Furthermore, values between 0. The Content validity ensures that the measure value of Chi — Square After includes an adequate and representative set of items examining the reliability and validity of the scale and that tap the domain of the concept Malhothra , testing appropriateness of data as above, Suitability of S, Warnakulasuriya, Communalities indicate the amount of variance b Factor analysis in each variable that is accounted for Table No - Factor analysis method has been employed to 03shows that initial communalities and extraction identify the dimension importance underlying communalities.
Initial communalities are estimates of the dimensions of job satisfaction of banking sector variance in each variable accounted for by all employees. Initial communalities are set as 1. Extraction communalities are Kasier — Meyer — OlKin KMO test assist to estimates of variance in each variable accounted for by measure sample adequacy.
The KMO statistic varies the factors in the solution. Accordingly, all items are fit between 0 and 1. A value close to 1 indicates that to the factor solution. Because, extraction value is more patterns of correlation are relatively compact and so than 0. This is a somewhat arbitrary procedure. Its In this test, eigen values are plotted against the application to this data set led to the conclusion that the factors arranged in descending order along the X- axis.
The number of factors that correspond to the point at Table 5 : Rotated Component Matrix Table No - 05 show that factors were divided significant Pal, ; Pal and Bagi, ; Hari, into the three groups. Each of three job satisfaction Anderson, Tatham, and Black, The higher a factors listed in table no is labelled according to the factor loading, the more would its test reflect or measure name of the value that loaded most highly for those job as job satisfaction Pallant, Actually in this study, satisfaction.
It is worth declaring out here that factor minimum factor component loadings of 0. The job 0. The rotated Varimax the title of each factor of job satisfaction. For parsimony, only those like factors with loadings above 0.
Happy to work Group- II Achievement consists of three factors Payment Promotion such as achievement, Appreciation and participation of Subordinate- decision making with loadings ranging from 0. Achievement Achievement Following table clearly exposed the Appreciation determinants factors and these groups Participation in decision making Proud to work proud to work Enough description V.
Conclusion Results of the study confirmed that job satisfaction can be determined by ten variable such as payment, happy to work, promotion, subordinate supervisor relationship, direction of supervisor, achievement, appreciation, participation in decision making, proud to work and enough description.
Journal of Occupational were divided into three groups. Groups -1 consist of five and Organizational Psychology, 67, — This group is called as Payment group. Group — Judge, T. This group is named as G. The job satisfaction-job performance Achievement. Final group consists of two factors. They relationship: A qualitative and quantitative review. Psychological Bulletin, , — Jurgensen, C. Journal of Applied 1.
Ali, N. Kohler, S. Year 2. Bracken, D. Benchmarking employee attitudes. Kovach, K. Employee motivation: 3. Brayfield, A. Journal of Applied Psychology, 35, performance. Employment Relations Today, 22, 93— 28 2 — Brief, A. Attitudes in and around Locke, E,A What is job satisfaction? Thousand Oaks, CA: Sage.
Cascio, W. Managing human resources: 4, — Productivity, quality of work life, profits. New York: Oxfordn university press, P Ramayah, T. McGraw Hill lnc. Cooper, C. Smith, P. Journal of Occupational Psychology, Vol. Chicago: Rand McNally. Drever, J. A dictionary of psychology.
Penguin Staw, B. Stability in the midst Books, Middlesex. Emsley, David Thiruchelvam s j and Velnampy t , Vol. Organizational environment and employee 9. Hellriegel, D. Velnampy T. Job Attitude and Employees Vol. Performance of Public Sector Organizations in Herzberg, F. Job Attitudes: Review of Management, Vol.
Research and Opinion. Pittsburgh: Psychological Iaffaldano, M. For comparison and testing purposes a random effects ordered probit is also estimated. We find that using the fixed effects approach that clearly rejects the random effects specification , considerably reduces the number of key explanatory variables. In addition to wages, good health and being a public sector employee are particularly important in explaining individual differences in job satisfaction.
Moreover, the impact of being employed on a temporary contracts or working in the public sector differs between the genders. Work psychologists have for a long time been arguing that for most people jobs cannot only be characterised by the pay and hours of work associated with them, as standard economic analysis does, but also by job and workplace features like promotion and other career prospects, job security, job content and interpersonal relationships; see Warr for a comprehensive survey.
In fact, when the employees are asked, as in e. Economists talk about utility all the time, but do not seem to have hope of ever observing it this side of heaven.
