levels of in? ence, social support, and information are at their worst, individual rewards are at their best, and all other variables are at their mean. The results of this regression are shown in Appendix D. Table 4 is analogous with Table 3. The results in Table 4 are consistent with the results in Table 3, take out that receiving the highest level of leader recognition now seems to compensate privately employed for a low level of social support. 5. Discussion The wa y work environmental and socio-economic factors related to job satisfaction was not only in terms of either increasing job satisfaction or not, i. e. eing motivational factors or not. Thus in line with Herzberg et al. s (1959) theory some job factors also function as maintenance factors that are only being capable of making employees dissatis? ed with the job. In addition to this, some factors only had the impact of cloggy the likelihood of being highly satis? ed with the job. These could be characterized as inconvenience factors with an unsettling effect on the motivation factors. 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 0. 713 0. 618 0. 727 0. 736 0. 563 0. 754 0. 598 . 798 0. 825 0. 881 0. 812 P(High JS) when X at its worst and the rest of the factors at their means 0. 838 0. 879 0. 829 0. 520 0. 721 0. 556 0. 804 0. 727 0. 814 0. 821 P(High JS) when Pay is at maximum, X at its wor st, and the rest of the factors at the means 0. 701 0. 848 0. 730 0. 817 0. 743 0. 827 0. 834 P(High JS) when Leader Recognition high, X at its worst, and the rest at the means 0. 717 0. 858 0. 746 0. 815 0. 741 0. 825 0. 832 P(High JS) when emerging Opportunities are high, X at its worst, and the rest at their means Probability of high Job Satisfaction for private employees when all variables at their mean 0. 901. Probability of high Job Satisfaction for public employees when all variables at their mean 0. 8052. Leader recognition is at its highest when the employee has answered To a very high degree when asked Is your work acknowledged and appreciated by the caution? and future opportunities are maximized when the employee has answered To a very high degree when asked are the future prospects of your job good? . Private sector Noise Information Social support Role con? ict Public sector Information Social support In? uence P(High JS) when X is optimal and the rest of the facto rs at heir means Table 3. Probability of a high level of Job Satisfaction (JS) for varying levels of dissatisfaction factors and rewards (X) 14 Lea Sell Bryan Cleal 0. 268 0. 320 0. 338 0. 408 0. 161 0. 396 0. 467 P(High M) when X at its worst and the rest of the factors at their means 0. 474 0. 532 0. 507 0. 380 0. 453 0. 187 0. 299 0. 353 P(High M) when Pay is at maximum, X at its worst, and the rest of the factors at the means 0. 443 0. 518 0. 230 0. 414 0. 476 P(High M) when Leader Recognition high, X at its worst, and the rest at the means 0. 448 0. 523 0. 233 0. 356 0. 415 P(High M) hen Future Opportunities are high, X at its worst, and the rest at their means Notes Motivation is at its highest when the employee has answered Yes, indeed when asked Do you feel motivated and engaged in your work? 39. 2% of the private employees and 46. 3% of the public employees answer Yes, indeed. Private sector Information Social support Public sector Information Social support In? uence P(H igh M) when X is optimal and the rest of the factors at their means Table 4. Probability of a high level of motivation (M) for varying levels of dissatisfaction factors and rewards (X)Job Satisfaction, Work Environment, and Rewards 15 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 16 Lea Sell Bryan Cleal While adding to the credibility of results, many respondents unfortunately are lost when using conditional likelihood estimation as those with none changing characteristics are dropped from the analysis. When comparing the results of the ordinary regression analyses with the results using conditional likelihood estimation it did not seem that controlling for ? xed effects alters results in regard to the subjective measures used.A possible explanation is that most answers are put as frequencies of motion-picture show during working hours leaving less room for misconceptions of the questions. About two-thirds of the results on work environment variables were common for public- and private-sector employees, with effects of just about the same size. Common factors were odd work positions and role con? ict, both factors lowering the probability of having a high level of job satisfaction, and information on decisions that concerns the work place and social support, of which higher levels predicted being highly satis? d with the job and lower levels predicted job dissatisfaction. Factors being speci? c for the private sector were noise and a combination of low control and high demands, whereas picture to aggression at the work place and level of in? uence only seemed to have an effect on public employees. Being exposed to violence, threats of violence or teasing, and having a job with low control in combination with high demands are examples of maintenance factors as the extent of their impact is con? ned to negative outcomes.In accordance with our results, public employees have been shown to have an increased risk of experiencing con? icts , teasing, or threats of violence at work (see Hoegh, 2005) whereas jobs with low control and high demands are typically found on industrial work sites within