In micro-economic theory, almost every model is built on utility functions of some kind. In empirical work little attempt is made to measure this all- pervasive concept. The concept is considered to be so esoteric as to defy direct measurement by mortals. Still, in a different role, viz. Rather, as is eloquently discussed in Lazear , economic analysis is concerned with the study of situations where the intrinsic motives are taken as given and the aim of the analysis is to examine the influence of extrinsic motives and especially the trade-offs economic agents face at the margin.
A substantial body of research has been built showing that job satisfaction is strongly correlated with several mental physical health indicators. In parallel, a growing number of studies focusing on life satisfaction or financial situation have appeared in economics; see e. Frijters et al. It should be noted that the concept of job satisfaction has certainly not been a cornerstone in the economic analysis of the labour market. Rather, in many analyses job satisfaction is more or less absent.
Nevertheless, in many countries firms and employers pay close attention to the subjective well-being of their employees and to how these perceive their current jobs. Job quality is certainly a multi-dimensional concept. Related to this, some authors distinguish between the economic contract and the psychological contract.
In the economic contract the focus is on the relationship between effort and reward, while in the psychological contract the interest is mainly in the working conditions. Because of this multi- dimensionality, the possibility of using one dimension to classify jobs according to their quality is often rejected. A similar approach has been taken by the EU. We therefore think that job satisfaction can be considered, at least to some extent, a good proxy for job quality.
Earlier research has typically found that consistent with economic theory, pay and work hours are positively and negatively, respectively related to job satisfaction. These studies have, however, been mainly based on cross-sectional data, and moreover to a surprisingly high extent on a single data set — the British Household Panel Survey BHPS. In that case, the statistical model predominantly used has been the ordered probit with random effects see Butler and Moffitt ; Frechette, Denmark has been shown to have among the most satisfied workers in the world; in Europe only Austria and Ireland reach similar levels see Blanchflower and Oswald undated and Sousa-Poza and Sousa-Poza, The same trend is confirmed by charts 1 and 2 that report job-satisfaction over the Member States countries for years and The approach in our paper differs from that in previous analyses in that we make use of longitudinal job satisfaction data while at the same time preserving the ordered nature of the information in the fixed effects approach.
For comparison and testing purposes we also estimate a random effects ordered probit model. We therefore attempt to make use of the panel element of the ECHP to deal with the problem of unobserved heterogeneity.
That is, certain reported levels of overall individual job satisfaction may be recorded because underlying unobservable characteristics, which vary across individuals, may increase the probability that a certain level of job satisfaction is reported as opposed to another. The remainder of the paper is organised as follows. Section two provides a brief review of the economic literature on the topic of the paper. Section three outlines the data used and section four discusses the empirical strategy adopted.
Section five gives the results. The sixth section summarises our conclusions. This literature differs in at least three respects from how economists have approached essentially the same data sets and closely related questions. First, the dependent variable in analyses aiming at understanding the factors underlying differences in job satisfaction across individuals has usually been constructed by averaging the ordinal responses to the questions concerning satisfaction.
Thus, satisfaction is implicitly assumed to be cardinal. This follows naturally from the implicit assumption that the job satisfaction responses are cardinal. Moreover, they do not account for the fact that the dependent variable is bounded. Typically, the literature has little discussion both of measurement errors in the dependent variable and of what is subsumed in the error term.
Most of the psychological as well as the economic research have been based on cross-sections. As a consequence, little attention has been paid to the importance of individual differences in baseline job satisfaction levels, which in a longitudinal framework could be modelled as individual-specific fixed effects. Clark and Oswald discuss the role of fixed effects in studies of well-being. Other factors that can affect self-reported job satisfaction measures and which should be taken into consideration in thinking about what goes into the error term include circumstances — local or of business cycle type — and aspirations.
The latter may for instance be cohort- specific. For an analysis of business cycle influences, see Gerlach and Stephan A previous study recognising the potential importance of controlling for fixed effects is Winkelmann and Winkelmann , in which the effects of changes in labour force status on life satisfaction are examined. The same approach is adopted by Hamermesh Finally, as pointed out in Clark and Oswald , unlike economists, scholars in psychology and related fields have not entered working hours as an explanatory variable.
In view of the fact that there are many surveys available that contain job satisfaction questions3 which furthermore are quite similar across surveys , the data of which are frequently used by economists, surprisingly few economic studies have been carried out. Next, we give a brief review of the work carried out by economists in the area.
Some of the key characteristics of the studies surveyed are collected in Table 1. The early contributions to the economic job satisfaction literature are from the late seventies. Hamermesh is to the best of our knowledge the first to develop and test a theory of overall job satisfaction, whereas Freeman and Borjas examined the relationship between unionism and job satisfaction where the latter is adopted with the motivation that it is a measure that captures other aspects of the workplace, which are not reflected by conventional objective variables.
This line of research of the effects of unionism has been picked up in several later studies see e. The s witnessed a renewed interest in job satisfaction research among economists spawned by a series of papers by in particular, Clark and Oswald.
Clark makes use of three different measures of job satisfaction obtained from the first wave of the British Household Panel Survey BHPS and carries out an ordered probit analysis of the importance of individual and workplace characteristics in explaining reported differences.
The satisfaction model is estimated by ordered probit for two dependent variables: overall job satisfaction and pay satisfaction. The focus in their paper is on relative versus absolute income as a determinant of job satisfaction and in the case of the former on different comparison groups. Specifications with relative income clearly outperform those with absolute income as a regressor.
They use the same variables as in Clark — obtaining similar results — plus two new ones: dummies for whether the employee is employed on a temporary contract and for whether she is in a managerial or supervisory position.
The dummies attach negative and positive coefficients, respectively. A third paper published the same year, using the same data set and methods is Clark, Oswald and Warr , which contains a more detailed analysis of the relationship between the age of the employee and her job satisfaction level.
The relationship is found to be U-shaped, reaching its lowest level at about the age of forty. Estimation is mainly by ordered probit, but the robustness of the results is checked by principal components, too.
See also Sloane and Williams, for similar conclusions from a study based on data from the British academic labour market5. His conclusions is supported by Sanz de Galdeano , who uses the waves of the BHPS and employs both Heckman selection models and propensity score methods in order to correct for differences in personal and job characteristics by gender as well as for potential sample selection problems. All four above mentioned studies are based on a single cross-section from the BHPS.
A more recent study by Gardner and Oswald makes use of the panel data for the years that can be obtained from the BHPS. The aim of their analysis is to explain the behaviour of two dependent variables: one is the GHQ12, which is a widely used measure of subjective well-being with a considerable weight put on mental health and the other is a simple index running from 1 to 6 based on answers to the question about overall job satisfaction.
The GHQ12 scores are analysed by OLS regression, whereas the job satisfaction data are once again modelled by ordered probit as a matter of fact by ordered logit in the update. Only the analysis of the GHQ12 makes use of the longitudinal character of the data by including individual fixed effects.
In the job satisfaction analysis where the dependent variable is ordered, the data are treated as annual cross-sections. The relationship is U-shaped with respect to age and job tenure. The authors also find a discernable negative time trend in job satisfaction, which is particularly pronounced among public sector employees.
This is quite remarkable considering the fact that the nineties was a period of strong economic growth. Two recent papers from the UK have utilized different data sources. In a study on self-employment, Blanchflower and Oswald estimate ordered probit models on two samples; one from the National Child Development Study when the respondents were 23 years of age, and the other from the same source ten years later.
They find that females, married and part-time workers were more satisfied than males, non-married and full-time employees. Union membership changed from having a negative impact in to a positive one in In both observation years the self-employed are observed to be significantly more satisfied than wage earners. Lydon and Chevalier examine two cohorts from and of graduates from UK higher education institutions using data from the Higher Education Funding Council for England Survey carried out in At the time of the survey cohort members were on average and 31 years of age, respectively.
The key question addressed in the paper is the effect of the potential endogeneity of wages in job satisfaction studies. In effect, the authors find that the direct wage effect is doubled once endogeneity is controlled for. All of the more recent studies of the factors underlying differences in individual job satisfaction we have discussed so far have used data from Britain.
This is no coincidence as the bulk of research has been carried out in that country. Outside Britain, investigations on the topic have been rather thin on the ground. For the U. For both countries Hamermesh only enters wage variables as explanatory variables in his logit models. This battery of questions is identical to that in the much used BHPS. The model is estimated under the assumptions that the effects of the individual and job characteristics are the same across countries, whereas there may be differences between countries and years.
The two differences are picked up by year and country dummies. The results are quite similar to those reached by the studies on the BHPS.
The same applies also for the results when the model is estimated separately for men and women. Kaiser estimates probit models for high versus low satisfaction on both pooled and individual country data. He finds that in most of the countries under study fixed-term contracts are associated with lower reported job satisfaction levels.
Satisfaction levels appear to differ little between employees working part- and full- time. Of course, both findings beg the question whether these features of the employment contracts are exogenous. In fact, the level is found to be lower and statistically significantly so for the two latter countries.
For comparison purposes and as a robustness check we also use the estimator proposed by Das and Van Soest Symmetrically assuming a logistic distribution leads to the ordered logit model Both these models have been often used with cross-sectional data in analyses about well-being and satisfaction. The ordered probit model has also been used in longitudinal studies.