the private sector. In testing the ameliorative capability of rewards to compensate for the negative effects on job satisfaction deriving from exposure to (primarily psychosocial) hazards in the work environment, our results indicated only a limited effect for this type of compensating differential.In particular, rewards could not neutralize the effects on job satisfaction when employees have low levels of information on decisions that concerns the work place, social support, or, as a result for public employees only, in? uence. Most previous studies searching for evidence of compensating wage differentials for work environment hazards have been concerned with observable occupational health hazards (see Rosen, 1986), an exception being for very stressful work (French and Dunlap, 1998). The results were duplicated and even more pronounced when the analysis was repeated substituting job satisfaction with motivation.Where the same member of public employees and private employees reported being highly satis? ed with the job, there was a discrepancy among the two sectors when comparing the fraction of employees reporting to be highly motivated. Thirty-nine per cent of the private employees and 46 per cent of the public employees reported to be the highly motivated. These results also correspond to the result that more public than private employees report that the most important aspect of their work was that the work interested them (66 per cent versus 53 per cent).The differences are small but the results support the theory that public employees should have higher intrinsic motivation (Benabou and Tirole, 2006). As wages did not show any signi? cant impact on the level of job satisfaction for public employees and neither had any signi? cant compensating value in regard to certain hazards at the job, the resu lts also point to that publicly employed workers are less motivated by high pay and place a higher value on the intrinsic rewards as also seen in Karl and Sutton (1998) and Houston (2000).Very low probabilities of having a high level of job satisfaction (0. 56) and being highly motivated at the job (0. 16) were evident for public employees with the lowest level of in? uence. This clearly suggests that lack of in? uence can demotivate public employees and points to that 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd Job Satisfaction, Work Environment, and Rewards 17 intrinsic motivation can be undermined if people feel controlled, and have little autonomy and freedom in performing work tasks (Deci and Ryan, 1985).Moreover, in the long run, lack of autonomy can pose a threat to value congruence between the employees and the organization, as suggested by Ren (2010). In regard to the results concerning gender differences, job security showed a general positive eff ect on job satisfaction as well as a gender-speci? c effect for employees in the public sector, suggesting women pursue job security more than men. For private employees, any effect of job insecurity would be dissatisfaction with the job and the size of the effect was just about the same for the two genders.In a study by DAddio et al. (2003), job security was found to have the same effect for men and women after adjusting for ? xed effects. Without adjusting for ? xed effects, men seemingly valued job security the most. In the study by Clark et al. (1998), they ? nd that the extent to which women or men pursue job security varies among countries and that the differences are relatively small. These other studies have split the analyses on gender, which complicates comparison, and the differing time span of years over which the observations are made most ikely has an effect too. Clark et al. (1998) also ? nd that women report having good relations at work more often than men. Whereas Sloane and Williams (2000) ? nd that good social relations are most important for women. This is consistent with our ? nding that among private employees, women value social support more than men. The impact on job satisfaction from wages may also re? ect an effect of satisfaction with the job that derives from increased total expenditure opportunities as the question on job satisfaction in our study is one that re? cts overall job satisfaction. The results may also be dependent on the given wage structure as both wages and wages scatter are lower within the public sector than within the private sector in Denmark at the time (Wadensjo, 1996). Finally, the impact on job satisfaction from the unemployment rate is large. DAddio et al. (2003) found a similar negative correlation between job satisfaction and the rate of unemployment. In both the study by DAddio et al. (2003) and our study, this relation is only signi? cant after controlling for ? xed effects.That is, apart from the res ult when making a separate analysis on gender and sector. It is noteworthy that the unemployment rate has these clear derived effects on the subjective feelings towards the job. According to the studies by Akerlof et al. (1988), a low unemployment rate makes it possible for unsatis? ed employees to change to jobs with more desired characteristics. Appendix A hark of work environment variables 1. Noise Two levels according to answer to the below 3/4 or more of the work day being exposed to noise that high that one must raise the voice to be able to speak with others. . Odd work positions A pull in with a one point increase when respondents have marked a positive answer to the following questions 3/4 or more of the working hours the work entails work with 1. The back heavily bended forward with no support for hands or arms. 2. The body twisted or bended in the same way several times an hour. 3. The hands lifted to shoulder height or higher. 4. The neck heavily bended forward. 5. Sq uatting or kneeling. 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd 18 Lea Sell Bryan Cleal 3. In? uence Four levels Can you plan your own work? 4.Low controlhigh strain In? uence Four levels Can you plan your own work? Job variation Four levels Is your job varied? Time pressure Recoded into two levels 1995 Does your work entail that you have to work under time pressure in order to get certain pieces of work do? 2000 Is it necessary to work very fast? Mental demands Does your work demand all your attention and concentration? 5. Job security Two levels according to (1995) Certain or pretty sure of keeping the job the close 12 months. (2000) The present job is not a ? xed-term appointment with less than 12 months left. . Information Four levels Are you informed about decisions that concern your work place? 7. Unclearness of role and con? icting demands Two levels according to the comply or not of either of two statements It is clear what my responsibility. I experience con? icting demands in my work. 8. Social support (four levels No support, always support from colleagues but not always from superiors, always support from superiors but not always from colleagues, always support from colleagues and superiors) 1995 Do you receive help and encouragement from your superior/colleagues? 000 How often do you receive help and support from superior or colleagues? 9. Con? icts, teasing, unwanted sexual attention, threats, or violence (two levels) 1995 Are you exposed to any form of unpleasant teasing, unwanted sexual attention, threats of violence, or violence at your work place? (Not reporting any incidents constitutes a no) 2000 Have you been exposed to unpleasant teasing, unwanted sexual attention, threats of violence, or physical violence at your work place within the last 12 months? (Not reporting any incidents constitutes a no) 10.Flexibility of work memorial Four levels according to the time space within a respondent can vary the daily working schedule without giving further notice. Can you change the placing of your working hours from day to day without making prearrangements, e. g. check at work late or leave work early? 11. Recognition Four levels Is your work acknowledged and appreciated by the counsel? 12. Future opportunities Four levels Are the future prospects of your work good? 2011 CEIS, Fondazione Giacomo Brodolini and Blackwell Publishing Ltd Job Satisfaction, Work Environment, and Rewards 19Appendix B Estimating high job satisfaction on the 2000 cross-sectional data. Divided on gender Men Private (Reg. 1) Women Public (Reg. 2) Private (Reg. 3) Public (Reg. 4) Coef. Cohabitinga Number of children High school or lessa Short further education Job tenure in years Leader statusa Unemployment rate 1. Noisea 2. Odd work positions 3. In? uence 4. Low controlhigh demand 5. Job security 1 yeara 6. Information 7. Role con? ictsa 8. Social support 9. Exposed to aggressiona 10. Flexible hours Monthly pay. Ln k r ideal error Coef. Standard error Coef. Standard error Coef. Standard error 0. 258 -0. 067 0. 237 0. 437* 0. 010 0. 181 -0. 011 -0. 587* -0. 176 0. 244* -0. 658 0. 087 0. 475* -0. 626* 0. 371* -0. 294 0. 175* 0. 639* 0. 1896 0. 0728 0. 1999 0. 1916 0. 0086 0. 2451 0. 0295 0. 2142 0. 1062 0. 0906 0. 4282 0. 3488 0. 0926 0. 1441 0. 0678 0. 2811 0. 0528 0. 2705 0. 1700 -0. 078 -0. 638* 0. 060 -0. 004 0. 743* 0. 010 0. 104 -0. 493* 0. 395* -0. 919 -0. 292 0. 759* -0. 578* 0. 314* -0. 732* 0. 143 0. 066 0. 2248 0. 0860 0. 2592 0. 1890 0. 0093 0. 3642 0. 0223 0. 2853 0. 1827 0. 1190 0. 6509 0. 2948 0. 1206 0. 1632 0. 0742 0. 1916 0. 0598 0. 2794 0. 307 -0. 026 0. 286 -0. 481* 0. 006 0. 348 0. 021 -0. 529(*) -0. 26 0. 121 -0. 991 0. 469* 0. 607* -0. 435* 0. 459 -0. 348* 0. 171* 0. 611* 0. 2327 0. 1004 0. 2709 0. 2177 0. 0109 0. 5299 0. 0369 0. 2808 0. 1476 0. 1252 0. 5711 0. 3909 0. 1390 0. 2099 0. 0856 0. 2922 0. 0685 0. 2802 0. 167 -0. 015 -0. 117 -0. 1656 0. 010 -0. 267 -0. 029* -0. 0 44 -0. 380* 0. 247* -0. 003 0. 369* 0. 623* -0. 542* 0. 362* -0. 335* 0. 104* -0. 092 0. 1477 0. 0612 0. 1804 0. 1349 0. 0070 0. 2914 0. 0139 0. 1963 0. 1126 0. 0880 0. 5338 0. 1888 0. 0896 0. 1212 0. 0508 0. 1397 0. 0471 0. 2195 a Dichotomous variables. CI 95% con? dence interval. Signi? cance levels (*) 0. 05 < p < 0. 10, * 0. 000 < p < 0. 05. Number of observations Reg. 1 = 1,356, Reg. 2 = 959, Reg. 3 = 728, Reg. 4 = 1,754. -log (Likelihood) Reg. 1 = 639. 3, Reg. 2 = 483. 2, Reg. 3 = 363. 1, Reg. 4 = 907. 1. Pseudo R2s Reg. 1 = 0. 17, Reg. 2 = 0. 18, Reg. 3 = 0. 17, and Reg. 4 = 0. 13. Appendix C Estimating high job satisfaction on the 2000 cross-sectional data (Reg. 1) (Reg. 2) Private (N = 2,057) Public (N = 1,296) OR Cohabitinga Number of children High school or lessa Short further education Job tenure in years Leader statusa Unemployment rate 1. Noisea 2. Odd work positions 3. In? uence 4. Low controlhigh strain . Job security 1 yeara 6. Information P>z CI lower CI higher OR P>z CI lower CI higher 1. 358 0. 934 1. 361 0. 653 1. 016 1. 252 1. 006 0. 628 0. 845 1. 121 0. 464 1. 186 1. 430 0. 042 0. 263 0. 064 0. 004 0. 024 0. 323 0. 796 0. 008 0. 058 0. 139 0. 033 0. 535 0. 000 1. 011 0. 829 0. 982 0. 488 1. 002 0. 802