In that case unobserved heterogeneity has been dealt within the random effects approach. The fixed effects approach has been rarely followed owing to the lack of suitable econometric methods. However, some authors have adopted it by transforming the ordinal variable into a binary one that takes the value of one above or under a specific threshold.
Assuming that the error term is logistic yields the ordinary logit model that can be estimated using standard likelihood methods see Andersen, ; Rasch, and Chamberlain, Two recent studies have proposed new estimations methods that can handle the original rankings of the dependent variable in the fixed effects approach.
Those models have the particularly appealing property that no particular correlation is assumed between the fixed individual effects and the error term. However, while the former are related to observed factors, the latter are not, i. Several studies have shown that reported general satisfaction levels are likely to perform well in predicting the underlying concept of welfare. This in turn implies that self-reported job satisfaction levels can be used as proxies for the well-being in the job sphere.
The second assumption is more strictly related to the statistical properties of our model. Through it, we assume that all relevant time-varying factors are observed and the remaining fixed unobserved factors affect the levels of other variables and not their changes. More precisely, we observe a sample of Danish workers over the years In addition to their self-reported job satisfaction levels, the data set includes many individual and job-related characteristics for each survey year, some of which will be used as explanatory variables in our analysis.
Individual heterogeneity is unobserved; therefore to obtain the unconditional log- likelihood we need to integrate the conditional log-likelihood. The integration is done with the Gauss- Hermite quadrature 25 points were chosen ; see Frechette, a , b , Butler and Moffit , and Greene In that case a spurious correlation between that dimension and those unobserved characteristics may arise and thereby bias the estimated coefficients.
While the random effects ordered probit can to a certain extent indicate the direction of the effects of some determinants of job satisfaction, the above-mentioned spurious correlation is most likely to be present.
In that case a fixed effects approach seems to be more appropriate. The model assumes that the intercepts are increasing i. However, it does not assume ordinal comparability. The model is estimated by maximum likelihood. Their estimator is based on a weighted average of the Chamberlain estimator for each k. This implies that when there is not enough variation over the categories, those thresholds cannot be used, and the corresponding categories should be dropped. This happens in our estimation for the low satisfaction values reported by men only values higher than 2 could be used for them.
Moreover, it should be noted that this estimator requires stricter regularity conditions than the one examined in the previous section. The Das and van Soest estimator is based on the following procedure. In the absence of effects related to fixed individual characteristics, we expect that the coefficients should be very similar. Conversely, when using the fixed-effects estimator only a sub-sample of individuals is used. Thereby, these two models do not share the same normalization.
See Frijters et al. As has been pointed out by Frijters et al. First, one needs to re-estimate the model to re-fit the unrestricted parameters of the model. Consequently, rejecting the null at the lower bound implies that the true statistic will also reject it. Thanks to the panel character of the data, they provide unique information about the dynamics of social change and individual behaviour. Concerning non-response and attrition the reader is referred to the paper by Nicoletti and Peracchi In general the non-response rates in the satisfaction question are found to be very low.
The job satisfaction questions in the ECHP ask the individuals to give an integer response on a scale from 1 to 6 which best describes how satisfied or dissatisfied they are with specific job facets: wages, job security, type of work, working hours, working time, working conditions and commuting time.
By selecting only people employed in the survey year, we can see their reported satisfaction level as an overall evaluation of the satisfaction in multiple job spheres. We will focus on that in our analysis.
A response of 1 represents the lowest level of satisfaction and 6 the highest. Table 2 and Chart 3 show the means for our dependent variable for men and women, separately. We may note that the means are remarkably constant during the five year-period under study. Furthermore, the scores are high: close to or slightly below 5. In each year, only about 38 35 per cent of the male female respondents report the same levels as in the preceding year.
Though the period considered— 5 years — is too short to shed light on whether changes are temporary blips or dips , or of a more permanent character, this result is especially noteworthy in view of the fact that our subsequent econometric analysis exploits the longitudinal aspects of the data. Although there is no trend in the average satisfaction levels, there is for each year a larger proportion of individuals whose satisfaction level is higher than in the previous year than there are employees whose satisfaction level has decreased compared to the year before.
Finally, we can se the patterns are very similar for the male and female employees. These show indeed that they are positively correlated, but the correlations are far from perfect. The trend of the two structural indicators over the observation period along with average satisfaction levels for men and women respectively are reported in Chart 4. Only intrinsically12 time-varying variables have been introduced in the fixed effects ordered logit. Before plunging in on the empirical estimates we refer to Tables 4 and 5 which give some descriptive information about the data used.
0コメント